Stable-Optimum Gain Tuning for Designing Mobile Robot Controllers Using Incest Prevented Evolution

Abstract

We present an evolution strategy (ES) algorithm - incest prevented evolution strategy (IPES) enhancing our novel evolution strategy (NES) algorithm. Validity of NES and IPES algorithms is compared with other evolutionary algorithms (EAs) and relative performances and also compared with test function results. The IPES algorithm shows the highest balance between exploration and exploitation over the NES algorithm on these test functions by achieving high-precision global results. Both algorithms are applied to solve stabilizing optimum gain tuning problems in mobile robot controllers. Two optimal servocontrollers are considered for a mobile robot with two independent drive wheels. A bidirectional fitness (cost) function is constructed for these controllers so that stable but optimum gains are tuned automatically evolutionarily instead of using a traditional algebraic Riccati equation solution. Two trajectory tracking control examples (straight line and circular) are considered for controllers. The superiority of the IPES algorithm over the NES algorithm is repeated in the application domain and the effectiveness of evolutionary gain tuning demonstrated by simulation results.