Active Heat Flow Sensing for Robust Material Identification

Thermal properties are significant for recognizing an object’s material but cannot be determined via visual and stiffness (or tactile)–based recognition techniques. Most studies have used temperature as a complementary part of multimodal sensing; however, the thermal signal is an unexplored capability that can be beneficial for recognizing target objects. Since changes in thermal responses can result from both material properties and initial temperature, realizing robust and high-accuracy recognition in different environments is a challenging issue. To tackle the issue, this paper proposes a novel strategy for material identification that can actively measure heat flow by heating and cooling a robot gripper, enabling the extraction of the thermal properties of contact materials regardless of the object’s initial temperature variation (referred to as “active heat flow sensing”). We use a robotic task as an example of one possible application of the proposed strategy. For this, we developed a gripper pad embedded in a temperature control system and heat flow sensor to monitor the thermal exchange during contact with a target object. The paper conducted some experiments divided into two scenarios. The first experimental results show that active heat flow sensing is realized within 0.4 sec from first contact for 100% classification of four heated materials. The second experimental results show that the three materials, whose thermal properties are largely different, can be classified within 0.7 sec from first contact using different initial temperatures of the training and test data. These results suggest robustness against environmental change, which has been difficult using conventional temperature-based methods.