Developing Robot Drumming Skill with Listening-Playing Loop.

Reacting according to external sounds is an important ability in multi-robot and human-robot collaboration. Although network might be the first choice to connect multi-agents in the robot world, the unexpected connection snap would be a disaster to the whole system. Utilizing sounds is a feasible and supplementary way to transfer information between agents, which is also a smart and robust way to support swarm intelligence. In this paper, under the scenario of a robot band, the issue how each robot member achieves its performance ability is focused. Unlike most of the previous researches, we emphasize that robot's performance ability is achieved all by itself in an autonomous way. And an approach of Listening-Playing Loop (LPL) is proposed, where the developmental learning is involved. With a simple drumming robot, the proposed approach is evaluated. Experimental results show the proposed approach is effective, and via transferring raw audio data to the motion control, the robot successfully develops the drumming ability.