Multi-objective Genetic Algorithm for System Identification and Controller Optimization of Automated Guided Vehicle

This paper presents a multi-objective genetic algorithm (MOGA) with Pareto optimality and elitist tactics for the control system design of automated guided vehicle (AGV). The MOGA is used to identify AGV driving system model and optimize its servo control system sequentially. In system identification, the model identified by least square method is adopted as an evolution tutor who selects the individuals having balanced performances in all objectives as elitists. In controller optimization, the velocity regulating capability required by AGV path tracking is employed as decision-making preferences which select Pareto optimal solutions as elitists. According to different objectives and elitist tactics, several sub-populations are constructed and they evolve concurrently by using independent reproduction, neighborhood mutation and heuristic crossover. The lossless finite precision method and the multi-objective normalized increment distance are proposed to keep the population diversity with a low computational complexity. Experiment results show that the cascaded MOGA have the capability to make the system model consistent with AGV driving system both in amplitude and phase, and to make its servo control system satisfy the requirements on dynamic performance and steady-state accuracy in AGV path tracking.