Improving Event-to-Sink Throughput in Wireless Sensor Networks

Abstract

Disaster relief is an important application of sensor networks, in which bursting data needs to be collected in a short period to the sink through a multi-hop wireless network. In general, packets containing the reported data have few correlations among each other, such that meaningful information can be inferred from partially received packets. For better understanding of monitored events, it is more important to capture the total number of unique reports rather than to reliably deliver each individual packet. Therefore, in the case of monitoring disaster filed with bursting data, communication channel throughput has higher priority than the channel reliability. Under this circumstance, we revisit the sensor network transport protocols, which use hop-by-hop recovery to provide reliable data transmission over unreliable links. We found that the complex recovery mechanism, while assuring high reliable individual packet delivery, reduces channel throughput when measured data is reported at high rate. To provide optimal data transport in a sensor network with bursting data generation, we propose a light weight sink centric transport protocol, which maximizes the channel throughput by minimizing the interference of packet recovery process. We implement the sink centric transport protocol in TinyOS and evaluate its performance in the TOSSIM simulator. The comparison shows that our proposed approach outperforms the hop-by-hop recovery approach in terms of event reporting throughput and transmission costs.