An action pool architecture for multi-tasking service robots with interdependent resources

We present a novel control architecture for execution of multiple tasks simultaneously. Traditionally, the task control architecture of a service robot has been focused on executing one task at a time. This has mainly been due to the history of mobility being the only function and tasks being so simple that just one or two abstract commands have been sufficient. During the last decade, these requirements have changed drastically and a more versatile operation is clearly required. Service robots commonly have physical subsystems such as manipulators and directable perception sensors that are not in use all the time. These task execution resources are dependent on the pose of the platform to which they are attached. Our architecture manages the concurrent use of these resources in order to achieve parallel task execution. We verify the architecture with experiments on two different robot platforms while performing concurrent tasks such as finding object, greeting people and taking pictures of a designated object. The proposed architecture can receive an arbitrary number of tasks to be executed at any time instant within limitations of computing resources.