The future is almost here when it comes to autonomous vehicles. And yet like a traffic jam on the freeway a few hundred feet from the exit, the wait continues. Easing that path appears to be highly dependent on resolving key business challenges, such as liability and cost control, while embracing new technologies like simulation.
On the business front, “the biggest barrier to autonomous vehicles is liability,” noted Dr. Derek Riley, associate professor in the Electrical Engineering and Computer Science Department at MSOE in Milwaukee. “Currently, there is not a clear model for how liability will be determined when an autonomous vehicle is involved in a collision. And regardless of the capabilities of the AI, an accident is inevitable. Companies willing to assume the liability will be first to market. But they also take a major risk in doing so.”
Another financial challenge could be the pricing of autonomous vehicles.
“Given the unsure legal and regulatory, environment, though, it can be very difficult to figure out what the true cost of delivering autonomous vehicles will be,” said Ryan McMaken, Editor of the Mises Wire and The Austrian at the Mises Institute in Alabama. “What will the cost of legal protections and insurance?” There are also behavioral questions as well, he added. “How quickly will consumers be willing to get into a car with no human driver?
“The industry no doubt thinks it can overcome these hurdles. But this is far more complex than simply coming up with a new way to deliver a new clothing line or a new appliance at an affordable price. To speed up the process, the industry will need to find ways to put insurers and lenders at ease while also convincing consumers to trust the product.”
That “trust” will likely come when consumers feel like they can believe in the technology.
This involves “ensuring that the devices work for all contingencies that present themselves on the roads and highways,” said Dr. Sharlene McEvoy, a Professor of Business Law at Fairfield University. “So far, the vehicles are confined to specific designated areas for testing. True tests will come when the vehicles are actually moving in real traffic conditions and in adverse weather conditions that may hamper the ability of the cars to navigate properly.”
Such diverse testing scenarios will be helped by simulation technology.
Celite Milbrandt is the founder and CEO and Austin-based monoDrive, which has developed a simulation technology that has been embraced by some of the largest car manufacturers in the world. monoDrive’s Ultra High Fidelity (UHF) Simulator reduces the need for using costly human drivers until it is absolutely necessary.
Thus far, simulation technology has been viewed as more of a stepping stone “leading to real-world testing,” according to Dr. Riley. “Simulations allow big errors to be found early, but they don’t replace real-world testing. The real challenge is testing an autonomous vehicle to unexpected/challenging scenarios, and these ‘edge cases’ are likely the hardest to identify and simulate comprehensively.”
Milbrandt believes that paradigm has already shifted. “Actually, we can put a vehicle in a testing scenario that you would never put a human test driver in, such as approaching a semi crossing the road as you are going 80 MPH with the sun suddenly blaring in your eyes,” he said.
And the financial implications are clear.
“Computer simulators are useful since the cost of using human testers can be significant (using a computer simulation requires significantly less resources compared to actual physical cars + human testers),” said Dr. Alexander Wyglinski, Professor of Electrical and Computer Engineering at Worcester Polytechnic Institute. “Moreover, computer simulators can test scenarios consisting of 100s and 1000s of vehicles, while for human testers we are limited by the number of vehicles we can deploy on the road at any given time. Computer simulators can also provide results that are reproducible given the same scenarios.”
The biggest benefit of simulation technology, said Milbrandt in reiterating his earlier point, is that you can put autonomous vehicles in ”unexpected and challenging scenarios. These are the hardest to identify and model comprehensively. Furthermore, testing scenarios with complex electrical (for radar) and lighting (for Lidar and Camera) conditions is either too expensive or not possible in the real- world environment.” Simulation technology addresses these limitations with comprehensive physically accurate testing scenarios.
Testing in a simulated world enables automatic labeling of “ground truth”, for which perception systems for autonomous vehicles can be directly verified, according to Milbrandt, and ultimately building trust in the AV’s ability to perceive and react appropriately in the real world.
And that will speed the adoption of AV, according to DINESH C, Lecturer for the Entrepreneur Program at University of California Berkeley. He added that industry must instill “confidence in consumers that AV is ready. At this point, there are too little info to demonstrate that AV is working well. Data such as mileage driven and alerts captured do not translate well to people’s confidence. There needs to be some standardize parameters that can be shared among all parties and that is easily understood by the general public.”
Simulation technology may satisfy that need, given the accuracy and comprehensiveness of the data that can be collected in a testing environment.