Generalised Steering Strategy for Vehicle Navigation on Sloping Ground

Abstract

An autonomous tractor guidance system for vehicle navigation on sloping ground was developed by extending a previous study in which navigation was limited to a specific gradient. A previously developed neural network vehicle model was trained along the contour lines of 0, 5, 11 and 15° sloping terrain. Using the trained models and a cost function, optimal steering angles for different ranges of deviations were sought by genetic algorithm and the optimal values were arranged in separate matrix-form reference table for each slope. The optimal steering values of the four reference tables were expressed by a third-order polynomial equation so that the equation can provide an optimal steering value for any inclination. The coefficients of the equations were tabulated in a matrix form, which is designated as a ‘general reference table’. If the slope inclination and vehicle attitude can be determined, real-time navigation of the tractor is possible using this table. An autonomous travel test was conducted on inclined surface. It was found that the developed guidance system could precisely navigate the tractor along a rectilinear contour path on 20° sloping terrain, where the mean offset was 0·05 m and its standard deviation was 0·044 m.