A cache-genetic-based modular fuzzy neural network for robot path planning

We propose a modular fuzzy neural network (MFNN) based on cache genetic learning process. In this model, we use cache genetic algorithm to generate the possible patterns of the structure and the parameters directly for the MFNN. Using the proposed cache genetic algorithm, small population can be held to speed up the genetic process from cache pool and keep chromosomes fresh by extracting new blood from auxiliary pool. A modular fuzzy neural network is able to learn the set of simpler functions faster than a multilayer perceptron can learn the undecomposed function in complex systems. Combined with the cache genetic algorithm and the modular neural network to synthesize the fuzzy logic controller, the performance is better than usual fuzzy neural networks. We use the proposed model to solve the problem of the robot path planning and compare it with the other methods to realize its performance by considering four factors: safety factor, smoothness factor, length factor, and time factor. Experiments results show the proposed model is superior than other approaches.