It was a landmark year in the development of automated driving systems in 2018, but not necessarily in the way that many advocates of the technology had been hoping. The number of pilot deployments of automated vehicles (AVs) increased and several companies even began revenue-generating services. At the same time, there is a growing recognition and acknowledgement that true wide-scale deployment of AVs will take far longer than the most optimistic projections of 1-2 years ago.
Companies including Aptiv, May Mobility, and Waymo are all generating revenue by providing rides in vehicles that are nominally self-driving. Others such as Voyage have expanded pilot programs while still more, including GM Cruise, plan commercial launches in 2019.
Automated Driving Tech Faces Challenges
However, there were also human fatalities related to automated driving technologies. An Uber test vehicle in Arizona struck and killed a pedestrian in March 2018. Days later, a Tesla driver in California died when the AutoPilot assist system ran the vehicle into a highway barrier after he failed to respond to alerts to take control.
Despite having what is widely regarded as the most advanced automated driving system in the business, even Google spin-off company Waymo is facing its own challenges. The widely touted launch of the Waymo One service in Chandler, Arizona in December turned out to be much smaller in scale than anticipated and still features safety drivers. At the summer National Governors Association conference, Waymo CEO John Krafcik proclaimed that AV deployments will “take way longer than you think.” At another conference in November, Krafcik told the audience that AVs that can operate anywhere and anytime (aka level 5) would likely be at least 10-15 years away and may never arrive.
The technical requirements for these systems (from sensing to computing power) will also be a bigger challenge than anticipated. Two years ago, Tesla launched its AutoPilot V2 hardware platform with the proclamation that it would provide level 5 capability within about 2 years with just software updates. Since then, Tesla has already upgraded its compute platform once and plans another order of magnitude upgrade in 2019. Nvidia and Intel have similarly ramped up computing expectations for the various levels of automation.
Despite how negative this harsh dose of reality may seem, it’s actually a very good thing. While the number of fatalities per 100 million miles driven has fallen by more than 80% since the early 1970s, the current rate of injuries, fatalities, and the associated societal costs remain too high. However, we are currently still far from proving that AVs can actually do better than humans.
A Measured Approach for the Best Results
Rather than the startups that sprouted in recent years in hopes of a wholesale takeover of the auto industry, engineers are now turning their attention to addressing specific problems within this larger space. This includes numerous companies focused on areas like simulation, development of more efficient neural network algorithms, and prediction engines that read human body language to better anticipate what pedestrians are likely to do.
Rushing headlong to deploy technology that is not ready for the real world risks turning off the public for a much longer period if it causes more accidents and deaths. Taking a more measured approach that increasingly factors in how humans react to and interact with technology may take a bit longer, but will likely pay off in the end with greater adoption and improved safety for everyone.
https://guidehouseinsights.com/news-and-views/avs-are-fully-into-the-trough-of-disillusionment-but-there-is-a-light-ahead
2019-01-16T00:00:00+00:00