NASA's Jet Propulsion Laboratory, Pasadena, California
A decentralized direct adaptive control scheme for a six-jointed industrial robot has been demonstrated in the JPL Robotics Research Laboratory. In this scheme, each joint is controlled via a local feedback controller at a high sampli ng rate. This largely decentralized scheme eliminates that part of the overall computational burden that would be imposed by a centralized controller and that would degrade the performance of the robot by reducing the sampling rate. In the experimental system, the sampling and computation cycle is only 7 ms long. The higher and intermediate-level functions of the control algorithm (e.g., planning trajectories, evaluation of actual trajectories, and communication of control information to and from the robot) are executed by an external digital control computer. Another computer that is part of the robotic equipment acts as an input/output device between the external computer and the joint motors; this computer includes: (1) one microprocessor that serves as the local controller for each of the six joints and (2) a microcomputer that passes data and commands back and forth between this computer and the external control computer (see figure). The adaptive feedback controller for each joint implements a proportional/integral/derivative position-control algorithm with position/velocit y feedback and with gains updated in real time. The adaptive feature of the local control law implemented by each joint controller compensates for any static and dynamic cross-couplings between that joint and the other joints and contributes to accurate tracking of the commanded trajectory. When all joints move rapidly simultaneously, the interjoint couplings can sometimes cause instabilities under the decentralized control scheme. In such cases, the controller-adaptation laws are modified slightly to obtain stability in the presence of unmodeled cross- coupling effects. The control and controller-adaptation laws are based entirely on the observed performance of the manipulator: there is no need to model the dynamics of the robot, nor is there any need to specify the inertial parameters of the payload.
Thus, the adaptive controllers can cope with uncertainties and variations in the robot and the payload, including friction (which cannot be modeled accurately) and the mass of the payload (which can vary substantially).
More details can be found in:
Seraji, H., Lee, T., and Delpech, M.: "Experimental study on direct adaptive control of a PUMA 560 industrial robot," Journal of Robotic Systems, 1990, 7(1), pp. 81-105.
Point of Contact:
Homayoun Seraji,
Mail Stop 198-219
Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena, CA 91109
seraji@telerobotics.jpl.nasa.gov
Telerobotics Program
page
Please email the site webmaster
with any comments, criticisms or corrections
for this page.
Maintained by: Dave Lavery
Last updated: May 10, 1996