This technology is a tactile sensing system that uses thermoelectric elements and temperature sensors attached to a robot gripper's fingertips to acquire time-series data on thermal conductivity changes during object contact, which is then processed by a 1D-CNN deep learning model to identify and classify objects.
Existing robot recognition systems based on pressure or force sensors often struggle with limited classification accuracy, as they fail to provide sufficient information regarding the unique physical properties of an object's texture or material.
This technology proposes a method where the thermoelectric element heats the fingertip above room temperature before contact; the temperature sensor then measures the temperature changes caused by the object's thermal conductivity, and the deep learning model classifies the data. This allows for precise object recognition that incorporates material properties. It can be applied to logistics sorting, recycling, and service robot object handling, providing a new means of perception that can distinguish objects that are difficult to identify using visual information alone.
This invention was developed with support from the Ministry of Science and ICT for the development of electro-hydraulic actuator-based soft robot modules.
N/A
