Collaborative data processing in wireless sensor networks.

Abstract

Wireless Sensor Networks (WSNs) have been used in many application domains, such as target tracking or environmental monitoring. Due to limitations of power supplies, power management and power efficient target tracking techniques have become more and more critical. In this dissertation, systematic approaches are proposed to address the above problems. In particular, efficient energy-aware architectural design aspects of a sensor network are developed, with the goal to reduce the control scheduling algorithm complexity and the power consumption of various components while maintaining the data quality and performance requirements. Research results on an efficient error-bounded sensing scheduling algorithm, a novel collaborative global error implied assisted scheduling algorithm(CIES) and fast target localization for mobile wireless sensor network are presented. Dynamic scheduling management in wireless sensor networks is one of the most challenging problems in long-lifetime monitoring applications. In this thesis, we propose and evaluate a novel data correlation-based stochastic scheduling algorithm, called Cscan. Our system architecture integrates an empirical data prediction model with a stochastic scheduler to adjust a sensor node’s operational mode. We demonstrate that substantial energy savings can be achieved while assuring that the data quality meets specified system requirements. We have evaluated our model using a light intensity measurement experiment on a Micaz testbed, which indicates that our approach works well in an actual wireless sensor network environment. We have also investigated the system performance using Wisconsin- Minnesota historical soil temperature data. The simulation results demonstrate that the system error meets specified error tolerance limits and up to a 70 percent savings in energy can be achieved in comparison to fixed probability sensing schemes. Building on the results obtained from CScan, we further propose and evaluate a collaborative error implication assisted scheduling algorithm, called CIES. This computationdistributive system integrates an implied-error based prediction model together with a stochastic scheduler to adjust neighboring sensors’ operational modes during the occurrence of rare or unusual sensing events. We demonstrate that substantial energy savings can be achieved while also satisfying a global error constraint. We have conducted extensive

simulations to investigate the system performance by using realistic Wisconsin-Minnesota historical soil temperature data. The simulation results demonstrate that the system error meets the specified error tolerance and produces up to a 60 percent energy savings compared several fixed probability sensing references. In order to manage data link quality, a distributed sensor network with mobility provides an ideal system platform for surveillance as well as search and rescue applications. We consider a system design consisting of a set of autonomous robots communicating with each other and with a base station to provide image and other sensor data. A robot-mounted sensor which detects interesting information will coordinate with other mobile robots in its vicinity to stream its data back to the base station in a robust and energy-efficient fashion. The system is partitioned into twin sub-networks in such a way that any transmitting sensor will pair itself with another nearby robot to cooperatively transmit its data in a multiple-input, multiple-output (MIMO) fashion. At the same time, other robots in the system will cooperatively position themselves in such a way that the overall link quality is maximized and the total transmission energy in minimized. We efficiently simulate the system’s behavior using the Transaction Level Modeling (TLM) capability of SystemC. Our results demonstrate the efficiency of our simulation approach and provide insights into operation of the network. Finally, a fast target acquisition algorithm without the assistance of a map, call GraDrive, is introduced for search and rescue applications. We evaluate a novel gradientdriven method, which integrates per-node prediction with global collaborative prediction to estimate the position of a stationary target and to direct mobile nodes towards the target along the shortest path. We demonstrate that a high accuracy in localization can be achieved much faster than with random walk models, without any assistance from stationary sensor networks. We evaluate our model through a light-intensity matching experiment using MicaZ motes, which indicates that our model works well in a wireless sensor network environment. Through simulation, we demonstrate almost a 40% reduction in the target acquisition time, compared to a random walk model, while obtaining a small error in the estimate of the target position.