A Deep Reinforcement Learning Approach for Spectrum Management in Tri-Band Wi-Fi Networks

The increasing reliance on high-speed, lowlatency wireless communication necessitates a shift toward intelligent network planning. Conventional strategies struggle to adapt dynamically to real-time variations in network traffic, interference, and environmental factors. This study introduces an artificial intelligence (AI) -driven optimization approach that integrates Deep Reinforcement Learning (DRL) with an Overset Grid simulation model to enhance Access Point (AP) placement and spectrum management in Tri-Band Wi-Fi networks. Simulation results, conducted within a two-dimensional $10 \times 10$ grid environment using a Deep Q-Network (DQN) agent, validate the potential of the proposed hybrid methodology. The AI-optimized model was benchmarked against a traditional heuristic-based planning method. Evaluation was based on key performance indicators, including network coverage, interference reduction, and Quality of Service (QoS), revealing substantial improvements across all metrics. Performance evaluation shows that coverage improved from $75.2 \%$ to $92.7 \%$, interference reduction increased from $12.5 \%$ to $34.8 \%$, and QoS metrics rose from $68.4 \%$ to $85.9 \%$. The findings a scalable, adaptive foundation for next-generation Wi-Fi deployments.