Gradient-Based Time-Sequential Potential Field Method for Path Planning in Infrastructure-Based Cooperative Driving Automation

Abstract

This study proposes a GTS-PF (Gradient-based Time-Sequential Potential Field)-based path generation method for real-time reference-path planning at infrastructure-based RSUs in cooperative driving automation environments. Conventional path planning approaches exhibit limitations in computational lightweight characteristics or responsiveness to dynamic environments, which restrict their suitability for negotiation-oriented reference-path generation and dissemination. To address these limitations, the proposed GTS-PF framework interprets the prediction time horizon as a sequence of updated temporal layers, enabling adaptive responses to dynamic obstacle variations. The method is formulated based on potential field principles to allow efficient computation while incorporating diverse interaction effects. A key feature of the proposed approach is the separation of direction planning and speed planning for obstacle avoidance, wherein a candidate acceleration set is generated based on future risk evaluation. Simulation results in an overtaking scenario involving a low-speed preceding vehicle demonstrate that the proposed method satisfies predefined safety and path-quality criteria. Moreover, the computation time was reduced by 81% compared to the baseline method, confirming computational lightweight feasibility for RSU-level implementation and demonstrating applicability in infrastructure-led cooperative driving automation.