计算机科学
运动规划
启发式
人工智能
路径(计算)
计算机视觉
实时计算
领域(数学)
钥匙(锁)
工作(物理)
全球定位系统
数据挖掘
弹道
作者
Enrique Aldao,Fernando Veiga-López,C. P. C. Chanel,Y. Watanabe,H. González-Jorge
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-05-08
卷期号:: 283-288
标识
DOI:10.1109/cai68641.2026.11536384
摘要
In recent years, UAVs have been increasingly deployed across various applications, including environmental monitoring, infrastructure inspection, and logistics. While many of these operations are still conducted by human remote pilots, the current trend is towards the development of fully autonomous solutions, which require efficient and robust control and path-planning systems. Traditionally, the most widely used path-planning algorithm is A*due to its deterministic behaviour and computational efficiency. This approach discretises the navigation environment into a grid with adjacent-node connectivity, thus constructing a graph, and minimises the path length through a graph search guided by heuristic functions. Nevertheless, the applicability of this approach is often limited to shortest-path problems where admissible heuristic functions are available. In other optimisation scenarios, such as when accounting for factors like operational risk, identifying heuristics that are both admissible and capable of efficiently guiding the search can be highly challenging. Another drawback of A*approaches lies in their discretisation, which restricts UAV movement to adjacent nodes. This limitation reduces vehicle manoeuvrability and compromises both path smoothness and optimality. To address these challenges, this paper presents a system based on machine-learning based heuristics for navigation within complex risk maps. The system is integrated into a customised A*framework that employs a k-Non-Aligned Nearest Neighbours (k-NANN) approach to compute smooth and optimal paths while maintaining low computational cost, deterministic behaviour, and interpretability.
科研通智能强力驱动
Strongly Powered by AbleSci AI