CCKS: Consensus-based Communication and Knowledge Sharing

2026-06-10 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

The Consensus-based Communication and Knowledge Sharing (CCKS) framework addresses challenges in action-advising-based knowledge sharing for cooperative Multi-Agent Reinforcement Learning (MARL) within Decentralized Training and Decentralized Execution (DTDE) systems. Current approaches often lead to excessive advising and suboptimal performance due to a lack of teacher-student compatibility. CCKS enables agents to smartly adopt recommendations based on consensus-derived constraints, balancing exploration with learning from experienced teachers. The framework constructs consensus models using contrastive learning based on local observations during the training phase. Agents then score and choose actions considering this consensus and shared knowledge. Designed as a plug-and-play solution, CCKS integrates with existing DTDE algorithms. Experiments in Google Research Football and StarCraft II Multi-Agent Challenge environments demonstrate significant improvements in cooperation efficiency, learning speed, and overall performance compared to current DTDE baselines. The code is available at https://github.com/yuanxpy/CCKS, published on 2026-06-10.

Key takeaway

For Multi-Agent Reinforcement Learning engineers developing cooperative DTDE systems, integrating the CCKS framework can significantly enhance agent cooperation and learning efficiency. By enabling agents to intelligently filter teacher advice through consensus, you can overcome issues of excessive advising and suboptimal stability. Consider adopting CCKS as a plug-and-play solution to improve performance in complex multi-agent environments like StarCraft II, leveraging its proven benefits.

Key insights

CCKS improves MARL cooperation by enabling agents to smartly adopt advice based on consensus, balancing exploration and learning.

Principles

• Consensus-based advice improves teacher-student compatibility.
• Balance exploration with learning from experienced teachers.
• Contrastive learning can build consensus models from local observations.

Method

CCKS constructs consensus models via contrastive learning on local observations during training. Agents then score and select actions based on this consensus and shared knowledge.

In practice

• Integrate CCKS into existing DTDE algorithms.
• Apply in Google Research Football environments.
• Use for complex StarCraft II MARL challenges.

Topics

• Multi-Agent Reinforcement Learning
• Decentralized Training
• Knowledge Sharing
• Consensus Models
• Contrastive Learning
• Action Advising

Code references

• yuanxpy/CCKS

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Robotics Engineer

Related on AIssential

• KD-MARL: Resource-Aware Knowledge Distillation in Multi-Agent Reinforcement Learning — KD-MARL distills expert coordination into lightweight, critic-free agents for efficient MARL deployment in resource-constrained settings.
• Interview with Xinwei Song: strategic interactions in networked multi-agent systems — Strategic interactions in networked multi-agent systems can be managed through mechanism design and reputation-aware reinforcement learning.
• Learning to Communicate Locally for Large-Scale Multi-Agent Pathfinding — LC-MAPF uses learnable local communication to enhance multi-agent pathfinding coordination and scalability.
• LLM-Guided Communication for Cooperative Multi-Agent Reinforcement Learning — LLMs can iteratively refine communication protocols in MARL to improve agents' shared state awareness.
• Bridging MARL to SARL: An Order-Independent Multi-Agent Transformer via Latent Consensus — CMAT uses a latent consensus mechanism to enable order-independent, simultaneous multi-agent actions within a SARL framework.
• Decentralized Diffusion Policy Learning for Enhanced Exploration in Cooperative Multi-agent Reinforcement Learning — Diffusion policies enable multi-modal exploration in MARL, outperforming restrictive Gaussian policies.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.