sainavaneet / LMPC

Focused on a project related to LMPC (Learning Model Predictive Control) sampled data control and H2 optimal control.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Introduction

For assessing the performances of LMPC sampled data control and H2 optimal control, we compared the quadratic costs of them. The constraints for the cost function include:

Parameters

  • Total Time (( T )): ( T = 10 ) s

  • State Weight Matrix (( Q )): ( Q = \text{diag}([1, 1, 1]) )

  • Control Weight Matrix (( R )): ( R = \text{diag}([0.1, 0.1]) )

  • Sampling Time (( \gamma )): ( \gamma = 0.1 ) s

  • LMPC Control Input Limits:

    • ( v_{\text{max}} = 0.3 ), ( v_{\text{min}} = 0.0 )
    • ( \omega_{\text{max}} = \frac{\pi}{4} ), ( \omega_{\text{min}} = -\frac{\pi}{4} )
  • Gain Matrix (( k )):

    [ k = \begin{bmatrix} 0.4662 & 0.2978 & 0.2067 \ 0.8341 & 1.2165 & 1.3492 \end{bmatrix} ]

Quadratic Cost Function

The quadratic cost function for a single timestep is given by:

[ J_{\text{total}} = \sum_{k=1}^{N} (\mathbf{x}_k^T \mathbf{Q} \mathbf{x}_k + \mathbf{u}_k^T \mathbf{R} \mathbf{u}_k) ]

Where ( N = 2 ).

Results

  • Total Cost Value for LMPC (( J=10 )): 8.98516268901749
  • Total Cost Value for H2: 18.775862036811883

About

Focused on a project related to LMPC (Learning Model Predictive Control) sampled data control and H2 optimal control.

License:GNU General Public License v3.0


Languages

Language:Python 87.6%Language:MATLAB 12.4%