A research group led by Prof. Hirofumi Tanaka and Prof. Hakaru Tamukoh at the Research Center for Neuromorphic AI Hardware, Kyushu Institute of Technology (Director: Hirofumi Tanaka), in collaboration with former Prof. Takuji Ogawa at Osaka University and Prof. Gimzewski at the University of California, Los Angeles (UCLA) has applied a random network of single-walled carbon nanotube (SWNT)/porphyrin (Por)-polyoxometalate (POM) composite as a reservoir computation (RC) device, a kind of artificial intelligence device.
They have succeeded in classifying the grasping object by using the tactile signals from the hand part of the robot arm (Fig. 1). This is one of the world's first examples of robot control using an "in-materio" reservoir device, which allows nanomaterials to carry out a computation, and is expected to be used in the future for energy-saving AI systems and situational awareness functions for autonomous robots working in the home. The results of this research were published online in the German scientific journal Advanced Intelligent Systems on 4/Jan/2022.
Research Points
- Success in developing a material-based physical (in-materio) reservoir arithmetic element.
- The first application of in-materio reservoir computation to the recognition of grasping objects.
- The developed reservoir computing device is expected to be applied to complex AI problems such as time series prediction and voice recognition in the near future.
Professor Hirofumi Tanaka's Comment
Prof. Tanaka, Director of Research Center for Neuromorphic AI Hardware (Neumorph Center), Kyushu Institute of Technology (Kyutech), JAPAN.
Neumorph Center at Kyutech is a research center where researchers from a wide range of fields, including four research divisions, materials, electrical and electronic circuits, numerical modeling, and AI robotics, working together to develop next-generation AI systems. In order to reduce the power consumption of present AI systems, it is necessary for a paradigm-shift of software to hardware, and also to shift from digital drive to analog drive since there is a limit to the power consumption reduction of hardware.
In this process, AI device fabrication based on materials engineering plays an important role. We are focusing on the fact that the chemical reactions and dynamic properties of materials themselves can be used for AI processing, and are developing AI devices that work with completely new materials and principles. Neuromorphic AI hardware research is a budding field, and the Neumorph Center is the first research organization in Japan in this field. With the success of the application of the In-materio reservoir device to AI robotics, we will continue to work vigorously to spread this field, which has a small population of researchers all over the world.