Safe Model Predictive Control Approach for Non-holonomic Mobile Robots in Dynamic Environment

This is a MPC project for motion planning & control of non-holonomic mobile robots

1. Video & Paper

our project video can be found here

our report can be found here

2. How to run the files:

1. In order to run the files, please make sure you have FORCESPRO installed and in the python PATH (tested on FORCESPRO 5.1.1).

2. There are multiple simulations you can run:

./controller/MPC_method1.py contains an MPC controller + simulation for the global model of the unicycle, linearised about the trajectory. You can choose parameters such as trajectory, horizon and Q/R costs for the controller. You can also set it to avoid a static obstacle (how to do that is explained in the code comments) using the two linear constraints.

./controller/LQR.py is the LQR controller, which is unconstrained and solves the DARE to find the K that follows the pre-defined trajectory.

./controller/MPC_method2.py is the two robots simulation where one of the robots avoids the other one by using a velocity constraint. Run this file to see how the controller performs. You can change the parameters for the trajectories to evaluate the performance on a shared or intersecting path.

./model/MPC_utils.py Open this file, uncomment the section at the bottom and run the file to see an example of how the outer approximation of the terminal set is calculated and plotted.