LPV Control of an Autonomous Vehicle

Abstract

There are four main procedures that an Autonomous car-like vehicle must perform in a real-time dynamic complex environment: Perception and Modelling, Localization and Map Building, Path Planning and Decision-Making, and Motion Control. The Motion Control procedure translates the decision taken by the Path Planning and Decision-Making module to specific commands for the different actuators of the vehicle. The Motion Control issue treated in this Thesis is the Path Tracking problem. The goal of this Thesis is to approach the problem in the simplest possible way whilst being accurate and developing an intuitive design procedure. To achieve this goal, the bicycle-like kinematic model and a switched state feedback LPV controller, which has been designed using the polytopic transformation via nonlinear embedding and bounding box approach in order to perform the control of the system in an integrated way (lateral and longitudinal control at the once), have been chosen. Consequently, two different controllers have been synthesized, compared and proven fully adequate to confront the Path Tracking Problem. Both controllers show very good results and seem very consistent in the non-conservative simulation tests performed. The approach proposed in this Thesis is simple, intuitive and has produced successful simulation results.