A Modular and Scalable Framework for Autonomous Actuation and Emergency Handling with Behavior Trees for Unmanned Aerial Vehicles

Abstract Unmanned Aerial Vehicles (UAVs) require reliable and adaptable autonomy frameworks to perform critical tasks such as infrastructure inspection, where safety, mission adaptability, and real-time responsiveness are of prime importance. This paper presents a modular framework for autonomous inspection missions using Behavior Trees (BTs) which can be applied to heterogeneous UAVs. This framework builds on prior efforts to integrate customized commercial UAVs within an open-source system, ensuring adaptability across various mission requirements. By leveraging BTs within the Robotic Operating System (ROS) framework, the architecture emphasizes modularity, scalability and real-time adaptability, overcoming the rigidity of traditional state machine approaches. A clear separation between decision-making and actuation is maintained, with BTs dedicated to execute well-defined tasks and responding to emergencies, whereas high-level decision logic is managed externally. The system supports essential mission tasks while incorporating robust emergency detection and response mechanisms to enhance operational safety. We validated this system in both simulated and real-world scenarios, demonstrating improved mission adaptability and safety across various inspection applications.