Hacker News Reader

Comment by jstanley | Hacker News Reader

If you have a car only a few inches behind you and you spot a problem and need to emergency brake, you can't actually start braking until the car behind has started braking, otherwise the car behind will crash into the back of you.

If you have 3 cars in a row, and the first car spots a problem, it can't start braking until after the 2nd car has started, but the 2nd car can't start braking until after the 3rd car has started.

It gets linearly more dangerous as more cars are added.

Not saying it can't be done, but the problem is more subtle than just working out how to stop the cars from crashing into each other at speed: they also need to be able to brake in emergencies without crashing into each other.

Possibly the way to do it would be to mechanically couple the cars like a train, so that it is safe for the one in front to start braking earlier than the one behind, albeit with reduced performance.

View on HN

The idea is predicated on autonomous control and presumably the vehicles communicate in real time...they all brake together at the delta V of the least capable member.

The "real time" control loop between the sensor on one car and the brakes on the same car is much tighter than that between the sensor on this car and the brakes on another car and then feedback from the other car that braking has started.

I don't actually know how long it takes between the computer deciding to brake and the car actually starting to slow down, but I expect you can't just throw out a message saying "I'm braking immediately" and start braking without any kind of confirmation that the message was received, and expect it to be OK.

But maybe you can afford to wait for a confirmation? The speed of light is pretty fast, and you're not transmitting that many bytes.

Especially if you consider that a human could easily take 200 milliseconds to react to something even if it's really paying attention. You could request a braking confirmation from the other side of the Atlantic and still come out ahead!

There could be some leeway to initiative braking on following cars quicker with control loop similar to existing emergency brake assist systems [1]. These currently apply the brakes quicker and more powerfully than a driver’s foot when the car thinks it’s in an emergency situation. The same signal could be sent to following cars by the time the driver sufficiently engages the pedal.

[1] https://en.wikipedia.org/wiki/Emergency_brake_assist

Braking delta V depends on many variables such as mechanical condition, component temperature, road surface friction, etc and is thus difficult to predict to that level of precision. Even with real-time communications, if the vehicle in front slows less than expected due to hitting a pothole or something then the vehicles behind are going to have to react damn fast to avoid a collision. It might be achievable on certain roads but it will be a tremendous engineering challenge. Don't expect to see this in passenger vehicles any time soon.

Yeah all good points, and just to add, there's also the challenge of maintaining perfectly reliable and low-latency communications over any sort of wireless medium in a variety of unpredictable outdoor conditions.

Reply on news.ycombinator.com