Global Path Planning Based on A* Algorithm with Neural Network Enhancement for Mobile Robots

As a widely adopted pathfinding approach, the A* algorithm still faces persistent challenges in extensive operational environments, including prolonged computation duration, excessive directional changes, and jagged trajectory outputs. Aiming at overcoming these problems, an improved A* algorithm using neural networks is proposed in this work. First, a prioritized search mechanism employing an adjusted heuristic weighting accelerates path detection. Then, a neural network model forecasts navigational probabilities across grid nodes to reduce computational latency during obstacle interactions. Furthermore, Dubins curve integration enables smooth path generation through geometric modeling, crucial for mobile robot kinematic constraints. Comparative simulations against conventional A*, Weighted A*, bidirectional A* and Neural A* implementations demonstrate superior performance in search efficiency and path quality metrics. Practical validation through Ackermann-steering vehicle experiments confirms the algorithm’s operational feasibility.