Neuro-adaptive motion control with velocity observer in operational space formulation

In this paper, a neuro-adaptive motion control with velocity observer is developed and validated in real-time experiment using a 6 DOF PUMA 560 robot. The controller is constructed for the operational space formulation, such that dynamic terms and the generalized force descriptions in this algorithm are expressed in the task space. The proposed strategy assumes no prior knowledge of the robot dynamics, and is formulated without assuming the availability of joint velocity feedback. As such, the controller takes only position feedback. This is an important feature as industrial robots are often fitted only with joint displacement sensors, not joint rate sensors. Stability analysis of the algorithm is analysed and presented in the paper. Real-time experiments on a 6 degrees of freedom (DOF) PUMA 560 manipulator were carried out to evaluate the effectiveness of the proposed NN adaptive control strategy with velocity observer and to compare the performance with the conventional inverse-dynamics control and a similar NN adaptive strategy using filtered velocity feedback, obtained through the backward difference of the displacement feedback. It should be noted that the experimental platform, PUMA 560, provides only joint displacement feedback through its joint mounted encoders. The results show a comparable results between the proposed strategy and the inverse-dynamics control law, without the need to perform dynamics identification procedures.