关键词: 动态行为树, 未知环境, 自适应决策, 改进Tangent Bug

Abstract: To address the challenges of environmental dynamics and complex real-time decision-making encountered by Unmanned Aerial Vehicles (UAVs) during autonomous exploration in unknown, GPS-denied environments, this paper proposes an autonomous exploration algorithm based on Dynamic Behavior Trees (DBT). The front-end of this algorithm constructs a lightweight time-varying voxel probability map model and integrates an improved Tangent Bug real-time path planning algorithm guided by the lightweight map, achieving safe and efficient navigation through optimized environmental perception. The core of the back-end employs an adaptive decision-making mechanism based on DBT, innovatively introducing dynamic weight-driven topological reconfiguration. It utilizes a hierarchical design and modular management to endow the system with algorithm scheduling capabilities and the intelligent decision-making ability to proactively respond to environmental changes. Simulation experiments conducted on the ROS2/Gazebo platform demonstrate that the DBT approach exhibits superior performance in perception, decision-making, and responsiveness. Compared with the finite state machine method, the exploration coverage of DBT algorithm increased by 2.33%~12%, and the average exploration coverage increased by 1%~13%, significantly enhancing the UAV's efficiency, robustness, and intelligence level in autonomous exploration within unknown environments.

Key words: dynamic behavior tree, unknown environment, adaptive decision making, improved tangent bug