Submitted by Prof. D. Cebon on Wed, 03/02/2016 - 21:41
Qiheng (Matt) Miao's PhD dissertation 'Vision-based path-following control of articulated vehicles' has been approved by the university.
The research project developed and tested a vision-based system, with one or two cameras, for accurately measuring the motion of the vehicle relative to the road surface and using this information to control the path-following trailer steering system. It has applications in situations when the road surface is slippery (for example off-highway) when conventional trailer steering systems fail to work because they lose track of their position due to lateral and longitudinal wheel slip. The work was performed in collaboration with members of the Cambridge Vehicle Dynamics Consortium www.cvdc.org.
Summary
The dissertation describes research into the use of a vision-based active steering system on the trailers of articulated heavy goods vehicles. This system improves manoeuvrability at low speeds by demanding a follow point (the trailer rear end) to follow the path of a lead point (the tractor 5th wheel). It is designed to operate in slippery, off-road conditions, in which conventional path-following trailer steering systems fail to work correctly.
A parametric study was performed to investigate the effects of road camber, grade and adhesion on the path-following performance of an existing active steering system. It was found in the study that road camber, grade and low adhesion could lead to large sideslip at tractor 5th wheel and longitudinal slip on drive wheels. These resulted in erroneous off-tracking estimates of the ‘follow point’ which is a key input to the active steering controller. Path-following performance therefore degraded under such conditions.
In order to provide accurate off-tracking signals for path-following steering control, two prototypes of vision-based navigation systems were developed. They both used feature-point-based techniques to obtain the desired navigation data with the camera(s) looking downwards at the ground, so they are denoted ‘Ground-Watching Navigation Systems’ (GWNS). These systems are less subjected to errors due to road cambers and grade so that the effects of sideslip and longitudinal slip on off-tracking errors under such conditions can potentially be eliminated.
A modified feedback controller was devised to generate demand trailer axle steer angles based on the off-tracking signals measured by the GWNSs. Simulations were conducted to determine open loop performance of the vision systems and closed loop performance of the overall active steering system. It was found in the simulations that such vision systems had the potential to improve substantially, the path-following performance at low speeds.
A test vehicle fitted with two ground-watching cameras and the active steering system was employed to investigate the practical performance of the developed systems. Vehicle testing was conducted for a series of manoeuvres in open loop and closed loop. The two vision systems were both found to provide accurate measurement signals in the open loop tests. Fed with navigation signals from the vision system, the path following steering system successfully controlled the trailer to follow the desired path in the closed loop tests.