About The Authors

Friday, July 11, 2014

Google's Autonomous Car: Don't Drink the Kool-aid

Google's senior executives are busily touting the wonders of autonomous vehicles. There's the technological marvel, at least in the eyes of Silicon Valley. There are the economic benefits - no more congestion, no more accidents. Wonder of wonders! – and great for the Google empire, and for its stock price.

The PR machine is a marvel to behold, and the gullibility of the audience – well, it's Google! Is their part in this really that much of a marvel? Will economic benefits be as great as they claim? Will they even be a player in future vehicle technologies? Their PR machine is not paid to probe such issues, much less point out that alternative technologies may bring almost all of these benefits more quickly and at a very modest cost.

First, the core innovations necessary for an autonomous vehicle are already on the road, the result of decades-long engineering efforts alongside which Google's investment and expertise pale in comparison. Blindspot detection, lane departure warnings, backup "assist" (outside the US that is surely called a safety feature) and adaptive cruise control are all necessary for an autonomous car. Now some of these aren't cheap, but they're falling in price. So we don't have to await an entirely new generation of vehicles to begin reaping the benefits. Crucial to Google's vision is that these are all partial solutions. However, I am not at all convinced that what Google offers will be a sufficiently big increment to offset the additional costs of full autonomy. Nor is it at all obvious that Google will have any role short of autonomy – their presence is not needed for these existing tools.

Second, Google's is not the only approach. In particular, connected vehicle technologies promise most of the benefits at a far lower price point and with a faster rollout. Such systems are inexpensive because they can use the copious computing power already in car, while the hardware consists of inexpensive RFID transponders (though not as inexpensive as the tags retailers use to deter shoplifting). The pieces of such systems are now being tested on the road, with a large test facility – the Michigan Mobility Transformation Center – an artificial cityscape – under construction in Ann Arbor, adjacent to the University of Michigan Transportation Research Institute. Such systems don't require the panoply of sensorts of an autonomous vehicle. Indeed the core components could be sold as an aftermarket item, albeit with lower functionality. Such connectivity could be rolled out in the course of years.

In contrast autonomy will require decades. First "real" vehicles need to be out in sufficient numbers to overcome regulatory fears and start the path to consumer acceptance. I can't see that happening before 2020, given the lead time for vehicle development when new systems are involved. It's not just that the hardware and software have to be integrated into existing vehicles, it's also that test procedures need to be developed for both the hardware and the software. Then production capacity has to be ramped up, while successive generations of vehicles are designed. That's another 10 years. And then the fleet will gradually turn over; with the average age of vehicles now over 11 years, that's another decade for half of all vehicles to be autonomous. We're thus looking at 2040. A combination of aftermarket and designed-in RFID systems could be on every vehicle by 2025, offering varying levels of collision avoidance and traffic flow smoothing.

Google likes to trumpet the elimination of accidents and the end of congestion. Perhaps. However, the restructuring of where people live versus where they work is a process that will take decades; in many cities, particularly in the US, the housing stock is widely dispersed, and so we really won't be able to get rid of all those cars, another claimed benefit of autonomy. Will connected vehicles deliver all of the same benefits? No, at least not initially; any aftermarket device could only offer warnings, nor take over steering and braking. Still, the price point and the time horizon are quite different from the drink Google wants us to imbibe.

Will there be a role for autonomous vehicles? Certainly truck trains are one application, and off-road uses in mining and construction. The Department of Defense is surely a potential customer. Don't look to these to drive Google's stock valuation, or you will have drunk the Kool-aid.