In this paper, a quantitative ground risk assessment mechanism is proposed in which urban ground features are extracted based on high-resolution data in a satellite image when unmanned aerial vehicles (UAVs) operate in urban areas. Ground risk distributions are estimated and a risk map is constructed with a multi-layer method considering the comprehensive risk imposed by UAV operations. The urban ground feature extraction is first implemented by employing a K-Means clustering method to an actual satellite image. Five main categories of the ground features are classified, each of which is composed of several sub-categories. Three more layers are then obtained, which are a population density layer, a sheltering factor layer, and a ground obstacle layer. As a result, a three-dimensional (3D) risk map is formed with a high resolution of 1 m × 1 m × 5 m. For each unit in this risk map, three kinds of risk imposed by UAV operations are taken into account and calculated, which include the risk to pedestrians, risk to ground vehicles, and risk to ground properties. This paper also develops a method of the resolution conversion to accommodate different UAV operation requirements. Case study results indicate that the risk levels between the fifth and tenth layers of the generated 3D risk map are relatively low, making these altitudes quite suitable for UAV operations.

With the rapid development of unmanned aerial vehicles (UAVs), their applications in urban firefighting, rescue, and military missions continue to expand. However, maintaining formation stability and achieving reliable obstacle avoidance in complex low-altitude urban environments remain challenging. To address this issue, this paper proposes an Adaptive Cooperative Potential Field (ACPF) algorithm for UAV formation control. The algorithm is built on the artificial potential field framework. It includes three components. First, a cooperative control mechanism combines formation force and neighbor coordination force to improve spatial consistency. Second, a dynamic flight weight adjustment mechanism adaptively balances attractive, repulsive, and formation forces. Third, a 3D obstacle avoidance and altitude constraint strategy ensures safe flight by incorporating terrain information and safety margins. Simulation results show that ACPF outperforms APF, VS-DWA, and CLF-CBF-QP in formation maintenance, obstacle avoidance, and trajectory efficiency. For the regular pentagon formation, the total path length is reduced by 5.4% on average, and the convergence time is shortened by 63 s. For the rhombus formation and V-shaped formation, the path-length reductions reach 7.2% and 7.4%, while the convergence time decreases by 53 s and 66 s, respectively. Statistical experiments and parameter sensitivity analyses further demonstrate the robustness and reliability of the proposed algorithm. These results demonstrate the effectiveness of ACPF for UAV formation control in complex urban environments.