It’s Time for the Car of the Future

[begbert2]

@Mangetout

Another way to smooth the transition would be for new cars to be equipped with full carbrains that control the car completely, but for inexpensive passive ‘beacon’ carbrains to be issued (probably under subsidy) for retrofitting into all other existing road vehicles, so the full carbrains would at least be able to get information about the position and vector of the older vehicles, even though they would not be able to negotiate on a full two-way basis with them

@DSeid

The context was in a situation of limited acceptance. The transponder is cheap enough to require placement in a legacy vehicles and allows all other fully equipped carbrains to benefit from knowing what amounts to very detailed and timely traffic reports.

Sounds like you guys are on the same page! Transponders make sense, though I still :dubious: over the specific idea of having them all broadcast to a central traffic authority. (If nothing else that would require a much more powerful transmitter.) Local-area “I am here” broadcasters are a viable notion, and would be possibly of some use, but smartcars would still have to be able to deal with untagged vahicles on the road. I mean, if your car plows over bicycles with abandon, that’s probably a bad thing, and if it can track a bicycle, I wouldn’t think any reasonable designer would exclude the ability to track the much-larger and more visible (taillights!) car.

[Sunspace]

In the book 2081 (piblished in 1981), the cars had the option of self-drive, and the carbrains read local map information from passive transponders embedded in the middle of the road every fifty metres or so. They sounded very similar to what we would now call ‘RFID chips’ but on a slightly-larger and more robust scale.

If we developed a standard format for describing the features of the road (‘curve with radius of 100 metres for next 50 metres’, ‘this lane merges left after 300 metres’, etc), I think we could simplify the requirements of the carbrain quite a bit.

I understand that the DARPA project’s goal is now fully-autonomous navigation over existing roads and in existing traffic. The 2081-style carbrain would still have to keep track of neighbouring vehicles, but at least it wouldn’t have to continuously keep recognising the features of the road.

Would the requirement to continuously keep watch for animals, children, etc, on the road require enough complexity that adding road recongnition wouldn’t be a huge further leap?

[begbert2]

@Sunspace

Would the requirement to continuously keep watch for animals, children, etc, on the road require enough complexity that adding road recongnition wouldn’t be a huge further leap?

Yes and no. Roads are defined by visible lines, notably a clear center line with an additional line to delineate each additional lane on each side of the line. Compared to recognizing virtually anything else visually, this is cake; just follow the lines. Heck, you can make a robot out of legos that can follow a line.

It does occur to me that there are probably better ways to recognize obstructions than visually, though; sonar, radar, or IR of some type that gives multidirectional ‘distance-to-obstruction’ would probably be a great deal simpler to parse for useful information and it’d tell you nearly as much as vision. If that’s what they use, the lines are of course invisible to it.

And of course none of this solves the problem of navigation; just because it recognizes the line-pattern of an intersection doesn’t mean that it knows to take that left at Albuquerque.

I’ve mused a bit to putting something in the road striping paint to make it easily detectable by cars at a distance (would make it compatible with the sonar solution mentioned above), but imbedding information-loaded nodes sounds even better, assuming they could be made durable against failure through age and weather, and were placed redundantly. (The paint solution still wins on ‘ease of application’, though; it doesn’t have to be custom-programmed per location.)

[Mangetout]

@begbert2

Sounds like you guys are on the same page! Transponders make sense, though I still :dubious: over the specific idea of having them all broadcast to a central traffic authority. (If nothing else that would require a much more powerful transmitter.)

Broadcasting to a central register could be achieved by the installation of static roadside network points, but I don’t see the central control thing as absolutely essential - it would be useful to be able to gather and distribute traffic density and flow information (but this can be (and is) done anyway - just by counting vehicles - without knowing who is where). Centralised monitoring of vehicle movement would be of great interest to law enforcement, of course, but that would be a terrible reason to implement it.

Local-area “I am here” broadcasters are a viable notion, and would be possibly of some use, but smartcars would still have to be able to deal with untagged vahicles on the road. I mean, if your car plows over bicycles with abandon, that’s probably a bad thing, and if it can track a bicycle, I wouldn’t think any reasonable designer would exclude the ability to track the much-larger and more visible (taillights!) car.

I think all road users would have to have the beacons - which is a serious obstacle to the implementation of any such scheme (there’s also the non-trivial question of what would happen when one of the passive beacons just fails to operate). Any automated driving system would have to incorporate a degree of visual sense - because there will always be occasions when objects are on the road and are only reasonably detectable by sight.

The sort of thing I’m talking about though, would absolutely require more than just visual data; a very good mock-up of the concept is illustrated in this TV commercial.