The Center for Automotive Research and KPMG have published a new report predicting the self-driving cars may be on the market as soon as 2019–if, however, the government takes action aimed at improving auto safety.
The report notes there are two approaches to self-driving cars. One, which it calls the “sensor-based solution,” is represented by the Google car and requires that each car have all sensors on board to detect everything in its surroundings. The other approach, which the report calls “connectivity based solutions,” relies on car-to-car (C2C) and car-to-infrastructure (C2I) communications to help cars navigate.
The report suggests that sensor-based solutions are “not cost-effective for mass market adoption” and require far better maps of streets and highways. It is true, as previously noted here, that the the “light detection and ranging” (LIDAR) device mounted on top of the Google car and other self-driving cars currently costs about $70,000. But that cost may come down, and Google seems committed to mapping the nation, state-by-state, to standards that self-driving cars will require.
The connectivity solution, the report continues, doesn’t “work with pedestrians, bicyclists,” and other moving objects that won’t carry their own C2C transmitters. More important, it will require “significant infrastructure investment” by state and local highway and street owners. This creates, as I’ve noted before, a chicken-and-egg problem: who will buy a car that has C2I connectivity before there is infrastructure to communicate with; and what states will install that infrastructure before there are cars to communicate with it?
The report recommends a mixed solution that would provide enough infrastructure to allow sensor-based cars to navigate without a $70,000 laser beam on top. The report recommends that the government invest in research aimed at trying to bring these costs down, but in the meantime, it suggests that a mixed solution can be more cost-effective.
To make this happen, the report suggests that the National Highway Traffic Safety Administration should issue a rule requiring that all new cars be sold with V2V communications. While it may not be possible to retrofit older cars so that they are completely driverless, they should all be retrofitted with V2V communications that can at least let fully driverless cars know they are there.
While highway safety might be a sound justification for such government mandates, I am not convinced it is necessary, nor am I confident that the government would get it right. Automakers are correctly skeptical about whether government will make the infrastructure investment needed for either the connectivity or mixed solution. I suspect that the high cost of the sensors will drop rather rapidly when Google and other companies are finished developing the software to run self-driving cars.
The main thing government needs to do is get out of the way. Anything else that it does could create more problems than solutions in the long run by locking us into technologies that are unnecessarily expensive.