Optimised Q-learning for Dynamic Slot Assignment in Medium Access Control Protocol for Wireless Body Area Networks

Abstract

Wireless Body Area Networks (WBANs) enable continuous health monitoring through implanted and wearable sensors, but their performance hinges on an efficient Medium Access Control (MAC) scheme. Conventional protocols struggle to balance throughput, latency, and energy use, key requirements for medical data delivery. This paper introduces QLDSA-MAC, a Q-learning-driven dynamic slot-allocation MAC protocol that continuously adapts time slots to current traffic conditions. The agent maintains a Q-table of state–action values and selects slot assignments that maximize a composite reward reflecting throughput, delay, and energy consumption. Extensive simulations compare QLDSA-MAC with Time Division Multiple Access (TDMA-MAC), Concurrent MAC (C-MAC), and the IEEE 802.15.6 standard. Results show that QLDSA-MAC consistently delivers the highest throughput and the lowest packet delay across a range of traffic loads. It also reduces energy consumption, extending node lifetime in power-constrained scenarios. These gains demonstrate that reinforcement-learning (RL) methods can address WBAN challenges more effectively than fixed-rule MAC designs. Overall, QLDSA-MAC offers a practical path toward reliable, low-latency, and energy-efficient communication in healthcare WBAN deployments.

Keywords