Learning rate control for downlink shared channel in WCDMA

The paper considers bit rate control algorithms for the downlink shared channel in WCDMA (wideband code division multiple access) network. The goal of the rate control is to maximise the data throughput while satisfying quality constraints for the voice users. A learning algorithm, based on Reinforcement learning, is proposed. The algorithm facilitates frame acknowledgments to make decisions on the rate change. The control channel power is used as the state variable and a probabilistic decision policy is applied. Performance of the algorithm is evaluated and compared with performance of a simple algorithm without learning. Data throughput and voice quality cost function are used as performance measures. The learning algorithm achieves higher throughput and lower voice quality cost than the reference algorithm.