Acquisition of Autonomous Distributed Rule for Posture Control of Variable Geometry Truss by Genetic Programming

Abstract

An acquisition technique of autonomous distributed rule for posture control of Variable Geometry Truss (VGT) is suggested by applying the Genetic Programming (GP) method. Each control rule produced by GP consists of hierarchical tree structures. The main tree structure contains condition trees and rule selection trees as function nodes, and control trees as terminal nodes. The acquisition system trains individual rule by the posture control simulation for the randomly given initial postures, and evaluates the rule by the simulation of given test postures. The validity of the acquisition system is confirmed numerically by applying to the stiffness maximization of ten-module VGT.