A lane decision algorithm for autonomous vehicles considering potential driving risks

Aiming at the lane decision problem of autonomous driving (AD) in normal driving scenarios on structured road, with the goal of high safety and high efficiency, a lane decision algorithm for autonomous vehicles (AVs) considering potential driving risks is proposed. In this algorithm, by simulating the dimensions of human drivers’ attention to the driving environment (DE), the DE of AV is characterized from uncertainty, driving efficiency, and regional driving risk, based on the proposed potential risk field model, right-of-way (ROW) field model, and regional performance risk field model. Furthermore, by mining the behavioral patterns of human drivers’ lane decision based on their natural driving behavior data, the proposed algorithm has the ability to simulate the human drivers’ lane decision behavior. This is beneficial for integrating AVs into practical driving scenarios and will not cause discomfort to surrounding vehicles (SVs). In a continuous complex dynamic scenario composed of eight sub-scenarios, the performance verification based on two baseline algorithms and the performance comparison with six advanced algorithms are carried out. The results show that the proposed algorithm can balance driving safety and efficiency and select an optimal lane for AV.