2021 Fiscal Year Final Research Report

Development of Model Predictive Interactive Intelligence and Its Application to Autonomous Drive

Research Project

PDF

Project/Area Number	19H00763
Research Category	Grant-in-Aid for Scientific Research (A)
Allocation Type	Single-year Grants
Section	一般
Review Section	Medium-sized Section 21:Electrical and electronic engineering and related fields
Research Institution	Nagoya University
Principal Investigator	Suzuki Tatsuya 名古屋大学, 工学研究科, 教授 (50235967)
Co-Investigator(Kenkyū-buntansha)	山口拓真名古屋大学, 工学研究科, 助教 (30745964) 奥田裕之名古屋大学, 工学研究科, 准教授 (90456690)
Project Period (FY)	2019-04-01 – 2022-03-31
Keywords	モデル予測型知能 / パーソナルモビリティ / インタラクション / 他者モデル / 自動運転
Outline of Final Research Achievements	In this project, a model predictive intelligence was developed, which exploits the model of others, behavior prediction based on others' model and real-time optimization. The proposed control architecture was implemented on a small personal mobility. The others' model embedded in the proposed architecture played an important role to realize an interactive intelligence between the vehicle and others. In this project, a model of decision making was particularly focused on and a new cost function of decision entropy of others has been defined. Since the decision entropy is a measure of the uncertainty of the decision of others, a natural human-like interaction between vehicle and others has been achieved by minimizing the decision entropy of others. The usefulness of the proposed architecture has been demonstrated by implementing on a real personal mobility which has interaction with pedestrians.
Free Research Field	システム制御
Academic Significance and Societal Importance of the Research Achievements	本研究では、知能化機械に求められる高次の知能として、他者とのスムーズなインタラクションを行う知能の実現を目指した。そのため、他者モデルを明示的に組み込んだモデル予測型のアーキテクチャを提案し、他者の判断のエントロピーをできるだけ小さくするという制御スキームを提案した。この考え方はいわゆる「説明可能なAI」の一翼を担う考え方となる。提案した手法を小型のパーソナルモビリティに実装し、歩行者とのインタラクションを伴う検証実験によりその有用性を検証した。この成果はパーソナルモビリティにとどまらず、今後は一般道でのレベル４を目指す自動運転の制御アーキテクチャの基本的な考え方になると期待される。