A heuristic trajectory planning method considering long-short term prediction-planning collaboration

A reliable and efficient trajectory planning method is crucial for safe and efficient autonomous driving. One promising approach involves incorporating dynamic changes in the surrounding environment into trajectory planning. Although current trajectory prediction methods can predict vehicle trajectories, there is still a need for further exploration in effectively utilizing these predicted trajectories with uncertainties in trajectory planning. This paper proposes a decoupling trajectory planning method considering long-short term prediction-planning collaborative strategy. In the path planning process, the uncertainty of the predicted trajectory is considered in the short term and not in the long term, based on time interval sampling. The advantage of this method is that the predicted trajectories of other vehicles can provide inspiration for the path planning part and improve the robustness of the planned trajectory in extreme scenarios. Finally, we conducted simulations based on typical scenarios, specifically focusing on multi-lane overtaking scenarios that prioritize traffic efficiency and emergency avoidance scenarios that emphasize safety. The simulation results demonstrate that our approach outperforms the comparative method in terms of both efficiency and safety.