A Neural Network Based Handover for Multi-RAT Heterogeneous Networks with Learning Agent

International audienceThe wireless communication networks continue to evolve with multiple radio access technologies (LTE, WIFI, WIMAX..) included in the same device. Thus, the end-nodes have the flexibility to decide which radio technology to select according to their availability and location. In this case, they must be made intelligent and autonomous enough in order to select automatically the best available wireless technology. Neural Networks represent a promising machine learning technique to be used to make such decisions. The proposed system is implemented on a System on Chip (SoC) integrating both processors and FPGA logic fabrics on the same chip with fast interconnection between them. This allows designing Software/Hardware systems with optimized performance. The adopted SoC offers the partial reconfiguration (PR) feature on the FPGA, which adds more flexibility to these high performance devices. In this paper, we propose a Neural Network based system for heterogeneous networks. The system is based on reinforcement learning that aims to make the system intelligent enough when deciding to perform vertical handover between multiple wireless systems. The switching between the wireless systems standards is implemented using PR technique. © 2018 IEEE