Swarm intelligence algorithm for optimality discovery in distributed constraint optimization

The distributed constraint optimization problem (DCOP) is known as a basic problem of multiagent systems. This problem seeks optimal solutions using cooperative behavior between agents. However, because DCOP is NP-hard, the computational load becomes massive as the number of agents increase. What are needed are high-speed algorithms that can solve problems without increasing the computational load. In recent years, methods that intentionally limit the number of cooperative agents, such as k-optimality and p-optimality, have been proposed as approximate algorithms. They are drawing attention as new approximate algorithms because they can obtain deviations from the optimal solutions. The value of optimality is an artificially established partition of state space, and there is a need to change the value for different network topologies. In this paper, we propose a method that uses swarm intelligence as a new DCOP algorithm. Each agent evaluates the state space using swarm intelligence. The available state is determined not based on values decided by each agent. Instead, a cooperative space is produced based on heuristic values to determine the optimal solutions. We also show that a new criteria of optimality can be obtained using swarm intelligence.