Subproblem allocation in distributed production systems

Partitioning and allocation of subproblems to processors is an important issue in parallel and distributed Artificial Intelligence, Besides its NP Complete nature, possible interaction between subproblems in these domains requires special consideration. In distributed production systems for monitoring applications, distribution of rules itself becomes problem distribution, The present study proposes a heuristic for allocating subsets of knowledge to agents based on the sizes of subsets, inter-subset coupling, capacities of agents and the inter-agent distances. Il is assumed that subproblems are available according to the capacities of agents, An initial estimate of the subset coupling can be had from the partitioning process itself and can be revised dynamically based on the run time statistics.