A motion-planning system for a domestic service robot

Service robots are intended to help humans in non-industrial environments such as houses or offices. To accomplish their goal, service robots must have several skills such as object recognition and manipulation, face detection and recognition, speech recognition and synthesis, task planning and, one of the most important, navigation in dynamic environments. This paper describes a fully implemented motion-planning system which comprehends from motion and path planning algorithms to spatial representation and behavior-based active navigation. The proposed system is implemented in Justina, a domestic service robot whose design is based on the ViRBot, an architecture to operate virtual and real robots that encompasses serveral layers of abstraction, from low-level control to symbolic planning. We evaluated our proposal both in simulated and real environments and compared it to classical implementations. For the tests, we used maps obtained from real environments (the Biorobotics Laboratory and the Robocup@Home arena) and maps generated from obstacles with random positions and shapes. Several parameters were used for comparison: the total traveled distance, the number of collisions, the number of reached goal points and the average execution speed. Our proposal performed significantly better both in real and simulated tests. Finally, we show our results in the context of the RoboCup@Home competition, where the system was successfully tested.