The Effective Zone Routing Protocol Design Using Deep Recurrent Neural Network for The Next Generation Wireless Network

Abstract

This study proposes the usage of LSTM-RNN to allow ZRP to adjust the value of zone radius to the environment by sizing each node’s routing zone based on network performance input metrics such as Routing Overhead, Energy Consumption, Throughput, and User Usage. Those input metrics were used as a dataset, and split into 500 as data training, and 100 as data testing to get the zone radius value as an output value in the simulation. The proposed algorithm was tested in two scenarios: a static node environment and a mobility node environment using MATLAB as a simulator. The bandwidth capacity used in this study is 300 Mbps, which meets the requirement of next-generation wireless networks (5G and beyond). Furthermore, the proposed algorithm’s (LSTM-RNN ZRP) results are compared to conventional ZRP in both scenarios. The range of zone radius for mobile node environments is wider than for static node environments, with a range of 2-6 for LSTM-RNN ZRP and 2-7 for conventional ZRP. Meanwhile, the range for mobile node environments is 1-7 for both LSTM-RNN ZRP and conventional ZRP. According to the relationship between input metrics and zone radius determination, the proposed algorithm is more effective when used in a static node environment. However, both algorithms are acceptable for application in a static and mobile node environment.