Delay-constrained power-efficient data aggregation framework in sensor-actor networks

• Main Author: Xu, Chendong
• Format: Others
• Language: en
• Published: University of Ottawa (Canada) 2013
• Subjects: Computer Science.
• Online Access: http://hdl.handle.net/10393/28260; http://dx.doi.org/10.20381/ruor-19162

In data aggregation problems, sensor measurements from the whole sensory field are collected at the data sink periodically or on-demand as a single report using functions such as average, maximum, minimum, counts, deviation, etc. This thesis is to design a data aggregation framework applicable for real-time sensor-actor networks. Our goal is to set up a reporting tree that will minimize power consumption at individual nodes while preserving delay requirements. Existing solutions to data aggregation problem usually use hop count as the energy cost metric and/or operate in a centralized fashion. The only known delay bounded power efficient localized algorithm [MGPA] sets up an initial power efficient tree (regardless of delay), and then dynamically changes the tree based on measured delay in ongoing traffic, with speed-ups and slow-downs achieved by using maximal/minimal transmission ranges at some nodes. We show here that the initial tree is closer to hop count than power optimal while the energy consumption per node is apparently unbalanced. We propose to construct a power optimal delay bounded data aggregation tree, assuming delay is proportional to hop count, which is a reasonable approximation in low traffic scenarios. Our desired hop progress (DHP) scheme constructs a data aggregation tree rooted at a sink/actor using only edges of localized minimal spanning tree (LMST) over all sensors, if delay along this tree is acceptable and power consumption is to be near optimal, like in [TKS]. Otherwise, hop selection (along LMST) made at each step is subject to the ratio of potential delay to message lifetime. The main idea is to reduce the network's overall energy consumption and balance energy consumptions at nodes by applying approximately equal hop lengths along the tree. There are two variants of DHP: DHP with Area coverage algorithm (DHPA), and DHP with Area coverage algorithm and CDS construction algorithm (DHPAC). In DHPA algorithm, area coverage algorithm is applied first to select a subset of active sensors that monitors the same area as the original set, thus allowing the rest of sensors to sleep. Then DHP is applied on the set of active sensors. In DHPAC protocol, a connected dominating set (CDS) is constructed over active sensors. Active sensors not in CDS report to their nearest CDS neighbors (related delay counted at the parent node), and DHP is then applied over LMST of CDS nodes and links in CDS. Experimental results indicate that DHP can totally save up to 75% energy and extend up to 123% network lifetime in comparison with [MPGA]. Meanwhile, DHPA and DHPAC also present significantly better energy efficiency than the variant of [MPGA], where area coverage algorithm is also implemented.