Neuro-fuzzy control with a novel training method based-on sliding mode control theory: application to tractor dynamics

Accurate automatic guidance of agricultural vehicles is essential for gaining the ultimate benefit in agriculture. As an agricultural vehicle, tractors have more than one subsystem interacting each other, e.g. yaw dynamics, longitudinal dynamics, implement dynamics, etc. Instead of modeling the subsystem interaction prior to model-based control, we have developed a control algorithm which learns the interactions by using the measured feedback error. In this study, two PD controllers and two fuzzy neural networks are combined for controlling the yaw and traction dynamics. While the former ensures the stability of the related subsystem, the latter learns the system dynamics and becomes the leading controller. The interactions between both subsystems are not taken into account explicitly, but are considered to be disturbances in the control of each individual subsystem. A novel sliding mode control theory-based learning algorithm is used to train the fuzzy neural networks, and the convergence of the parameters is shown using a Lyapunov function