Bionic lane driving of autonomous vehicles in complex urban environments: Decision-making analysis

A tactical-level lane-driving and decision-making model that considers multisource information in a complex and dynamic urban environment is critical for the development of autonomous vehicles. A key challenge in operating an autonomous vehicle robustly in the real world is to deal with dynamic and uncertain information. In the real world, drivers are capable of making accurate and timely decisions that should also be required of autonomous vehicles. In this work, the authors describe the development of an algorithm for autonomous lane change functions in which information was extracted from the decision-making process of human drivers to support the decision making of autonomous vehicles. First, a virtual urban traffic environment was built with PreScan, which is a simulation environment for the development of advanced driver assistant systems and intelligent vehicle systems. The vehicle dynamics were simulated by a dynamic model, with 6 degrees of freedom, based on Simulink, and driver decision rules were extracted through the concept of rough set theory. After that, an algorithm was presented for lane-driving decision making at the tactical level when velocity control operation was desirable and feasible. The development of the algorithm was based on driver experience, safety thresholds, and acceptable gap theory. The algorithm was proved to provide satisfactory velocity control actions as well as to safely decide whether to change lanes in a real urban environment. Finally, the reliability and effectiveness of the model was validated by both simulations and real road experiments. The findings from this study can provide a theoretical basis for the in-depth study of driving decision making in complex and uncertain environments.