Calculating Optimal Decision Using Meta-level Agents for Multi-Agents in Networks

In spatial graphs with a vast number of nodes, it is difficult to compute a solution to graph optimisation problems. We propose using metalevel agents for multi-agents in a network to calculate an optimal decision. The network contains nodes and arcs wherein the agents are information carriers between the nodes and, since there is one agent per arc, the agents are statically located. These agents, operating at a ground level, communicate with a comprehensive agent, operating at a meta-level. The agents at the meta-level hold information computed by the ground-level agents, but also include groundlevel agents' special conditions. As an example, we apply the work to the travelling salesman problem and use a map, with cities and roads, constituting the network where the information about the roads is carried in the meta-level agents. For multi-agents in maps, we use parallel computing.