Output details
11 - Computer Science and Informatics
University of Edinburgh
Scaling Up Multiagent Planning: A Best-Response Approach
<22> Originality: This is the first work to allow efficient, scalable, equilibrium computation in planning games, and has broken new ground in terms of applying existing state-of-the art planning algorithms to strategic decision-making problems.
Significance: The paper introduced game-theoretic ideas into heuristic single-agent planning, enabling a different algorithmic methodology from previous approaches. We later used this algorithm to plan joint travel sharing routes using the entire real-world UK public transportation system.
Rigour: The paper was published at ICAPS 2011, the highest-rated international conference in the field. Our algorithm significantly outperformed the best existing algorithms on a large set of benchmark problems.