ARCHIVES

In the era of 6G communication, optimizing wireless network performance through intelligent beamforming has become a key research focus. This work presents a reinforcement learning (RL)-based framework for beam direction optimization, aimed at enhancing channel capacity and minimizing interference in dynamic network environments. Users can interactively position transmitters and receivers in a 2D simulated environment, while RL algorithms Q-Learning, SARSA, Expected SARSA, and Double Q-Learning—are employed to determine the most efficient beam configurations. The system visually represents beam alignments and signal paths, offering realtime feedback on signal-to-noise-plus-interference ratio (SNIR) and overall network capacity. Experimental results demonstrate that Double Q-Learning achieves superior stability and performance, with improved convergence compared to other methods. The proposed system not only validates the potential of RL techniques in adaptive beam selection but also serves as an educational and analytical tool bridging AI concepts with wireless communication design