Adaptive Retransmission Layer for Duty Cycle Wireless Networks

In large-scale industrial wireless sensor networks (IWSNs), retransmission strategies are widely adopted to guarantee the reliability of multihop forwarding. However, keeping retransmitting over a lossy link may consecutively fail. Moreover, the retransmission will also be useless over the back-up links that are spatially correlated with the previously failed link. Thus, it is necessary to design a unified retransmission strategy by considering both temporal and spatial link properties to further improve data delivery performance in IWSNs. In this paper, we propose RxLayer, a practical and general supporting layer for data retransmission. By using conditional probability models, RxLayer captures the temporal and spatial link properties without inducing noticeable overhead. According to RxLayer, data forwarding protocol could assess the success rate of the next transmission over the currently used link. When the currently used link is seriously degraded, RxLayer will select an optimal back-up link that temporally has the best link quality to retransmit. RxLayer can be transparently integrated with existing data forwarding protocols. The prototype of RxLayer is implemented in TinyOS. We have evaluated its performance on both indoor and outdoor testbeds. Experimental results show that RxLayer can improve data delivery reliability and energy efficiency in various scenarios compared with the state-of-the-art retransmission strategy. On average, data delivery performance is improved by up to 7.96%, and the number of transmissions is reduced by 43.6%.