Path planning and obstacle avoidance optimization for delivery robots in complex traffic environments

To address the problems of blind sampling and chassis mechanical wear in modern smart logistics park delivery robots operating in complex traffic environments, this paper proposes an edge-cooperative path optimization framework that integrates multi-level potential field-guided global planning with a spatiotemporal collaborative local game theory-based improved dynamic window method. First, a local artificial potential field is used to provide spatial geometric guidance for the extended tree of the global path search. Combined with fast collision avoidance detection based on discrete point angles and quadratic B-spline smoothing, the globally optimal reference path is generated. Second, multi-dimensional angle threshold conflict judgment and adaptive speed game strategy are introduced into the local dynamic obstacle avoidance underlying framework, reconstructing the trajectory evaluation model and endowing the robot with the ability to accelerate and avoid oncoming high-risk vehicles and smoothly stop and yield to pedestrians crossing laterally. Results show that in an extreme congested environment with 50% obstacle density, the improved global planning algorithm has a single computation time of only 0.95 s. Under a full load of 100 kg, the maximum lateral tracking error is controlled at 8.5 cm, and the comprehensive energy consumption for a single full-load delivery is only 25.2 Wh. In dynamic interactive obstacle avoidance tests, when faced with a sudden lateral pedestrian, the robot's linear velocity exhibits a smooth U-shaped curve, with an angular velocity adjustment pulse peak of only 0.25 rad/s. This demonstrates that the research method effectively eliminates mechanical structure wear caused by high-frequency, large-angle steering, ensures high real-time performance of underlying control commands and smooth electrical execution, and provides reliable theoretical support for the safe operation of heavy-duty park delivery robots.