Planning for Multiple Robots in a Smart Environment

We consider configuration planning in a real assistive system that consists of multiple robots embedded in a smart environment. Our configuration plans are fine-grained action plans which specify the causal, temporal, resource and information dependencies between individual sensing, computation, and actuation components. Components reside on-board the robots, like cameras or manipulation components, or they may be distributed in the environment, like presence sensors or elevators. Our configuration planner exhibits closed-loop performance, thanks to the fact that plans are represented as constraint networks, which are shared between the planner and the executor and updated during execution. This mechanism allows our system to react to contingencies at execution time. We illustrate our system on both simulated and real experiments performed in the context of a large EU project on robotics for elderly assistance.