Frame based back-off for Q-learning RACH access in LTE networks

Abstract

This paper introduces a novel back-off scheme to improve the performance of a Q-learning based RACH access (QL-RACH) scheme. A Frame-based Back-off with QL-RACH (FB-QL-RACH) scheme is proposed. The scheme reduces the probability of collision between Human-to-Human (H2H) and Machine-to-Machine (M2M) users when sharing the same frame for both the initial access and the back-off. Also the probability of idle slots caused by M2M back-off in the QL-RACH scheme is eliminated. In addition the paper also considers the interaction of Poisson and Periodic traffic models in sharing the RACH. Simulation results show that the new back-off scheme improves the overall RACH throughput by around 70%. The results also show that both traffic distributions can be controlled using QL-RACH.