Parameter learning and compliance control using neural networks

Abstract

The problem of identifying uncertain environments for stable contact control is considered. For this purpose, neural networks originally developed using terminal attractor dynamics are utilized. In the sequence, neural networks are used for learning the dynamics of an environment with which a robot establishes contact. In particular, system parameters are identified under the assumption that environment dynamics have a fixed nonlinear structure. A robotics research arm, placed in contact with a single degree-of-freedom electromechanical environment dynamics emulator, is commanded to move through a desired trajectory. The command is implemented using a compliant control strategy, where specified motion is biased with a compliance signal generated based on the error between desired and actual forces. The desired force is computed using neural network identified parameters and desired motion trajectory.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>