A lot of things seem like they have PID control in them, but in most cases what you’re seeing is actually just hysteresis control. To take the thermostat example, the furnace comes on at X-2 degrees, but shuts off at X+4 degrees (for set temperature X). That’s a hysteresis of 6 degrees. This simple technique avoids hunting, chasing, and thrashing in systems with noisy data and a lot of interia. Another example is altitude control in airplanes. The first lesson they teach pilots is “don’t chase the altimeter” for this reason. The plane takes longer than you think to change altitude, and the altimeter itself is a noisy, slow instrument that takes time to reflect where you are.
Without a hysteresis, your furnace would click on at X-0.5 degrees, run for 1 minute until it hit X degrees, then shut off. However air currents in the room would shift the temperature around a little and suddenly the area around the thermostat would be X-0.5 again the furnace would click on again instantly. Even worse, the thermostat might be slow to detect the change due to air currents, and the furnace might shoot us up to X+2 and then the A/C would kick on to frantically try and get back down to X. It would then overshoot again and the heat comes on again at X-1.The whole system would be constantly thrashing itself, chasing X in a very inefficient way.
PID control is one level above this, but is only used when hysteresis isn’t enough (such as in multi-rotors). The advantage of PID is that it can react extremely quickly and copes well with imperfect data without doing too much hunting. What it’s really doing is pushing the hunting down to a smaller domain. In the thrashing furnace example, instead of hunting to hold ±2 degrees, it will hunt to hold 0.1 degrees. For systems that can react quickly, like the RPM of a small brushless motor, this is ideal. You get a stable result with minimal perceptible hunting. The downside PID control is that while it compensates for slow noisy inputs (eg altimeters) it requires a lot of CPU and fast-reacting positive-feedback outputs (or precise inputs for slow, noisy outputs). Great for BLDC motors, not so great for many other things (eg. ICE drones use different control systems for this reason). Mostly though it just isn’t necessary in most applications where a hysteresis is sufficient (and can be implemented with no CPU at all- old furnaces do it with a bimetallic strip holding a mercury switch. Reliable, no power required, and maintenance free for 100 years).
What Nissan is doing is PID control with better software, basically. If you know the exact physical properties of inertia in the thing (in this case a bowl of liquid) and have good accelerometers, you can PID your way to optimal deceleration while minimizing forces on that substance (ie. humans). ABS brakes are actually doing the same thing in a very primitive way. By maintaining wheel roll and not allowing them to lock up, you keep deceleration at maximum in a controlled way (if a wheel locks, it goes into static friction and deceleration decreases). The real benefit of ABS is really that you maintain steering control in a panic stop (most people don’t know this though and don’t try to steer around the object like they should) so it’s not a great analogy.
The catch with this is that it requires precise output control, like the BLDC motors again. So it isn’t going to work in non-perfect road conditions and probably works best with electric cars that can achieve very consistent braking control through regeneration. This is why high end trains do it, as observed above- regenerative braking and consistent predictable traction conditions exist. But this Nissan system isn’t going to work on black ice or while hydroplaning, for example.
I dunno if that answered your question. Process and motion control is not my field of expertise, but I’ve done some adjacent things like inverse kinematics and know enough to be dangerous and/or sound smart on the internet.
