AIF: An Artificial Intelligence Framework for Smart Wireless Network Management

To solve the policy optimizing problem in many scenarios of smart wireless network management using a single universal algorithm, this letter proposes a universal learning framework, which is called AI framework based on deep reinforcement learning (DRL). This framework can also solve the problem that the state is painful to design in traditional RL. This AI framework adopts convolutional neural network and recurrent neural network to model the potential spatial features (i.e., location information) and sequential features from the raw wireless signal automatically. These features can be taken as the state definition of DRL. Meanwhile, this framework is suitable for many scenarios, such as resource management and access control due to DRL. The mean value of throughput, the standard deviation of throughput, and handover counts are used to evaluate its performance on the mobility management problem in the wireless local area network on a practical testbed. The results show that the framework gets significant improvements and learns intuitive features automatically.