Under the mixed traffic of manned and unmanned driving, due to the high uncertainty of human drivers' driving behaviors, unmanned vehicles only slowed down or stopped to avoid conflict which would cause low traffic efficiency. Aiming to solve the interaction problem between manned and unmanned vehicles at intersections, this paper adopted FSM(Finite-state machine) and Multi-objective optimization method to build a behavior decision-making system. Firstly, the Gauss process regression model was used to predict the conflict vehicles high-precision trajectory, and then the collision at intersections was judged by PET (Post Encroachment Time) parameter. Secondly, when the unmanned vehicle arrived the intersection area, the dynamic decision-making mechanism for unmanned vehicles was established based on FSM and then select an action for unmanned vehicles by maximizing the expected benefit, the benefit function was constructed with the safety, speediness and comfort indicators. Finally, the validity of conflict avoidance model for unmanned vehicles was verified by PreScan and MATLAB/Simulink simulation platform, the results showed that this method could make unmanned vehicle dynamically cooperate with other vehicles, the average time of conflict resolution was reduced by 6.5%, and no accidents occurred in the test, indicating that the model could not only ensure safety but also improve the traffic efficiency at intersections.

CICTP