Context aware architecture for distributed robotics

Multi-robot applications in industrial scenarios have to cover a large spectrum of tasks reaching from low level control functions on the machine level up to high level coordination tasks. The configuration of these tasks related to single or multi-robot algorithms depends on the individual requirements of the scenario (i.e. number of entities, communication bandwidth, failure assumptions). Therefore, the developer may decide to run a path planning algorithm individually for each robot to avoid an additional exchange of information. For a scenario with a large communication bandwidth, a planner in the global context would be more suitable in order to receive an optimal result. In this paper we present an architecture that implements this decision online. Our context-aware task controller evaluates the communication situation and triggers a context switch (global, local, individual) of the task execution. By shifting the decision from design-time to run-time, we close the gap between supervision and swarm approaches and provide an optimized behavior.