No Joy Ride Yet for Autonomous Vehicles
Artificial Intelligence is emerging as an important tool for improving the safety of autonomous vehicles, thereby paving the road to adoption.
In a much-publicized judgement earlier this year, California regulators permitted the autonomous vehicle fleets of General Motor’s Cruise and Alphabet Inc’s Waymo to run paid taxi services at any time of day or night. But San Francisco is only one of three cities, the other two being Phoenix and Austin, that are allowed a driverless robotaxi service. That said, autonomous vehicles – which were slated to grow in value from $12.27 billion in 2018 to $31.17 billion in 2028 – have encountered many obstacles on the road to growth.
Need to dismantle barriers
A combination of technical, market and regulatory barriers stand in the way of vehicles transitioning from the current Level 1/2 autonomy – where “advanced driver assistance” helps people with steering and braking/ accelerating to different degrees – to full automation, where the vehicle drives itself.
Messagepoint Announces Generative AI Capabilities for Translation and Plain Language Rewrites
Doubts around technology:
Ironically, technology is one of the reasons why driverless vehicles haven’t progressed faster. The LiDAR sensors, radar systems, cameras and artificial intelligence software used in AVs must be capable of “sensing” the environment – traffic signals, road signs, other vehicles, and above all, pedestrians – and navigating it with high accuracy, even in difficult weather and traffic conditions. In areas with “mixed” or “unpredictable” traffic – where not just vehicles, but also pedestrians, kids at play, or even stray animals are on the streets – the performance of these technologies is yet to prove reliable.
Stalled legislation:
In the U.S., a six-year-old impasse that has blocked key AV legislation in Congress shows no signs of getting resolved. Questions around legal liability – who is to blame if an AV is involved in an accident – and insurance coverage continue to elude legislative consensus. Also, there is an urgent need to implement universally accepted standards for in-vehicle security and automated driving systems safety, such as UNECE WP.29 and ISO 22737.
Public hesitancy:
Resolving these two barriers will mitigate the third challenge, namely the lack of public acceptance, to a great extent. It is worth noting that consumers’ enthusiasm for all things digital doesn’t quite extend to driverless vehicles. In a 2020 survey of nearly 1,200 Americans, 48 percent said they would never travel in an AV, and 20 percent said these vehicles would never be safe. But 58 percent thought AV safety was about ten years away; hence improving the technical, safety, and security aspects of the vehicles would go a long way in building trust and adoption.
The U.S. needs to speedily address these issues because growth in AVs is beneficial for individuals, the automobile industry, and society in general. For drivers and passengers, these vehicles make the on-road experience safer, easier, and more enjoyable. More than 40,000 people die in road accidents every year in the U.S., of which more than 90 percent are caused by human error. AVs can make a huge difference to safety since they are 50 percent less likely than humans to be involved in a collision. People can use the time inside the car to work, connect with others, or entertain themselves. Since they can be equipped with sophisticated features, self-driving vehicles could be a very good option for the elderly or those with mobility challenges. Indeed, with autonomous driving systems evolving rapidly towards Level 3 and Level 4 autonomy, the extra features, components, and related commercial solutions could add $300 billion – $400 billion to the passenger car market by 2035.
Top AI Pain Points that Challenge UK-based Marketers in 2023
Also, because self-driving vehicles are powered by electric or hybrid engines, they lower greenhouse emissions and further help the environment through their optimized driving behavior, energy-friendly acceleration, and lower fuel consumption.
AI can help in Fixing the Safety in Autonomous Vehicles
Artificial Intelligence is emerging as an important tool for improving the safety of autonomous vehicles, thereby paving the road to adoption.
Safety testing is a highly complex activity, requiring massive financial resources and time; testing a vehicle’s performance in a natural driving environment, reflecting “real” conditions, may require hundreds of billions of testing miles. But now, breakthrough research suggests that using AI to train vehicles could potentially slash the testing miles requirement by 99.99 percent. The problem in testing AVs is that safety-critical data is a minuscule proportion of the overall massive data, which makes safety-critical events very rare during testing; to address this, researchers isolated the small-sized safety-critical data and used only that to train neural networks, dramatically accelerating the learning (and therefore, testing) process. Although the study focused on road geometry and moving objects, the model can be extended to test for extreme weather events and even for complex road environments (many highways, intersections, etc.) which are beyond the scope of existing testing methods.
Autonomous vehicles are the future of mobility; to seize it, the U.S. must address the technical, regulatory, and consumer barriers to adoption. AI can be a key enabler of this agenda.
Comments are closed.