Model predictive control for dynamics in autonomous cars

Abstract

Car accidents are quite common in several countries, most of these accidents are due to human error, these cover a wide range, from hitting a pole to hitting a pedestrian and this can lead to fatal accidents as well, also the high time in traffic is due to the lack of attention or expertise on the part of the driver. To reduce these accidents and this time lost in traffic, the investigation of autonomous vehicles is chosen, this thesis is focused on the control system for lateral, longitudinal, and yaw angle position using a predictive control model.

This study analyzes the performance of the model predictive control (MPC) against PID by controlling the steering and acceleration of a vehicle, it was used Matlab to create the waypoints in the driving scenario and to make the simulations in linear MPC and adaptive MPC blocks, it was analyzed the performance in each control to see how well the position of the vehicle follows the reference line for a double change lane maneuver and compare the results against a PID to see which one performs better. The parameters used were measured from a golf cart.

These results suggest that the MPC has a better performance when the tuning is done correctly also the data obtained from the simulation showed that speed was significant to lateral position and yaw angle when lane changes were made. We conclude that Model Predictive Control has better performance against PID also the MPC can be adjusted to work as we wanted it had a lot of maneuverability and is quite flexible.