Knowledge Configured Vehicle - A layered artificial cognition based approach to decoupling high-level UAV mission tasking from vehicle implementation properties
Michael Kriegel, Stefan Brüggenwirth, Axel Schulte · 2011
This work proposes the concept of a so-called Knowledge Configured Vehicle (KCV). In our current UAV research we investigate approaches, which enable vehicles to perform missions on a higher level of autonomy by AI-based onboard decision functions working in cooperation with a human operator. In order to achieve platform independence of such UAV mission management functions and at the same time providing full access to low-level system functions, we developed and flight tested a prototype following the design concept of such a Knowledge Configured Vehicle. A KCV is characterized by the separation of high level mission capabilities related knowledge from the physical vehicle implementation related knowledge. Therefore, we use our Cognitive System Architecture (COSA). COSA generates system behaviors on the basis of explicit knowledge models. In a KCV a mission-specific, but vehicle independent knowledge package is combined with a vehicle specific, but mission independent knowledge package, each of which interchangeable as according to the vehicle implementation and the mission needs. A field prototype of a KCV based upon our small helicopter UAV has been developed and demonstrated in a real-world flight trial performing a simple sensor deployment mission. As a result, the flight with the cognitive agent running on an on-board mission computer provides a proof of concept and shows the maturity of the implemented prototype.