Modeling, analysis, and adaptive neural modified-backstepping control of an uncertain horizontal pendubot with double flexible joints

Abstract

Position control of an underactuated uncertain horizontal pendubot with double flexible joints (PDFJ) that has only one actuator is a new and challenging topic. This study establishes the dynamics model of the PDFJ, analyzes its characteristics, and presents an adaptive neural modified-backstepping method to achieve its position control. Through theoretical analysis, we obtain the relationship between underactuated system dynamics and the kinetic energy derivative of the two links. On this basis, an adaptive neural modified-backstepping method is proposed, which deals with the effects of parameter perturbation, model uncertainty, and external disturbance and makes the PDFJ reach the target position. This method relaxes the strict restriction of conventional backstepping methods and dynamic surface control methods that demands that the relative degree of each virtual subsystem must be one, and it can still alleviate a differential explosion problem. Finally, we demonstrate its effectiveness through simulation and experimental results.