Obstacle Avoidance using Learning Model Predictive Control

Obstacle Avoidance using Learning Model Predictive Control

This project builds on the Learning Model Predictive Control (LMPC) framework applied to the autonomous racing problem. The controller uses the data from the previous laps to learn  how to perform overtaking maneuvers. The data of the from the previous lap are used to build a collision free safe set which allows the controller to drive safely around the track avoiding obstacles and improving the lap time.

The proposed strategy has been tested on the Berkeley Autonomous Race Car (BARC) platform. Experimental results show that the controller is able to lear the overtaking maneuver improving the lap time. It is interesting to notice that the local information about the obstacle position influence the vehicles behavior in the whole track.

Credits:
     MS Thesis of Francesco Ricciuti
Supervisor:
     Ugo Rosolia
     Jon Gonzales

Navigation in Tight Environments

Navigation in Tight Environments

Self-Driving Car Testing, with Unity, VR and a big open space

Self-Driving Car Testing, with Unity, VR and a big open space