Over the last year or so, I’ve seen more and more news stories to the effect of, “Yes we told you we’d have fully autonomous vehicles on the market by 2020 [or whenever], but it’s going to take longer. A lot longer.”
Even the New York Times felt the need to weigh in with a recent story titled, Driverless Cars Are Taking Longer Than We Expected. Here’s Why.
The recurring theme in these stories is, Of course it will take longer, driving is hard! I’m reminded of the aphorism, “The first 90% of the code takes the first 90% of the time. The last 10% of the code takes the second 90% of the time.”
But is there another reason? Is it because V2V adoption is stalled?
Two years ago, I attended the ITS annual conference in Montreal. I made a point of asking this question in a couple of panel Q&A sessions, and to various individual experts working on autonomous vehicles:
“Can you imagine a future in which a large percentage of the total vehicle fleet is autonomous, but not connected?”
Every time I asked that question, the answer was unequivocal: V2V technology so simplified the challenge of full autonomy that V2V should be thought of as a prerequisite to higher levels of autonomy.
At the time, only Tesla appeared to be working towards ‘autonomous autonomy’. Tesla’s objective seemed to be to fake human driving, which is to say, by relying on visible spectrum data, and using AI to anticipate the behavior of other actors on the road.
All the other OEMs seemed to envision a much easier road to full autonomy (not easy, but easier.) That is, achieving that ‘driverless’ status in an environment where most vehicles were connected, with many vehicles communicating intent and the vast majority communicating at least location & vector. Again, to be clear, the experts I spoke to all assumed that increasing levels of autonomy would come in an increasingly connected environment.
Today, I get the feeling that most OEMs are pushing towards ‘Tesla-style’ autonomy — albeit with a more diverse sensor suite — in which each vehicle operates on its own, without direct input from surrounding vehicles.
What changed? Well, the proponents of 5G effectively outplayed DSRC in the U.S. market, and pushed the timeline for widespread V2V connectivity back by several years. There are a number of reasons why the grandiose claims of various CEOs, (cough, Ghosn, cough) haven’t come true: Hubris was a part of it, but we haven’t put enough of the blame on the failure to deploy V2V, which would not only improve safety and augmented driving systems in the medium term, but would lower the bar for full autonomy, long term.