Path-speed decoupling planning method based on risk cooperative game for intelligent vehicles

In order to reduce the solving dimension of trajectory planning process and improve the solving efficiency, the path and speed are often decoupled in the planning process of intelligent vehicles. However, under the existing serial decoupling planning framework, that is, the path is planned first and then the speed is planned, the collaborative relationship between path and speed is weak, and usually speed planning can only be carried out under fixed path planning. The first fixed path cannot be adjusted according to the speed state, making it difficult to balance the overall performance of the trajectory. Therefore, in scenarios that require close coordination between the two, such as steering to avoid obstacles, the synthesized trajectory often lacks safety or smoothness. Hence, this paper proposes a path-speed decoupling planning method based on risk cooperative game (DP-RCG). It performs decoupled planning of path and speed at large time intervals, and obtains a series of candidate coarse-grained path and speed curves. Then, a coarse-grained trajectory satisfying the demand is obtained by the static risk cooperative game of path and speed. Finally, the convex optimization method is used to generate a fine-grained smooth trajectory that can be well executed and tracked by the vehicle. Moreover, an omnidirectional risk assessment model is built to assess the risk of collision. Simulation results in different scenarios show that DP-RCG can ensure the safety and comfort of the planned trajectory. Meanwhile, compared with the existing methods, it has better computational efficiency and real-time performance.