Quantum genetic algorithm to evolve controllers for self-reconfigurable modular robots

Purpose The purpose of this study is to explore using real-observation quantum genetic algorithms (RQGAs) to evolve neural controllers that are capable of controlling a self-reconfigurable modular robot in an adaptive locomotion task. Design/methodology/approach Quantum-inspired genetic algorithms (QGAs) have shown their superiority against conventional genetic algorithms in numerous challenging applications in recent years. The authors have experimented with several QGAs variants and real-observation QGA achieved the best results in solving numerical optimization problems. The modular robot used in this study is a hybrid simulated robot; each module has two degrees of freedom and four connecting faces. The modular robot also possesses self-reconfiguration and self-mobile capabilities. Findings The authors have conducted several experiments using different robot configurations ranging from a single module configuration to test the self-mobile property to several disconnected modules configuration to examine self-reconfiguration, as well as snake, quadruped and rolling track configurations. The results demonstrate that the robot was able to perform self-reconfiguration and produce stable gaits in all test scenarios. Originality/value The artificial neural controllers evolved using the real-observation QGA were able to control the self-reconfigurable modular robot in the adaptive locomotion task efficiently.