A brain-inspired AI Hardware acquiring memory of individual episodes obtained from bodies and its hardware implementation

2022 Fiscal Year Final Research Report

• Project/Area Number: 21K21318
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 1002:Human informatics, applied informatics and related fields
• Research Institution: Kyushu Institute of Technology
• Principal Investigator: Yuichiro Tanaka 九州工業大学, ニューロモルフィック AI ハードウェア研究センター, 助教 (70911288)
• Project Period (FY): 2021-08-30 – 2023-03-31
• Keywords: 脳型人工知能 / 身体性 / ロボット / ハードウェア

Outline of Final Research Achievements

This study developed an artificial intelligence (AI) model that acquires memory of individual experiences of robot bodies and its hardware. The AI model is based on reservoir computing so that the computational cost for training is low and therefore, the training can be done with a small amount of robot experience. For example, a robot can memorize tactile information when grasping some object and judge whether the robot has grasped the object or not referring to its memory. Additionally, this study implemented the AI model on a reservoir computing device by nanomaterial toward low-power implementation.

Free Research Field

ソフトコンピューティング

Academic Significance and Societal Importance of the Research Achievements

現行主流の統計的AIは大量のデータによる学習に基づいているため，少量の経験に基づいた記憶獲得の実現は困難である．本研究が開発に取り組んだ少量の経験に基づくAIは，現行AIが対応していない機能を備える新しいタイプのAIである．このAIは例えば，ホームロボットがその固有の環境で得られる経験の獲得などに活用できると考えられ，本研究はホームロボットの社会実装に貢献するものと考える．