< back to main site

Publications

Investigations using the Wireless Sensor Network demonstrator testbed.

Collett, M; Adamson, D B; Esward, T J (2008) Investigations using the Wireless Sensor Network demonstrator testbed. NPL Report. TQE 3

[img] Text
TQE3.pdf

Download (7MB)

Abstract

Wireless Sensor Networks (WSNs) are set to become a significant enabling technology in many areas of measurement. An increasing range of applications including health monitoring, environmental monitoring and military surveillance are beginning to require WSN based solutions. Traditional metrology concentrates on the calibration of single instruments. WSNs represent a new type of measurement system that employs distributed sensors with both local and centralised processing. Calibration of such distributed measurement systems and the validation of their outputs present significant new metrology challenges.
In this report we summarise initial work carried out at NPL to develop a practical understanding and capability in WSN-based measurements and to demonstrate this expertise internally and externally to establish a basis for future work in this field.
The first stages of this exploratory project involved researching the existing work and industrial needs in the area of WSNs. Through a combination of a literature review, seminar attendance and visits to key collaborators a number of important measurement issues were identified. These included radio link quality, networking, sensors, onboard processing and high-level processing. It was also found that there were two main barriers to adoption of these technologies. Firstly, the need for expertise in software and electrical engineering to set up any network has proved to be prohibitively expensive and time-consuming for many potential users, this demonstrates that the technology is, in many ways still in it's infancy. Secondly, the outputs of such systems are generally not fully understood. The accuracy and reliability of WSN measurements needs to be more fully characterised for cost-benefit or risk analysis to be carried out.
In order to investigate some of these measurement issues, as well as provide a demonstration platform, a testbed of MICAz nodes was set-up. These operate under the Zigbee protocol, and were applied to some measurement tasks.
Of the key measurement issues, radio link quality was chosen as one of the areas on which to concentrate. As this is important to all WSN applications, any experimental findings would have a broad impact. The importance of this issue is increased by the likelihood that WSNs will be deployed in harsh environments and be subjected to a range of interferences. Severe degradation of the wireless links could make any system completely inoperative. Using a metric of packet loss-rate the effects of a number of environments and interference types were investigated. It was found that highly reflective environments could cause the radio links to fail entirely, and the level of signal quality was highly position-dependent. When looking at the effects of external interference it was found that relatively low levels of white noise could jam the system and prevent the nodes from transmitting. When a simulated Bluetooth signal was used, it was found that this jamming didn¿t occur, but the receiver was eventually blinded. The fact that the types and magnitudes of interference effects observed was dependent on the type of signal (rather than just the power) means that new methods of measuring and predicting compatibility will be required.
Another important finding of the review of work in the area of WSNs was that a large proportion of the research carried out elsewhere concentrated on individual aspects of the systems, such as novel sensors or new networking protocols. WSNs are highly integrated systems. The components of sensors, onboard processing, power supply and radio communications are all closely linked. In order to observe all the possible effects, a `system-based' approach was required. For this reason it was decided to look at the performance of the testbed when performing the resource-hungry application of acoustic ranging. By measuring the time of flight of an acoustic pulse the nodes can make estimates of their distances from each-other. This requires capabilities including local data processing, local data validation and synchronisation. It was found that the standard uncertainty in a single ranging estimate was approximately 3.5 cm and that different node pairs had different offsets. The nodes were all nominally the same, so these differences in measurement offset must have resulted from manufacturing differences in the hardware. It was also found that the onboard software was very unstable, so minor differences in hardware could have resulted in large differences in ranging estimate. Despite these limitations and problems, it was possible to calibrate the system to remove the offsets. With the aid of the Largest Consistent Sub-Set technique developed within the National Measurement System's Software Support for Metrology programme it was possible to use the sensors for a collaborative measurement task which achieved uncertainties of 0.27 cm. This was valuable as it demonstrated that this data analysis technique could be applied to WSNs and greatly improve the system performance.
This project identified a number of important areas where further work is required. Node self-localisation is vital to a large number of WSN applications, but the current techniques of acoustic ranging are not of sufficient accuracy or robustness. Methods based upon radio interferometric and signal strength measurements show more promise and warrant further work. Interference between different radio protocols and the influence of signal digitisation is becoming an increasingly important issue as the electromagnetic spectrum becomes more crowded. The need for calibration of networks made up of usually cheap, poor-quality sensors is vital for any meaningful measurements to be made. This becomes an increasingly challenging task as networks grow in scale and deployments are made in harsh environments. Techniques for performing distributed calibrations within a network are required. Synchronisation is also an important capability which is in need of improvement for these resource-constrained systems, as it is necessary for many applications and networking protocols. The experimental work within this project to look at radio interference and the resource hungry application of acoustic ranging has not only enabled the above issues to be identified, it has also helped develop vital practical experience and expertise which will help drive continued work in the future.

Item Type: Report/Guide (NPL Report)
NPL Report No.: TQE 3
Keywords: wirless sensor network, testbed, communications, interference, acoustics
Subjects: Electromagnetics
Electromagnetics > RF and Microwave
Last Modified: 02 Feb 2018 13:15
URI: http://eprintspublications.npl.co.uk/id/eprint/4194

Actions (login required)

View Item View Item