Due to the enormous increase of users, Internet websites and online services, the next generation networks are becoming complex. Therefore, optimization of these complex networks is a major challenge these days. In this paper, we consider a network model which uses reinforcement learning to develop some significant features like user cell association, enhanced QoS, capacity and coverage leading to ultra-high data transfer rates. We present three possible Q-learning algorithms based solutions that determine the best factor intrinsic to the learning algorithms which result in utmost throughput of the network. Simulation results show that our developed algorithms assign the channel appropriately and work for spatial reuse, allocate the resources by extending the range of small cells and finally ensure user's satisfaction by maintaining the QoS.
展开▼