Generally I am "against" the idea of automated vehicles following magnets buried in the road. I am not against it in that I think it's a good way to do guidance. I simply think it's a distraction from the real problem, which is to get a system that works on existing roads without having to upgrade them. Otherwise you build a system where the robocars, like transit, can only go on specially prepared tracks. Even if it becomes cheap to embed magnets in all the roads (with a machine that essentially has a giant punch) you want the complete flexibility humans have to change lanes, move around obstructions etc. You don't get that if you have to follow a track.

But snowplows are an interesting problem because all the visual cues go away and not even humans can see where the road is reliably, at least on these mountain roads. The more common technique is to have tall posts on the snow roads, but magnets can work as well, perhaps better.

It may also be possible to have localized radio transmitters that allow centimeter-accurate positioning to supplement GPS. This, combined with inertial guidance and the odometer, can give highly reliable positioning.

A plow however also has to identify obstacles in the road, and plow around them, and possibly call for a tow-truck on them. LIDAR is good at seeing bumps in the snow indicating something is buried, which could allow the plow to send out probes to find out what it is, and if it's something to plow around.

One of my goals, however, with the concept of adapting idle vehicles to use as plows is to never let the snow get very deep. With so many vehicles available, you can make sure it's never so bad you can't see the shape of the road.

And frankly, not just the idle vehicles. No reason that vehicles taking people on trips can't also be plowing, if there was time for them to fetch a plow-head at the depot. The limiting factor is the number of plow-heads available. Note that for light snow, plow-heads might be something cheap and light; not necessarily a giant scoop of heavy steel. More like a large version of the snow shovel that everybody already owns.

And let me add that I am of two minds in another way on magnetic guidance. While I think it is a dead end, it may have a role in enabling what I call the whistlecar. My thought is that we could deploy whistlecars well in advance of people-carrying passenger robocars. To do that we would need them to be able to move at reasonable speed, but only along roads that have been specially prepared for them -- perhaps with magnets though more probably with optical markings.

Then the whistlecars, when they get to the arterial path closest to their real destination, must be able to leave the marked path and travel a short distance on any ordinary road. To do this safely, they would do it more slowly than passengers would tolerate. Then the car is delivered to the summoner, who drives it anywhere within a modest distance of the whistlecar arteries.

Remember the plan is for using this more in very light snow, and doing it more frequently, so little can be buried in the snow. And remember that because robocars use LIDAR, they would see bumps in the snow as well as they would see objects on the road. And then pause and ask for a human to look at the HD video feeds and resolve anything they can't resolve.

Cell towers and some TV stations already give out such signals, and there are variants of GPS with very high accuracy.

Robocars can see lane markers just fine already, so no magnets are needed. The only question is in the snow, when even humans can't see the lane markers, but still work it out OK. They can see the sides of the road.

That people would steal snowplows. They sit in parking lots too, they don't get stolen and they are not even moving. Robocars will of course have cameras everywhere (they need it to drive) and if threatened, your crime will be on live video, and the police (and police robots, which won't shoot you but will block you) will know exactly where in real time.

There will be design errors and malfeasance of course. But tech like this will not be deployed until the level of it is well below the level with human drivers. Of course people will try to game things, they try to game everything.