JST ASPIRE : AI-Physical Systems
Project Overview
This project (2025.12 - 2029.03) seeks to bring together researchers from Osaka Metropolitan Unviersity, University of Toronto, and MIT in control theory, formal methods, AI, game theory, and robotics to address one of the emerging challenges in autonomous systems: Ensuring safety while maintaining flexibility in AI-controlled multi-agent systems.
🎯 Mission
Develop a systematic framework for AI-Physical Multi-Agent Systems that can interpret ambiguous natural language commands and execute them safely in the physical world, with mathematically provable guarantees.
🔬 Innovation
Uniquely integrate Large Language Models, Supervisory Control Theory, and Game-Theoretic Optimization to bridge the gap between AI reasoning and physical system safety.
🌏 Impact
Contribute to the Society 5.0 vision through applications in smart agriculture, maritime safety, and beyond, while fostering international brain circulation and next-generation researcher development.
Research Thrusts
Thrust 1: Feedback Reprompting Design with Supervisory Control
Develop algorithms to verify safety of LLM-generated formal specifications and create feedback reprompting loops. When unsafe specifications are detected, the system generates safety violation analysis and/or counterexamples as prompts to guide LLMs toward safe solutions.
Thrust 2: AI-Physical Layer Feedback Replanning Design
Transform discrete action sequences into continuous control signals for multi-agent systems using game theory. When unexpected changes in the operating environment are detected in execution, the information is fedback for replanning or reprompting.
Thrust 3: Integration and Use Cases
Integrate feedback loops into a unified architecture and validate through real-world applications: smart agriculture (multi-robot farming) and maritime traffic control (autonomous vessel coordination).
News
January 09, 2026: ASPIRE Seminar
Gioele Zardini from MIT visit us and delivered a seminar. LinkedIn
December 16, 2025: MIT Visit
Two-day visit at MIT with Co-PI Chuchu Fan and her students/postdocs. LinkedIn
December 15, 2025: OMU Research News Featured
Our project has been highlighted in Osaka Metropolitan University's news. English | Japanese | LinkedIn
December 15, 2025: JST Press Release
Our project is listed on JST's official website. Japanese
December 01, 2025: Project Launch
Project officially starts!