Traffic scheduling, network slicing and virtualization based on deep reinforcement learning

The revolutionary paradigm of the 5 G network slicing introduces promising market possibilities through multi-tenancy support. Customized slices might be provided to other tenants at a different price as an emerging company operators. Network is difficult deliver higher performance and cost-effective facilities render resources utilisation in alignment with customer activity. Therefore, this paper, Deep Reinforcement Learning-based Traffic Scheduling Model (DRLTSM), has been proposed interact environment by searching for new alternative actions reinforcement patterns believed encourage outcomes. DRL situations addresses power control core priority-based sizing involves radio resource. This paper aims develop three main blocks i) traffic analysis slice forecasting, (ii) admission management decisions, (iii) adaptive load prediction corrections based on calculated deviations; Our findings suggest very significant possible improvements show that DRLTSM dramatically improving its efficiency rate 97.32%, scalability compatibility comparison baseline.