Forward kinematic analysis of a parallel ankle rehabilitation robot using modified sugeno inference system

Forward kinematics (FK) mapping for parallel mechanisms is difficult as it involves highly coupled nonlinear motions. Since FK is a key element in closed loop position and force control of parallel manipulators, an accurate prediction of end pose from the available link lengths is essential. This paper proposes a modified Sugeno Inference System (SIS) to accurately predict the forward mapping between joint and Cartesian coordinates. The SIS algorithm has been implemented on a new parallel robotic device, designed for flexible ankle rehabilitation treatments. To increase the accuracy of the SIS, antecedent fuzzy variables are tuned using a modified genetic algorithm and consequent crisp membership functions are trained with a gradient descent algorithm. It has been found that the proposed system provides better accuracy and is also time efficient, which makes it a better candidate for real time control applications.