Hybrid State Observer Design for Estimating the Hitch Angles of Tractor-Multi Unit Trailer
Sangwon Han, Kyusang Yoon, Geonyeong Park, Kunsoo Huh
Abstract
In this paper, a new approach of state observer is proposed for the tractor with multi unit trailers. In the case of tractor with two articulated trailers, the dynamic characteristics of the trailers are dominantly determined by the two hitch angles connecting each trailer. A novel estimation system for the hitch angle is introduced by combining the Kalman filter with the deep learning network and transfer function techniques. The Gated Recurrent Unit (GRU) network is constructed to calculate hitch angles and these values are used as the virtual measurement in the Kalman filter. The dynamic characteristics of two trailers with respect to the hitch angles are expressed as the transfer function models and these models are used to calculate the hitch angles as the virtual measurement. The virtual measurements from the two methods are integrated separately into the Kalman filter design. The estimation performance of the hitch angle with the proposed hybrid observer is validated in simulations with various curvature scenarios.