AUTO-GENERATION MODEL/CONTROL FUZZY IMAGE MOBILE ROBOT SYSTEMS DESIGN

The evolutional particle swarm optimization (PSO) learning algorithm with self-regulated parameters and an auto-configured fuzzy model machine were applied to efficiently generate the mobile robot control systems. The omnidirectional image sensor was mounted on the mobile robot platform to capture objects surrounding the mobile robot with smart image processing technology to approach the desired traveling path. The generated kinematics mobile robot represents the behavior of the mobile robot in the visual traveling space. The appropriate fuzzy control rules of a mobile robot can be automatically extracted by the direction of the flexibly defined fitness function. The proposed self-learning algorithm can simultaneously avoid obstacles, approach the shortest path, and select the required fuzzy rules numbers. Based on the parameters of the self-generation procedure, the appropriate fuzzy rules were derived to guide the mobile robot toward the desired targets as soon as possible. Six examples of nonlinear mobile robot control problems were applied to demonstrate the adaptability of the self-generated learning algorithm. In the simulated examples, several blocks of various sizes (20, 30, and 40), various locations, and unusual initial and targeted positions were considered to test the adaptation of the learning scheme. Two types of evolutionary PSO learning algorithms were applied to achieve the desired results: one algorithm generates fuzzy rules with an adaptive procedure, and the other algorithm generates fuzzy rules with a random scheme. A comparison of the simulation results of the adaptive PSO (APSO) and random PSO (RPSO) learning algorithms showed that the appropriate mobile robot fuzzy systems were automatically generated by the APSO to form the required fuzzy rules, and detect and escape the obstacles within the desired and shorter traveling path when the initial environments changed.