Learning-Based Resource Allocation for Data-Intensive and Immersive Tactile Applications

Abstract

The immersive tactile applications that are emerging in the entertainment, education and health industries are anticipated to be available for mobile users in the close future. These applications are data-intensive and delay-sensitive due to the nature of information that is being exchanged. With today’s mobile networks, the throughput and latency challenges are the major roadblocks for mobile users. In this paper, we propose a resource allocation technique with the aim of increasing throughput and reducing latency of Data Intensive Devices (DIDs). We consider the coexistence of DIDs with traditional User Equipments (UEs) on a two-tier, densely deployed network of Small cell Base Stations (SBSs) and eNBs. We propose a Q-learning-based resource allocation scheme, namely, Throughput Maximizing Q-Learning (TMQ) that learns the efficient resource allocation of both SBSs and eNB. The proposed technique is compared with well-known Proportional Fairness (PF) algorithm in terms of average throughput, delay, and fairness. Simulation results show significant improvement in throughput, 80% reduction in delay, and 6% increase in fairness.