A Generalized Model Predictive Path Planning System for Autonomous Vehicles on Structured Roads

Autonomous vehicles process the data received from their sensors to recognize the road and the obstacles in their perception module, and determine the desired route via the decision making module. Then, in the path planning module, they plan a path so that the vehicle follows the route while it observes the rules, avoids obstacles, and keeps the vehicle stable. Two common path planning methods, potential fields and model predictive controllers, have been combined in this project to develop a path planning module that is general, optimal, and predictive. The developed module has been simulated and showed an appropriate performance. In this project, the module is implemented on the autonomous vehicle of the host lab to validate the simulations. The path planning module is modified to be compatible with other modules used in the autonomous vehicle (the perception and decision making modules). It is also tuned for the vehicle, so that the experimental planned paths are as good as the simulation results.

Faculty Supervisor:

Amir Khajepour

Student:

Yadollah Rasekhipour

Partner:

Discipline:

Engineering - mechanical

Sector:

University:

University of Waterloo

Program:

Globalink Research Award