Regulating path control of industrial robot - computes torque by multiple layered forward aligned neuronal network from parameter variables of regulating system for controlling three axes industrial robot

摘要

The neuronal network is a three layer preceptron, by means of which, the connections of the preceptron which connect the input with the hidden layer are sigmoid. The connections between the hidden layer and the output layer are linear. The neuronal network has input junctions for angular, rotational rates and correction acceleration of the robot and output junctions for the drive torques of the robot. ADVANTAGE - Analytical compilation of dynamic equations are no longer necessary. Extremely computer efficient. Physical system data are automatically learnt and time fluctuations are considered. Compensation of effects depending on actual load can be directly trained.