An Integrated Control Framework for Long-Term Autonomy in Mobile Service Robots

This paper describes an integrated framework for the long-term task-driven control of mobile service robots. The core components of the framework are: a high-level task executor that manages execution, for example by reacting to failures, or adding extra tasks required by the end-user on-the-fly; a task scheduler that schedules sets of tasks throughout the day, taking into account travel times between locations and task durations, while satisfying the time constraints associated with each task; and a probabilistic topological motion planner that provides time-dependent optimal navigation policies and expected navigation times between task locations. We illustrate the overall framework by reporting on a three-week deployment in a real-world office environment, and use the data collected during the deployment to validate and illustrate the capabilities of the framework to adapt itself to the different travel time expectations throughout the day.