An Architectural Approach for Enabling and Developing Cooperative Behaviour in Diverse Autonomous Robots

The paper introduces an architecture for robot-to-robot cooperation which takes into consideration how situational context augmented with peer modeling fosters cooperation opportunity identification and cooperation planning. The presented architecture allows developing, training, testing, and deploying dynamic cooperation solutions for diverse autonomous robots using ontology-based reasoning. The architecture operates in three different worlds: in the Real World with real robots, in a 3D Virtual World by emulating the real environments and robots, and in an abstract Block World that enables developing and studying large-scale cooperation scenarios. We describe an assessment practice for our architecture and cooperation procedures, which is based on scenarios implemented in all three worlds, and provide initial results of stress testing the cooperation procedures in the Block World. Moreover, as the core part of our architecture can operate in all the three worlds, development of the robot cooperation with the architecture can regularly accommodate insights gained from experimenting and testing in one world as improvements in another. We report our insights from developing the architecture and cooperation procedures as additional research outcomes.