In the forthcoming 6G era, network infrastructure must achieve unprecedented energy efficiency, adaptability, and ultralow latency. This paper proposes a novel Deep Q‐Network (DQN)‐based routing framework that leverages network‐controlled repeaters (NCRs) to optimize routing paths dynamically. Unlike prior studies that remain largely simulation‐only, we develop a realistic emulation environment in GNS3, integrated with a custom reinforcement‐learning testbed compatible with OpenAI Gym. The DQN agent learns to activate optimal NCR configurations based on traffic and channel conditions, thereby minimizing unnecessary energy expenditure. Experimental results show that the proposed approach achieves up to 34% energy savings, 21% faster convergence, and a 7%–10% improvement in packet delivery ratio (PDR) compared to baseline and classical energy‐aware protocols. Scalability experiments confirm consistent performance in networks of up to 50 nodes, while preliminary vehicular mobility tests using SUMO demonstrate robustness under dynamic conditions. By bridging reinforcement learning (RL) with emulation‐driven evaluation, this work offers a scalable, efficient, and deployment‐oriented framework for designing future 6G networks. The simultaneous improvement in both metrics is enabled by the dual‐benefit nature of NCR activation, which reduces per‐link energy cost while shortening route convergence time, coordinated through a multiobjective reward function with carefully tuned weighting coefficients.
Mourin et al. (Thu,) studied this question.