Hacker News Reader

Comment by mike_d | Hacker News Reader

Motors know nothing. The encoders know how much current is being input to the motor, but do not know the rotational output of the motor or the friction coefficient between the tires and the ground. All three are things you'd need to emulate a locking differential in software. You could fake the last two by having a reliable way to measure the travel of the vehicle from a source other than the wheels (GPS, downward camera, etc).

View on HN

Err, sorry but the encoders don't measure current. An absolute position encoder knows the exact speed/position of the motor. By that you can extrapolate the wheel speed position which you can just match 1:1 with the one on the other side. It won't be perfect as the reaction time of the system if wheel A breaks away and starts spinning but you don't need GPS or a downward camera to mimic a locking diff.

Sorry I was trying to rewrite for clarity and said encoder instead of controller.

If you naively emulate (as Telsa surely would) a 1:1 fully locking diff in software it will suck to drive. Electronic limited slip diffs already exist, like the Jeep FD/AD systems, which will use tire rotation sensors to determine forward momentum.

Now that I think about it, the Ford is probably already doing this.

Technically, you can sometimes measure the motor position without a position sensor by measuring the back EMF (although I doubt Tesla is doing this, IIRC it's hard to get good starting torque out of this setup)

Reply on news.ycombinator.com