CDMA admission control based on reinforcement learning
Boris Makarevitch · 2003
The paper describes the application of reinforcement learning (RL) to CDMA admission control. Admission control based on load or power thresholds does not adapt to the environment and may lead to low resource utilisation or bad quality. By using RL near-optimal performance for various environments and thus a guarantee of high utilisation and good quality can be achieved. An algorithm, based on Sarsa learning with linear function approximation, as well as performance evaluation results are presented.