Explainable Rational Synthesis in Multi-Agent Systems: A Compositional Approach

Abstract

We are currently developing novel algorithms for explainable rational synthesis in multi-agent systems (MAS) with Linear Temporal Logic (LTL) objectives. Our work addresses the challenge of computing Nash equilibrium strategies in concurrent graph games, a problem known for its double-exponential complexity. I have created a compositional approach that enhances the performance and explainability of rational synthesis by converting the game into a suspect game, solving it as a parity game, and optimizing strategies for multi-agent coordination. Additionally, my research includes the practical implementation of a tool that significantly outperforms state-of-the-art LTL synthesis methods in various case studies, with a focus on improving system performance and understandability in applications such as autonomous systems and robotics.

ArshiA Akhavan

part time Software Engineer, part time Computer Science enthusiast

My research interests include computer systems, parallel computing and distributed systems, programming languages and verification, high performance computing, operating systems, computer architecture, and software engineering.