Last month, Sung Jo, a representative for Nuvation, Sunnyvale, Calif., sent me news about new efforts to get autonomous vehicles on the road. Our sister publication Electronic Design covers autonomous vehicles and the company regularly — in articles about its work on batteries, robotics, and other splashy announcements. (Here's one more picture of Nuvation's technology on display late last year.)
Now, the company has signed a four-year R&D agreement with researchers from the University of Waterloo and the Waterloo Autonomous Vehicle Laboratory to address some barriers to autonomous-vehicle design.
According to Nuvation, within the next five years, fully autonomous vehicles could become as common as hybrid and electric cars are today. Why bother? Well, it seems human error is an epic source of waste. (Shocking, I know.) A study by the Eno Center for Transportation (a think-tank that "promotes policy innovation") estimates that if 90% of U.S. vehicles were autonomous, it would save 724 million gallons of fuel (kind of an extension of how some of today's automatic transmissions can get better fuel economy than manual) and prevent 4.2 million accidents annually. True, many consider the idea of a car driving on its own using sensors and complex algorithms through traffic to be bizarre and scary … but indeed, all the technology needed for driverless vehicles already exists. It’s merely a question of how soon autonomous design engineers can advance the hardware and software.
Among the myriad details for engineers and technicians to work out include algorithms for platooning, emergency braking, and the ability to detect background objects.
Platooning is when vehicles travel together in a synchronized form, even in congested traffic — usually by communicating by wireless signals and sensors to maintain good spacing. Good platooning maximizes the number of vehicles on a road to improve safety and decreases the consumption of fuel. The main algorithm for coordinated platooning is called Cooperative Adaptive Cruise Control (CACC). Through it, vehicle-to-vehicle communication and sensors work together to optimize car spacing and coordinate smooth lane changes and merging. In fact, CACC algorithms are still being tested to increase autonomous vehicle stability.
According to Nuvation, another challenge for engineers to overcome is that of emergency braking. The system currently under the most testing is an Autonomous Emergency Braking System (AEBS), which includes driver warnings and brake-assist functions. Program and hardware detect imminent front-end collisions, judging the severity based on speed and distance. Then brakes engage automatically if danger is detected. Without the help of constant vehicles-to-vehicle communications, however, AEBS can't precisely detect the severity of an oncoming incident.
The last challenge is Object Detection. Modern detection systems used during experimenting with autonomous vehicle design utilize both LIDAR-based sensors, which implements laser reflections to detect objects, and vision-based optical hardware to differentiate streetlight signals and stop or yield signs. Such systems can have trouble seeing unexpected objects, and bad weather sometimes garbles reliability.
But once these systems are perfected, look out! — for perfectly under-control driverless cars.