부산대학교 도서관

Heuristic approaches are recently prevalent. Although they cannot be explained in mathematics, its performances are also impressive. In this paper, comparing with Genetic Algorithm (GA), we make the vehicle avoid obstacles in simulation by applying Grey Wolf Optimizer (GWO) to train the Artificial Neural Network (ANN). This method is unsupervised. Its inputs are the distances between vehicle and obstacle which are collected by rangefinders, while its outputs are the two steering angles that decide which direction the vehicle turns to. The fitness is lifetime of the vehicle. Finally, the experimental results show that the convergence speed of GWO is faster than that of GA.