EXPERIMENTAL IMPLEMENTATION OF AN ARTIFICIAL IMMUNE SYSTEM ON A ROBOTIC MANIPULATOR

Author： Williams Keith Garbanzo Flavio Karr Charles

ISSN： 1087-6545

Source： Applied Artificial Intelligence, Vol.24, Iss.3, 2010-03, pp. : 175-193

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

An artificial immune system (AIS) has been used to realize robust control of a robotic manipulator. The AIS recognizes “self” and “non-self” operation of a closed-loop system, where self is defined as a condition where controller gains are appropriate for a given manipulator configuration. As configuration changes occur, the changing performance of the system indicates a transition to non-self. When non-self operation is first detected, the corresponding dynamic response is defined as a receptor and a genetic algorithm (GA) is called to optimize the controller for the new configuration. A library of receptors is built as additional configuration changes are experienced. For subsequent self to non-self transitions, new and recorded receptors are compared. In the event of a high correlation between the receptors, previously determined controller gains are implemented without calling the GA. The system is agile and robust and can recognize and respond to recognized receptors within a single reference step.