April 3, 2018
Drive to quality in autonomous vehicle market
by Paul Carroll
Another week, another controversy on autonomous vehicles. The latest involves the crash that killed a Tesla driver on March 23 while the car was in Autopilot mode. After being silent for several days, Tesla on Friday wrote a blog post assigning much of the blame, if not all, to the driver. In a blog post, the company said he had received several visual warnings and one audible warning that he needed to retake the wheel but had not done so. The post said the driver had five seconds to react and 150 meters of unobstructed view of the concrete divider that he crashed into.
In the wake of these recent crashes, I think there will be a flight to quality, which doesn’t bode well for either Uber or Tesla in the short term. Uber, which has been known as a cowboy, including in driverless cars, has been scrambling since a car in autonomous mode struck and killed a pedestrian in Tempe, AZ, on March 18. It’s not clear to me that there’s a fundamental flaw in Uber’s technology, but the company seems to be moving too fast. It will have to scale back testing and increase security (there was just one safety driver, rather than the usual two, in the car that struck the pedestrian). Uber will also have to pound away at testing a lot more, both on simulators and in the real world, before putting any significant numbers of cars on the road.
Tesla may need to reboot even more. It claims it can get to full autonomy without using lidar, which the rest of the industry is treating as essential. The rotating lidar bulbs on the top of driverless cars surely don’t fit the sleek design esthetic at Tesla—maybe lidar becomes acceptable once it can be delivered via just a couple of chips built into bumpers. But solid-state lidar is at least a few years away from being possible in production models, and I don’t believe that Tesla can get close to full autonomy in the meantime. Tesla also needs to back off its belief that drivers can easily toggle between autonomous mode and full engagement with driving. Others aren’t counting on drivers, who seem to need 10 seconds or more to reengage, and I think Tesla must change its thinking.
Companies taking a more deliberate approach will have to contend with the negative publicity from the recent Uber and Tesla fatalities. But I see no signs that Google’s Waymo or GM’s Cruise are backing off their plans to launch ride-sharing AV services this year and next in limited, well-mapped environments.
Vivek Wadhwa published an article with us that argues for a moratorium of at least two to three years on any testing around pedestrians. He owns a Tesla and is comfortable using it in Autopilot mode on highways but says local streets are too complicated for now.
Our own Guy Fraker conducted a thought-provoking webinar on the myths and realities of autonomous vehicles that you can listen to here. He provides great perspective both on how far we’ve come and on how far we still have to go.
Have a great week. (And drive safely.)