The Center for Automotive Research (CAR)’s annual Management Briefing Seminars opened yesterday in Traverse City, Michigan. The afternoon sessions included a robust discussion of “Connected and Automated Vehicles – The Future is Now (Almost).” This title reflected the ambivalence and widely varying viewpoints of how close fully autonomous vehicles are to becoming a reality, with predictions ranging from one year to 15 years. As someone who is closely tracking developments in this area, this was a head scratching moment that made me ask myself: Why are the trees not growing to the sky anymore?
The panelists included moderator Richard Wallace, Director of Transportation Systems, CAR; Kevin Kerrigan, Director of the State of Michigan’s Automotive Office; Kirk Steudle, Director of the Michigan Department of Transportation; Brian Daugherty, CTO of the Motor and Equipment Manufacturers Association, Andreas Busse, Senior Architect of Driver Assistance Systems, Electronics Research Lab, VW; Praveen Singh, Vice President, Connectivity, Lear Corporation; Gareth William, Director of Advanced Development, Mitsubishi Electric; Ralf Lenninger, Senior Vice President, Interior Electronics Solutions, Continental AG; and Danny Shapiro, Senior Director of Automotive, NVIDIA.
Even in the midst of all the recent Connected and Autonomous Vehicle positive developments, including the recent announcement of the American Center for Mobility at the former 335-acre Willow Run (aka “Arsenal of Democracy”) facility near Ann Arbor and the planned 350 miles of “Smart Corridors” to be installed in southeast Michigan by 2019 that were noted by Kevin Kerrigan and Director Steudle, the tone and predictions of the panel appeared to be somewhat muted by the recent Tesla automated vehicle crash, which is receiving a lot of regulatory scrutiny.
Brian Daugherty of MEMA was quick to discuss “Autonomous” vs. “Automated” vehicles, with the latter being a critical and growing part of the automotive parts industry and the former being the subject of excess “media hype” and a “noble goal.” Daugherty noted that
fully autonomous vehicles are a long way off, except in controlled environments.
Daugherty stressed the need for reliability and redundancy, at scale. On the other hand, automated vehicle development including Dedicated Short Range Communications (DSRC) protocols are low cost systems that help drivers in 80% of all non-impaired crashes, he noted.
Andreas Busse of VW began the second half of the panel with a detailed and technical discussion of the “Deep Learning for Piloted Driving” that VW is researching, using artificial intelligence and “neural networks” to develop “image segmentation,” speed up data acquisition and assimilation, and help connected vehicles assess “the full context of potential scenes” they may encounter, while acknowledging that “cornering” cases are the most challenging. Use of such methods will speed development in an area that requires extremely robust field testing that cannot be done by any one company, he noted, with estimates of the required amount of testing ranging from 1 billion miles to 13 billion miles!
Praveen Singh of Lear drew historical parallels between “V2X” technology and the development of Bluetooth, noting that law cost, ubiquitous deployment and ease of connection were keys to Bluetooth’s success and will be key to autonomous vehicle development, in addition to being “open source” and “standards based.”
Gareth Williams of Mitsubishi Electric shared important information on a recent “end user study” on Advanced Driver Assistance Systems (ADAS) that the company conducted. The study found that automatic braking was the most desirable feature, although only trusted by 64% of the respondents. The highest level of trust (73%) was in adaptive cruise control, with lane keep assist being trusted by only 9% of the respondents. The study also revealed the shortcomings of the current ways that safety features and the deployment thereof are communicated in the car, including visibly, audibly and tactilely (e.g. seat vibration), with 55% of the respondents noting that improvements in communication are essential to building trust. Other important factors include perceived safety, consistency and predictability.
Ralf Lenninger of Continental gave a thoughtful presentation of the three “waves” of using enhanced electronics in the vehicle: Wave 1 which added electronics to a component (like ABS), Wave 2 which used electronics for system integration of numerous components, and Wave 3 which will use electronics and digital services to connect cars, which he described as the “current dream” that the industry is developing, with the accumulation of all three waves ultimately leading to the self-driving car by 2025.
Danny Shapiro of NVIDIA closed out the panel with an optimistic and dizzying discussion of the developments in Artificial Intelligence (AI), including the ability to use software to “train the vehicle in the cloud and bring it into the car,” which is the holy grail vs. the bulky GPU’s that fill the trunks of many autonomous test vehicles today. He noted that VC funds have recently plowed as much as $5 billion into various AI platforms, many of whom intend to operate in the automotive space.
In the Q&A session, several panelists opined that regulatory developments will lag technological ones, and have the prospect of slowing down autonomous vehicle deployment. Mr. Lenninger noted that each critical country to autonomous vehicle development ideally will have a “national harmonization board,” which will make the ultimate international harmonization of standards more efficient as well.
Is the industry tapping the brakes on full autonomy? No, but it is safe to say that this incredibly complex undertaking will be evolutionary and will not proceed on a straight line. Hold on for the ride! (Or, at least be prepared to intervene in the ride in certain circumstances!)