I Would If I Could: Reasoning about Dynamics of Actions in Multi-Agent Systems
A new logic framework models how agents grant and revoke actions in real-time.
Researchers Rustam Galimullin, Hermine Grosinger, and Munyque Mittelmann introduced ATL-D (Alternating-time Temporal Logic with Dynamic Actions) to address a gap in multi-agent system (MAS) reasoning. Standard logics like ATL model state- or history-dependent behavior but rarely handle dynamic action availability or agents' knowledge of required actions. ATL-D models the process of granting and revoking actions, while its extension ATEL-D captures how such updates affect agents' knowledge, enabling more realistic adaptive behavior in autonomous agents.
The paper, set to appear in KR 2026, provides several technical results: expressivity analysis relative to ATL, connections to normative systems, and complexity results for computational problems. This work advances MAS reasoning by allowing agents to adapt dynamically, crucial for applications like autonomous driving, robotics, and distributed AI systems where action availability changes in real-time.
- ATL-D models granting and revoking actions dynamically, unlike standard ATL.
- ATEL-D extension captures how action updates affect agents' knowledge.
- Paper includes expressivity analysis, links to normative systems, and complexity results.
Why It Matters
Enables more adaptive autonomous agents in real-world multi-agent systems like robotics and distributed AI.