Commande d'un syst eme non lin eaire par une m ethode hybride

Abstract

Moving a suspended load along a pre-speci ed path is not an easy task when strict speci cations on the swing angle need to be satis ed. However, to minimize the swing angle, adaptive control laws are essential and especially in case the dynamics of systems subject to uncertainties. In this work we propose a neuro-fuzzy controller for o -line control of the Motoman robot in simulation as well as a decoupled controller neuro- fuzzy controller based on the sliding mode theory for the control of 3D crane system. With the latter, we used two structures of the hybrid controller ; the rst is the neuro- fuzzy-type 1 and the second is the neuro-fuzzy type-2 controller. The considered 3D crane system involves a plan movement in conjunction with a lifting movement. It has three control inputs only (trolley and hoisting forces) with ve controlled variables (the trolley position in the XOY plane, the length of the lifting cable, and the two angles of swing). The interactions between each control subsystem are not taken into account explicitly, but are considered to be disturbances in control of each individual subsystem. In the proposed approach, a conventional controller (PD) is used in parallel with the neuro- fuzzy controller, the PD controller ensures the asymptotic stability in compact space, the parameter update rules of the fuzzy neural network are derived, and the proof of the online learning algorithm is veri ed by using the Lyapunov stability method. Ex- perimental results are given to solve the crane position control problem of 3D crane system laboratory equipment. Abstract Moving a suspended load along a pre-speci ed path is not an easy task when strict speci cations on the swing angle need to be satis ed. However, to minimize the swing angle, adaptive control laws are essential and especially in case the dynamics of systems subject to uncertainties. In this work we propose a neuro-fuzzy controller for o -line control of the Motoman robot in simulation as well as a decoupled controller neuro- fuzzy controller based on the sliding mode theory for the control of 3D crane system. With the latter, we used two structures of the hybrid controller ; the rst is the neuro- fuzzy-type 1 and the second is the neuro-fuzzy type-2 controller. The considered 3D crane system involves a plan movement in conjunction with a lifting movement. It has three control inputs only (trolley and hoisting forces) with ve controlled variables (the trolley position in the XOY plane, the length of the lifting cable, and the two angles of swing). The interactions between each control subsystem are not taken into account explicitly, but are considered to be disturbances in control of each individual subsystem. In the proposed approach, a conventional controller (PD) is used in parallel with the neuro- fuzzy controller, the PD controller ensures the asymptotic stability in compact space, the parameter update rules of the fuzzy neural network are derived, and the proof of the online learning algorithm is veri ed by using the Lyapunov stability method. Ex- perimental results are given to solve the crane position control problem of 3D crane system laboratory equipment.