Ko Jakyung, Yang Inchul
Department of Civil and Environmental Engineering, University of Science and Technology, Daejeon 34113, Republic of Korea.
Department of Highway and Transportation, Korea Institute of Civil Engineering and Building Technology, Goyang-si 10223, Republic of Korea.
Sensors (Basel). 2025 Sep 8;25(17):5601. doi: 10.3390/s25175601.
This study proposes a real-time path generation method called the Gradient-based Time-extended Potential Field (GT-PF) for cooperative autonomous driving environments. The proposed approach models the road environment and dynamic obstacles as a time-variant potential field and generates safe and feasible paths by tracing the negative gradient of the field, which corresponds to the direction of steepest descent. In contrast to conventional sampling-based or optimization-based methods, the proposed PF framework enables lightweight computation and continuous trajectory generation in spatiotemporal domains. Furthermore, a velocity-oriented bias is introduced in the PF formulation to ensure that the generated paths satisfy the vehicle's kinematic constraints and desired cruising behavior. The effectiveness of the proposed method is verified through comparative simulations against a sampling-based Rapidly exploring Random Tree (RRT) planner. Results demonstrate that the GT-PF approach exhibits superior performance in terms of runtime efficiency and safety. The system is particularly suitable for RSU (Roadside Unit)-based infrastructure control in real-time traffic environments. Future work includes the extension to complex urban scenarios, integration with multi-agent planning frameworks, and deployment in sensor-fused cooperative perception systems.
本研究针对协同自动驾驶环境提出了一种名为基于梯度的时间扩展势场（GT-PF）的实时路径生成方法。所提出的方法将道路环境和动态障碍物建模为时变势场，并通过追踪该场的负梯度（对应于最陡下降方向）来生成安全可行的路径。与传统的基于采样或基于优化的方法相比，所提出的PF框架能够在时空域中进行轻量级计算并生成连续轨迹。此外，在PF公式中引入了速度导向偏差，以确保生成的路径满足车辆的运动学约束和期望的巡航行为。通过与基于采样的快速扩展随机树（RRT）规划器进行对比仿真，验证了所提方法的有效性。结果表明，GT-PF方法在运行时效率和安全性方面表现出卓越的性能。该系统特别适用于实时交通环境中基于路侧单元（RSU）的基础设施控制。未来的工作包括扩展到复杂的城市场景、与多智能体规划框架集成以及部署在传感器融合的协同感知系统中。
