The thesis developed and tested an autonomous reversing system for multiply-articulated heavy goods vehicles. The experimental system was shown to work well with up to vehicles up to three trailers (See video). The work was funded by EPSRC and Volvo Trucks with support from other members of the Cambridge Vehicle Dynamics Consortium www.cvdc.org.
This dissertation describes research on automated reversing of multiply-articulated heavy vehicles. Two aspects are addressed: planning collision-free trajectories and controlling the vehicle to track a desired path in reverse.
The project motivation, background, literature review and research questions are presented in Chapter 1. There is a strong case for automated reversing on multiply-articulated vehicles. Relevant examples of path-tracking controllers were found in the literature, but these lack practical considerations such as sensor noise and controller parameter tuning techniques, and none have been implemented on a full-size, multiply-articulated vehicle. Many of the existing path planners for articulated vehicles have limitations such as calculation runtime, repeatability and vehicle geometry assumptions.
In Chapter 2, two general vehicle models are derived using kinematic and dynamic equations of motion. Both vehicle models are linearised and a method is outlined for calculating the swept path of a vehicle.
Path planning is discussed in Chapter 3. A ‘car track’ path planning strategy is presented which uses a set of path segments, each with a pre-computed vehicle swept path. An empirical relationship between vehicle parameters and trailer path curvature is developed and used to generate feasible path segments. An optimisation algorithm is used to connect the segments together to form a collision-free path.
Three controllers for reversing articulated vehicles are described in Chapter 4: a simple, stabilising controller and two controllers for path tracking in reverse. Linear analysis and controller tuning approaches are discussed and simulation results are presented for all three controllers. The path-tracking controllers are shown to be robust to sensor noise and vehicle parameter variations.
Chapter 5 describes the experimental vehicle testing setup including test vehicles, instrumentation, actuation, cameras and global controller software. Chapter 6 outlines the vehicle testing results from implementing the controllers developed in Chapter 4 onto a Tractor-Semitrailer, a B-double and a B-triple test vehicle (with one, two and three trailers respectively). The B-double is able to track paths with offsets at the rear trailer axle of less than 0.15m. Simulation and experimental results are compared. The experimental results for the path-tracking controllers exhibit some persistent oscillations which do not exist in the simulation data. An investigation into the oscillations is presented in Chapter 7, the root cause is identified and the possibility of reducing them through controller tuning is investigated.
Conclusions and recommendations for future work are given in Chapter 8.