Risk-Aware Enabled Path Planning for Drones Flight in Unknown Environment

Under unknown environments, drones should always maintain vigilance to address potential threats. In fact, unknown obstacles suddenly moving and blocking the way could generate great flight safety risks. Besides conventional static and moving obstacles, addressing such unknown malicious obstacles is crucial for enhancing drone safety, yet relevant research is scarce. In this work, we propose a systematic planning framework for drones with switchable obstacle avoidance strategies based on risk estimation of unknown obstacles. When the risk value in the unknown environment is low, the drone adopts a global planning strategy. However, when encountering high-risk obstacles that move suddenly, the drone switches to a reactive obstacle avoidance strategy. Firstly, an online dynamic point cloud recognition method is employed to identify dynamic and static obstacles in unknown environments. Obstacle trajectories are then predicted based on historical positions, without the need for predefined motion models. A risk estimation function based on field theory is devised to assess the potential risk caused by static obstacles in unknown environments. To accommodate different obstacle threats, a gradient-based global path planning method is utilized to avoid conventional static and dynamic obstacles, while a reactive avoidance strategy is promptly activated to avoid high-risk malicious obstacles that move suddenly. Extensive simulations and real flight tests validate the efficacy of the proposed approach. The reaction time from detecting the sudden movement of a static obstacle to planning a safe trajectory is less than 3 $$\textrm{ms}$$ .