CN100499572C - Method and apparatus for packet scheduling - Google Patents

Abstract

Packet scheduling apparatus schedules packets of data for transmission from a transmitter such as a Note B to a plurality of receivers such as a plurality of UEs via at least one channel, for example in a high-speed downlink packet access system (HSDPA). A first measures producing unit produces first measures of scheduling performance in at least two different aspects. A decision unit employs a weighted combination of the first measures to decide the receiver(s) to which packets are to be transmitted. The weighted combination is produced by combining the first measures according to respective corresponding weights. A second measures producing unit produces at least one second measure of scheduling performance. A classifying unit classifies the weights for the or each second measure into at least two different classes of weights according to a probable influence the weight is expected to have on the second measure concerned. A weight adapting unit employs the or each second measure, together with the classification of the weights for the or each second measure, to adapt the weights.

Description

The method and apparatus of packet scheduling

Technical field

The method and apparatus of the packet scheduling that the present invention relates to for example in wireless communication system, use.

Background technology

Fig. 1 shows the part of wireless communication system 1.This system comprises a plurality of base stations 2, only shows one among Fig. 1.The sub-district is served in base station 2, has a plurality of isolated users in the sub-district.Each user has independent user equipment (UE).In Fig. 1, only show user equipment (UE) 2, UE11 and UE50.Each UE for example is portable terminal (mobile phone) or portable computer.

As everyone knows, in code division multiple access (CDMA) system, use different channel code (channelisation code) to distinguish (to be also referred to as " Node B ") signal that sends to different UEs from the base station.In so-called third generation wireless communication system, high-speed downlink packet access (HSDPA) technology has been proposed, be used in down link direction (from the base station to UE) transmission data.In this technology, a plurality of channels can be used for sending data.These channels have different channel codes.For example, in the sector of given sub-district or sub-district, may exist 10 different channel C1 and can be used for HSDPA to C10.In HSDPA, downlink transmission is divided into a series of Transmission Time Intervals (TTI), on each different available channel, data packet transmission is arrived selected UE.Which which can in each TTI, carry out by the new selection of channel for UE service.

Fig. 2 shows the HSDPA technology in the operation example of a series of Transmission Time Interval TTI1 to the TTI9.As shown in Figure 2, in TTI1, determine and to send two data groupings to UE50, send four data groupings, send four data groupings to UE2 to UE11.Thereby, be that UE50 distributes two channels, for UE11 and UE2 respectively distribute four channels.Therefore, as shown in Figure 1, to UE50 allocated channel C1 and C2, to UE11 allocated channel C3 to C6, to UE2 allocated channel C7 to C10.

At next Transmission Time Interval TTI2, send a grouping to new user equipment (UE) 1, all the other UE of appointment continue to receive grouping among the TTI1.

Therefore, the HSDPA system effectively utilizes a plurality of parallel shared channels data are sent to different UE with the form of dividing into groups from the base station.For example wish to be used to support world wide web (www) to browse this system.

In order to determine in each TTI, should to have adopted the packet scheduling technology for which UE service on which channel.Routinely, the dispatching technique of having considered two kinds of fundamental types is to be used for HSDPA: repeating query (RR) is dispatching technique and maximum Carrier interference ratio (maximum C/I) technology (round-robin).

Basic round-robin technique at first compiles the tabulation that has data UE to be sent such as to locate at transmitter (base station) at present.For each TTI, last UE has the limit priority of next TTI in the tabulation.Therefore, serve UE in the repeating query mode.In the simplest Round Robin technology, the UE that takes to have limit priority uses the mode of all channels.Yet the round-robin technique of packet-weighted also is known.Its relative data amount according to different UEs is distributed available channel to one group of user in each TTI.In this packet-weighted technology, the UE with more data to be sent is distributed more channel.

Aspect allocation of radio resources, the fairness between the UE that the Round Robin technology is emphasized to compete.Yet the data total throughout that this technology provides is often relatively poor relatively.

Except that the Carrier interference ratio of reporting according to each UE in each TTI (C/I) sorted to the UE tabulation of pending datas such as having, maximum C/I dispatching technique and Round Robin technology type were seemingly.C/I is the index (measure) of channel quality.By UE tabulation being sorted according to C/I, to channel quality preferably UE give higher selected chance.In the simple version of this technology, all channel allocation are given UE with pending datas such as having of the highest C/I.Also can adopt the packet-weighted modification, in this packet-weighted modification, not to select to have the single UE of the highest C/I, but select one group of UE, and in this group UE, cut apart available channel according to the relative quantity of the data to be sent that these UE had with the highest C/I value.

Maximum C/I dispatching technique is the total throughout of maximum data often, but this is cost with the fairness.As can be seen, for example because apart from the base station far away or because near have many other to disturb UE and reported that the UE of poor C/I value only has very little selecteed chance.Therefore, these UE suffer unacceptable long-time delay possibly when receiving grouping.

At " A proposal of all-IP mobile wireless network architecture-QoS packet scheduler for base stations ", Masahiro Ono etc., NECCorporat ion, Technical report of IEICE, MoMuC2002-3 (2002-05) discloses another kind of dispatching technique among the pp.13-18.Thereby the target of this packet scheduling technology is to require and maximized system capacity by the Quality of Service that foundation radio link quality and required QoS rank distribute time slot to satisfy different UEs adaptively.The hierarchy that this technology adopts a plurality of different independent schedulers to constitute comprises maximum C/I scheduler, equitable proportion (PF) scheduler, weighted round ring (WRR) scheduler and priority repeating query (PRR) scheduler.Require the grouping that will dispatch is presorted by the different QoS of grader according to different UEs.The grouping that will belong to a different category subsequently is applied to the different schedulers in the first order scheduler in the hierarchy.Operation dispatching in the second level of this hierarchy and the third level in addition.In this way, use a plurality of independently schedulers to satisfy the different requirements of different business classification.Yet, because dynamically fast the changing of wireless channel, with fixed form or with the dynamical fashion of slow change scheduling is handled to divide being proved to be difficulty and poor efficiency.The possibility of result is that the scheduler that is assigned to a part (class of service) will bear very big pressure required QoS rank is provided, and it is under-utilized and have an idle capacity to be assigned to the scheduler of another part (class of service).The efficient of also finding the scheduler that these are divided out obviously worsens along with the increase of the shared bandwidth amount and the scope of business with different demands with performance.In practice, recover in this subregion scheduler some the loss efficient trial just increased computation complexity and supplementary costs.

Require among the UK Patent Application No.0303859.3 of its priority in the application, the inventor has proposed a kind of grouping scheduling method, and this grouping scheduling method for example is used in the HSDPA of cordless communication network system dispatching through the packet that at least one channel sends to a plurality of receivers from transmitter.At least two different aspects to scheduling performance (for example quality of service, delay and fairness) assign weight.For each receiver independently,, generate scheduling performance at least one combined index at different aspect according to the weight of being distributed.Employing is used for each combined index of different receivers and determines the receiver that grouping will send to.The PCT that submits on the same day with the application applies for that [acting on behalf of reference signs: P84315PC00] is corresponding to GB 0303859.3.Here the full content of having incorporated GB 0303859.3 and this PCT application by reference into, and submit the copy of these applications to the application.

This grouping scheduling method makes can carry out data dispatch as single group with all receivers, and need not receiver is divided into different scheduling classification, for example different service types.This method also makes this scheduling can need not very high computation complexity promptly can consider some different aspect of performances.

In the method, consider that desired traffic carrying capacity sight comes weight assignment.For example, carried out various simulations, each the simulation respectively with the analog service amount environmental correclation that wherein always has 40 UE.Suppose that 10 UE among these UE attempt to receive real-time video, suppose that all the other 30 UE attempt to carry out the WWW browsing session.In a simulation, suppose that the total of traffic load goes up basicly stable in the sequence (10 seconds time periods) that 5,000 TTI form.In another simulation, suppose that WWW user is activated with interlace mode, thereby on scheduler, cause variable input load.In the 3rd simulation, supposing provides mechanism at a slow speed by the HSDPA system for WWW browses.This at a slow speed mechanism prevent from before fully successfully downloading last session, to download new session.

Find that from these simulations for different traffic carrying capacity sights, a plurality of weights that must be applied to the different aspect of scheduling performance need obviously different.In view of the above, wish that the operator of scheduler can adjust weight, thereby when desired traffic carrying capacity sight changed, weight can change also.In UK Patent Application No.0303859.3, also wish between the device on-stream period, can adjust weight automatically.

At S.Abedi, S.Vadgama, " Hybrid Genetic Packet Scheduling andradio Resource Management for High Speed Downl ink Packet Access ", WPMC 2002 Conference, Hawaii, among the pp.1192-1196, the inventor has also proposed to adopt the packet scheduling technology of genetic algorithm.In this technology, for each TTI generates a plurality of candidate's scheduling schemes.Each candidate scheme is corresponding to the individual of genetic algorithm and be defined among the relevant TTI which channel allocation to give which receiver with.Determine the grade of fit of each candidate's scheduling scheme.This grade of fit can be considered the several different index of relevant scheduling scheme performance, for example throughput, delay and fairness.When determining the grade of fit of each candidate scheme, can be weighted each performance index.According to the fitness value of corresponding candidate scheme, in the present age, select individuality as parents.It is higher that better suited scheme is chosen as parents' chance.By selected parents according to such as intersect (crossover) and make a variation (mutation) generate child's (follow-on candidate scheme) for operator.By this way, by a series of iteration (generation), genetic algorithm extracts a plurality of candidate's scheduling schemes, up to being that the TTI that is considered has chosen a preferred plan at certain some place.

With compare with the routine techniques of maximum C/I scheduling such as Round Robin, this genetic algorithm provides the possibility that realizes much better scheduling performance.

The inventor is at S.Abedi, S.Vadgama, " A Radio Aware RandomIterative Scheduling Technique for High Speed Downlink PacketAccess ", VTC 2002, Fall, vol.4, pp.2322-2326,24-28 has proposed another kind of grouping scheduling method among the Sept.2002.This method also generates a plurality of candidate's scheduling schemes for each TTI.Generate at least one candidate scheme at random.Can determine the grade of fit of each candidate scheme by the mode identical with the aforesaid mode that is used for being proposed based on the dispatching method of genetic algorithm.

The selection of the suitable weight that when determining the grade of fit of each candidate scheme each performance index is weighted is also greatly depended in the success of these grouping scheduling methods.Particularly, need different weights to handle different business amount sight.

Under different situations, distribute the same problem of suitable weight also to appear in wireless network, using in the packet scheduler of other type of conceiving.For example, at people's such as R.Agrawal " Class and Channel Condition Based Scheduler for EDGE/GPRS ", Proc.of SPIE, vol.#4531, Aug 2001, and A.Jalali, R.Pankaj; " Datathroughput of CDMA-HDR a high efficiency-high data rate personalcommunication wireless system ", VTC 2000, Spring, vol.3, pp.1854-1858 has described so-called equitable proportion (PF) scheduler among the May 2000.The PF scheduler also need distribute suitable weight under different business amount load and channel condition.

Dynamically the mobile cellular environment makes that suitable weight is set becomes difficult task.At first, between channel condition, may there be great variety.

The curve chart of Fig. 3 (A) schematically shows under first channel conditions, and how the C/I value of four different UEs reports changes on a series of N TTI.In this case, the C/I value is relatively stable, but has constant C/I value difference between a UE and next UE.Some UE have relatively poor relatively channel conditions on the sequence of a whole N TTI, and other have relative good channel situation on whole sequence.Under these channel conditions, above-mentioned a lot of dispatching methods often are difficult to realize the fairness between the different UEs.The others of scheduling performance as entire throughput, then unlikely become problem.

Fig. 3 (B) shows under the second channel situation that is different from first channel conditions (situation of Fig. 3 (A)), the variation of the C/I value of four different UEs.In this case, the variation of the C/I value of each single UE is than big a lot of among Fig. 3 (A), and in the different moment on the sequence of this N TTI, each different UEs all has good channel conditions.In this case, the realization fairness is not difficult.But, the scheduling performance of total throughout that very difficult realization is high or desired others.

Effectively grouping scheduling method should be handled this variable channel situation of describing with reference to Fig. 3 (A) and 3 (B), and dynamically (variable input channel) and distribution (profile) are moved along with it and become the use different business and fast-changing UE also must to handle incoming traffic amount load.Particularly, UE can move to another sub-district from a sub-district apace.In this case, the grouped data that must be buffered in the current Node B is sent to new Node B.If there are many this transmission, then may there be the variation of frequent similar step function in the total load of being handled by packet scheduler at given Node B place.In some cases,, find that when weight fixedly the time, this load variations causes what is called " out of control " phenomenon and the instability of packet scheduler from experience as explanation after a while in this manual.Empirical studies described below shows that also for one or more aspects of desired scheduling performance, great input channel changes may cause significantly decline of performance appearance.

Decreased performance may only influence one type business, for example, and real time business, and can not influence the business of other kind, as non-session service.Some schedulers (as the PF scheduler) are claimed by according to type of service scheduling being divided and distributing different fixed weights to solve this problem to different piece.But, facts have proved that this is a poor efficiency, this is because use the UE quantity of each type service to change, and causes fixed weight that good QoS can't be provided for related business.For example, fix if distribute to the weight of different business, and the high bit rate non real-time number of users of downloading the WWW session significantly increases, then Node B will be born very big pressure and be come to provide acceptable QoS for the session of real-time video dialog mode.

Therefore, expectation provides a kind of grouping scheduling method, and its feasible weight that can adjust the different indexs that impose on scheduling performance automatically is so that scheduler can both effectively turn round under various different business amount sights.

Summary of the invention

According to a first aspect of the invention, a kind of grouping scheduling method is provided, this grouping scheduling method is used for dispatching from the grouped data that transmitter sends to a plurality of receivers by at least one channel, this method may further comprise the steps: generate first index of scheduling performance at least two different aspects, and the incompatible specified data of the set of weights of using first index grouping receiver that will send to, described weighted array is to generate by making up first index according to each respective weights of distributing to different aspect; Generate at least one second index of scheduling performance; At each second index, for relevant may influencing of described second index weight is categorized as at least two kinds of different weight types according to desired weight; And use the classification of the weight of each second index and each second index to adjust described weight.

Different receivers weighted array separately in the method, can every single receiver ground generates the weighted array (as in the method for GB 0303859.3) of first index, so that for example can compare the receiver that grouping will be sent to definite mutually.Alternatively, can every candidate's scheduling scheme ground generate the weighted array of first index, described at following document: S.Abedi, S.Vadgama, " Hybrid Genetic Packet Scheduling and radio ResourceManagement for High Speed Downlink Packet Access ", WPMC 2002Conference, Hawaii, pp.1192-1196 and S.Abedi, S.Vadgama, " A RadioAware Random Iterative Scheduling Technique for High Speed DownlinkPacket Access ", VTC 2002, Fall, vol.4, pp.2322-2326,24-28, Sept.2002.In these cases, each weighted array that generates for different candidate's scheduling schemes can be compared, to select the optimal candidate scheduling scheme, perhaps (under the situation of genetic algorithm) selects a plurality of candidate's scheduling schemes to evolve to the next generation.

Preferably, at least one described first index is independent of at least one the described aspect that influences another described first index.This makes that these indexs are separate, makes it possible to realize accurate control to scheduling performance by the control weighting.

At least one described first index may be assigned to a kind of business of offering receiver (as, WWW browses, video) the influence of priority.

Preferably, adjust at least one described first index to reduce not matching of itself and another described first index.Can adjust these indexs so that all indexs all in same scope, as 0 to 1, thereby identical weighting has same effect for different indexs.

Can carry out normalization to one or more described first indexs, for example can will be used for the value normalization of any one receiver with respect to the summation of the value that is used for all receivers.

In a preferred embodiment, generate described one or each weighted array by the product that forms described first index for correlation receiver.

May there be many channels that can be used for data are sent to receiver, as in the HSDPA system.In this case, specifically implementing a method of the present invention also comprises: be each described single receiver generation first and second this weighted arrays separately; According to its first weighted array separately these receivers are carried out rank, and form the receiver tabulation according to rank order; According to the receiver of the high rank of its second weighted array separately from tabulation to the receiver allocated channel in the tabulation.

In this embodiment, the judgement of the receiver that will send to for packet is subjected to the influence of first and second weighted arrays, and this dispatches the judgement greater flexibility.

May wish to consider the scheduling performance of a lot of different aspects.These aspects may comprise: successfully still be failure when in tolerable delay threshold value data being sent to receiver; Channel quality between transmitter and the receiver; Estimate at how many data and can successfully be transferred to receiver; In transmitter, there are how many data to be sent to receiver waiting; With the delay of transfer of data to receiver; And the fairness between the different receivers.These aspects must not operate on it.For example, can only consider commercial aspect, for example provide being benefited of operator in the business procedure to receiver.

Preferably, the sequence scheduling moment (for example TTI) is repeated this method.In one embodiment, to each described scheduling constantly,, and make and which receiver to send the new decision of packet to for described receiver generates new weighted array.

Described transmission can be a wireless transmission, and described transmitter can be the part of the base station of wireless communication system, and each receiver can be the part of the subscriber equipment of this system.

According to a second aspect of the invention, a kind of packet scheduling apparatus is provided, described packet scheduling apparatus is used for dispatching to the packet that a plurality of receivers send from a transmitter by at least one channel, described device comprises: generate scheduling performance in first index of at least two different aspects and use the device of the receiver that the incompatible specified data grouping of set of weights of first index sent to, described weighted array generates by making up first index according to each respective weights; Generate the device of at least one second index of scheduling performance; May influence the device that weight be categorized as at least two different weight types according to desired weight for relevant second index at each second index; And use each second index and weight to adjust the device of described weight at the classification of each second index.

According to a third aspect of the invention we, provide a kind of transmitter, having comprised: the packet scheduling apparatus of specifically implementing the aforementioned second aspect of the present invention; And dispensing device, be operably connected with described packet scheduling apparatus, and can carry out the operatively feasible receiver of determining by described packet scheduling apparatus that grouping is sent to.

Description of drawings

Below will be in the mode of example with reference to accompanying drawing, in the accompanying drawings:

Fig. 1 shows a plurality of parts (illustrating in front) of the wireless communication system of the HSDPA technology that is used for the down link transmission;

Fig. 2 shows the operation example of the HSDPA technology in the system of Fig. 1;

The curve chart of Fig. 3 (A) and 3 (B) is used to be illustrated in the variation of the C/I value of four UE under the first and second different channels situations;

Fig. 4 shows the block diagram of concrete enforcement packet scheduling apparatus of the present invention;

Fig. 5 shows the flow chart of the operation of the preferred embodiments of the present invention;

Fig. 6 shows the schematic diagram of the example of the channel allocation process of carrying out in the embodiment of Fig. 5;

Fig. 7 is the schematic diagram that is used for explaining in the mixed service flow scene of HSDPA system;

Fig. 8 is the schematic diagram that is used for explaining the cellular environment of HSDPA system;

Fig. 9 is the schematic diagram of the transmitted in packets activity in the HSDPA system that represents in the first simulated environment example;

The curve of Figure 10 is shown in the performance of the maximum C/I scheduler of conventional FIFO weighting in a series of Transmission Time Intervals in first simulation example;

The curve chart of Figure 11 still shows the performance in the initial period among Figure 10 in more detail corresponding to Figure 10;

The curve of Figure 12 is shown in the performance of scheduler in a series of TTI of the fixed weight of describing among the GB 0303859.3 in first simulation example with rough distribution;

The curve chart of Figure 13 still shows the performance in the initial period among Figure 12 in more detail corresponding to Figure 12;

The curve chart of Figure 14 shows in first simulation example performance of packet scheduling apparatus in a series of TTI according to an embodiment of the invention;

The curve chart of Figure 15 shows in first simulation example performance of packet scheduling apparatus in a series of TTI according to another embodiment of the present invention;

The curve chart of Figure 16 shows the variation of the Throughput Variance of the different UEs in a series of TTI of first simulation example that is used for specifically implementing packet scheduling apparatus of the present invention and other scheduler;

The curve chart of Figure 17 shows the variation of QoS of the UE of the reception real-time video traffic in initial period of TTI of first simulation example that is used for specifically implementing packet scheduling apparatus of the present invention and other scheduler;

Figure 18 (A) is used for concrete enforcement packet scheduling apparatus of the present invention and other scheduler of first simulation example are compared at the different aspect of one group of TTI internal performance to the curve chart of Figure 18 (F);

Curve chart shown in Figure 19 (A), 19 (B) and 19 (C) shows the cumulative density function (CDF) in all UE throughput in time of first simulation example that is used for specifically implementing packet scheduling apparatus of the present invention and other scheduler;

Figure 20 (A) corresponds respectively to the curve chart of Figure 18 (A) to Figure 18 (F) to the curve chart of Figure 20 (F), but about second simulation example;

Figure 21 shows the schematic diagram of the transmitted in packets activity in the HSDPA system in the 3rd simulation example;

The curve chart of Figure 22 (A) to 22 (F) corresponds respectively to the curve chart of Figure 18 (A) to Figure 18 (F), but about the 3rd simulation example.

Embodiment

Fig. 4 shows the block diagram of concrete enforcement packet scheduling apparatus 10 of the present invention.Device 10 is used to dispatch will be by at least one channel from the packet of transmitter to a plurality of receivers transmissions.For each continuous scheduling moment (for example each TTI) is made the scheduling decision.Transmitter for example is the base station (Node B) in the wireless communication system.In this case, a plurality of receivers are different UEs that the base station is served.

Device 10 comprises the first index generation unit 12, and the described first index generation unit 12 receives the performance-relevant data with dispatching device.These data can comprise that for example, carrier wave is to interference ratio (C/I) report and the information relevant with the filling rank of source formation, and in transmitter, the data that will send to different receivers were buffered in the formation of described source before sending.

According to the data that received, the first index generation unit 12 generates one group of first index of scheduling performance.This first index is relevant with at least two different aspects of scheduling performance.For example, as subsequently will be in greater detail, one of first index may be relevant with throughput, and another first index may with postpone relevant.According to the type (with reference to hereinafter) of employed dispatching algorithm in the device, can each receiver (UE) or one group of first index of each candidate's scheduling scheme ground generation.

Device 10 also comprises decision unit 16, and described decision unit 16 receives one group of first index that is generated by the first index generation unit 12.Decision unit 16 also receives one group of weight from weight adjustment unit 20.Each first index has the weight of independent correspondence.Decision unit 16 is by making up the weighted array that first index generates first index according to each respective weights.

Decision unit 16 adopts the weighted array of first index to decide the receiver that constantly packet will be sent in the scheduling of being considered.For example, when each receiver when having generated one group of first index (as among the GB 0303859.3), decision unit 16 can each receiver ground generates the weighted array of first index, and each weighted array of different receivers is compared the receiver that decides grouping to send to.Alternatively, when each candidate's scheduling scheme ground generates one group of first index (as the situation of the method in two pieces of papers of the people such as S.Abedi that quoting in the foreword), decision unit 16 can each candidate's scheduling scheme ground generation weighted array.This weighted array can be the fitness function that for example is used for the different candidate schemes of comparison.Then, adopt the weighted array of different candidate schemes, for example by selecting single optimal candidate scheduling scheme or by selecting to be used for some candidate's scheduling schemes of gene evolution, the receiver that decides grouping to send to by dispatching algorithm.The weighted array of employing first index of the present invention decides the mode of the receiver that grouping will be sent to be not limited to above-mentioned example.

Device 10 also comprises above-mentioned weight adjustment unit 20, the second index generation unit 24 and weight taxon 28.The second index generation unit receives and to offer the first index generation unit 12 and performance-relevant some or all data dispatching device.Though not shown among Fig. 4, the second index generation unit 24 can also receive one or more first index that the first index generation unit 12 generates.The second index generation unit 24 generates at least one second index of scheduling performance.Being correlated with in a certain respect of each second index and whole scheduling performance (that is, generally considering all receivers).For example, be described in more detail as following, second index may relate to the average throughput and the fairness of whole QoS, entire throughput, different receivers.

For each second index, weight taxon 28 is divided at least two different weight classifications according to the above-mentioned weight that will offer decision unit 16 to may influencing of second index that associated weight is supposed to.For example, weight can be divided into favourable, unfavorable and neutral classification.Favourable classification produces positive each weight (if present) that may influence by being supposed to second index and forms.Unfavorable classification can produce negative each weight (if present) that may influence to second index by being supposed to and form.Neutral classification is by each weight (if present) of the influence of second index uncertain (not expecting for absolute positive or absolute negative) is formed.

Weight adjustment unit 20 receives each second index from the second index generation unit 24, also receives the information relevant with the classification of weight from weight taxon 28.Weight adjustment unit 20 adopts each second index and weight to adjust weight at the classification of each second index.For example, if second index is reducing or do not change, then can increase each weight in the favourable classification of second index.If second index is reducing or do not change, then can reduce each weight in the unfavorable classification of second index.Can not consider the variation in second index and keep each weight in the neutral classification of second index constant.

Device 10 operationally links to each other with transmitting element 32, scheduling that decision unit 16 made the scheduling decision for it occurring constantly the time, and this transmitting element 32 makes and divide into groups to send to selected receiver.

Next detailed description is suitable for the embodiments of the invention that in the HSDPA system, use.This embodiment consider such as service quality (QoS) but, optimize the performance of HSDPA system aspect channel quality transmission of data packets, the data of in transmitter, waiting for and the index that postpones to distribute.

In the present embodiment, generate one group of five first index of scheduling performance for each UE.First of these five scheduling index is the QoS index.In order to generate this index, for available miscellaneous service definition tolerable in the HSDPA system postpones threshold value Tolerance_Delay.For example, for real-time video traffic, suppose that this tolerable delay threshold value is 100ms.For the world wide web (www) browsing session, suppose that tolerable delay threshold value is 1.5s.Usually, the HSDPA system wishes in defined tolerable postpones threshold value packet as much as possible to be transferred to each UE from transmitter (Node B).

Suppose that N is the sum of the UE that grouping will send in current TTI.OctR _EceivednBe the quantity that success (zero defect) is transferred to eight hytes of n UE.Eight hytes of these error free transmission can be divided into eight hytes that satisfy QoS and do not satisfy QoS.Eight hytes that do not satisfy QoS are eight hytes transmitting outside tolerable delay threshold value.Therefore, for each UE, the quantity of eight hytes that satisfy QoS of being received can be defined as:

Oct _{Received_Satisfied_QoS n}＝Oct _{Received n}-Oct _{Received_Failed_QoS n} n＝1…N (1)

Oct wherein _{Received_Satisfied_QoS n}Be the quantity that satisfies eight hytes of QoS, Oct _{Received_Failed_QoS n}Be quantity for n eight hytes that do not satisfy QoS that UE received.

In addition, Node B can be found its transmission packets that is transferred to each UE delay according to each UE sends it back Node B after receiving grouping affirmation message ACK.For example, in our No. the 0216245.1st, common unsettled UK Patent Application, provide further information about these acknowledge messages.

For each UE, the ratio that satisfies the throughput of qos requirement can be defined as:

$\mathrm{Throughput}\_\mathrm{Satisfy}\_\mathrm{Qo}{S}_n=\frac{{\mathrm{Oct}}_{\mathrm{Received}\_\mathrm{Satisfied}\_\mathrm{QoS\; n}}}{{\mathrm{Oct}}_{\mathrm{Arrived\; node}\_\mathrm{B\; n}}},n=1\·\·\·N---\left(2\right)$

Wherein, Oct _{Arrived_Node_Bn}Be the quantity of original transmission to eight hytes of the source formation that is used for n UE of Node B.

There is not packet to arrive under the situation of Node B for n UE, suppose for this UE:Throughput_Satisfy_QoS _n=0.

Subsequently this value is carried out non-linear conversion, thereby

Ratio_Satisfy_QoS _n＝1/(1+Throughput_Satisfy_QoS _n)，n＝1…N(3)

QoS index definition with each UE is then

$\mathrm{Metric}\_\mathrm{Qo}{S}_n=\mathrm{Ratio}\_\mathrm{Satsfy}\_\mathrm{QoSn}/\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Ratio}\_\mathrm{Satisfy}\_\mathrm{Qo}{S}_i,n=1\·\·\·N---\left(4\right)$

Last index is adopted this normalization, this QoS is referred to target value is mapped to the scope between 0 and 1.

Second in five first indexs among this embodiment based on the C/I value reporting that receives from UE.For the C/I index with each UE is mapped between scope 0 and 1, with the C/I value of each report C/I value sum bi-directional scaling with respect to all reports:

$\begin{array}{cc}\mathrm{Metric}\_C/I_n=C/I_n/\underset{i=1}{\overset{N}{\mathrm{\Σ}}}C/I_i& n=1...N\end{array}---\left(5\right)$

Wherein, C/I _nIt is C/I value for n UE report.

The 3rd index of five first indexs in the present embodiment is related to the estimate amount Eff_Oct of eight hytes that each UE can effectively transmit _nIn this case, suppose that the HSDPA system adopts adaptive modulation and coding (AMC) technology, makes transmitter (Node B) can select different modulation and/or encoding scheme under different channel conditions.In the AMC technology, each UE generates the index of the channel quality of the down link that is experiencing between it and the base station, and this index is reported to Node B.This index for example is the C/I value of downlink channel.Node B adopts the reporting channel index of each UE subsequently, and with system constraint and available modulation and the relevant information of encoding scheme (MCS) rank, be used for the MCS rank of the full blast of this particular UE with identification.Therefore, have better channel or near be positioned at the Node B UE can adopt higher MCS rank, and therefore enjoy than higher transmission rate.For example, can by force C/I value threshold value (for example-8dB ,-2dB ,+4dB) select, thereby move to next MCS rank.The channel quality that the result is based on each UE has carried out effective classification to the transmission rate.

In the present embodiment, suppose each UE in each TTI report C/I value, and Node B can be provided with new MCS rank for each available channel in each TTI.

The another one factor of the quantity of eight hytes that influence can effectively be transmitted for each UE is to handle the mode of the wrong grouping that receives.In the present embodiment, suppose this wrong grouping (failed packet) the employing so-called tracking combined treatment (chase combing process) that receives.In following the trail of combined treatment, retransmit failed packet by Node B, all copies of UE same packets that (for example using the high specific combination) received carry out " soft " combination subsequently.Thereby effectively Carrier interference ratio (C/I) is the C/I sum separately of these two groupings of being combined.Therefore, this tracking combined treatment has improved the C/I of the packet that sends.

Suppose that also all re-transmissions all have than transmitting higher priority for the first time.This represents all retransmission packet are given the chance of transmission before sending new grouping for the first time.

For estimating at first to determine UE to the quantity of effective eight hytes transmitting of UE _nThe MCS rank MCS of full blast _nSecondly, determine UE _nBe in sending mode (being ready to receive the new grouping that sends for the first time) or re-transmission pattern (wait node B retransmits the packet that had before been received by the UE mistake).

If UE is under the re-transmission pattern, the quantity Oct of eight hytes that can be used for sending according to formula (6) decision then _n

Oct _n＝min(Oct(MCS _n)，Oct_Retransmit _n)，n＝1...N(6)

Wherein, Oct (MCS _n) be the maximum quantity of eight hytes can transmitting for n the selected MCS rank of UE, Oct_Retransmit _nBe the quantity (just in Node B, waiting for the quantity of eight hytes that retransmit to this UE) of eight hytes can be used for retransmitting.

If UE is under sending mode, the quantity of eight hytes that can be used for sending based on formula (7) decision then.

Oct _n＝min(Oct(MCS _n)，Oct_Waiting_in_Queue _n)，n＝1...N(7)

Wherein, Oct_Waiting_in_Queue _nThe quantity of eight hytes that send to n UE in the formation of expression Node B.

The maximum quantity of eight hytes that can send by arbitrary UE according to formula (8) decision subsequently.

Oct _{n max}＝max(Oct _n)，n＝1...N (8)

Wherein, n _MaxExpression has the numbering of the UE (" preferably UE ") of eight hytes that can be used for sending of maximum quantity, Oct _{N max}It is the maximum quantity that can send to eight hytes of this best UE.

Determine eight hytes of transmitting according to formula (9) subsequently for effective estimation of this best UE.

Eff_Oct _{n max}＝Octn _{n max}·(1-FER _{n max}) (9)

FER wherein _{N max}Be n _MaxThe FER (Floating Error Rate) of the estimation of individual UE.

Relative effective quantity of transmitting eight hytes of n UE of this maximum normalization subsequently, thus the 3rd indivedual indexs become:

Metric_Oct _n＝Oct _n·(1-FER _n)/Eff_Oct _{n max}，n＝1...N (10)

The four-index of five first performance index in the present embodiment is that how many data of expression are just waiting the index that sends to each UE in Node B.This is an inverse correlation with throughput to the data of each UE.

For n UE, throughput is defined as:

${\mathrm{Th}}_n=\frac{{\left({\mathrm{Oct}}_{\mathrm{Received}}\right)}_n}{{\left({\mathrm{Oct}}_{\mathrm{Arrived}\_\mathrm{Node}\_B}\right)}_n},n=1...N---\left(11\right)$

Then, with etc. the ratio Metric_Waiting_Ratio of data to be transmitted _nBe defined as:

Metric_Waiting_Ratio _n＝1—Th _n，n＝1...N (12)

The last index of five performance first indexs in the present embodiment is relevant with the delay of each UE experience.At first, consider not transmit eight hytes, and determine these eight retardations that hyte stands in the maximum duration of the medium UE of being sent to of Node B.

Delay _n＝M·TTI-Arrival_Time_Earliest _n，n＝1...N (13)

Wherein, MTTI represents the current time, and Arrival_Time_Earliest _nEight hytes that the expression maximum duration is not transmitted arrive the time of the Node B source formation that is used for n UE.

Then, use relevant with the first index as mentioned above same tolerable of setting up to postpone threshold value, calculating the worst delay that UE bore and the distance between its tolerable delay threshold value as different business.

Delay_Distance _n＝Delay _n-Tolerance_Delay _n，n＝1…N(14)

Wherein, Delay_Distance _nIt is the distance of n UE and threshold value.

Subsequently the delay distance of all UE is mapped on the occasion of.For carrying out on the occasion of mapping definite at first so minimum distance:

Delay_Distance _min＝min(Delay_Distance _n)，n＝1…N (15)

Subsequently the delay distance through adjusting is defined as:

Adjusted_Delay_Distance _n＝Delay_Distance _n-Delay_Distance _min，

n＝1...N(16)

Carry out Nonlinear Mapping then.The result is:

Mapped_Delay _n＝β _n/(1+Adjusted_Delay_Distance _n)，n＝1…N(17)

Wherein, Mapped_Delay _nBe the mapping length of delay of n UE, and β _nIt is the parameter of priority of pointing out to offer the business of n UE.Here, for example, business is that WWW browses or real-time video.β _nHigh more, professional priority is high more.

The last index relevant with delay is confirmed as:

$\mathrm{Metric}\_D{\mathrm{elay}}_n=\mathrm{Mapped}\_D{\mathrm{elay}}_n/\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Mapped}\_{\mathrm{Delay}}_n,n=1...N---\left(18\right)$

This has guaranteed index Metric_Delay _nBe mapped as the value between 0 and 1.

In the present embodiment, calculate first weighted array of first index of each UE by the weighted product of the single performance index of formation as described below:

Ranking_Metric _n＝(1+W ₁(m)·Metric_QoS _n(m))·(1+W ₂(m)·Metric_C/I _n(m))·(1+W ₃(m)·Metric_Oct _n(m))·(1+W ₄(m)·Metric_Waiting_Ratio _n(m))·(1+W ₅(m)·γ(m，j)·Metric_Delay _n(m))， (19)

Wherein, W _kFirst group of k weight (rank weight) and index a when (m) being m TTI represent n UE, Metric_QoS _n(m), Metric_C/I _n(m), Metric_Oct _n(m), Metric_Waiting_Ratio _n(m) and Metric_Delay _nEquation (1) when (m) being m TTI arrives defined five first finger target values in (18).

γ (m, j) be distribute to have the same delay tolerance limit UE group j's and be used for the weight relevant of accurate adjustment service priority with the variable delay tolerance limit.Be described in more detail as following, by weight adjustment unit 20 adjust γ (m, j) and weights W _k(m).The Metric_Delay of definition in the formula (18) _n(m) parameter beta that is comprised in is the fixed value of specifying service priority.As mentioned above, for each UE group j, with professional weight beta with same delay tolerance limit _jDistribute to the UE that adjusts.

According to UE _nAffiliated group is selected γ (m, value j) of use in the formula (19).Therefore, if UE27 belongs to group 2, then be used to calculate Ranking_Metric ₂₇Last of formula (19) become (1+W ₅(m) γ (m, 2) Metric_Delay ₂₇(m)).

Calculate second weighted array of first index of each UE in the present embodiment as follows:

Channel_Allocation_Metric _n＝(1+V ₁(m)·Metric_QoS _n(m))·(1+V ₂(m)·Metric_C/I _n(m))·(1+V ₃(m)·Metric_Oct _n(m))·(1+V ₄(m)·Metric_Waiting_Ratio _n(m))·(1+V ₅(m)·Metric_Delay _n(m)) (20)

V wherein _kSecond group k weight (channel-assignment unit) when (m) representing m TTI.

Fig. 5 is the flow chart of operation example of decision unit 16 that is used for apparatus of the present invention of key-drawing 4.For each scheduling of considering moment (for example TTI) is carried out sequence of steps shown in Figure 5.

In first step S1, this decision unit 16 calculates first weighted array (rank is measured (ranking metric)) of first index of UE for each relevant UE according to formula (19).When generating this first weighted array, use first group of weight (rank weight) W _k(m) to the first index weighting.

In step S2, decision unit 16 is measured according to each rank UE is carried out rank.This decision unit generates the UE tabulation by rank, the top that the UE that rank is the highest is tabulating.This tabulation only comprises such UE: the data that are sent to this UE that considers such as have in the formation of the source of transmitter.

At step S3, second weighted array (channel allocation is measured) that decision unit 16 calculates first index of this UE according to formula (20) for each relevant UE.When generating second weighted array, use second group of weight (channel allocation weight) V _k(m) first index is weighted.The channel allocation weight can be different from the rank weighting.

In step S4, decision unit 16 is with reference to the UE tabulation that generates in step S2.From the tabulation top, describe in detail as reference Fig. 6, decision unit 16 is measured according to the channel allocation of each UE and is given these UE with channel allocation.The processing of proceeding among the step S4 will distribute up to not remaining channel.

This current scheduling processing of (TTI m) constantly that is through with.Repeat this processing for next scheduling moment (TTIm+i) then.

Begin to carry out the channel allocation process the step S4 from the top of rank UE tabulation, and according to the relative value allocated channel of the channel allocation index that generates in the formula 20.

In Fig. 6, schematically show this processing.

Suppose that the UE tabulation comprises five UE altogether, next UE 30 is UE2,9,11 and 17 at the top of tabulation.Simultaneously, suppose to be respectively 1.5,4.3,1.2,2.0 and 3.2 for the channel allocation index that these UE generate by formula 20.

In first step, calculating to be 12.2 to the channel allocation index summation of all UE (i.e. all UE in this case) of its allocated channel.To be UE 30 to the UE of high rank of its distribution.Then its channel allocation index 1.5 is carried out " normalization " with respect to all channel allocation indexs, and become 0.12.Hypothesis always has 10 available channels in this example.Therefore, obtain 1.23 with the 10 normalization channel indexs that multiply by UE 30.Use " floor function " that it is truncated to generate final result then, promptly UE 30 is distributed a channel.Remaining like this 9 channels are used for the distribution of subsequent step.Incidentally, in the present embodiment, if final result less than 1, then is made as 1 with it, thereby minimum channel is assigned as 1 channel.

In second step, recalculating remaining will be 10.7 to the channel allocation index summation of the UE of its allocated channel.The channel allocation index 4.3 of the UE of high rank that to be left is then carried out normalization with respect to new summation, becomes 0.40.With the residue number of channel, promptly 9, multiply by this value, generate the result of 3.62 channels, it is truncated finally distributed 3 channels.Stay 6 channels and be used for the next step distribution.In third step, carry out this process once more, the result, UE 9 has distributed a channel, stays 5 channels.In the 4th step, calculate and also should distribute 1 channel, at remaining 4 channels of this stage to UE 11.At last, in the 5th step, UE 17 is distributed 4 channels, and processing finishes.

Should be appreciated that in embodiments of the present invention and also can adopt other channel allocation process.

In the present embodiment, generate one group of four second scheduling index.First of these four second scheduling index is total QoS index.

For each UE, defined the throughput ratio that satisfies the qos requirement of relevant UE in the superincumbent equation (2).Then will second index relevant be defined as mean value simply for the single value of the Throughput_Satisify_QoS of different UEs with total QoS that scheduling is handled.

$\mathrm{Total}\_\mathrm{Qos}\left(m\right)=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Throughput}\_\mathrm{Satisfy}\_{\mathrm{QoS}}_i\left(m\right)---\left(21\right)$

Here, the same with above formula (19) with (20), the parameter value when m TTI only represented in term " (m) ".

In these four second indexs second relevant with total throughout at all UE of m TTI.Calculate this total throughout index as follows

$\mathrm{Total}\_\mathrm{Th}\left(m\right)=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\left({\mathrm{Oct}}_{\mathrm{Received\; i}}\left(m\right)\right)}{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\left({\mathrm{Oct}}_{\mathrm{Arrived}\_\mathrm{Node}\_B_i\left(m\right)}\right)}---\left(22\right)$

Oct wherein _{Received i}(m) for when m TTI, being transferred to the sum of eight hytes of i UE.

The 3rd index in four second indexs of in the present embodiment this is relevant with the average throughput of m TTI different UEs constantly.This value is the mean value of the single throughput that calculates in the above formula (11):

$\mathrm{Avg}\_\mathrm{Th}\left(m\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{Th}}_n\left(m\right)---\left(23\right)$

In these four second indexs last is relevant with fairness.In the present embodiment, think that fairness is relevant with the scope of the Throughput Variance that arrives single UE.According to above formula (11), the vector of single throughput is

TH(m)＝{Th _n(m)}，n＝1…N (24)

The vector variance of single throughput is defined as

var_Th(m)＝var(Th(m)) (25)

Fairness is defined as

$\mathrm{Fairness}\left(m\right)=\frac{1}{\mathrm{var}\_\mathrm{Th}\left(m\right)}---\left(26\right)$

Understand easily, by the decision unit 16 employed rank weights W of handling a part as scheduling _k(m) and channel weight V _k(m) different aspect for scheduling performance has different may influencing.According to channel conditions and traffic load, some weights are believed to be helpful in the possibility height that improves certain second index, and think that other weights help to reduce the possibility height of certain second index.By independent increase or reduce in five weights one and simultaneously all other weights are made as 1 and study and respectively organize each may influence in five weights in the weight in the present embodiment for these four second indexs.Represented this result of study in the table 1 below.

Table 1

In table 1,, then this weight is categorized as the favourable classification (" F ") of the weight that belongs to this second index if the increase of a weight is considered to cause the possibility height of one second index improvement.If think when possibility that the increase of a weight causes one second index to reduce is high, then this weight is categorized as the unfavorable classification (" H ") of the weight that belongs to this second index.With the not high weight of possibility that is considered to make one second index to improve or to reduce be categorized as belong in the middle of classification (blank column in the table 1).

By considering how pairing first index of these weights (different aspect of scheduling performance) and first index may influence each second different index and come evaluation weight may influence for second index.

For example consider the second index Total_Th (m) relevant with total throughout.

Carrier-in-interference (C/I) according to current report recently defines the first index Metric_C/I _n(m), and according to the effective quantity of transmitting eight hytes that is used for n UE define Metric_Oct _n(m).If the historical record display performance of second index relevant with total throughout descends, then by in rank and channel allocation process, all increasing the weights W relevant with these first indexs ₂(m), W ₃(m), V ₂(m) and V ₃(m), recover total throughout possibly.Be categorized as these weights favourable thus.On the contrary, Metric_Waiting_Ratio _n(m) hinder total throughout probably.Weights W that will be relevant with this first index ₄(m) and V ₄(m) be categorized as for total throughout unfavorable.Reduce the high weight of these possibilities and will recover throughput.In fact, in two TTI, handle W in this way ₂(m), W ₃(m), V ₂(m), V ₃(m), W ₄(m) and V ₄(m) will make the behavioural trend of decision unit in the behavior of maximum C/I packet scheduler.

Adjust or regulate weight by the classification of using second index and weight, dispatching device can be become more efficiently scheduler, that is, carry out different operating in response to different situations.By using a plurality of second different indexs to generate and adjust weight, can realize the equilibrium of a plurality of aspect of performances that they are measured.

The example of the mode that can adjust weight is described to (64) below with reference to formula (24).

Keep and use two values for each weight (rank weight and channel allocation weight).First value is " reality " value, and it offers decision unit 16 to be used the receiver that decides packet to send to by decision unit for weight adjustment unit 20.This first value satisfies the particular specification condition of appointment.For example, in the present embodiment, its must be always more than or equal to zero, and must can not surpass a certain predetermined maximum.This maximum is provided one group of initial weight in the time of can activating this device for the first time by the operator of for example this device.Suitable initial weight group is disclosed in GB 0303859.3.

Second value of each weight is " void " value, and it does not offer decision unit 16 by weight adjustment unit 20 inner uses.This second value is that weight adjustment unit 20 is used for adjusting weight.In each TTI, calculate new empty value according to the previous void value in the weight adjustment unit 20.These empty values are not applied to the restriction of real-valued rated condition, can arbitrarily adjust.Use mathematic(al) manipulation to handle to derive from the void value real-valued, this mathematic(al) manipulation is handled and is guaranteed that this real-valuedly satisfies rated condition.

The real-valued of rank weight among m TTI is

W(m)＝{W ₁(m)，W ₂(m)，W ₃(m)，W ₄(m)，W ₅(m)} (24)

The real-valued of channel allocation weight among m TTI is

V(m)＝{V ₁(m)，V ₂(m)，V ₃(m)，V ₄(m)，V ₅(m)} (25)

Similarly, the void value Virtual_W of the rank weight of current m TTI be expressed as Virtual_W (m, s)={ Virtual_W ₁(m, s), Virtual_W ₂(m, s), Virtual_W ₃(m, s), Virtual_W ₄(m, s), Virtual_W ₅(m, s) }

(26)

The void value Virtual_V of channel allocation weight is expressed as

Virtual_V(m，s)＝{Virtual_V ₁(m，s)，Virtual_V ₂(m，s)，Virtual_V ₃(m，s)，Virtual_V ₄(m，s)，Virtual_V ₅(m，s)}

(27)

Set up the one-one relationship between real-valued and empty value Virtual_V (m) and the V (m).For example, if W ₁For Total_QoS (m) is favourable, then Virtual_W ₁For Total_QoS (m) also is favourable.

In the present embodiment, in each TTI, in the sequence of four steps, carry out the adjustment of the void value of weight.Index s in formula (26) and (27) represents step number, Virtual_W _i(m, s) (or Virtual_V _iWeights W among the step S of m TTI of (m, s)) expression _i(or V _i) the void value.Describe the adjustment of carrying out in these four steps below in detail.

Step 1:

For first Total_QoS (m) in this second index, the historical record that adopts this desired value to change is assessed last L the index trend among (as L=100) TTI:

$\mathrm{Decisivie}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}[\mathrm{Total}\_\mathrm{QoS}(m-l+1)-\mathrm{Total}\_\mathrm{Qos}(m-l)]---\left(28\right)$

Trend and the weight in the table 1 based on first index Total_QoS (m) in second index classified in step 1 then, adjusts final step (the void value of the weight that step 4) kept of previous T TI (m-1 TTI).

For the rank weight that helps Total_QoS (m), we have

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,1)=\mathrm{Virtual}\_W_i(m-\mathrm{1,4})+\mathrm{\ϵ}& \\ -\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)\≤0,\\ \mathrm{Virtual}\_W_i(m,1)=\mathrm{Virtual}\_W_i(m-\mathrm{1,4})& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)>0\end{array}\right.---\left(29\right)$

i＝1，4，5

Wherein ε is the normal real number of value between 0.001 and 0.01.

For the channel allocation weight that helps Total_QoS (m), we have

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,1)=\mathrm{Virtual}\_V_i(m-\mathrm{1,4})+\mathrm{\ϵ}& \\ -\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)\≤0,\\ \mathrm{Virtual}\_V_i(m,1)=\mathrm{Virtual}\_V_i(m-\mathrm{1,4})& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)>0\end{array}\right.---\left(30\right)$

i＝1，4，5

For the rank weight that is unfavorable for Total_QoS (m), we have

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,1)=\mathrm{Virtual}\_W_i(m-\mathrm{1,4})-\mathrm{\ϵ}& \\ +\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)\≤0,\\ \mathrm{Virtual}\_W_i(m,1)=\mathrm{Virtual}\_W_i(m-\mathrm{1,4})& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)>0\end{array}\right.---\left(31\right)$

i＝2，3

For the channel allocation weight that is unfavorable for Total_QoS (m), we have

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,1)=\mathrm{Virtual}\_V_i(m-\mathrm{1,4})-\mathrm{\ϵ}& \\ +\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)\≤0,\\ \mathrm{Virtual}\_V_i(m,1)=\mathrm{Virtual}\_V_i(m-\mathrm{1,4})& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{QoS}\left(m\right)>0\end{array}\right.---\left(32\right)$

i＝2，3

As can be seen, as the result of formula (29) to (32), if this second index reduces or be constant, then the void value to the favourable weight of Total_QoS (m) increases, if this second index increases, then should the void value remain unchanged.On the other hand, if second index reduces or be constant, then the void value to the disadvantageous weight of Total_QoS (m) reduces, if this second index increases, then should the void value remain unchanged.

In the step 1 of this example, there is not neutral value.

Step 2:

In this step, in the end among L TTI once more in response to the trend of the second index Var_Th relevant with fairness, adjust the void value of weight.Judge this trend based on following formula

$\mathrm{Decisivie}\_\mathrm{var}\_\mathrm{Th}\left(m\right)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}[\mathrm{var}\_\mathrm{Th}(m-l+1)-\mathrm{var}\_\mathrm{Th}(m-l)]---\left(33\right)$

Adjust the void value of the weight that helps fairness as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,2)=\mathrm{Virtual}\_W_i(m,1)+\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)\&GreaterEqual;0,\\ \mathrm{Virtual}\_W_i(m,2)=\mathrm{Virtual}\_W_i(m,1)& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)<0\end{array}i=\mathrm{1,4}\right.---\left(34\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,2)=\mathrm{Virtual}\_V_i(m,1)+\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)\&GreaterEqual;0,\\ \mathrm{Virtual}\_V_i(m,2)=\mathrm{Virtual}\_V_i(m,1)& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)<0\end{array}i=\mathrm{1,4}\right.---\left(35\right)$

Adjust the void value of the weight that is unfavorable for fairness as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,2)=\mathrm{Virtual}\_W_i(m,1)-\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)\&GreaterEqual;0,\\ \mathrm{Virtual}\_W_i(m,2)=\mathrm{Virtual}\_W_i(m,1)& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)<0\end{array}i=\mathrm{2,3}\right.---\left(36\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,2)=\mathrm{Virtual}\_V_i(m,1)-\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)\&GreaterEqual;0,\\ \mathrm{Virtual}\_V_i(m,2)=\mathrm{Virtual}\_V_i(m,1)& \mathrm{Decisive}\_\mathrm{var}\_\mathrm{Th}\left(m\right)<0\end{array}i=\mathrm{2,3}\right.---\left(37\right)$

All the other weights are neutral, and remain unchanged

Virtual_W _i(m, 2)=Virtual_W _i(m, 1), the neutral weight (38) of i ∈

And

Virtual_V _i(m, 2)=Virtual_V _i(m, 1), the neutral weight (39) of i ∈

Above formula (34) to (37) is different from the respective formula of using in the step 1 (29) and arrives (32).This is because the variation of the second index Var_Th and throughput is relevant, that is, and and itself and fairness inversely related.

Step 3:

In this step, assess the trend of the second index Avg_Th relevant as follows with average throughput:

$\mathrm{Decisivie}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)=\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}[\mathrm{Avg}\_\mathrm{Th}(m-l+1)-\mathrm{Avg}\_\mathrm{Th}(m-l)]---\left(40\right)$

Adjust the void value of the weight that helps average throughput as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,3)=\mathrm{Virtual}\_W_i(m,2)+\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_W_i(m,3)=\mathrm{Virtual}\_W_i(m,2)& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)>0\end{array}i=4\right.---\left(41\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,3)=\mathrm{Virtual}\_V_i(m,2)+\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_V_i(m,3)=\mathrm{Virtual}\_V_i(m,2)& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)>0\end{array}i=4\right.---\left(42\right)$

Adjust the void value of the weight that is unfavorable for average throughput as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,3)=\mathrm{Virtual}\_W_i(m,2)-\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_W_i(m,3)=\mathrm{Virtual}\_W_i(m,2)& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)>0\end{array}i=2\right.---\left(43\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,3)=\mathrm{Virtual}\_V_i(m,2)-\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_V_i(m,3)=\mathrm{Virtual}\_V_i(m,2)& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Th}\left(m\right)>0\end{array}i=2\right.---\left(44\right)$

The residue weight is neutral, and remains unchanged

Virtual_W _i(m, 3)=Virtual_W _i(m, 2), the neutral weight (45) of i ∈

And

Virtual_V _i(m, 3)=Virtual_V _i(m, 2), the neutral weight (46) of i ∈

Step 4:

In this step, assess the trend of the second index Total_Th relevant as follows with total throughout:

$\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)=\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}[\mathrm{Total}\_\mathrm{Th}(m-l+1)-\mathrm{Total}\_\mathrm{Th}(m-l)]---\left(47\right)$

Adjust the void value of the weight that helps total throughout as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,4)=\mathrm{Virtual}\_W_i(m,3)+\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_W_i(m,4)=\mathrm{Virtual}\_W_i(m,3)& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)>0\end{array}i=\mathrm{2,3}\right.---\left(48\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,4)=\mathrm{Virtual}\_V_i(m,3)+\mathrm{\&epsiv;}& \\ -\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_V_i(m,4)=\mathrm{Virtual}\_V_i(m,3)& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)>0\end{array}i=\mathrm{2,3}\right.---\left(49\right)$

Adjust the void value of the weight that is unfavorable for average throughput as follows

$\left\{\begin{array}{cc}\mathrm{Virtual}\_W_i(m,4)=\mathrm{Virtual}\_W_i(m,3)-\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_W_i(m,4)=\mathrm{Virtual}\_W_i(m,3)& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)>0\end{array}i=4\right.---\left(50\right)$

And

$\left\{\begin{array}{cc}\mathrm{Virtual}\_V_i(m,4)=\mathrm{Virtual}\_V_i(m,3-\mathrm{\&epsiv;}& \\ +\mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right),& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)\&le;0,\\ \mathrm{Virtual}\_V_i(m,4)=\mathrm{Virtual}\_V_i(m,3)& \mathrm{Decisive}\_\mathrm{Total}\_\mathrm{Th}\left(m\right)>0\end{array}i=4\right.---\left(51\right)$

The residue weight is neutral, and remains unchanged

Virtual_W _i(m, 4)=Virtual_W _i(m, 3), weight (52) in the middle of the i ∈

And

Virtual_V _i(m, 4)=Virtual_V _i(m, 3), weight (53) in the middle of the i ∈

In this stage, use empty value to upgrade real-valued.For this reason, determine minimum empty value at first as follows

Virtual _min＝min(Virtual_W _i(m，4)，i＝1...5，Virtual_V _i(m，4)，i＝1...5) (54)

The void value is carried out further adjustment, just to be:

$\left\{\begin{array}{cc}\mathrm{Virtual}2\_W_i\left(m\right)=\mathrm{Virtual}\_W_i(m,4)-2\&CenterDot;{\mathrm{Virtual}}_{\min },\mathrm{Virtual}2\_V_i=\mathrm{<figure-callout\; id="2"\; label="Virtua"\; filenames="C200480004588E00441.png,C200480004588E00451.png"\; state="\{\{state\}\}">Virtua</figure-callout>}2\_V_i(m,4)-2\&CenterDot;{\mathrm{Virtual}}_{\min }& i=1...5,{\mathrm{Virtual}}_{\min }<0\\ \mathrm{Virtual}2\_W_i\left(m\right)=\mathrm{Virtual}\_W_i(m,4),\mathrm{Virtual}2\_V_i=\mathrm{Virtual}\_V_i(m,4)& i=1...5,{\mathrm{Virtual}}_{\min }\&GreaterEqual;0\end{array}\right.---\left(55\right)$

Virtual2_W wherein _i(m) and Virtual2_V _i(m) be through adjust on the occasion of.

Next, find these through adjust on the occasion of in maximum:

Virtual _max＝max(Virtual2_W _i，i＝1…5，Virtual2_V _i，i＝1…5) (56)

As mentioned above, in the present embodiment, real-valued the surpassing for example by the weight limit in the initial weight group that the operator applied that does not allow weight:

max_initial_weight＝max(W _i(1)，i＝1…5，V _i(1)，i＝1…5) (57)

Then, real-valued for weight, we have

W _i(m)＝Virtual2_W _i(m)·max_initial_weight/Virtual _maxi＝1…5 (58)

V _i(m)＝Virtual2_V _i(m)·max_initial_weight/Virtual _maxi＝1…5 (59)

The mathematic(al) manipulation that introduce from (54) to (59) has limited the real-valued growth of weight, and prevents to provide one of QoS or some performance index difference to occur.These formula have been produced by practical studies widely.

Refer again to formula (19) and (20), as can be seen, decision unit 16 also adopted adaptive another weight γ relevant with delay tolerance (m, j).For each UE group j with same delay tolerance limit, all exist such adaptive weighting γ (m, j).Weight adjustment unit 20 also can adjust these at every group adaptive weighting γ (m, j).For each the group j, calculate as follows the average service quality that in m TTI, is experienced by the UE that belongs to this group index Avg_Group_QoS (m, j).

$\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m,j)=\frac{1}{N_j}\underset{i=1}{\overset{N_j}{\mathrm{\&Sigma;}}}\mathrm{Throughput}\_\mathrm{Satisfy}\_{\mathrm{QoS}}_i,j=1\&CenterDot;\&CenterDot;\&CenterDot;J---\left(60\right)$

N wherein _jBe the UE quantity among the group j, calculate Throughput_Satisify_QoS for each UE in this group according to above formula (2).

Assess the trend of the average service quality of group among the last L TTI then as follows for each group j:

$\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m,j)=\underset{i=1}{\overset{L}{\mathrm{\&Sigma;}}}[\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m-l+1,j)-\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m-l,j)]---\left(61\right)$

Adopt then this trend come following calculating parameter ω (m, j)

$\left\{\begin{array}{cc}\mathrm{\&omega;}(m,j)=\mathrm{\&gamma;}(m-1,j)+\mathrm{\&epsiv;}& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m,j)\&le;0\\ -\mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m,j),& \\ \mathrm{\&omega;}(m,j)=\mathrm{\&gamma;}(m-1,j)& \mathrm{Decisive}\_\mathrm{Avg}\_\mathrm{Group}\_\mathrm{QoS}(m,j)>0\end{array}\right.---\left(62\right)$

Upgrade as follows then every group Adaptive QoS weight γ (m, j)

γ(m，j)＝ω(m，j)/ω _min(m)，j＝1…J (63)

ω wherein _Min(m)=min (ω (m, j)), j=1 ... J (64)

Guarantee from (62) to the conversion of (64) all UE have on the occasion of group QoS weight γ (m, j), and the weights with group of minimum service priority are 1 always.This conversion also can avoid organizing QoS weight γ (m, j) the unrestricted growth.Preferably, these mappings are used to prevent the difference of the performance index relevant with the QoS of each UE, and the difference of the performance index relevant with total QoS.Derive and proved these conversion by experience.

In this stage, upgrade all weights for m TTI.These weights can be applied to decision unit, so that its rank and the channel allocation that can be created on definition in (19) and (20) are measured.Then can be by the step in the flow chart of decision unit execution graph 5.

Next, will the various analog results of the preferred embodiment of the present invention be described.These analog results and the analog service amount environmental correclation that a plurality of UE are arranged.Suppose that first group of UE attempts to receive real time video data, suppose that second group of UE attempts to carry out the WWW browsing session.In Fig. 7, schematically show this scene.As shown in Figure 7, WWW is downloaded grouping by the internet and by in the application receiver buffering area of HSDPA system transmissions in the UE.Directly real-time video transmission is arrived the application receiver buffering area of this UE by the HSDPA system.

Real time video data is made up of real-time protocol (rtp).In this simulation, adopt to meet ITU video traffic amount model H.263.This traffic model concentrates on the traffic carrying capacity characteristic relevant with the RTP transmitted in packets with video coding.This model per second kind produces 7.5 frame of video, and the target bit rate of its outputting video streams is 32kbps.These frame of video of big young pathbreaker according to frame of video are divided into one or more RTP groupings.

The WWW browsing session is believed to comprise as at TR 25.848, ver 1.0.0, RP-010191, TSG-RAN#11, a series of packet calls of definition among the March 2001.Suppose the WWW browsing session has been used mechanism at a slow speed.This at a slow speed mechanism form before fully successfully downloading last session, download new session.For example, if first webpage links with second webpage, then mechanism does not allow to download second webpage before successfully downloading all first page at a slow speed.

Identical channel code (channel) in real time video data and the WWW data sharing HSDPA system.

In Fig. 8, schematically shown the cellular environment that adopts in this simulation.In Fig. 8, suppose the 6km that is spaced apart between the adjacent base station (Node B) 30.There is a base station each sub-district.Each is divided into three sectors for a short time, and the border of these sectors as shown in Figure 8.As shown in Figure 8, suppose all to be evenly distributed in the sub-district when UE begins, they move around then.

Suppose relatively that with cell environment Node B has fixing transmitting power.For fully loaded (fully-loaded) HSDPA system that does not adopt power control, this is the hypothesis of a reality.Neighbor cell interference is the result who sends owing to from a plurality of adjacent node B.Determine the rank of neighbor cell interference according to the constant power rank of the Node B that sends.In addition, suppose to exist path loss and its to influence signal quality.Also be assumed to be each sector a concrete packet scheduling apparatus of the present invention in real time is provided, to handle the data of all UE in the sector.

For the influence of modeling Rayleigh (Rayleigh) decline, adopt ETSI 6-path Rayleighvehicular A channel model.The speed of supposing each UE is 3.6km/h.Following table 2 shows the relative delay and the average power in 6 different paths that constitute channel.

Table 2

Suppose that shade has logarithm normal distribution (log-normal distribution).Listed the Shadow Parameters of being supposed in the table 3 below.

Table 3

Parameter	Value
		The standard deviation of lognormal shadow fading	8dB

The decorrelation distance	20m
		Correlation between the website	0.5
Correlation between the sector	1

Fig. 9 shows the grouped data of the arrival different UEs of being supposed in first simulation example.UE1 to 10 is video user.UE 11 to 90 is WWW users.In first example, suppose that total traffic load is roughly a constant in one group of 30,000 TTI (60 second cycle).As can be seen from Fig. 9, each video user (UE1 is to UE10) has continuous substantially arrival stream of packets.For WWW user (UE 11 to UE 90), it is discontinuous that packet arrives.But when data arrived, instantaneous data rates may be more a lot of greatly than the video data rate of 32kbps.Thus, suppose that total duty ratio that offers WWW user offers the big of video user.In other words, video user has narrow pipeline, and WWW user has wide pipeline.

Suppose that the channel estimating that each UE carries out is good, and feedback signal is error-free.Simultaneously, minimum report postpones to be 3TTI.

Simultaneously, to postpone threshold value be 1.5 seconds in the tolerable of supposing the WWW session.For video traffic, suppose that it is 100 milliseconds that tolerable postpones threshold value.Carry out and follow the trail of combination, and the maximum quantity that grouping sends is 6.If can not in 6 times retransmit, transmission one divide into groups, then abandon this grouping.Suppose that video user belongs to service groups j=1, WWW user belongs to service groups j=2.Below in all simulations, with the service priority parameter beta of video user ₁Be fixed as 10, WWW user's service priority parameter is fixed as 1.Also suppose, consider the trend of second index among L=100 the TTI, and ε=0.001.Suppose that (m is 1 for all UE j) to γ.

Below with reference to Figure 10 in the explanation of Figure 22, the performance of the scheduler through considering in advance with fixed weight described in the performance of specific implementation dispatching device of the present invention and maximum C/I scheduler of conventional FIFO weighting and the GB 0303859.3 is compared.

At first, Figure 10 is relevant with the performance of maximum C/I scheduler under the traffic carrying capacity scene shown in Figure 9 with 11.How four second indexs that Figure 10 shows formula (21), (22), (23) and (26) definition change in the simulation cycle of 30,000 TTI.Figure 10 also shows the variation of the group index Avg_Group_QoS of the video traffic of formula (60) definition and WWW business.

The curve chart of Figure 11 is corresponding to Figure 10, but only shows the variation among preceding 2,000 TTI (4 seconds) of simulation cycle.As can be seen, because all real-time video users and WWW user share same channel, so aspect the QoS and fairness of real-time video, all videos when the HSDPA systematic function begins in response to simulation cycle the most at the beginning and WWW call out entering suddenly and collapse.It can also be seen that it is about 0.6 that the QoS of real-time video rises at first, but performance is reduced to about 0.3 subsequently.Performance recovery subsequently, but for video user, this " out of control " phenomenon is not satisfied.Can not occur this " out of control " phenomenon in WWW user's QoS, its stable performance increases in simulation cycle.

Next, Figure 12 and 13 has represented the information identical with Figure 10 and 11, but for the scheduler that illustrates among the GB0303859.3, has the fixed weight group of following rough distribution:

W ₁(1)＝0.1，W ₂(1)＝10，W ₃(1)＝0.1，W ₄(1)＝0.1，W ₅(1)＝0.1

V ₁(1)＝0.1，V ₂(1)＝0.1，V ₃(1)＝0.1，V ₄(1)＝10，V ₅(1)＝0.1

These values are not disclosed optimum values among the GB 0303859.3.But they are elected to be nonoptimal value specially, because in fact, best fixed weight initial value can not be arranged in this scheduler.

As shown in figure 13, in initial 1,000 TTI (or two seconds), also there is runaway effect.In Figure 13, WWW user's QoS and real-time video user's QoS obviously has two opposite trend.

In addition, in first two seconds of simulation cycle, the Throughput Variance steady growth.As can be seen, the recovery among Figure 13 is than fast among Figure 11 (maximum C/I scheduler).But in some cases, for the real-time video user, the final degree of this restore cycle and the stability realized may be unacceptable.

Figure 14 shows the performance of the embodiment of the invention, wherein begins to adjust weight according to the thick initial value that is provided with above.Show the performance index identical with Figure 10 to 13.But, in this example, when generating weight real-valued from the void value, do not carry out the conversion of formula (54) to (59) and the conversion of (62) to (64).Thus, effectively the void value of unrestricted adjustment is directly applied to decision unit 16 among Fig. 4.

As can be seen from Figure 14, in preceding 30 seconds of simulation cycle, WWW user's QoS is more much lower than the QoS of video user.The rank of the QoS that WWW user is smooth was illustrated in this initial period, almost all sent to video user, and had blocked to WWW user and transmit data.This serious input bit pipeline disequilibrium has directly influenced the output bit rate that is transmitted, and this bit rate is greater than 500kbps and less than maximum C/I scheduler of conventional FIFO weighting and the scheduler of GB 0303859.3 with fixed weight of rough distribution.For this reason, preferably, in the mixed service scene, carry out the mathematic(al) manipulation (normalization) of formula (54) to (59) and (62) to (64), between all different aspects of provided QoS, to realize balance.

Figure 15 shows when having used the conversion of formula (54) to (59) and (62) to (64), specifically implements the performance of scheduler of the present invention.Suppose to be used to generate Figure 12 and result's shown in Figure 13 same rough set of weights and come this scheduler of initialization.Adjust these weights by scheduler then.

Obviously, adjusting controlling mechanism manages performance is brought up to than the high a lot of grade (referring to Figure 12) of disclosed fixed weight scheduler among the GB 0303859.3.

The Throughput Variance that Figure 16 is compared as follows three kinds of situations is over time: (a) the maximum C/I scheduler of FIFO weighting (b) has the fixed weight scheduler of GB0303859.3 of the weight of rough distribution, and (c) specifically implements scheduler of the present invention.As can be seen, specifically implement scheduler of the present invention and have efficiently response fast for out-of-control phenomenon.

The curve chart of Figure 17 will compare over time for the QoS of three kinds of different schedulers (a) to the video user of (c).Clearly, specifically implement scheduler of the present invention and still have efficiently response fast for information out of control.

Figure 18 (A) to other performance between three schedulers of 18 (F) expression (a) to (c) relatively.

The represented curve chart of Figure 18 (A) is used for the total throughout (WWW and video) of these three schedulers of comparison.These lines are represented the cumulative density function (CDF) of the throughput of all UE.

The curve chart of Figure 18 (B) is used for the average throughput of these three schedulers of comparison.These lines are represented the average CDF of throughput of the UE1 to 10 of receiving video service.

The curve chart of Figure 18 (C) still is used for the average WWW throughput of three schedulers of comparison corresponding to Figure 18 (B).These lines represent to receive the average CDF of throughput of the UE 11 to 90 of WWW business.

The curve chart of Figure 18 (D) is used for the QoS of the video user of three schedulers of comparison.

The curve chart of Figure 18 (E) is used for the WWW user's of three schedulers of comparison QoS.

The curve chart of Figure 18 (F) is used for the total throughout of all UE of three schedulers of comparison.

Following table 4 is represented to compare for the performance of first simulation example with the form that quantizes.Performance comparison index is transmission bit rate (60 seconds total throughouts are removed by 30,000 TTI), average throughput, average retardation, the QoS of WWW service and the QoS of Video service.Average throughput is the mean value by the single throughput of formula (23) definition.Average retardation is each mean value that postpones that the grouping successfully transmitted stands.Each professional QoS is eight hytes that are transferred to relevant UE of zero defect in the tolerable delay tolerance of related service and the ratio of eight hyte sums that arrive Node B.

Table 4

Scheduler	(a)	(b)	(c)
				Transmission bit rate	1.969Mbps	2.186Mbps	2.1921
Average throughput	0.9127	0.9731	0.998
				Average retardation	4.5966 second	1.4563 second	0.6071
The QoS condition (WWW) that satisfies	0.6331	0.7949	0.8758
				The QoS condition (video) that satisfies	0.4958	0.6814	0.9783

How concrete enforcement scheduler of the present invention (c) has improved all system performance characters effectively as can be seen from Table 4.Even in short time period, it also can be successfully to the fair output of all wireless terminals.

The empty weights group of simulation cycle last moment is:

Virtual_W ₁(30000)＝4.79，Virtual_W ₂(30000)＝0.4881，

Virtual_W ₃(30000)＝38.4451，Virtual_W ₄(30000)＝9.6，

Virtual_W ₅(30000)＝30.87

Virtual_V ₁(30000)＝4.79，Virtual_V ₂(30000)＝-9.411，

Virtual_V ₃(30000)＝38.4451，Virtual_V ₄(30000)＝4.965，

Virtual_V ₅(30000)＝30.87

The real power value group of simulation cycle last moment is:

W ₁(30000)＝4.12，W ₂(30000)＝3.37，W ₃(30000)＝10，W ₄(30000)＝4.96，

W ₅(30000)＝8.68

V ₁(30000)＝4.12，V ₂(30000)＝1.64，V ₃(30000)＝10，V ₄(30000)＝4.96，

V ₅(30000)＝8.68

Figure 19 (A) shows three schedulers (a) respectively to (c) performance aspect the CDF of the throughput of each independent UE in simulation cycle to 19 (C).In 19 (C), dotted line is represented the UE of video user at Figure 19 (A), and solid line is represented WWW user's UE.As can be seen, implement in the scheduler of the present invention (c) concrete, it is fast a lot of that transmitted in packets is wanted.

Next, will describe second simulation example, in second simulation example, the condition of the assumed condition and first simulation example is similar substantially.But, in second simulation example, supposed one group than employed more suitable " well " initial weight group in first simulation example.In GB 0303859.3, disclose when to the WWW browse application at a slow speed when mechanism employed such one group of " well " set of weights, that is:

W ₁(1)＝1，W ₂(1)＝10，W ₃(1)＝1，W ₄(1)＝1，W ₅(1)＝1

V ₁(1)＝1，V ₂(1)＝1，V ₃(1)＝1，V ₄(1)＝10，V ₅(1)＝1

In following result, suppose that " well " set of weights is for being applied to the fixed weight group of disclosed scheduler (b) among the GB 0303859.3.

The Figure 20 (A) that is used for second simulation example is to the situation of 20F corresponding to the Figure 18 that is used for first simulation example (A) to 18 (F).As can be seen, by using " well " set of weights (fixed weight) to improve the performance of the scheduler (b) among the GB 0303859.3 greatly.But,, specifically implement scheduler of the present invention (c) and still make performance obtain further raising, specifically referring to Figure 20 (D) by adjusting weight even in this case.

Following table 5 is similar with table 4 above, has compared scheduler (b) and the performance (c) that is used for second simulation example with the form that quantizes.

Table 5

Scheduler	(b)	(c)
			Transmission bit rate	2.32Mbps	2.1921
Average throughput	0.9869	0.9982
			Average retardation	0.5697 second	0.6172
The QoS condition (video) that satisfies	0.8994	0.9590
			The QoS condition (WWW) that satisfies	0.8658	0.8535

From Figure 20 (E) and table 5 as can be seen, in order to provide better service quality to video user, specific implementation scheduler of the present invention (c) has reduced the quality of services for users to WWW.This is because offer the fixed service priority parameters (β of video traffic ₁=10) than the priority parameters (β that offers the WWW business ₂=1) high a lot.Figure 20 (D) illustrates the QoS decline how concrete enforcement scheduler of the present invention successfully prevents video traffic.

Next, consideration relates to the 3rd simulation example of the scene of traffic load variation.In the 3rd example, always have 50 UE, wherein UE1 to 10 is video user, UE 11 to 50 is WWW users.Figure 21 illustrates the arrival of the grouped data of different UEs in the 3rd simulation example.In the beginning of simulation cycle, admit all 10 video user, admit a WWW user (UE50).After 10 seconds, allow preceding 19 WWW users to begin the WWW browsing session.After 30 seconds, allow residue WWW user to begin the WWW browsing session.

Figure 22 (A) to Figure 22 (F) show three schedulers (a) to (c) in the 3rd simulation example results of property.These curve charts correspond respectively to be used for first simulation example Figure 18 (A) to 18 (F).

From Figure 22 (E) as can be seen, when admitting new WWW user after 10 seconds, the performance of scheduler (a) aspect QoS significantly descends.The concrete scheduler of the present invention (c) of implementing has been avoided this situation, and provides the scheduler (b) than fixed weight stronger robustness.

The variance that Figure 22 (A) shows single throughput in whole simulation cycle has reduced and the fairness transmitted has improved, and can observe the balanced action of concrete enforcement scheduler of the present invention in Figure 22 (A).In addition, the average throughput curve among Figure 22 (B) and 22 (C) has been confirmed to compare with fixed weight scheduler (b), adjusts the performance that weight has improved.These curve charts also show with scheduler (b) and compare, and scheduler (c) can be video traffic and the better QoS of WWW business realizing simultaneously.Figure 22 (F) illustrates concrete enforcement scheduler of the present invention (c) how better instantaneous total throughout successfully is provided in all simulation cycles.

Similar with top table 4 and table 5, following table 6 compares with the performance that the form that quantizes shows the 3rd simulation example.

Table 6

Scheduler	(a)	(b)	(c)
				Transmission bit rate	1.074Mbps	1.2487Mbps	1.2861Mbps
Average throughput	0.9203	0.9960	0.9973
				Average retardation	1.3508 second	0.483 second	0.2639 second
The QoS condition (WWW) that satisfies	0.7276	0.9283	0.9592

The QoS condition (video) that satisfies

0.8322

0.9662

0.9770

The 3rd simulation example shows under the traffic carrying capacity scene of heavy load, specifically implements scheduler of the present invention and can improve total throughout, the QoS of business and the average transmission of grouping simultaneously, increases the robustness of scheduler for the traffic load variable effect simultaneously.In dynamic mobile cellular environment, a plurality of new calling with different QoS requirement and delay tolerance can be admitted in the base station.Therefore, as time passes, the characteristic acute variation of incoming traffic amount load.The required scheduler of operation is faced a plurality of input channels under this environment, and aspect bit rate, some input channels are scape calibrations, and some then change.No matter successful packet scheduler is the variation of these input bit pipelines in the mixed service environment, can provide stable output quality (with in the wired connection relative, realize with wireless environment similar) scheduler.Result shown in Figure 22 (A) to (F) has confirmed that concrete enforcement scheduler of the present invention can provide stable " wired " output quality in the whole transmission cycle.

In the above in the embodiments of the invention of Xiang Xishuominging,, then each weighted array of different receivers is compared mutually the receiver that will send to the specified data grouping for each different receiver generates the weighted array of first index.But, in other embodiments, the present invention can also be applied to the grouping scheduling method described in the following document: S.Abedi, S.Vadgama, " HybridGenetic Packet Scheduling and radio Resource Management for HighSpeed Downlink Packet Access ", WPMC 2002 Conference, Hawaii, pp.1192-1196 and S.Abedi, S.Vadgama, " A Radio Aware Random IterativeScheduling Technique for High Speed Downlink Packet Access ", VTC2002, Fall, vol.4, pp.2322-2326,24-28, Sept.2002.For example, in one list of references of back, generate a plurality of candidate's scheduling scheme η ₁To η _nEach candidate scheme η has indicated at least in the scheduling of being considered (as TTI) packet constantly will send to receiver.In these candidate schemes at least one generates at random.For each candidate scheme η, according to grade of fit function calculation " unitized grade of fit index "

$f\left(\mathrm{\&eta;}\right)=\frac{W_{1E}+W_{2E}\&CenterDot;u(\mathrm{Eff}\_\mathrm{Oct})}{W_{1E}}\&CenterDot;\frac{W_{1D}+W_{2D}\&CenterDot;v(\mathrm{Delay}\_\mathrm{Profile})}{W_{1D}}.$

$\frac{W_{1F}+W_{2F}\&CenterDot;x\left(\mathrm{Fairness}\right)}{W_{1F}}\&CenterDot;\frac{W_{1R}+W_{2R}\&CenterDot;y(\mathrm{Ratio}\_\mathrm{Waiting}\_\mathrm{Oct})}{W_{1R}}.$ (65)

Wherein u (.), v (.), x (.) and y (.) are mapping functions, W _1E, W _2E, W _1F, W _2F, W _1D, W _2D, W _1RAnd W _2RBe weight coefficient, Eff_Oct, Delay_Profile, Fairness and Ratio_Waiting_Oct are first indexs in this case.If the first index Eff_Oct is to having selected scheme η then the estimation of eight hyte quantity will transmitting.The first index Delay_Profile is the index of the delay that does not send eight hytes experience the earliest of transmission not by arriving transmitter but as yet.The first index Fairness is the index of expectation candidate scheme η to the contribution of the fairness of packet scheduling processing.The first index Ratio_Waiting_Oct is the index of the ratio of eight hyte sums arriving in medium eight the hyte quantity to be sent of transmitter and the formation of Node B source in the correlation candidate scheme.

Find in the paper about the inventor of the more information reference in the above of the mode that how to obtain these first indexs, incorporate it by reference in full at this.

Should be appreciated that first index in the present embodiment is not that each receiver ground calculates.But each candidate's scheduling scheme ground calculating.Usually, this expression must be considered the aspect more than the performance of a UE when generating first index.For example, may need to consider the many aspects of the scheduling performance of all UE in candidate's scheduling scheme.

From equation (65) as can be seen, the grade of fit function is for each candidate's scheduling scheme rather than be the weighted array of first index that generates under the situation of each receiver.In case generated fitness function separately (weighted array) for different candidate's scheduling schemes, then they compared mutually to select a best candidate scheme η according to comparative result _Best

At S.Abedi, S.Vadgama, " Hybrid Genetic Packet Scheduling andradio Resouce Management for HighSpeed Downlink Packet Access ", WPMC 2002 Conference, Hawaii has used similar grade of fit function in the grouping scheduling method of describing among the pp.1192-1196.In this case, grouping scheduling method is based on genetic algorithm, and as the part of the generation parents' that select to have child of future generation processing, the fitness function that will be used for different candidate schemes compares mutually.Thus, in this case, in a series of generations, use this grade of fit function, to arrive the optimal candidate scheduling scheme.Yet, also can adopt fitness equation (weighted array) to determine the receiver that grouping will send to.The details that adopts the mode of grade of fit equation is provided in inventor's paper, has incorporated it by reference in full at this.

Though toply described example of the present invention relatively, should be appreciated that the present invention also can be applied to wherein occur any other network of scheduling problem with the wideband CDMA network with asynchronous packet pattern.These networks can be other cdma network such as the IS95 network, perhaps can be the network according to these networks adjusted.These networks also can be for not using other mobile communications network of CDMA, the perhaps network that adjusts of these networks for example uses the network of the many access techniques below one or more: time division multiple access (TDMA), wavelength division multiple access (WDMA), frequency division multiple access (FDMA) and space division multiple access (SDMA).

Though described embodiments of the invention with a plurality of " unit " with difference, but it will be appreciated by those skilled in the art that all or some function that to use microprocessor or digital signal processor (DSP) to realize base station (Node B) in the embodiment of the invention and/or subscriber equipment in the practice.

Claims (37)

1, a kind of grouping scheduling method is used for grouped data is dispatched to the transmission of a plurality of receivers by at least one channel from transmitter, and this method may further comprise the steps:

Generate first index of scheduling performance at least two different aspects, and the receiver that uses the incompatible definite grouping of set of weights of first index to be sent to, described weighted array generates by according to each respective weights first index being made up;

Generate at least one second index of scheduling performance;

At each described second index, relevant may influencing of second index is categorized as at least two kinds of different weight classifications with weight according to the weight of expecting; And

Use each described second index and adjust described weight at the weight classification of each second index.

2, method according to claim 1 wherein, is by may influencing the favourable classification that constitutes for each positive weight to the described of described second index at a described weight classification of each second index, if present.

3, method according to claim 1 wherein, is by may influencing the unfavorable classification that constitutes for each negative weight to the described of described second index at a described weight classification of each second index, if present.

4, method according to claim 1 wherein, is by may influencing the neutral classification that uncertain each weight constitutes to the described of described second index at a described weight classification of each second index, if present.

5, method according to claim 2 wherein, if second index reduces, then will increase at the weight in the described favourable classification of this second index.

6, method according to claim 2 wherein, if second index increases, then will remain unchanged at the weight in the described favourable classification of this second index.

7, method according to claim 2 wherein, if second index is constant, then will increase at the weight in the described favourable classification of this second index.

8, method according to claim 3 wherein, if second index reduces, then will reduce at the weight in the described unfavorable classification of this second index.

9, method according to claim 3 wherein, if second index increases, then will remain unchanged at the weight in the described unfavorable classification of this second index.

10, method according to claim 3 wherein, if second index is constant, then will reduce at the weight in the described unfavorable classification of this second index.

11, method according to claim 4, wherein, any variation in second index all will remain unchanged at the weight in the described neutral classification of this second index.

12, require each described method among the 1-11 according to aforesaid right, further comprising the steps of: as to use the historical record of each second desired value value in time to determine that this second index still is constant in increase, reduction.

13, require each described method among the 1-11 according to aforesaid right, wherein:

Each described weight has first value and second value separately, and described first value is used to generate the weighted array of first index and is subjected to one or more predetermined restricted, and described second value is not subjected to described restriction; And

Come to calculate the second new value based on described classification and each described second index according to the second previous value; And

Come to obtain the first new value to guarantee the mode that described first value satisfies described predetermined restricted according to second value.

14, method according to claim 13, wherein, a described restriction is that each first value must be more than or equal to zero.

15, method according to claim 13, wherein, maximum first value that described restriction is any described weight must be not more than predetermined maximum value.

16, method according to claim 15, wherein, described predetermined maximum value is maximum first value in the initial weight group.

17, require each described method among the 1-11 according to aforesaid right, wherein, at least one described first index is independent of another at least one the described aspect in described first index of influence.

18, require each described method among the 1-11 according to aforesaid right, wherein, influence described first index at least to the priority of a kind of traffic assignments that offers receiver.

19, require each described method among the 1-11 according to aforesaid right, wherein, generate the weighted array of at least one this first index at each receiver, and use the receiver that incompatible definite grouping will send at different receivers set of weights separately.

20, method according to claim 19, further comprising the steps of:

Generate first weighted array and second weighted array of this first index separately at each described receiver;

First weighted array is separately carried out rank to receiver according to receiver, and forms the receiver tabulation according to the receiver rank order; And

According to receiver second weighted array separately, the receiver of the highest rank from described tabulation is to the receiver allocated channel in the tabulation.

21, method according to claim 20, wherein, generate first weighted array by first index being made up according to corresponding first weight separately, and generate second weighted array by first index being made up according to corresponding second weight separately, and described first weight and/or described second weight classified, and it is adjusted according to described second index and described classification.

22, according to each the described method in the claim 1 to 11, may further comprise the steps:

Generate a plurality of candidate's scheduling schemes, each scheduling scheme is the designated packet receiver that will be sent at least;

Generate this first index and this weighted array at each candidate scheme; And

To the different receiver of candidate scheme use weighted array separately to determine that grouping will be sent to.

23, method according to claim 22, wherein, at least one candidate scheme is to use genetic algorithm to generate.

24, method according to claim 22, wherein, at least one described candidate scheme generates at random.

25, method according to claim 22 wherein, can use a plurality of channels will divide into groups to be sent to receiver from transmitter, and how each described candidate scheme specifies the receiver to appointment to distribute available channel.

26, method according to claim 22, wherein, described weighted array is the grade of fit function.

27, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is that in tolerable postpones threshold value data to be sent to correlation receiver be success or failure.

28, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is the channel quality between transmitter and the correlation receiver.

29, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is to estimate at how many data can successfully be sent to correlation receiver.

30, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is to have how many data waiting at the transmitter place to be sent to correlation receiver.

31, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is that data are sent to delay in the process of correlation receiver.

32, require each described method among the 1-11 according to aforesaid right, wherein, an aspect in the described different aspect is commercial aspect.

33, require each described method among the 1-11 according to aforesaid right, a series of scheduling are carried out constantly iteratively, wherein,, generate new weighted array, and will send to which receiver to dividing into groups and newly adjudicate for each described scheduling moment.

34, require each described method among the 1-11 according to aforesaid right, wherein, described transmission is wireless transmission.

35, require each described method among the 1-11 according to aforesaid right, wherein, described transmitter is the part of the base station of wireless communication system, and each described receiver is the part of the subscriber equipment of this system.

36, a kind of packet scheduling apparatus is used for grouped data is dispatched to the transmission of a plurality of receivers by at least one channel from transmitter, and this device comprises:

Generate scheduling performance in first index of at least two different aspects and use the device of the receiver that the incompatible definite grouping of set of weights of first index will send to, described weighted array generates by according to each respective weights first index being made up;

Generate the device of at least one second index of scheduling performance;

May influence the device that weight be categorized as at least two kind different weight classifications according to the weight of expection to relevant second index at each second index; And

The device that uses each second index and adjust weight at the weight classification of each second index.

37, a kind of transmitter comprises:

Packet scheduling apparatus as claimed in claim 36; And

Dispensing device operationally is connected with described packet scheduling apparatus, and can operate grouping is sent to the receiver that packet scheduling apparatus is determined.

Images