Rivian - distributed computing [on] wheels

This week Fully Charged posted an episode on Rivian: https://youtu.be/J-falgJE1xg About 5 1/2 minutes in it started to click: Rivian thinks about its vehicle as an embedded, distributed computer. They are using software for suspension control and corner-by-corner roll control. (They have a separate motor driving each wheel, so why not?) It was weird to hear terms like "ethernet backbone" when talking about the design of a vehicle.

... because they're four separate motors you can then drive the torque individually per wheel ... including in the opposite direction to achieve things like tank turn." "... we can do perfect torque vectoring... torque vectoring can do two things: it can stabilize the vehicle... it can make the vehicle more agile.

This bit starting at 09:33 made me think "machine learning" -- but no doubt they can achieve all of this with a straightforward control loop:

[one other thing] we can do in terms of traction control... optimizing the slip of the wheel with respect to the road...

... because we can use the knowledge of the current going into that motor to estimate the torque... we know very accurately what the torque is... we can optimize it immediately... we can determine how much torque is getting us how much acceleration and how much slip, so we know the µ [coeff. of friction] and that means we can optimize the slip of the wheel... for the grip that we have available...

[talks about the slow feedback loop for recovering grip in an ICE vehicle]

... [the Rivian feedback loop is] so fast... imperceptible to the driver...

What an interesting interview!