High-Throughput Real-Time Reliable Data Collection in Wireless Sensor Network: Implementation and Analysis

This article presents the implementation and analysis of a high-throughput, low-latency, reliable data collection protocol for a wireless sensor network. Insofar as reliable data collection is concerned, end-to-end packet loss recovery is demanded to improve the packet delivery ratio. Traditionally, end-to-end packet loss recovery consists of phases of connection formation and tearing down, which cause extra collection delay. To ease the burden while achieving end-to-end loss recovery, this article makes full use of the payload of tokens and proposes to piggyback lost packets' information on tokens to achieve end-to-end packet retransmissions. Meanwhile, high throughput can be maintained by pipelined burst data transfer and multichannel time-division multiple access (TDMA) communication. One token's piggyback capability is limited due to the fixed available payload; token-burst is further used to improve piggyback capability. For Token-Scheduled multichannel TDMA communication data collection in wireless sensor networks, existing research lacks quantitative analysis on the relationship between topology and collection capacity. Hence, quantitative analysis is conducted for the protocol primitive using the mathematical induction method. The analysis result shows that regardless of the shape of tree-based routing topologies, the collection capacity difference between Token-Scheduled multichannel TDMA and Optimized-TDMA is upper-bounded by 25%, which is verified by extensive simulations. The protocol stack is implemented on sensor nodes equipped with the CC2420 radio chip. The efficiency has been verified with 100% data collection on a deployed test bed consisting of 29 sensor nodes. In addition, extensive simulations are also conducted to investigate the protocol's performance in a fine-grained way.