Joint torque optimization for redundant manipulators using neural networks

One of the important applications for the resolution of redundant manipulators is torque optimization. To achieve this objective, finding out the most desirable configuration from the infinite number of possible configurations that satisfy the end-effector constraint is required. It has been previously shown that the pseudoinverse plays a crucial role in doing such calculations. In this work, the Tank-Hopfield (TH) network is adopted for pseudoinverse calculations and the connection weights of the network can be directly obtained from the known matrices at each sampling time. At acceleration level, the joint acceleration commands related to torque optimization are generated from the outputs of the network. Incorporating the TH network into the Null-Space (NS) algorithm allows a torque optimization to be implemented in real-time. Simulation results for a three-link planar manipulator are given to prove that the proposed scheme is efficient and practical.