Parametric Motor Control: A New Approach to the Control of Point-to-Point Manipulator Movements

摘要

We propose a control scheme for point-to-point movements of robotic manipulators, which is based on an established concept about human movement control. In this approach, actuator drive signals have an invariant shape and are parametrically modified to fit the requirements of different tasks. The appropriate parameter values are provided by an Artificial Neural Network before movement onset. Explicit solutions of the inverse kinematic and inverse dynamic equations, or multiple feedback loops, are not required for this approach. We found that Parametric Motor Control yields satisfactory performance, allowing the manipulator to discriminate between targets in a 5 x 3 grid and to generalize to new target locations. The new control scheme has potential benefits for scenarios where complex manipulators must be controlled in real time.