Scope & Objectives

Agents, devices and information sources connected in large scale networks have to share information in effective ways, so as the right information to reach the right agents at the appropriate time, for agents to integrate and interpret data to perform the necessary tasks.

The distribution, diversity, volatility of data, and, in many emerging applications ubiquity of information sources, make the information sharing task a challenging task. This is important in many real-world settings, where voluminous information from different sources need to reach distant agents.

The problem becomes even more challenging when agents have different “views” for the meaning of the information they share, when they have to manipulate heterogeneous data from different sources, or when they have to jointly control actuators for which they do not share a common representation. Also, sometimes, information by multiple sources needs to be pre-processed, before being propagated to the right agents: The later may need specific information to be, for instance, extracted, implied, abstracted, or somehow aggregated, in different ways.

In all the above cases, semantics play an important role.

Considering to be a decentralized control problem, information searching and sharing in large-scale systems of cooperative agents is a hard problem in the general case: The computation of an optimal policy, when each agent possesses an approximate partial view of the state of the environment and when agents’ observations and activities are interdependent (i.e. one agent’s actions affect the observations and the state of an other), is hard.

The above considerations, has resulted to efforts that either require agents to have a global view of the systems, to heuristics, to the pre-computation of agents’ information needs and information provision capabilities for proactive communication, to localized reasoning processes built on incoming information, to analytical frameworks for coordination whose optimal policies can be approximated for small (sub-) networks of associated agents, and to reinforcement learning algorithms for hierarchical peer-to-peer information retrieval systems. On the other hand, there is a lot of research on semantic peer-to-peer search networks and social networks many of which deal with tuning a network of peers for effective information searching.