Online evolutionary control using a hybrid genetic based controller

Servo system controlled by the proposed hybrid GA based controller is presented in this paper. Due to their powerful optimization property, genetic algorithms are currently being investigated for development of a learning controller. However, because of time consuming in evolutionary process and the unstable response during learning period, most of genetic based controllers are implemented on off-line learning problems. To investigate this controller for the online learning, in this paper, a hybrid control architecture that integrates bang-bang controller with a genetic based controller is proposed. Bang-bang control is applied in the case of the actual output is far away from the set point (high error range). In such mode, bang-bang control is required to avoid the unstable response. In medium and small error ranges, a genetic based controller is applied as a learning controller. Simulation results demonstrate the effectiveness and learning ability of the proposed controller.