Sliding Mode Adaptive Neural-network Control for Nonholonomic Mobile Modular

Abstract. A general mobile modular manipulator can be defined as a m-wheeled holonomic/ nonholonomic mobile platform combining with a n-degree of freedom modular manipulator. This paper presents a sliding mode adaptive neural-network controller for trajectory following of nonholonomic mobile modular manipulators in task space. Dynamic model for the entire mobile modular manipulator is established in consideration of nonholonomic constraints and the inter-active motions between the mobile platform and the onboard modular manipulator. Multilayered perceptrons (MLP) are used as estimators to approximate the dynamic model of the mobile mod-ular manipulator. Sliding mode control and direct adaptive technique are combined together to suppress bounded disturbances and modeling errors caused by parameter uncertainties. Simu-lations are performed to demonstrate that the dynamic modeling method is valid and the con-troller design algorithm is effective. Key words: adaptive control, neural network, nonholonomic mobile modular manipulator, sliding mode control