Are we seeing the slow death of Level 3? Despite the initial reaction to a car without a steering wheel, with more manufacturers coming around to their point of view, Waymo looks prescient to have championed Level 4 & 5 vehicles.
Back in 2014 Google Cars, now Waymo, launched their driver-less car, sans steering wheel or pedals. Its toy town looks didn’t endear it to petrol-heads. The leap of faith it embodied didn’t endear it to the critics. The radical change of paradigm did not lead major automotive manufacturers to endorse Google’s stance. The academics didn’t believe it was feasible.
Pretty much all of the organisations with a stake in the traditional car paradigm (owned privately, driven by a human) predicted a steady increase in autonomous functions and features as cars climbed up the evolutionary scale from zero (no intelligent features) to five (full autonomy). What started with anti-lock brakes was forecast to evolve into collision avoidance and self-parking, through intelligent cruise control and lane keeping to limited autonomy. Eventually and over a period of ten to twenty years, the self-driving car would arrive. The assumption was they’d be made by the existing car manufacturers, sold by existing dealers, owned by existing car buyers via existing financing methods and used on the existing traffic infrastructure. That was the generalised view from 2014.
How do things look in 2017?
Three years later and the idea of a step change from driver to driver-less doesn’t look as far fetched. Could there be a widespread jump from Levels 1 and 2 to Levels 4 and 5, missing out Level 3?
The SAE definition of Level 3 is: Conditional Automation – the driving mode-specific performance by an Automated Driving System of all aspects of the dynamic driving task with the expectation that the human driver will respond appropriately to a request to intervene.
In early 2017 Ford announced it would go directly to Level 4 vehicles by building mass produced autonomous vehicles for use in shared commercial fleets. They promise significant deployments by 2021.
BMW has been working with Intel and MobilEye to produce similar Level 4 vehicles. However this joint approach seems to have taken a bold step to set a larger goal, to deliver Level 5 vehicles by the same date.
As reported in several publications, in March 2016 Elmar Fickenstein, BMW’s senior VP for autonomous vehicles said that they are aiming at Level 3, 4 and 5 vehicles by 2021. As though to underline the strategy, Intel has bought its previous business partner, Israeli company MobilEye for $15bn. Interestingly Mobileye work with a range of vehicle manufacturers (for example Audi); they are not exclusive to BMW. Mobileye provided the technology behind Tesla’s early foray into autonomous driving (they have since gone their separate ways).
Volvo have also decided to jump Level 3 for similar reasons and will aim for Level 4 by 2021. Marcus Rothoff, a manager responsible for implementation of autonomous driving technology at Volvo said “We just haven’t found a solution to provide safety when we have this transition. You might not be ready to take over and we can’t push our customers into that scenario.”
Are there other signs Level 3 could be DOA?
There is a UK government committee (the House of Lords Science and Technology Committee) investigating the likely impact of autonomous vehicles. In late 2016 the chair of the committee acknowledged “… our inquiry is not just about autonomous vehicles but very much about connected vehicles, and we recognise that the government strategy is to follow two not necessarily compatible strands of development: one a leapfrog perhaps towards autonomous vehicles, but one that certainly has more credibility at the moment of connected vehicles and incremental improvements to existing technologies.”
Some of the expert witnesses to that committee disagree with Ford and BMW’s forecasts, for example Professor Natasha Merat who said in response to a question whether we should move directly to Level 5: “I am sorry, but we cannot for a long time. We are not capable of doing that. The technology is not there. The sensors and cameras, et cetera, are still developing. If there is a bit of rain and the sensors get wet then they will not work. Level 5 is a long, long way away“.
It is interesting however to read the minutes of such discussions, as it is clear that a lot of academic (in both senses) time is being spent researching and discussing the complex issues created by Level 3 vehicles. If those discussions were undertaken in the context of an Occam’s Razor-like “Level 5 or nothing” then it becomes a much simpler debate.
The Committee’s Report did not pull punches in concluding “We challenge the expected benefits of a level of automation at which a driver takes back control of the vehicle in an emergency situation. Given the evidence that reactions could be slow and poor in such circumstances, it may be that the risks associated with this are too great to tolerate and that a way should be found to bypass Level 3 where a driver does not need to monitor the dynamic driving conditions, nor the driving environment at all times, but must always be in a position to resume control.”
What will happen – a forecast for the next 12 months
Expect to see more commercial and government organisations questioning whether Level 3 is do’able. Particularly, is it safe? Safety is a big stick and it will get attention. By the end of 2016 proponents of Level 3, for example Audi, could find that the balance of the narrative has moved from “when will we get Level 3” to “should we allow Level 3“?
If this change in emphasis coincides with credible evidence and increasing confidence in manufacturers having the ability to reach Level Five in 4 to 5 years then Level 3 will be stillborn. It could cease to be a choice a manufacturer can make for themselves, It will be mandated by the authorities, fearful of the safety implications for inattentive drivers being called to take over in an emergency. Once the general expectation is of a move directly to Level 4 and 5 then naturally the weight of R&D will re-focus from adding autonomous features to removing the driver. VW, GM, Toyota, Honda and Nissan will find themselves behind bolder companies, such as Ford, Google, Volvo and BMW.