Re: AutoPilot, years ago
A really good idea
No, it really isn't. Relying on automated systems to "maintain[] safe distance", for example, would have a high failure rate and terrible failure modes.
In most of the world, private automobiles are maintained by their owners (or at the direction of their owners), and are not subject to frequent rigorous inspection. (In many US jurisdictions, they're not subject to inspection at all.) Transponders will fail when the vehicle is traveling at speed, reducing the amount of information other vehicles have about the failing one. Other sensor systems will fail. Driving mechanisms will fail. Cars will experience catastrophic failures of basic physical integrity – I've seen a rear bumper, a window, and a wheel all come off other vehicles on the highway. These events are very difficult even for very flexible and capable autonomous-driving systems to deal with; they'd be absolutely beyond the sort of thing described in Popular Science decades ago.
Automated systems are similarly troubled by intermittent road hazards. In the past five weeks I've driven through mountain passes six times; four of those times it was snowing briskly, resulting in a roadway that was alternately fine and slippery, due to snow drifting and other factors. One of those times was at night, and I could only determine where the lane was by my distance from the reflective markers on the side of the road. (Autonomous system could use a map and GPS? Right. You get a precise, accurate, continuous GPS lock while winding between mountain peaks at 60 mph with snow pouring out of the sky.) Around here, it's very common on a sunny winter day to be driving on clear roads and come around a corner where the road is shaded to the south, leaving a patch of slick ice. I've encountered oil slicks on roads, too. It's tough to create an AD system that handles those sorts of situations well.
People would modify their cars to override the automated systems so they could cheat, driving over the set speed when possible, passing in the wrong lane, etc.
What's the transition plan? Do we take already over-subscribed highways and make some of them automated-car-only? Yeah, that's going to work.
Cars aren't trains. If trains are what we want, or more precisely what we need, then that's what we should be investing in. It'd be technically feasible to replace a lot of private-passenger and cargo-transport long-distance car and truck use with a built-up rail network, and use autos (with car rental or some other process at the train station) for local transportation. An autocratic government could impose that sort of scheme. Wouldn't be popular. But the fix for highways isn't "smart highways", because they simply won't ever be smart enough.