Path Following Control for Autodocking

Abstract

This thesis describes the design and implementation of a path following controller that can be used for automated docking of tractor-semitrailers on yards. Two path following controllers based on either cascaded feedback control or nonlinear model predictive control are developed, which control the steering wheel angle of the truck in order to follow a given reference path.
The cascaded feedback controller is an updated version of the original controller developed by DAF and is referred to as the updated controller. The original controller was demonstrated successfully at one of DAF’s largest customers, but needed improvements for certain use cases. The problems of the original controller have been analyzed and reconstructed in simulation using a dynamic vehicle model. Improvements have been implemented in the original controller, resulting in an improved performance. Simulation results show that the implemented improvements have decreased the average absolute lateral error at the docking position.
Besides improving the original controller, new docking requirements have been identified. A new path following controller using a nonlinear model predictive control (NLMPC) approach is developed, which is selected because of its ability to explicitly include constraints and limitations, while also being able to control multiple vehicle states. This controller is referred to as the MPC controller.
Autonomous docking simulations in MATLAB Simulink have shown that both controllers are able to successfully dock a tractor-semitrailer, while staying within the prescribed error tolerances. Simulations show that the NLMPC controller achieves a better performance at the docking location than the updated controller, for the determined set of controller parameter values and the used reference path.
The updated controller has been implemented and tested on a real truck, at DAF’s test track in Sint-Oedenrode. These tests have shown a good docking performance for left side docking in terms of lateral error at the docking position. A large difference in performance is observed for right side docking. An analysis of the results has shown that this may be a result of localization errors caused by the localization algorithm. This problem needs to be addressed in more detail before the updated controller can be successfully used for autonomous docking applications on real trucks.

Date of Award	21 Dec 2021
Original language	English
Supervisor	Igo J.M. Besselink (Supervisor 1)