Recognition of surfaces based on haptic information using self-organizing maps

This paper proposes a new surface recognition method by a robot. Nowadays, demands for 'Real World Haptics' are increasing. Tactile sensation which is acquired by rubbing motion is important for haptics. Therefore, surface recognition based on haptic information is focused in this paper. First, the surface sensing is needed to collect feature values. There are some researches which focus on sensing and analysis of surface tactile information by using sensors with robots. However, there are some disadvantages of using sensors such as signal noise. From that reason, this paper proposes getting surface haptic information without sensors at the tip of the robot. Second, a pattern recognition method needs to be applied for surface recognition. There are some pattern recognition methods. Self-organizing maps (SOM) is one of the solutions. SOM is able to summarize high dimensional data to low dimension with preserving the topological properties of data. SOM is suitable for multi-class recognition. From these reasons, this paper proposes the surface recognition based on haptic information using SOM. Multi class surface recognition is achieved by the proposed method. The validity of the proposed method was confirmed through 7 surface recognition experiments.