Learning to Focus: CSI-Free Hierarchical MARL for Reconfigurable Reflectors

This project represents a niche application of Hierarchical Multi-Agent Reinforcement Learning (HMARL) to solve a specific physical layer problem in 6G/mmWave communications: the optimization of Reconfigurable Intelligent Surfaces (RIS). The core innovation is the 'CSI-free' approach, which bypasses the massive computational overhead of channel estimation. While the project has 5 forks in just 4 days—indicating strong immediate interest from the academic/research community—the defensibility is currently low (3) because it is a reference implementation of a paper rather than a production-ready tool or library. The 'moat' lies in the domain-specific reward shaping and hierarchical architecture, which is difficult to replicate without deep wireless engineering knowledge. Frontier labs (OpenAI/Anthropic) have zero interest in PHY-layer beamforming, but platform risk is 'medium' as major telecommunications equipment vendors (Huawei, Ericsson, Qualcomm) are the likely entities to either adopt or supersede this with proprietary implementations. The displacement horizon is long (3+ years) due to the slow cycle of wireless standardization and the nascent state of RIS hardware deployment.