Adaptive Retransmission for Wireless Sensor Nodes Under Bursty Error Conditions

Wireless sensor nodes, used in disaster management and industrial applications, typically require high reliability and low energy consumption. This reliability is often hindered due to bursty channel errors in practical wireless systems. The existing retransmission scheme used in IEEE 802.15.4 fails to capture the characteristics of burst channels as they are designed under the assumption of uniform error channels. Thus, they do not significantly improve the reliability under bursty channels while increasing the energy consumption with multiple retransmissions. Our proposed method utilises Reinforcement Learning (RL) to develop an adaptive retransmission strategy. This strategy is informed by analysing the transmission history of the sensor. The RL algorithm learns the optimal policy, which maximises energy efficiency and Packet Delivery Ratio (PDR) in networks with burst channel errors. Our RL retransmission strategy performed better in PDR and energy efficiency than the IEEE 802.15.4 and
Retransmission Adaptive Intelligence Medium Access Control (RAI-MAC) proposed in the literature. By adaptively modifying the retransmission strategy, we improved the energy efficiency and PDR by 16% and 10%, respectively, under a 10-15% packet error rate. The results highlight the importance of an adaptive retransmission strategy for resource-constrained wireless sensors.