Novel system identification method and multi-objective-optimal multivariable disturbance observer for electric wheelchair.

Electric wheelchair (EW) is subject to diverse types of terrains and slopes, but also to occupants of various weights, which causes the EW to suffer from highly perturbed dynamics. A precise multivariable dynamics of the EW is obtained using Lagrange equations of motion which models effects of slopes as output-additive disturbances. A static pre-compensator is analytically devised which considerably decouples the EW's dynamics and also brings about a more accurate identification of the EW. The controller is designed with a disturbance-observer (DOB) two-degree-of-freedom architecture, which reduces sensitivity to the model uncertainties while enhancing rejection of the disturbances. Upon disturbance rejection, noise reduction, and robust stability of the control system, three fitness functions are presented by which the DOB is tuned using a multi-objective optimization (MOO) approach namely non-dominated sorting genetic algorithm-II (NSGA-II). Finally, experimental results show desirable performance and robust stability of the proposed algorithm.