Unlocking Efficient Connectivity:Advanced AI-Driven Planning for Multi-Hop IAB Networks

In the evolution toward 6G and Open RAN architectures, cost-effective and high-performance network deployment has become increasingly important. Multi-hop integrated access and backhaul (IAB) networks that leverage millimeter-wave technology can reduce infrastructure costs by lowering fiber dependency, but optimal node deployment remains challenging in diverse urban scenarios and involves a large discrete action space. This paper proposes a framework that combines transfer learning with deep reinforcement learning (DRL) to address the node deployment problem under varying donor configurations. Extensive simulations show that the proposed approach reduces the required number of nodes by approximately 20% and cuts model training time by up to 50%, improving adaptability across different donor layouts. These results indicate that the proposed transfer-enabled DRL framework can streamline deployment, optimize coverage, and improve overall efficiency in dynamic settings, supporting scalable and sustainable next-generation wireless networks.