Abstract:With the rapid growth of the network applications, efficient routing technique in wireless networks appears as a key challenge. This is because traditional routing protocols do not learn from historical network features and performance to make optimal routing decisions. To this end, we propose the deep learning empowered QoS-aware adaptive (DLQA) routing algorithm based on the convolutional neural networks (CNN). In the proposed DLQA routing algorithm, we use the CNN to judge the network performance and guide the routing decisions. In addition, we include the network traffic features and link state characteristics in the network feature matrix, which makes the CNN’s evaluation of network performance more accurate. And, we take a variety of network characteristics into account , such as end-to-end delivery delay, packet loss rate, etc., when evaluating the network performance, in order to support the multiple QoS requirements of the transmission. The simulation results demonstrate that the proposed DLQA routing algorithm can adaptively adjust the routing strategy according to the varying network features to meet the QoS requirements of the transmission with low end-to-end delivery delay and packet loss rate.

Deep Learning Empowered QoS-aware Adaptive Routing Algorithm in Wireless Networks