Measurement of articulation angle by image template matching

Abstract

Articulated heavy goods vehicles and long combination vehicles offer improved efficiency and emissions but reduced low-speed manoeuvrability and high-speed stability compared to rigid lorries. Active control systems can help to address these issues but require additional instrumentation, particularly articulation angle sensing. Existing commercial and research solutions for articulation sensing are limited either in terms of measurement accuracy or practicality. In this paper, a camera-based articulation angle sensing concept using template-matching and an Unscented Kalman Filter is investigated, including detailed simulation and full-scale vehicle experiments. The effects of aliasing and trailer pitch angle are studied, and a correction model for trailer pitch is proposed. The sensing system demonstrates superior measurement accuracy compared to the published literature, with root mean square errors of 0.8°–1.8° for articulation angles up to 37° in full-scale experiments. The sensor solution proposed in this work addresses more of the practical criteria for this problem than the comparable solutions in the literature. It does so through the novel camera and image processing system which is unique to this application. The proposed sensor offers acceptable measurement accuracy in real-time, is non-contact, requires minimal knowledge of trailer parameters and is compatible with semi-trailers and rigid draw-bar trailers.

Camera-Based Articulation Angle Sensing for Heavy Goods Vehicles

“…The shape of the error function with respect to articulation angle was found Fig. 6: Tractor semi-trailer test vehicle and instrumentation (all dimensions in mm) [26] Fig. 7: Stereo initialisation and zeroing in field tests to be consistent in shape, but varied from test to test in its precise location which gave rise to the maximum errors manifesting at different articulation angles between tests.…”

“…In 2019, De Saxe and Cebon presented a further development on the rear-facing camera and template matching approach of Harris [26], incorporating an Unscented Kalman Filter to smooth the noisy template-matching measurements. The system relied on some knowledge of the trailer geometry (making the solution not fully tractor-based), but sensitivity to this was shown to be small.…”

“…Representative articulation angle signals were generated using a three degree-of-freedom 'bicycle model' of a tractor semi-trailer vehicle combination. (More details can be found in [26].) The model was subjected to a square wave steer input, yielding an approximately sinusoidal articulation angle signal.…”