Pure Pursuit Control Method Based on SVR Inverse-model for Tractor Navigation

Abstract

Considering the fact that uncertain tire sliding and uncertain pavement lead to trajectory of tractor could not be accurately described by kinematic model, a pure pursuit control method based on SVR inverse model was proposed for agricultural machine navigation. This paper analyzed and determined the main structure and technical parameters of the method. The inverse model of forward heading in tractor was established by using the method of granular support vector regression, and the corresponding relation function of kinematic theoretical curvature and actual curvature was obtained. The error of the output of the pure pursuit navigation model was corrected by inverse model, thus the adaptability and dynamic performance of pure pursuit control method were improved. The path tracking experiments carried out on navigation system of the tractor and the pavement experiment results showed that the maximum of linear tracing pitch yaw roll error was less than 0.0614m, when the speed of agricultural machinery was 1.2m/s and path length was longer than 125m. Compared with the method of conventional pure pursuit navigation model, the pure pursuit control method based on SVR inverse model has better linear racing performance. Field experiment results concluded that the maximum lateral deviation was 0.0887m, when the running speed of the tractor was 0.5m/s, and the proposed controller significantly improved precision of field experiment. Based on the path tracking experiments and field experiments results, the navigation control method could be applied to automatic row controlled operation of 2BFQ-6 type direct seeding combined dual purpose planter for rapeseed.

Keywords: tractor  navigation,  inverse-model,  support vector regression, pure pursuit model