Struct ModelBasedControl 

pub struct ModelBasedControl { /* private fields */ }

Learn a plant model from data, tune a PID inside it, and deploy on the true plant.

Implementations

impl ModelBasedControl

pub fn new(mass: f64, damping: f64, disturbance: f64, setpoint: f64) -> Self

A task on a hidden damped point mass, with the controller starting proportional-only.

pub fn identify(&mut self)

Stage 1 — identify ẍ = a·x + b·v + c·u + d from data by least squares.

pub fn learned_coeffs(&self) -> [f64; 4]

The learned model coefficients [a, b, c, d].

pub fn true_coeffs(&self) -> [f64; 4]

The true coefficients (for comparison): a=0, b=−damping/m, c=1/m, d=−disturbance/m.

pub fn gains(&self) -> [f64; 3]

pub fn deploy(&self) -> Vec<f64>

The step response on the true plant.

pub fn imagine(&self) -> Vec<f64>

The step response in the learned model (the “imagination”).

pub fn deploy_error(&self) -> f64

Steady-state error on the true plant.

pub fn train_step(&mut self, lr: f64) -> f64

Stage 2 — one Adam step tuning the PID by differentiating through the LEARNED model. Returns the cost.

pub fn train(&mut self, epochs: usize, lr: f64) -> f64

Train the controller in the learned model for epochs steps.