부산대학교 도서관

A multi-factor fusion risk assessment approach for lane shifting state is designed with the goal of increasing the safety of lane changing in the human-machine co-driving environment of intelligent vehicles. To begin, the urban road traffic risk factors are described at three levels: "human, vehicle, and road" to create the driving risk factor set; then, the key indicators are screened using principal component analysis; and finally, the vehicle lane change risk level is identified by combining fuzzy theory and Bayesian Network. The experimental and computational results show that the fuzzy Bayesian network model accurately reflects the impact of several factors on driving safety, with weather and acceleration having the most influence. Simultaneously, when compared to the traditional Bayesian network, this method can effectively compensate for the Bayesian network's inference precision shortcomings, accurately determine the degree of danger of lane-changing behavior, and determine the current driver's risk level as a potential risk (level IV), which provides a theoretical basis for when to make a safe switch of the driving right in the intelligent vehicle human-computer co-pilot system and how to select the lane-changing strategy.