Model Predictive Control and Infinite-Horizon Stochastic Optimal Control

Overview

This project utilizes Model Predictive Control (MPC) and Generalized Policy Iteration (GPI) to compute a control policy for an agent to follow a reference trajectory. Our agent is expected to avoid obstacles while having its motion be affected by noise.

Implementation

This was implemented in Python using NumPy. The code has been redacted, if you wish to see it, you may contact me at charles.lychee@gmail.com

Results

Model Predictive Control (MPC)

Time Horizon = 20

Time Horizon = 70

Generalized Policy Iteration (GPI)

Discount Factor = 0.9

Discount Factor = 0.99

Mathematical Approach (WIP)