Dynamic lane-changing trajectory planning based on spatiotemporal skewed driving risk field
TL;DR
A skewed driving risk field (SDRF) model is developed considering the spatial asymmetry of risk distribution induced by the relative motion, and a multi-objective optimal controller is proposed by incorporating spatiotemporal risk into the cost function to achieve risk-aware motion planning.
Abstract
In lane-changing trajectory planning, the relative motion between the ego vehicle (EV) and surrounding vehicles (SVs) poses dynamic collision risks. However, existing risk field models limit their applicability in complex traffic environments as they are insufficient to quantify the direction and dynamic evolution of risks. This paper proposes a risk-aware dynamic trajectory planning (RDTP) method for lane-changing maneuvers. Specifically, a skewed driving risk field (SDRF) model is developed considering the spatial asymmetry of risk distribution induced by the relative motion. To capture the temporal evolution characteristic of risk, the model is further expanded to a spatiotemporal SDRF by integrating EKF-based short-term motion predictions of SVs. On this basis, a multi-objective optimal controller is proposed by incorporating spatiotemporal risk into the cost function to achieve risk-aware motion planning. Moreover, a rolling optimization scheme is designed to ensure real-time responsiveness, particularly for the handling of emergency scenarios. Multiple simulations are conducted in different scenarios using the highD dataset, and results jointly demonstrate the effectiveness of the proposed RDTP method for optimal and risk-averse lane-changing trajectory generation.