Robust material discrimination by a soft anthropomorphic finger with tactile and thermal sense

Thermal sense plays an important role in object recognition. The idea to identify materials by heat flow from the skin to the touched materials have been introduced to artificial thermal sensors and proved effective. Moreover, the addition of tactile sensors is found to improve the performance of material identification by obtaining further properties of objects such as roughness and stiffness. However, the application of these sensors to autonomous recognition by a robot is not a well investigated topic. Contact with the objects will vary in such cases, but no previous sensor has shown the ability of material recognition robust to change in contact conditions with the required size and softness. The limitation is partly due to the design of current thermal sensors where the skin and the sensing devices are separated. This paper introduces a soft anthropomorphic finger with both thermal and tactile sensing elements embedded in soft polyurethane material. Experimental results show the ability to discriminate materials robust to variety in contact condition by virtue of the deformable tissue and compensation by the tactile sensors. Recognition robust to change of temperature is also realized by an additional thermal sensor.