A Driving Behavior Decision Method Based on Gradient Boosting Tree Algorithm

In order to reduce the incidence of traffic accidents and improve passengers’ driving experience, intelligent driving technology has attracted more and more attention. The core content of intelligent driving technology includes environment perception, behavior decision-making and control follow-up. Simulating driver’s behavior decision-making based on multi-source heterogeneous environment information is the key to liberate drivers and become the focus and difficulty of intelligent driving technology. Aiming at this key problem, this paper presents a design method of driving behavior decision maker based on machine learning after fuzzy classification of historical data. Firstly, 1000 sets of driving environment-decision results database are generated randomly according to driving rules and driving state. A fuzzy classification rule is established to classify driving environment information such as speed and relative distance. Then, a driving behavior decision maker is designed based on gradient lifting tree algorithm of machine learning. Finally, the driving behavior decision accuracy of the proposed method is 100% after running the simulation software, while the decision accuracy of the behavior decision maker based on the traditional BP neural network method is only 70.4%. Through the above research, it can be concluded that the driving behavior decision maker based on machine learning has obvious advantages, and its high accuracy and fast operation can also meet the requirements of real-time, fast and accurate intelligent driving, which provides a technical basis for the realization of unmanned driving technology.