Post written by
Nitin Kumar
It is commonly believed that level 5 autonomy is at least a decade away. But ADAS (advanced driver assist systems) with geofenced autonomous vehicles is on the horizon in the next two to three years. Hardware is advancing quickly, with the cost of LIDAR dropping and GPU capacity going up significantly over the last five years. Currently, the architecture of the autonomous vehicle is not converging, the public and private sector are working in a siloed manner, standards are not widely accepted, and infrastructure is well behind even present-day needs. All these things make it challenging for accelerated AV (autonomous vehicle) adoption.
During my management consulting days, I had the good fortune of working on a few strategies and operations in the autonomous driving arena, and I am summarizing some key findings from that life here.
data-param-cid="62cec241-7d09-4462-afc2-f72f8d8ef40a"
data-player-id="44f947fb-a5ce-41f1-a4fc-78dcf31c262a"
data-playlist-id=aff7f449-8e5d-4c43-8dca-16dfb7dc05b9
data-elements-player="true"
layout="responsive"
width="16"
height="9"
>
Challenges With Autonomous Driving Adoption
There are several unsolved challenges to the mainstream adoption of AVs, ranging from sensing and identification to image processing and decision-making. Most of them are rooted in the evolution and integration of artificial intelligence with physical and technical infrastructure. Business models here have yet to evolve, stabilize and scale, so the mass adoption of ADAS will create more opportunities for monetization, as opposed to a limited install base. Developing an ecosystem is critical for business model development, as a tight ecosystem between auto OEMs, infrastructure providers, physical/digital infrastructure, cities and hardware and software vendors will create customer willingness to pay.
In this article, we will examine some of the key drivers and enablers of physical and digital infrastructure a bit further.
Physical Infrastructure Considerations
Let’s start with some physical infrastructure considerations.
Public Sector Infra Upgrades
As infrastructure upgrades occur, the public sector will need to analyze developments in AV technologies and weave in key enabling upgrades into their execution, such as sensors, intelligent street signs, navigation interfaces, signals and more — all aligned with applicable V2I standards.
Safety By Design
Potholes, inadequate stripes and infrequent and ad hoc maintenance measures create safety issues and challenges for AV adoption. Restriping to enable AV testing or retroreflective imaging can minimize the risk of cars going off-road. Organizations in this space also must partner with mapping software companies to make sure dimensions are accurate to the sub-inch level.
New Facilities
Fleets of AVs will need dedicated areas for maintenance, charging, idle time staging and servicing. The economics around these need to be established. Would private sector organizations (fleet operators) pay for these, or would cities/states pay for them if AV fleets are an extension of public transport? Major CAPEX costs here manifest around EV charging infrastructure, and major OPEX costs include space rent; both will need to be factored into business models.
Urban Planning
The design, aesthetics and locations of this infrastructure build, along with its safety and long-term scalability, will require careful planning to avoid disruption to traffic, the environment and other civic concerns.
Curb Redesign
Today, curbs are abundantly used for parking and pick-up points — and are often prominently accident zones. Creative and intelligent curb marking can designate efficiency in its use (at different times of the day). Organizations can also help monetize the space using taxes or other dynamic space-time pricing models.
Financial Considerations
Public transport in the U.S. is already lacking people, resources and funds — owing to budget issues with federal and state governments. There is already a backlog of highway upgrades due to a lack of capital and AV acceleration. Infrastructure will create further stress on this gap. Tax revenue from hydrocarbon fuels, DMV revenues and more could further add to this funding gap. Governments will have to think of new revenue streams that encourage the adoption of AVs, while not risking delays in funding the infrastructure needs of the new world.
Digital Infrastructure Considerations
On the other side of the coin, let’s take a look at some digital infrastructure considerations.
Test Infrastructure
Given the high volume of AV drive-testing across the world, this “test” infrastructure is a big business of its own and has been for several years. If data is core to making cars learn, then attributes like location precision, road conditions, weather conditions, anomalies and more require large amounts of data. The ability to collect, analyze, store, process and archive all of this data will make AV infrastructure very different from what companies have today.
NAS Limitations
Much existing automotive infrastructure relies on network-attached storage distributed globally to store data. Given the volume and velocity of anticipated AV data, this approach will become unscalable. Cost, scale and performance limitations will soon surface as data from fleets completely consumes conventional infrastructure. Organizations will need to look for alternate solutions, such as edge infrastructure, to make things more efficient and feasible.
Data Movement
If the scalability of data at rest is an infrastructure limitation for AV adoption, data in motion is even more complex. AVs should ideally be located near the infrastructure they need to align with jurisdiction laws and enable real-time use cases. Data in these instances will need to move in huge volumes, and movement at this scale will require network, storage and compute resources at an unprecedented scale. Everyone will need to rethink their investments in future state infrastructure to de-risk the development and learning of AVs. Edge computing-enabled colocation might alleviate some trade-offs when it comes to cost, performance, speed of development and agility.
Centralized Data Governance And Security
Even with decentralized and distributed infrastructure, the security, privacy and governance of data must be centralized, which allows for the control, logging, monitoring and reporting of said data. If different teams across the globe are working on different projects — for example, auto-braking in the U.S. and lane changing in Germany — data quality, consistency, integrity, availability and synchronization will be critical. A fractured data footprint makes centralized policies and the management of privileges, traceability, audit trails and more harder to manage across multiple clouds and on-premise locations. Centralized governance here is the way to go.
The Future
Because there are several financial, technical and operational barriers to be overcome prior to mainstream adoption, the evolution of physical and digital infrastructure is vital. Private-public partnerships, new revenue streams, cohesive business models and the evolution of scalable edge computing infrastructure will be fundamental in laying strong foundations for the future of autonomous vehicles.