Cognitive UAV resource management allowing task-based mission execution under data link limitations

Successful UAV mission performance with limited or completely lost data link requires highly autonomous capabilities on-board the vehicle. These situations can occur either as a result of a planned action or be unpredictable. We present an approach using the capabilities of cognitive automation to manage on-board processing and communication resources for the sake of more flexible mission planning and execution. Therefore, static expert knowledge as well as dynamic system- and environment information are interpreted to create a plan for the deployment of the resources, which can be executed during the course of the UAV mission. The resources we consider in the first step are onboard functions for route planning and optimization as well as on-board functions for mass payload data storage, retrieval and transmission. The software concept is built as a hierarchical control system. The top-level node consists of a cognitive architecture, which enables knowledge based mission management by inference-based interpretation of the situation, goal-driven decision-making, search-based planning and pattern-based task execution. The underlying automation child-nodes such as data link, route-planners and the flight control system are realised conventionally and serve as subsystems to the toplevel node, able to handle given tasks. These subsystems' capabilities are explicitly represented as knowledge in the top-level node and used to generate solutions to fulfil the given UAV task. The first platform on which this technology is integrated is a fixed wing motor-glider research UAV technology demonstrator. The approach was evaluated in real flight. A human operator in the ground control station gave a route reconnaissance task to the UAV. The mission management node was able to sense and react to a data link loss. The mission was replanned, according to the current environmental situation. New subsystem-tasks like data recording over the target as well as data replay and transmission in communication range are derived and then executed to accomplish the given reconnaissance task.