Combining Probabilistic Contexts in Multi-Agent Systems

We propose an approach for modelling, integrating, and querying distributed probabilistic contexts in multi-agent systems. We assume each agent to be equipped with an independently acquired uncertain context. By taking advantage of established database technologies, we represent the uncertain context of each agent as a set of probabilistic facts conveniently stored in a probabilistic database. Members of a multi-agent system act autonomously and interact with each other. The interaction between agents consists of sharing access to their contexts with each other and allowing queries over the combined shared contexts. This amounts to the challenge of combining and querying distributed probabilistic databases. To combine probabilistic contexts, we define a context-matching operator that creates a joint probability distribution with given marginal probabilities. Furthermore, we propose a query answering method over combinations of probabilistic contexts.