Reactive Mission and Contingency Management for UAS: Implementation and Flight Testing

Increasing the autonomy of unmanned aerial systems (UAS) is a key challenge to open up new use cases, especially for Beyond Visual Line-Of-Sight (BVLOS) applications. The mission and contingency management system onboard the UAS serves as the backbone for any automated flight, as it is responsible for mission planning and execution in normal and abnormal conditions. Current solutions for mission management delegate decision making to a remote pilot on the ground. However, increasing levels of autonomy require more onboard deliberation capabilities, especially for missions in dynamic environments. This paper demonstrates an onboard mission and contingency management framework for a UAS. In the fast-paced UAS research community, autonomy software stacks should be modular and extensible to allow for evolving capabilities. Open-source software can provide an effective foundation for continuous capability growth. Nevertheless, combining these components into a versatile and adaptable autonomy framework for use in a safety-critical application is a complex task that presents its own set of challenges. In this paper, we derive requirements from an exemplary maritime surveillance use case for an autonomous mission and contingency management system. We showcase our integration of behavior trees using the BehaviorTree.CPP library within a ROS2 skills-based framework for drones using the PX4 autopilot and show how the selected tools fulfil the needs of the use case. We deploy the software stack on a Raspberry-Pi Compute Module 4 companion computer with a Pixhawk 6X flight computer and test the system in flight on a Holybro S500 multicopter. The results of the flight tests demonstrate the real-world applicability of our framework to dynamic and reactive missions. In future work, we plan to integrate more features, for example trajectory planning and following, allowing the execution of more complex missions.