A full implementation of PID control under the framework of (variational) free energy minimisation and active inference. This implementation produces interesting results for the long standing problem of tuning parameters in PID controllers.
The mathematical equivalence of PID control and active inference was first shown in A Probabilistic Interpretation of PID Controllers Using Active Inference. The details of the current model can be found in thew forthcoming "PID control as a process of active inference with approximate linear generative models", using a cruise control problem to show the basic workings of active inference with an example from Feedback Systems, An Introduction for Scientists and Engineers.
- Conor Heins @conorheins, for letting me know about some little bugs in the code