An DL-based approach for Packet Error Compensation using radio mobile network quality parameters in a rainfall scenario

This study proposes a new approach to guarantee the Quality of Service of mobile radio links via Packet Error Rate (PER) compensation at varying levels of rainfall. Compared to traditional radio channel degradation estimation methods, the proposed study uses a different set of parameters mainly related to the cell selection and handover mechanisms, more appropriate for rainfall scenarios. By means of a new model based on a Convolutional Neural Network (CNN), the proposed technique estimates the PER on two different radio links in such a way as to dynamically adapt the weights of a VPN (Virtual Private Network) Bonding algorithm based on a dual Subscriber Identity Module (SIM) system belonging to two different telephone operators. This guarantees a halved adaptation frequency when compared to traditional cellular bonding methods and avoids saturating the channel band during the estimation of the weights. The experimental results carried out in video surveillance applications for smart road scenarios show that it is possible to dynamically improve the radio link with an enhancement in the quality perceived by the user of about 200%.