Like many manufacturers, autonomous vehicle (AV) developers suffered through a 2021 that was rocked by semiconductor shortages, global supply chain disruptions, eroding customer confidence, and other challenges.
Heading into 2022, the AV industry’s biggest challenges will continue to be a disrupted supply chain, chip shortages, and a skeptical public. AVs rely on AI technology in the form of graphics processor units (GPUs) to handle deep learning and machine learning tasks. “Those chips are advancing with Qualcomm’s SnapDragon being a big one and NVIDIA in the space as well with the TX2/Jetson models,” says Chris Mattmann, CTIO at the NASA Jet Propulsion Laboratory. “With the supply chain crisis that includes chip manufacturing, getting these chips and many of them per vehicle is even more important in the autonomous vehicle industry than it is in the consumer sector.”
Moving into 2022, many AV manufacturers hope to build trust with increasingly skeptical lawmakers and consumers. Phil Koopman, a Carnegie-Mellon University associate professor with appointments in the department of electrical and computer engineering and with the Robotics Institute, believes that Tesla's use of vehicle owners as “beta testers” is reckless and damaging to the image of the entire autonomous vehicle industry. “Reckless, because [drivers] are running stop signs, running red traffic lights, and veering across centerlines on public roads,” he explains. “Tesla is using civilian drivers who are neither specifically trained in testing safety nor operating according to best practices for road testing safety.”
Koopman says that the stance taken by the entire AV industry “to push back hard against any requirement to follow safety standards” further erodes public trust. He notes that manufacturers face a choice in 2022 and beyond. “They can continue to take an adversarial approach with regulators and have a problem when a high-profile crash forces regulators to intervene, or they can take a cooperative approach now while they still have time.”
An excellent first step, Koopman says, would be for AV developers to voluntarily agree to follow the SAE J3018 standard for safe road tests. “The industry itself wrote that standard based on lessons learned from the Uber ATG testing fatality in Tempe, Arizona, but there is not a single AV company that will publicly pledge to follow that standard.”
Driving to Level 5
The biggest challenge for vehicle manufacturers and their technology partners is developing models that can deliver a true autonomous driving experience. Within the AV industry, full autonomy is referred to as Level 5 Advanced Driver Assistance Systems (ADAS). “At Level 5, there is no human intervention required and the vehicle is fully capable of driving itself,” says Matt Desmond, automotive principal industry analyst at business advisory firm Capgemini Americas.
None of the AVs marketed to be sold in the next few years will incorporate Level 5 ADAS. “Delivering a truly autonomous vehicle—without steering wheel, accelerator, or brakes—is a steep technological and safety challenge, and there are many significant hurdles to achieving fully autonomous solutions,” Desmond says. In the meantime, leading vehicle manufacturers and technology firms are investing massive sums in developing, testing, and refining AV systems in an effort to mitigate technical issues and deliver a robust technology foundation, he notes.
As things currently stand, Level 5 ADAS vehicles may not reach market for at least several years. “The reality is that the core technologies of ADAS needs to mature to a point where virtually any scenario can be identified and addressed safely by the autonomous software,” Desmond explains. He notes that vehicle manufacturers and technology providers have already driven AVs for thousands of hours to train the onboard software to learn various driving environments. “However, there's still much more work to do, especially in scenarios where there is inclement weather, such as snow, mud, sand, or rain, that can possibly interfere with sensors.”
Several key issues need to be resolved before Level 5 ADAS vehicles can become mainstream transportation technology. Besides addressing the core technical challenges presented by code complexity, network latency, and hardware gaps, numerous market- and legal-oriented matters must be settled. “Taken as a whole, the industry and the ecosystems of business, law, policy, and culture have a long way to go to provide solutions for the mass market deployment of autonomous vehicles,” Desmond says.
In a sense, AV developers are facing a “chicken and egg” scenario, since many potential ADAS challenges can't be fully vetted and resolved until production AVs are released to market, Desmond says. “As real production dates for autonomous vehicles are announced, we believe we will see real traction for resolution of these issues from the car and technology companies developing ADAS products, the insurance industry, and federal, state and local regulatory agencies.”
While fully autonomous vehicles won't be generally available in the near-term, autonomous features will continue to be added to conventional vehicles, observed Raj Rajkumar, a professor in Carnegie-Mellon University's department of electrical and computer engineering and co-director of the General Motors-Carnegie Mellon Vehicular Information Technology Collaborative Research Lab. “The endpoint of full autonomy will not be an overnight revolution, but the final stop in an evolutionary path of progress with multiple milestones,” he says.
What to Read Next: