Speech recognition based building semantic map method on Aldebaran Nao

Shallow domain knowledge, such as labels attached to objects and places is necessary for task planning of mobile robots. In order to save energy and protect environment for robots, a fast and effective semantic building map method based on sonars, vision and audio is proposed. Firstly, humanoid robot Nao builds the border map of an unknown environment by its sonars. And then Nao follows a red ball that controlled by a person to close to an obstacle in the environment. After that, Nao builds the border of the obstacle by its sonars autonomously. When finishing an obstacle border, Nao will get the name of obstacle by speech recognition. In this way, Nao builds the border map of all obstacles one by one until the map of the unknown environment is built. As illustrated in experimental results, the proposed algorithm is very fast and effective, and easy to implement.