| Summary: | This study presents a modified algorithm of the grey wolf optimizer to solve the problem of learning rate selection in the multilayer type-2 asymmetric fuzzy controller (MT2AFC). The improvements of our modified optimizer are: the best position of the swarm is memorized, thus making the alpha wolves only update when a better position appears in the next iteration; search performance is enhanced by giving more freedom to update the grey wolf position. The proposed optimizer algorithm is then applied to optimize the suitable learning rates for the proposed controller. The multilayer type-2 asymmetric membership function is used in the fuzzy control network to enhance the learning ability and flexibility of the designed network architecture. The gradient descent method is used to adjust the parameters of the proposed MT2AFC controller online. The stability of the system is guaranteed using the Lyapunov approach. Besides, the self-evolving algorithm is used to construct the network structure autonomously. Ultimately, the numerical simulations of the chaotic synchronization systems are carried out to verify the effectiveness of our proposed method.
|
|---|
