Assume feet and hands.
The federal highway transportation safety agency agrees with Google that the artificial intelligence system that controls its autonomous cars is the driver for purposes of federal vehicle safety rules. According to a letter sent to Google by the agency and posted on its website…
Google’s design choices in its proposed approach to the [self driving vehicle] raise a number of novel issues in applying the [federal motor vehicle safety standards (FMVSSs)]. Those standards were drafted at a time when it was reasonable to assume that all motor vehicles would have a steering wheel, accelerator pedal, and brake pedal, almost always located at the front left seating position, and that all vehicles would be operated by a human driver. Accordingly, many of the FMVSSs require that a vehicle device or basic feature be located at or near the driver or the driver’s seating position. For vehicles with an [artificial intelligence] driver that also preclude any occupant from assuming the driving task, these assumptions about a human driver and vehicle controls do not hold.
Google isn’t home and dry yet, though. Some rules – that cars have a foot or hand operated parking brake, for example – need to be re-written, since “the [self driving system] will have neither feet nor hands to activate brakes”. The agency was also skeptical about Google’s claim that no steering wheel, pedals or other controls are needed at all.
But the agency offered a solution: go through a (lengthy) process to change the rules, but ask for a waiver in the meantime. That doesn’t mean it’s approving Google’s design, just that it’s clearing the way for testing to see if an artificially intelligent car meets federal standards. That’s a better approach than the install a steering wheel anyway approach the California department of motor vehicles is taking while it figures out what to do with self driving cars.