Tutorial: closed-loop NCbT

This tutorial shows how to use NCbT_closed to estimate controller parameters from closed-loop data.

Load data

import numpy as np

u = np.load("u_t.npy")
y = np.load("y_t.npy")
r = np.load("r_t.npy")
t = np.load("time.npy")
Ts = 0.05

Choose reference model and controller structure

# Reference model M(q)
omega_bar = 10
alpha = np.exp(-Ts * omega_bar)
num_M = np.array([0, 0, 0, (1 - alpha)**2])
den_M = np.array([1, -2*alpha, alpha**2])

# Controller basis 
beta = [
    ([1], [1, -1]),
    ([0, 1], [1, -1]),
    ([0, 0, 1], [1, -1]),
    ([0, 0, 0, 1], [1, -1]),
    ([0, 0, 0, 0, 1], [1, -1]),
    ([0, 0, 0, 0, 0, 1], [1, -1])
]
l = 20

Estimate parameters

from pyncbt import NCbT_closed

estimator = NCbT_closed(u, y, r, num_M, den_M, Ts, t, l, beta)
rho = estimator.run()
print(rho)

The result is the vector of controller parameters.