Learning Automata Based Q-Learning RACH Access Scheme for Cellular M2M Communications

This paper introduces a novel Learning Automata (LA) scheme to improve the performance of a Q-learning based RACH access (QL-RACH) scheme, when applied for combined human and Machine-to-Machine (M2M) communication in a cellular network. The scheme employs an LA technique to eliminate the excessive punishment suffered by the M2M users by controlling the use of a penalty factor in the QL-RACH scheme. Simulation results show that the proposed LA-QL-RACH scheme improves the RACH throughput by 82% and reduces the access delay performance by 79%, thereby eliminating the repetitive collisions experienced in the dedicated M2M slots.