Path planning is crucial for autonomous flight of fixed-wing Unmanned Aerial Vehicles (UAVs). However, due to the high-speed flight and complex control of fixed-wing UAVs, ensuring the feasibility and safety of planned paths in complex environments is challenging. This paper proposes a feasible path planning algorithm named Closed-loop Radial Ray A* (CL-RaA*). The core components of the CL-RaA* include an adaptive variable-step-size path search and a just-in-time expansion primitive. The former enables fast path search in complex environments, while the latter ensures the feasibility of the generated paths. By integrating these two components and conducting safety checks on the trajectories to be expanded, the CL-RaA* can rapidly generate safe and feasible paths that satisfy the differential constraints that comprehensively consider the dynamics and control characteristics of six-degree-of-freedom fixed-wing UAVs. The final performance tests and simulation validations demonstrate that the CL-RaA* can generate safe and feasible paths in various environments. Compared to feasible path planning algorithms that use the rapidly-exploring random trees, the CL-RaA* not only ensures deterministic planning results in the same scenarios but also generates smoother feasible paths for fixed-wing UAVs more efficiently. In environments with dense grid obstacles, the feasible paths generated by the CL-RaA* are more conducive to UAV tracking compared to those planned using Dubins curves.

With the significant reduction in mechanical failures and the rapid development of Artificial Intelligence (AI) technology, Human Factors (HF) issues in Unmanned Aircraft Systems (UAS) are becoming increasingly serious. To enhance the overall performance of specialized UAS in the low-altitude economy, it is essential to address the key HF design technologies. This paper, based on literature research and industrial practical experience, identifies three key scientific problems: the formulation of systematic UAS HF standards, efficient and trustworthy human-AI collaboration, and life-cycle HF design. First, the unique HF challenges of UAS are outlined. Next, current status of four categories of HF issues is discussed: automation design, display-control design, staffing and collaboration, and selection and training. Finally, three key HF design technologies and their progress, HF standard architecture design for UAS, efficient human-AI teaming design, and human-system integration design, are highlighted.

For autonomous Unmanned Aerial Vehicles (UAVs) flying in real-world scenarios, time for path planning is always limited, which is a challenge known as the anytime problem. Anytime planners address this by finding a collision-free path quickly and then improving it until time runs out, making UAVs more adaptable to different mission scenarios. However, current anytime algorithms based on A* have insufficient control over the suboptimality bounds of paths and tend to lose their anytime properties in environments with large concave obstacles. This paper proposes a novel anytime path planning algorithm, Anytime Radiation A* (ARaA*), which can generate a series of suboptimal paths with improved bounds through decreasing search step sizes and can generate the optimal path when time is sufficient. The ARaA* features two main innovations: an adaptive variable-step-size mechanism and elliptic constraints based on waypoints. The former helps achieve fast path searching in various environments. The latter allows ARaA* to control the suboptimality bounds of paths and further enhance search efficiency. Simulation experiments show that the ARaA* outperforms Anytime Repairing A* (ARA*) and Anytime D* (AD*) in controlling suboptimality bounds and planning time, especially in environments with large concave obstacles. Final flight experiments demonstrate that the paths planned by ARaA* can ensure the safe flight of quadrotors.

Non-learning based motion and path planning of an Unmanned Aerial Vehicle (UAV) is faced with low computation efficiency, mapping memory occupation and local optimization problems. This article investigates the challenge of quadrotor control using offline reinforcement learning. By establishing a data-driven learning paradigm that operates without real-environment interaction, the proposed workflow offers a safer approach than traditional reinforcement learning, making it particularly suited for UAV control in industrial scenarios. The introduced algorithm evaluates dataset uncertainty and employs a pessimistic estimation to foster offline deep reinforcement learning. Experiments highlight the algorithm’s superiority over traditional online reinforcement learning methods, especially when learning from offline datasets. Furthermore, the article emphasizes the importance of a more general behavior policy. In evaluations, the trained policy demonstrated versatility by adeptly navigating diverse obstacles, underscoring its real-world applicability.