H.L. Li, Y. Xu, Y.X. Huang, X.K. Zeng, W. Xu, H.Y. Xia

pdf icon

Pages: 133-142

In addressing challenges associated with behavioural decision-making in intelligent driving, a hybrid algorithm, merging fuzzy classification with neural networks (termed the FC-NN Decision-Making methodology), has been introduced. Features from the driving environment, such as vehicle speed and relative distance, were systematically extracted. Using established traffic rules and vehicular performance metrics, a database linking the driving environment to decision outcomes was formulated. Through defined classification rules, fuzzy categorisation was applied, upon which training was subsequently conducted via NNs. This led to the design of an efficient Fuzzy-NN Decision-maker. Analyses demonstrated that the introduced method yielded an accuracy rate of 94%, markedly surpassing the accuracies of both the RBF NN decision-making at 56% and the direct NN approach at 76%.
Keywords: intelligent driving; behavioural decision-making; fuzzy classification; neural network; decision-making accuracy