A safety-prioritized trajectory planning method based on multimodal predictive risk field for ramp merging

Ramp merging areas present critical safety challenges due to complex vehicle interactions and significant speed differences, which frequently lead to elevated collision risks. With the rapid development of connected and automated vehicle (CAV) technologies, advanced trajectory planning has emerged as a crucial capability to ensure safe in complex traffic environments. However, traditional trajectory planning methods for CAVs often encounter bottlenecks in ramp merging scenarios. They primarily rely on deterministic prediction models, thereby lacking the necessary adaptability to highly dynamic and uncertain traffic flows. To address these limitations, we propose a novel trajectory planning framework specifically designed for ramp merging situations, with a primary focus on enhancing safety. Our work mainly includes two contributions. First, we construct a multimodal predictive risk field (MPRF) to better capture the uncertainties in vehicle behaviors. This involves two key steps: 1) developing a multimodal trajectory prediction model based on an modified interactive multiple model (IMM) Kalman filter, which integrates a priority mechanism and refined motion models to predict lateral and longitudinal movements more accurately; and 2) constructing and calibrating a dynamic risk field using real-world aerial trajectory data processed through YOLO-based computer vision methods, ensuring the risk field reflects diverse driving behaviors. Second, based on MPRF, we design two risk matrices: one for path planning and one for speed planning. This separation allows the trajectory planner to independently optimize path and speed, better handling multimodal prediction uncertainties and dynamic-static coupling. Simulation experiments show that our proposed method significantly improves safety performance while maintaining traffic efficiency.