Automatic tractor guidance on sloped terrain (part 1) - formulation of NN vehicle model and design of control law for contour line travel

摘要

An automatic tractor guidance system for contour line travel on sloped terrain has been developed. To actualize this system, four procedures were done. First, a neural network (NN) vehicle model was proposed to obtain relations between the steerage and behavior of tractors on sloped terrain. Second, a navigation planner for contour run is constructed by the use of NN based vehicle model. Third, an optimal control theory is applied to the design of a steering controller. As two variables lateral displacement and heading angle are important for contour line tracking control, a quadratic-form cost function of the two are minimized by genetic algorithm. Finally, the fourth procedure involves using an installed steering mechanism, sensors and a Total Station, in carrying out autopilot tractor tests on contour line travel. The overall performance of the developed navigation system for the contour line tracking tractor-like robot was successful. Paper 2 will discuss a navigation planner for a rectangular path on sloped terrain