The SelfFlow project addresses the flow of self-knowledge between intelligent autonomous agents in teaming contexts.
The project does not target robot teamwork or collaboration in general, but systems collaboration is the proper context for the project. The specific focus theme is the transfer of self-knowledge to be used for cooperative work of heterogeneous robot teams. Self-knowledge representation, query and transfer are the central issues.
An UGV is searching and collecting tennis balls in a tennis court (or other more cluttered environment to make the task harder). It has difficulties finding the balls due to limited vision capability (e.g. it only has short-ranged ultrasound or close cameras).
An UAV enters the scene. UAV and UGV interchanges info on tasks and capabilities. The UGV learns about the capability of imaging of the drone. It asks for collaboration in searching for balls. The UAV agrees and collaborates in the task. In case of the drone not having enough computing power to process the images, a third, static computer, may help in the search of balls by the processing drone images.
The work is to devise a representation/engine/protocol to store, exploit and interchange self-knowledg (self-models) for robots to cooperate.
- Investigate literature on heterogeneous robot teaming
- Investigate literature on capability representation
- Investigate literature on robot sel-awareness
- Investigate literature on robot mutual knowledge
- Define the self--knowledge that is of relevance
- Do SK interchange in planning-time/real-time
Robot cooperation. Agent cooperation.
