Unit 9: Feedback and PID Control

The robot misses the target. Now what?

Open-loop commands assume the same command will produce sufficiently similar behavior each time. Load, battery state, friction, and disturbances make that assumption fragile.

Why Open Loop Misses introduces measurement as part of control. P, I, and D Without the Magic gives each PID term a specific job. Tune with Evidence replaces random gain changes with controlled trials.

Think in response shapes

You care about more than the final value:

• rise time;
• overshoot;
• oscillation;
• steady error;
• saturation;
• noise sensitivity.

Keep a closed-loop response trace and a tuning record that shows what changed between trials.

A graph is useful, but a timestamped table is enough if it makes the response inspectable.

Simulation is not second-class here

Control is one of the best places to use simulation because you can change one variable and repeat the same initial condition. If physical hardware is available, use it to test where the model stops matching reality, not merely to make the Unit look more "hands-on."