Dynamic Replanning of Multi-drone Missions using Dynamic Forward Slicing

Unmanned Aerial Systems (UAS, typically known as drones) are useful in many application domains, such as logistics and precision farming, especially when they fly Beyond Visual Line of Sight (BVLOS). To effectively use multiple UAS for BVLOS missions, it is required to precisely plan the flight of each UAS involved in the missions, allocating the required airspace ahead of time. The dots Domain-Specific Language (DSL) can be used to plan and execute such missions but does not support the adaptation of missions when unexpected changes occur during the execution of these missions. Such dynamic replanning of missions to changing conditions is a crucial feature for the safe integration of UAS into the airspace since it allows the UAS to adapt to these changes, such as yielding to emergency response aircraft. In this work, we propose a program transformation technique inspired by program slicing to dynamically replan ongoing dots-specified multi-UAS missions using the same planner built for the dots language. Whenever dynamic replanning is needed, we compute a forward slice of the dots program based on the current runtime state of the mission. This generates a new program that obeys the original mission specification but only specifies the mission from the point in time where the replanning is needed. We validate our proposed technique by integrating it into ros-dots, a service-oriented architecture for autonomous UAS operations.