Ranked Communication Channel Confidence for Multi-Agent Reinforcement Learning