The authors propose a model-based multiobjective fuzzy control method which is optimized online via a novel multiobjective dynamic programming. The new multiobjective dynamic programming is guaranteed to derive a Pareto optimal solution. To estimate the effect of each candidate for control input in the dynamic programming procedure, we use state-value predictors of multiple objectives based on the plant model. Temporal difference learning and supervised learning are used for update of the predictors and the plant model. As the learning proceeds, the proposed method derives the compromised solution among multiple objectives. To show the effectiveness of the proposed method, some simulation results are given.
展开▼
