Hybrid position/force control of constrained robot manipulator based on a feedforward neural network

In this paper, the control of constrained robotic manipulators is addressed and the solution of a reduced order model is obtained through a nonlinear transformation. A set of differential-algebraic equations are first derived. Then controllers are designed for position and force control. The position control involves the position and velocity feedback of end-effector, while the force control is developed based on an artificial neural network. The weights of the neural network are updated online using the force error as the objective function. An example of a 2-DOF manipulator system is studied in detail. Comparisons between a conventional PID controller and the designed controller are made and a practical application is carried out. The results demonstrate the effective performance of the system.