Hybrid position/force controller of robot manipulators based on CMAC learning

We propose a stable hybrid position and force control scheme based on the CMAC learning algorithm. A learning scheme is used to adjust the memory values in the CMAC module online based on observations of the robot input-output relationship, in order to form an approximate dynamic model of the robot in appropriate regions of the state space. The CMAC module is used to predict the actuator torque required to follow a desired trajectory (force), and these torque are used as feedforward terms in parallel with a fixed-gain, linear PD feedback controller. The parameters of the PD feedback controller are strictly selected to ensure system dynamic stability. Finally, the simulation on a PUMA 562 manipulator shows that the CMAC-based hybrid force/position controller can provide dynamic stable control and fast learning speed. It is suitable for both repeated and non-repeated tasks.