The Design Of Robust Control System Based On Multi-objective

The designing result of robust control systems based on traditional multi-objective method is more conservative. This thesis mainly applies multi-objective genetic algorithm to obtain less conservative result for this kind of problem.Firstly, by combining genetic algorithm and local search method, the thesis designs an Adaptive Multi-Objective Genetic Algorithm with the Dimension Reduction and Scanning approach (DRSA-MOGA). By using the method of fitness function normalization, the dimension reduction, scanning and the adaptive adjustment of fitness function, the performance of DRSA-MOGA is improved in global optimization and the local search ability.Simulation results show that Pareto solutions obtained by DRSA-MOGA can approch Pareto nondominated front very well,at the same time can keep good diversity.The DRSA-MOGA is used to design robust controller of automobile four wheel steering system considering three objectives at the same time, and these are the position of closed-loop poles, dynamic performance in time-domain and H₂/H_∞ mixed objective. This is one kind of more difficlt problem in multi-objective control system design. The simulation results show that the performance of system is better than that obtained by LMI approach.Lastly, the thesis studies the application of genetic algorithm in the mixed sensitivity problem of H_∞ control. Firstly singular-objective genetic algorithm is used to solve the problem. Then, the DRSA-MOGA is utilized to obtain the solution of the mixed sensitivity problem. By simulation example, we find this result has satisfactory performance in sensitivity and complementary sensitivity.