On recent Uber’s driverless car kill, I asked two experts for comments:
– Vivek Wadhwa: Author of Driver in the Driverless Car, a distinguished fellow at Carnegie Mellon University Engineering at Silicon Valley (Menlo Park, CA)
– Dr. Qufei Wang: Founder of EVSoft, building sharable driverless car (Guangzhou, China)
Both Vivek and Qufei agree self driving will eventually save lives. Here goes their different arguments.
Vivek: The race to self-driving needs to stop
“It is clear that the technology was at fault. Neither Lidar nor radar — nor camera technologies have progressed to the point of perfection yet. It will take several years before they reach this point — but they will. It won’t be an easy transition and there will likely be many more accidents along the way. But then, tens of thousands of lives will be saved in the US alone and millions on a worldwide scale.
The lesson here is that the self-driving technology race needs to be stopped. Companies are taking unnecessary risks because they are competing with each other to be the first. It is better if we take 5 years longer and minimize the risk of casualty.
This gets back to the theme of my book, Driver in the Driverless Car: we have to make the difficult choices between the risks and rewards. I have advocated slowing down some technologies to make sure that we get these right.” — Vivek Wadhwa
Qufei: It’s Lidar’s fault
“All accidents reveal some technical problems. This one was about the inherent fault of Lidar. Lidar is static and it does not see color, so it is difficult for for Lidar-based autopilot (such as Uber’s) to track people… . Although Uber could easily set a higher margin for Lidar safety distance, however, doing so is ok on highway but will make the car unable to move in populated area, because Lidar sees an object as point cloud. In a crowded environment, Lidar can’t segment or track object, therefore can not predict a collision.
All accidents teach the human society a valuable lesson. The first Tesla fatal accident revealed the fault of monocular vision. This time it is Lidar.
Car industry is all about safety. Vision, Lidar and Radar all fused together is the correct way to go. The three technologies are complementary. What is easy in vision is difficult in the other two, and the vice versa. The Lidar and Radar design philosophies should be changed with consideration of vision.
Done right, driverless cars may be the biggest innovation of this century. We can benefit so much. — Dr. Qufei Wang