MKE Scheme for Planning and Control of Dual-arm Robotic System Aided with Recurrent Neural Networks

Dual-arm robotic systems have great advantages over the single-arm robotic system, for the arms can either run independently or cooperate to finish the task. In this paper, minimum kinetic energy (MKE) is concerned as a performance index for the planning and control of a dual-arm robotic system. First, two sub schemes are formulated on the left arm and right arm of the robot, which are constructed on velocity level with consideration of manipulators' physical constraints. Then, the two sub schemes are unified into one scheme, which is finally transformed into a quadratic program (QP) problem. For solving the formulated QP problem, a recurrent neural network (RNN) is devised based on the Lagrange multiplier method. At last, the simulations and simulative experiments on a dual-arm redundant robotic system named Baxter are carried out. The simulation results reveal that the devised RNN model has a good performance for solving the presented MKE scheme applying to the dual-arm robotic system.