Convolutional neural network-based channel allocation and resource optimization for D2D communications

In mobile communication systems, D2D techniques can significantly improve the throughput of the system. To overcome the problem that the sharing of spectrum resources in cellular networks can cause link interference between cellular and D2D users resulting in difficult resource allocation, this paper constructs a convolutional neural network (CNN) to approximate the traditional optimization method, which takes into account various constraints such as optimal system performance, minimizing interference and satisfying the minimum quality of service for users, satisfying the upper and lower power requirements as well as the demand variability. Experimental results show that the proposed adaptive approach is suitable for complex multi-cell scenarios, with a practical efficiency of around 1.5 times the original efficiency and low computational complexity.