Distributed coding in multiagent systems

This article describes using distributed coding of logical information in multiagent systems. First we identify failures of the agent platform in case of an unreliable underlying network. Next a simple, but inefficient way of distributing the agent state is given. Finally, we expose the way of solving failure problems by use of state snapshots and distributed coding. We develop a model of the network, based on the notion of erasure graph channel. We treat the properties of the erasure code encoding data efficiently with respect to a prescribed level of distortion. This work introduces information-theoretic ideas to multiagent middleware. It should enable applications of multiagent systems in environments where it is hard to establish a reliable communication infrastructure. The exposed work is a part of project Combined, funded by DECIS Lab, Delft.