Research On Low Latency Data Aggregation Problem In Wireless Sensor Networks

With the development of micro electronic technology, embedded technology, wireless communication and sensor technology, the purpose of producing low cost-, low power-, tiny-sensors has been come true^[1-5]. Wireless sensor network consists of large numbers of sensor nodes equipped with wireless communication equipment. It is widely used in many applications like data collection, environment monitoring, and target tracking. In these applications, statistic information is often quired by users, including the number of events and the number of active nodes. As the operation of data aggregation is very suitable for getting these informations, reducing the latency during the course of data aggregation is very important for improving the real-time ability of applications. Consequently, it is neccesary to research the problem of low latency data aggregation.The performance of data aggregation algorithms is affected by many factors distributed in different layers of wireless communication stack. Therefore, we consider the problem of low latency data aggregation from different layers of wireless communication stack. Our contributions of this paper are as follows:1) Firstly, we propose a collision free data aggregation algorithm using signal strength received in physical layer. As Unit Disk Model[7] limits wireless interference within a disk area, ingoring the fact that signal propagates in infinite space, traditional data aggregation algorithms based on Unit Disk Model cannot avoid collision in data aggregation. The algorithm porposed by us can overcome this problem. It is based on Physical Interference Model^[6] which models wireless interference in infinit space. At the beginning, a tree for data aggregation is constructred by forming minimum match sets and maximum independent sets. Thereafter, we design our collision free scheduling in this tree. Data are aggregated to the base station along this tree. The simulation results show that our algorithm can be used in large scale wireless sensor networks. The algorithm has a good performance in the environment with strong background noise.2) Secondly, we propose a data aggregation algorithm based on pre-aggregation scheme in MAC layer. This algorithm can sharply reduce the size of transmissions used in data aggregation, leading to low latency of data aggregation. In our algorithm, nodes overhear the transmissions from neighbors in open wireless channel. By taking pre-aggregation, nodes reduce the redundancy information in their packets and then send the results for aggregation. As the length of results is less then the raw packets, the size of transmission used in data aggregation is reduced, leading to a good performance on latency. The simulation results show that comparied with the WAN algorithm^[7], the latency is reduced by 65% while the energy consumption is reduced by 60%.3) Thirdly, we propose an opportunistic data aggregation algorithm with anycast in route layer. We consider the packet loss problem in data aggregation. As retransmissions are usually required when packet loss happens, resulting in more latency of data aggregation, we try to reduce the number of retransmissions in data aggregation by using anycast to transmit data packets. As the number of retransmissions is reduced, less energy are used in data aggregation. The simulation results show that the latency of data aggregation by our algorithm is reduced by 15% while the energy consumption is reduced by 10%, compared with SPT (Shortest Path Tree) algorithm, leading to a good performance on latency and energy.4) Lastly, we also propose another opportunistic data aggregation algorithm with energy harvesting scheme in route layer. We consider the wireless sensor networks in which sensors are equipped with energy harvesting equipment and can havest energy from environment. The algorithm proposed can avoid quick dying resulted from abusely participating opportunistic routing. It makes a good tradeoff between reducing the latency of data aggregation and improving the energy efficiency. The simulation results show that the latency of data aggregation by our algorithm is reduced by 45.8% and 28.9%, compared with Sketch^[8] and Tributary-Delta^[9], respectively.