Integrating CPO and Hybrid A* Algorithms for Efficient UAV Path Planning *

To address the drawbacks of the traditional A* algorithm, such as excessive search node expansion, high computational complexity, unsmooth paths, and over-sensitivity to heuristic functions, this paper proposes an improved path optimization algorithm, CPO-A* (Hybrid A* and Continuous Path Optimization). The algorithm combines the initial A* path with gradient descent optimization to minimize the squared distances between adjacent nodes, thereby smoothing the path and improving its quality. Initially, the algorithm generates a starting path using A* through a warm-start approach. It then uses gradient descent to update intermediate nodes, at the same time, the priority of the drone is set, and if possible collision is detected, adjust the speed and path planning strategy, ensuring no collisions. To further enhance smoothness and simplicity, a shortcut postprocessing technique removes redundant nodes by connecting non-adjacent ones. The algorithm is tested in a two-dimensional irregular obstacle environment, and compared with A*, PRM, and other algorithms. Experimental results show that CPO-A* produces the shortest path with the best smoothness and outperforms A* and PRM by 6.23% and 2.99% in efficiency, making it suitable for two-dimensional path optimization