Path Control for Autonomous Four-Wheel Steered Haulers

Abstract: Path following control is a fundamental part of vehicle autonomy and over the lastdecades extensive research has been conducted towards the development of moreadvanced algorithms for path control of autonomous vehicles. The primary interestin improving path following control for autonomous vehicles is to be able to reducethe tracking error relative to the vehicle's reference path while also increasing thespeed at which trajectory tracking is possible. In this thesis, path following controlwas considered for an autonomous electric four-wheel steered hauler. For thispurpose, two dierent control paradigms were tested. Non-linear feedback controland model predictive control. Both controllers were designed to stabilize the lateraloset of the hauler in the curvilinear coordinate system constituting its referencepath. The controllers were synthesized with regards to a kinematic model of afour-wheel steered vehicle in a curvilinear coordinate system, and tuned extensivelyversus a multi-body simulation of the hauler. The MPC was also implemented inexisting ROS code for a prototype of the real hauler and its real-time schedulabilitywas tested on the machine control hardware in the hauler prototype. It was foundthat both controllers met the required tracking precision, while the model predictivepath controller performed better but at a signicantly higher cost in terms ofcomputational complexity and coding overhead.