Output details

Graphical models for interactive POMDPs: representations and solutions

<22> This paper presents a probabilistic graphical model based framework, called interactive dynamic influence diagram (I-DID), to represent sequential multiagent decision making under uncertainty. It solves the problem from the perspective of individual agents thereby overcoming drawbacks of traditional game solutions (Nash equilibrium). The methodology is applicable to both cooperative and competitive multiagent settings, which significantly extends previous work on multiagent planning. In addition, the introduction of graphical models provides a more scalable technology, and seamlessly integrates research in uncertainty artificial intelligence and agent communities.