ISA Transactions 47 (2008) 247255 www.elsevier.

com/locate/isatrans

Development of an integrated wireless sensor network micro-environmental monitoring system

Xianghui Cao a , Jiming Chen a, , Yan Zhang b , Youxian Sun a
a State Key Lab. of Industrial Control Technology, Institute of Industrial Process Control, Zhejiang University, 310027 Hangzhou, PR China b Simula Research Laboratory, Norway

Received 30 April 2007; received in revised form 5 January 2008; accepted 4 February 2008 Available online 19 March 2008

Abstract Wireless Sensor Network (WSN) is increasingly popular in the eld of micro-environmental monitoring due to its promising capability. However, most systems using WSN for environmental monitoring reported in the literature are developed for specic applications without functions for exploiting users data processing methods. In this paper, a new system is designed in detail to perform micro-environmental monitoring taking the advantages of the WSN. The application-oriented hardware working style is designed, and the system platform for data acquisition, validation, processing and visualization is systematically presented. Several strategies are proposed to guarantee the system capability in terms of extracting useful information, visualizing events to their authentic time are also described. Moreover, a web-based surveillance subsystem is presented for remote control and monitoring. In addition, the system is extensible for engineers to carry their own data analysis algorithms. Experimental results are to show the path reliability and real-time characteristics, and to display the feasibility and applicability of the developed system into practical deployment. c 2008, ISA. Published by Elsevier Ltd. All rights reserved.
Keywords: Wireless sensor networks; Micro-environmental monitoring; Data processing; Visualization; Internet access

1. Introduction Recently, Wireless Sensor Networks (WSN) is attracting intensive research efforts due to easy deployment and a large variety of application scenarios. It can be widely applied in habit monitoring, environmental observation, structural detection, physiological tele-monitoring and even drug administration, etc. such as in [17,18,12]. To enable high performance of a WSN, there exist a number of challenges in research as well as in practice. Sensor is normally powered by a battery and has very limited power. Hence, one of the most important issues is about the hardware and protocol design on how to save the limited energy in a sensor node and sensor network as well. This will consequently prolong the networks lifetime. The other problems include enhancing the networks stability and robustness and enabling self-organization and self-healing capability. These requirements necessitate efcient
Corresponding author. Tel.: +86 571 87953762, fax: +86 571 87951879.

E-mail address: jmchen@ieee.org (J. Chen).

routing protocol as well as the network sensor node hardware design. Besides the issues above, sensory data processing becomes substantially increasing in a large-scale sensor networks. For instance, in a large-scale sensor network for environmental monitoring, long monitoring time or high frequency of data acquisition is required. In such a situation, the sensor networks will be ooded with very large amount of measured data [15,21]. The situation becomes even worse if the data are multimedia, e.g. monitored video. The data processing function will be intensive and consumes much energy. To alleviate this problem, advanced methods are needed in data transmission and processing. One approach is data compression in the sense that only the measured data within certain bounds will be transmitted to the Base Station (BS) for afterwards processing [25,26]. However, in this way, useful information of the sensor networks may be lost. An alternative is called distributed processing, which allows sensor nodes or distributed low-cost stations to analyze local sensory data and only deliver the results to the BS. In many cases, centric data management

0019-0578/$ - see front matter c 2008, ISA. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.isatra.2008.02.001

248

X. Cao et al. / ISA Transactions 47 (2008) 247255

and processing are required [21]. A third method is to improve the data processing efciency of the data processing center. Several software platforms, e.g. Matlab [16], are excellent in handling large amount of data with high efciency. As indicated, WSN has a variety of applications in practice, especially micro-environment. A micro-environment is the environment of a small area such as an ofce and an area of crops. Taking the tea garden for example, the growth of tea is highly sensitive to its micro-environment, including illumination, temperature and humidity. Its quality and productivity can be signicantly reduced by unexpected cold weather and insect pest. To prevent the tea from external harm, several traditional solutions are usually employed: (1) the rst depends on the large-scaled weather forecast or insects forecast, which could not provide accurate information of a small tea garden; (2) the second is based on regular harm checking-up or eld data sampling by manual apparatus. This method is accurate though, it cannot be a real-time monitoring system and any unusual event between two checking-ups cannot be detected; (3) newly developed tools based on manual information gathering are capable of data analysis and environmental modeling and predictions making. Nevertheless, lacking of ubiquitous monitoring and low automatization are their inherent limitations. In the literature, there are many researches carried out by using WSN in micro-environmental, particularly agricultural, monitoring in practice. In 2002, researchers deployed 32 motes on the Great Duck Island to monitor Petrels nesting behavior [11]. The self-constructed WSN continually measures environmental data including temperature, humidity, barometric pressure and light level in and around Petrels nesting burrows without disturbing the nesting birds. In [2], a WSN designed for the long-term study of rare and endangered species of plants is presented. The WSN monitors those plants and their environment through temperature, humidity, rainfall, wind and solar radiation sensors. According to [4], researchers have built up a WSN for soil moisture observation which can react with the environment according to the amount of rain falls. The Lofar Agro project measures the micro-climate in potato crops, and the information gathered about circumstances within each individual eld will be used to advise on how to combat phytophtora within a crop [13]. In 2004, a research group of Accenture Lab. deployed a WSN monitoring system for a large vineyard. Sensory data about environmental humidity, wind, water, and soil and air temperature were collected by the BS and then forwarded to a network server, by which multiple clients could query the environmental information [1]. At the same time, there are several specic systems and platforms dealing with WSNs observation and data visualization. MoteView, developed by Crossbow [7], is an end-user interface that provide tools for real-time charting, data logging and mote conguration. TinyDB [22] is a queryoriented platform for extracting WSN information. It provides a simple, SQL-like interface by which users are able to specify the data queried. However, the low extensibility of these tools limits their applications in scientic researches. SpyGlass [3] is another WSN visualizer, which was developed

with Java and the visualization part was implemented by Java2D framework. The jWebDust [5] is a generic and modular application environment based on Java, facilitating a wide range of WSN implementations. Users can query and monitor the execution of the WSN simultaneously as long as they could access the Internet. The TAG approach could provide end-users with efcient query expression and execution, which is derived for general mote-based wireless sensor networks [14]. A practicable structural monitoring WSN called Wisden is developed in [24] which emphasizes on reliable data transportation and time-stamping techniques. In this paper, we propose a generally applicable real-time system based on WSN. In terms of real-time, we mean that the events taking place in a remote environment can be promptly observed, monitored, visualized and recorded to their authentic time. The main technical challenges and features are as follows: (1) the sensor network shall be generally applicable in micro-environmental monitoring with open interface for data acquisition and processing; (2) the WSN can be monitored and controlled remotely via Webinterface. In particular, our system is primarily designed to be a practical applicable and automatized running system that can be applied to general micro-environmental monitoring such as tea garden micro-climate monitoring. It was also designed as an experimental platform to facilitate the data acquisition and fundamental visualization procedures. A user-friendly interface is embedded into the system to facilitate easy programming and data analysis. In addition, it can handle a large amount of data by integrating software platform, e.g. Matlab. To date, among the data processing platforms, Matlab has been mostly used in WSN researches due to its high ability in mathematical calculation and results visualization. However, the existing applications that use Matlab in WSN are mainly in the area of data analysis, calculation, algorithm evaluation [6] and simulation. For example, the Prowler project, running under Matlab, provides a generic and networks multi-layer simulation environment based on Berkeley MICA mote [19]. These applications are always post-processing simulation systems that usually perform off-line. Based on our design, Matlab could perform on-line and real-time. Besides, our paper emphasizes the realistic system design, analysis, implementation and test of a practical wireless sensor networks from the deployed environment-side components to the user-side interface development. The practice-oriented methodology in the paper is able to provide good reference and guideline for practical sensor networks design. In addition, we also have the following specic technical contributions: (1) a practically feasible system architecture which also enables Internet access to WSN data is proposed; (2) a signicant realistic issue on data acquisition, data validation and useful information extraction, data visualization in consideration of event time authenticity, and web-based monitoring are discussed; (3) illustrative numerical results are presented to indicate the system feasibility and efciency; (4) suggestions to apply this system to large-scaled WSNs are provided based on eld experiments.

X. Cao et al. / ISA Transactions 47 (2008) 247255

249

and easy deployment. Usually, they sense their surroundings and send data packets in three ways: Timer-based: sensor nodes sense and deliver data packets periodically with pre-dened timer. Event-driven: nodes send packets only when a specic event takes place, e.g. a re that causes the measurements change over certain extents. Requirement-based: the sensor nodes start to function or report data on the basis of user requirement and time.
Fig. 1. System architecture.

The remainder of this paper is organized as follows. Following an overview of the system architecture in Section 2, the hardware structures is discussed in Section 3. The design of the upper tiers is demonstrated in Sections 4 and 5. Section 6 illustrates the performance metrics with respect to path reliability and monitored data correctness. Finally, Section 7 concludes the paper. 2. System architecture Fig. 1 shows the system architecture, which comprises three basic tiers as described below. The infrastructure tier: This tier contains the wireless connected sensor nodes and the base station (BS). For environmental observation, sensor nodes are usually diffused into certain areas of interest either regularly for simple or irregularly as determined by geographical constraints or user requirements. These nodes can self-organize themselves into a WSN, which continuously generates and delivers to the BS the environmental data packets. The information may include temperature, humidity and light surrounding the sensor nodes. Issues about the manner of sensing, reporting and packet routing protocol are critical in designing this tier. The server tier: Based upon the infra-hardware, the Server tier captures each packet from the BS through the serial cable or the Internet, and forwards them to upper processing program for analysis, calculation and visualization. For more comprehensive afterwards data processing, some professional platform, e.g. Matlab, powerful in handling large amount of data could be introduced. Moreover, it keeps broadcasting packets rened beforehand over the Internet. In such a case, all the remote users (the third tier) can read the broadcasted packets. Due to its centralized manner of data processing and management and system maintenance, the Server should be resource rich and efcient. The user tier: As an important feature of the system, remote users can access the data using Web interfaces. Following this way, users are able to gather, visualize the data packets from the Server and consequently they are able to have a real-time monitoring of the system. In consideration of the WSN security, users are only provided with data surveillance but assumed having no authority to change parameters of the fundamental tier. 3. Hardware structure The sensor nodes are small and battery powered devices with features of wireless communication, long-time running

The recent result indicates that most of the nodes energy is consumed by packets transmission while only a little is used in coding and packet generating. Thus in applications with rare events,1 the second pattern is obviously energy-saving. As for the last pattern, the functions of the sensor network could be more controllable and exible. However, which pattern to choose is based on application-oriented service and dened by the network surveillance. In this study, as we want to provide a long term and automatically running service of environmental monitoring, the sensor nodes are programmed to periodically sense and report information of their circumference. In this way, we can get a knowledge about any node at any sampling time. As for the event-driven or the requirement-based mode, it is difcult to know whether a node is ceased already or dormant unless to spend extra energy in the purpose of retrieving a feedback. In our study, we programmed the nodes to periodically sense and report sensory packets while the period is dynamically adjustable. Hereby, the frequency of sending packets is called the reporting frequency Fr , which is approximately proportional to the internal timer frequency. In practise, the selection of Fr comes to a conict: a higher value of Fr indicates more energy consumption and shorter lifetime, while a lower value incurs less sensitivity to the environmental changes. In the case of micro-environmental monitoring, Fr should be as low as possible but to guarantee that considerable movements of the interested environmental variables could be traced. In spite of introduced complexity to the network, the multihop approach is able to signicantly prolong the networks lifetime. Additionally, since the sensory information is relayed to the BS, the scalability can be enhanced. Among all our known practicable multi-hop routing protocols, the Surge routing protocol [23] is selected for the nodes to build up packet paths from each leaf node to the BS. Packets are sent from one node to the next which is chose dynamically to cope with cases of node failures. 4. Data acquisition and processing within server The Server basically performs two tasks: acquiring and processing WSN data from the base station, and enabling accessing to those data from Internet users as will be discussed in Section 5. Fig. 2 shows the architecture and main components for the rst.
1 Here we refer events to specic emergencies for which the sensor nodes care in the event-driven case.

250

X. Cao et al. / ISA Transactions 47 (2008) 247255

Fig. 3. Two important steps in the procedure of packet validation.

Fig. 2. The Servers components that are responsible for data acquisition and processing.

4.1. Data acquisition The RPackets from the WSN are continuously read by this basic data acquisition component from either its serial port or the Internet. When sensory digital data arrive in the serial receive buffer, a Java program will be notied to partition out individual packets. At the same time, each packet will specify a packet time ( Pt ) as the time when this packet arrives the Server. Then Pt will be padded to a new structure DPacket. Whereafter, the DPacket will be passed to the data processing platform for further analysis. Based on the Java program of data acquisition, the Servers top level analyzes up-going packets, converts the measurements into engineering units, plots them in real time, shows out the networks topologies and saves those packets into diversiform les. 4.2. Packets validation and information extraction Due to wireless communication unreliability and hardware malfunction, there are sometimes abnormal packets which we call the bad packets. A bad packet can be different in bytes or bits from the original correct one. An efcient packet combining mechanism proposed by H. Dubois-Ferri` ere et al. could help to recover the original packet from multiple bad copies [9]. However, we contribute to extract the sensory information from a bad packet when no other versions of the same original packet captured. The detailed procedure will be described by taking for example a bad packet with one byte shorter than a normal one, while bad packets in other cases are to be similarly treated. Although the packet contains useful information, it is hard to decide to which information cell does that missed byte belong. To solve it, a ltering algorithm is embedded. The steps for a one-byte-missed bad packet Pb are (an example of the procedure of the steps 3 and 4 is shown in Fig. 3): Step 1. Check Pb s headers. Since the packet headers are of limited number of values (e.g., the original address for node ID, could only be certain values), the packets

with bad or missed header values could be directly selected out and the algorithm stops. Otherwise go to Step 2. Step 2. Obtain each information cells except for the last one pretending that Pb is a real correct packet. Go to Step 3. Step 3. Check out the very cell, if any, whose value of engineering unit is the most abnormal, i.e. its value exceeds the regular bound (e.g., for temperature, the bound may be [10 C, 40 C]) that is maintained as the Server starts. This cell will be treated as the byte missed cell and then go to Step 5. If there is no such cell and Pb has not been shifted before, then go to Step 4, otherwise go to Step 6. Step 4. Shift Pb forward by one byte and nd corresponding cells similar as in Step 2, but with the rst cell excepted, and then go to Step 3. Step 5. Preserve the correct cells as extracted information, while the one-byte-missed one will be assigned the previous measured value. Stop. Step 6. Drop this packet, But count it into the number of packets from its corresponding node. Stop. Finally, the information extracted will be reconstructed to MPackets to handle and process within the professional data processing platforms such as Matlab and Mathematica. Once the Server launched, a data structure will be initialized and begin to enroll every incoming packets. When applied for long-term environmental monitoring, the data structure grows larger and larger, thus the program will work under great burden. To maintain the quality of on-line service, Data will be ushed into certain les when its volume reaches certain threshold. 4.3. Data visualization After a packet has been received and registered, it is then transferred to the visualization procedure which (1) shows the information contained in this packet, (2) charts the measurements with engineering units, and (3) maps the sensor network topology. Although these are easy to be accomplished for many data processing platforms, dening the visualizing time system remains a big challenge in the view of monitoring to authentic time, since it should reect the real when the events were happening and without information disordering. Before continuing, the following variables should be claried: Denition 1. A packets receiving time Pt : the time when the corresponding RPacket arrives the Server, it is already dened in Section 4.1.

X. Cao et al. / ISA Transactions 47 (2008) 247255

251

Fig. 4. Steps for estimating T0 .

path, by the BS once received. When Ni receives the reply at t2 , it is able to estimate the path delay according to its own clock system by 1 1 2 (t2 t0 t1 ), where t1 relates to the receiving time, medium access time and sending time. Then Ni send the BS an acknowledgement and starts to periodically generate and report packets. On the BS side, when the acknowledgement is received, it makes estimation of the path delay as 2 1 2 (t3 t1 t1 t2 ). Note that these steps are scheduled after the network reaches stable state so that the invariance of each nodes routing path and thus the symmetrical path delays between node and BS could be assumed. The BS will check whether 1 and 2 are close enough (so that the path is stable) and decide to estimate the term T0 by Eq. (2). 1 (2) T0 t3 (1 + 2 ). 2 Although a forth-and-back message communication between the BS and the sensor node is able to give an estimation of the path delay and subsequently the T0 , it should be noticed that our proposed way could judge whether the path is stable at that time, i.e. whether the path delays are symmetric between the forth-and-back communicating. If 1 differs from 2 to a considerable extent, the BS should send a command to stop Ni and initiate the above steps some later. Examples of the results of the measurement charts are shown in Figs. 6 and 8. A nodes topological information is dened by three integers: (1) its group ID, which is only useful for cases with multiple WSNs; (2) its node ID, of course; and (3) its parent nodes ID, which might vary as it always selects the optimal route for data relay. The group ID never changes since it has been programmed up to the node. The topology map is plotted based on these three numbers. When a packet of i th node arrives, it will be investigated to see whether this node has changed its routing parent. If so, the data structure will update itself after having saved its old version as historical data. Consequently, the map also renews. The topology map in present version visualizes the logical relations between father nodes and child nodes. It can be easily promoted to a more meaningful topology map by assigning each node a location information (obtained offline after deployment). Our future version will build the dynamic topology map by on-line positioning, for example, using GPS. 4.4. Embedded function (1) The main purpose of providing the embedded function is to strengthen the systems extensibility by interfacing user dened functions into real-time data processing. The embedded function is embedded in the data processing procedure which could performs with a single thread. The introduction of the embedded function could simplify the researchers work on applying advanced algorithms, e.g. data ltering, or other convenient data analysis methods. It takes four parameters:

Denition 2. A packets Ps : the sequence number, viz., the number of packets has been reported by the corresponding node. Denition 3. A packets Tactual : the time when this packet is generated and reported by its original node. Taking the measurement charts for instance, there are two ways to dene the time system (the charts horizontal axes). Pt -Based: Every packet has been assigned a Pt which can be directly used as its horizontal coordinate. However, this Pt is different from the time Tactual . The gap between the two time systems relates to several factors: sending time, medium accessing time, propagation time, accepting time, and receiving time [20]. In particular, if the time delay grows to a signicant value, this Pt -based method is no longer reliable for charting and reecting the actual eld time. Another drawback of this method lies in the fact that when a packet is currently received, whether there are packets of the same node lost before it is unclear. Moreover, in the case that the time delay is accidentally longer than the time gap between two successive receiving, the Pt -based method may make mistakes. For example, if a node sent a packet p1 through a path with great delay to the BS, but sent its next packet p2 through a much shorter and more uent path. Then, p2 may arrive at the BS rst. Consequently, an increase in environmental temperature, for instance, may turn out to be a decrease in the charts. Ps -Based: Based on Ps , Tactual can be derived from Eq. (1), where Fr refers to the reporting frequency, T0 is the time when the corresponding node starts to function, and the last s term iP =1 i indicates the nodes timer drifts.
Ps

Tactual = T0 + Ps / Fr +
i =1

T0 + Ps / Fr .

The timer drift of a sensor node is typically 30 50 s in 1 s [10]. Therefore, this term could be neglected as long as Ps is not too large. In order to estimate T0 , an NTP-like [8] protocol is applied. Each node Ni should rst inform the BS for registration at some time t0 (as shown in Fig. 4); this message will be immediately replied, through the same

252

X. Cao et al. / ISA Transactions 47 (2008) 247255

Fig. 5. The components of both the Server and the user tiers that could enable Internet access to WSN data.

Fig. 6. The humidity measurements through one night.

(a) Measurement charts with the Pt -based coordinate system.

Fig. 7. The path reliability with packets Ps as the horizontal axes.

(1) the historically received packets, (2) current received packet in the form of MPacket, (3) a global data structure for recording calculated results, and (4) a global gure handle for results visualization. In this way user codes along with the main Server program are able to run in real time. Therefore, dynamic data analysis and other possible algorithms could be easily proposed and directly applied, resulting in reinforcing of the Servers extensibility. A typical application of the embedded function

(b) Measurement charts with the Ps -based coordinate system. Fig. 8. The light sensor readings by the small WSN.

concentrates on dynamic prediction. Based on historically received packets, the environmental condition about a node at time especially when this node is temporarily out of work or its packets get lost could be predicted.

X. Cao et al. / ISA Transactions 47 (2008) 247255

253

5. Internet access to WSN data In order to facilitate remote WSN monitoring, the Server provides remote users with real-time data accessing, by broadcasting rened packets over the Internet if requested. The detailed structure responsible for this task are illustrated in Fig. 5. 5.1. The server side In fact, the Server maintains a connecting list that only the users appear in it can receive data packets and any action of connection or disconnection with the Server will cause this list to be updated. Before being sent, an MPacket, which contains elds of integers, strings and oat numbers, will be transformed into an array of bytes (BPacket for short) for broadcasting over the Internet. In the opinion of data security, any user should rst get permission from the Server before receiving any packet. In our design, a user should rst send a request to the Server conveying its user name, password and its own Internet address. Then, this information will be checked by a small embedded program to see whether the name and password are right and whether the address is accredited. If it is permitted, an acknowledgement message will be sent back to this user. 5.2. The user side As an end-user interface, the user program generally accomplishes three functions: (1) interacting with the Server, (2) charting packets, and (3) listening and responds to user manipulations. Once the Server opens up for connections, a user is able to connect with it via the Internet if it has been permitted. Then, it receives all the BPackets transmitted from the Server, and parses them afterwards. In fact, the user rst checks out and simply drops the abnormal BPackets which may result from Internet unreliability. Then it translates the rest into MPackets for information extraction and plotting. Besides, a user also keeps listening to the actions by the user until a disconnection command is called. Before the user ceases, it will send a closed message back to the Server to erase its name from the connecting list. At the same time, the action that closes the Servers will shut down all the user connections imperatively. The user side is a passive program that it can only listen packets from the Server but could not manipulate it in terms like starting/pausing communication with the sensor network. 6. System implementation and test Among our known professional data processing platforms and tools to date, e.g. Maple, Matlab and Mathematica, we select Matlab to embody the concepts and frameworks discussed above. In our experiment, 5 Mica2 motes with one for the base node (all function under the same emitting power and the same reporting frequency Fr ) are involved. The MTS400CA [7] sensor boards are used to detect the environment. 3 PCs with one for the Server and the other two for the users are used to test the network ability.

6.1. Environmental measurements This experiment was carried at late December and all the equipments ran to monitor the both indoor and outdoor environment in the suburb of the city of Hangzhou. The reporting frequency of each node was set to 1/30 s. Fig. 6 expresses the variations of humidity,2 in which node 8 was placed outdoor and the rest were indoor where the humidity was obviously much lower. It can be seen that the humidity both indoor and outdoor ascended through the whole night. This is because the air in saturated state could contain less water as the temperature decreased in the night. 6.2. Performances evaluation In order to evaluate the whole system performances when it is applied for large-scaled WSNs, the nodes reporting frequency is enlarged so that similar amount of data will be transmitted over the network with only a few sensor nodes. In our experiment, 5 nodes are involved and are deployed carefully to form a WSN with maximum 4 hops. Each nodes Fr is set to 1/5 s. At rst, the nodes will communicate with each other for a short period to stabilize their routing paths. Then, they will run through the steps mentioned in Section 4.3 one after another node to let the BS estimate their T0 . All the following data analysis is done by the embedded function as mentioned in Section 4.4. We introduce the path reliability as the percentage of packets received over the number BS should received, to evaluate the multi-hop path quality. For each node, its path reliability at time t is dened in Eq. (3), where Ps (t ) is the biggest Ps known by the BS till t . As shown in Fig. 7, it is almost always the case that the more multi-hops, the less reliable of the packet routing paths. After a long time of communication, the packet losses for each node nally reach their stable states, which differ only a little from each other. This reects that instead of wireless channel unreliability, the processing resource competitions amongst injected and self-generated packets are responsible for packet losses. This experiment gives the insights that in largescaled WSN applications using the Surge routing protocol, the path reliability curve for each node will look similar to that is shown in Fig. 7, but will be quite different from each other depending on environmental conditions of wireless communications.
t

Packet received path reliability(t ) =

0

Ps (t )

(3)

Fig. 8 shows the environmental light intensity detected by the nodes, which is charted with the two coordinate systems, i.e. the Pt -based and the Ps -based systems. The quick drops in both subgures indicate the event that the corresponding nodes are shaded from daylight simultaneously. However, without consideration of packet delay, Fig. 8(a) under Pt
2 The relative humidity.

254

X. Cao et al. / ISA Transactions 47 (2008) 247255

clocks should be set asynchronous and wither longer periods to prevent network trafc bursts. The BS is now also an exception comparing with other nodes, which again advocates the high efciency of serial communication and of upper PC software data processing. 7. Conclusion In this paper, a system capable of environmental monitoring and data processing based on WSN is presented. Experiments making use of this system are then demonstrated, which shows the systems well performance and the convenience to employ Matlab into data analysis and results visualization. The system has made several achievements in: (1) unlike some simulators, this system is a practical environmental monitoring application where sensor nodes periodically sense their ambience and the information they got can be relayed eventually to the research center for upper software analysis. (2) sensory information is extracted and visualized accounting for its authenticity. (3) Packets from the Server can be forwarded to all the users over the Internet which makes remote monitoring of elds environment feasible. (4) The system is also extensible that professional data processing platforms could be embedded to carry out on-line algorithms such as digital lter, efcient statistical methods, accuracy-aware algorithms, etc.. Future work will employ more nodes and powerful base station to provide long-term monitoring of a tea garden. Besides, this system can be reinforced by introducing more robust and efcient hardware working styles and more reliable routing methods to make it less aware of environmental disturbances. One proposal concentrates on the routing way, i.e. to distribute some routing stations (RS) within the WSN. Any sensor nodes can be assigned an RS and its packets directly goes to this RS and then can be forwarded to the BS. Thereby, an increase in the network reliability and a decrease in energy cost when applied in practical usage are expected. The part of web-based monitoring can be strengthened by introducing the database technology, thus, users could query and retrieve data from the services of the Server. The network data security will also be taken into account in our future work. Besides, future work will integrate applications of other common sensor node species to make the system a more generic model. Acknowledgements This work is supported by the Joint Funds of NSFCGuangdong under grants U0735003, National Natural Science Foundation of China under Grant No. 60604029, National Natural Science Foundation of Zhejiang Province under Grant No. Y106384, the R&D Program of Zhejiang Province under Grant No. 2007C31038 and the Scientic Research Fund of Zhejiang Provincial Education Department under Grant No. 20061345. References
[1] Accenture. Pickberry vineyard: Accenture prototype helps improve crop management. http://www.accenture.com/xd/xd.asp?it=enweb\&xd= services%5Ctechnology%5Ccase%5Cpickberry.xml.

(a) Packets delay time viewed at their received time by the BS.

(b) The percentage of available message buffer size over the total buffer size, which is reported by each node periodically. The total buffers all are set to capable of accommodate 64 packets. Fig. 9. Figures showing the relationship between packet delay and nodes buffer congestion.

base coordinate system gives the illusion that the above event happened asynchronously; while in Fig. 8(b), the corresponding two curves behave almost the same. Although, there still exists a slight time difference between the 1hop node and 4hop node, which is related to the error of T0 estimations, asynchronous environmental detection and their timer drifts, etc.. From Fig. 9(a), we can see that the packet delay for each node is not constant but grows with time slowly. Among the many factors [20] that delay the packet transmission, it is the sending and receiving time (approximates to the time that a packet waited in buffers of both the sender and receiver sides for processing) that could vary greatly with time and thus dominates the packet delay. Fig. 9(b) illustrates how the nodes buffers changing with time. It is clear that the decrease in available buffers (more packets unprocessed and accumulated due to bursts of data arrivals and low-processing abilities at the nodes) is parallel to the increase in packet delay. The underlying suggestion for large-scaled WSN applications is that in order to shorten the delay or lessen that accumulations, the nodes

X. Cao et al. / ISA Transactions 47 (2008) 247255 [2] Biagioni E, Bridges K. The application of remote sensor technology to assist the recovery of rare and endangered species. International Journal of High Performance Computing Applications 2002;16(3). [3] Buschmann Carsten, Psterer Dennis, Fischer Stefan, Fekete S andorP, Kr oler Alexander. Spyglass: A wireless sensor network visualizer. ACM SIGBED Review 2005;2(1):16. [4] Cardell-Oliver R, Smettem K, Kranz M, Mayer K. A reactive soil moisture sensor network: Design and eld evaluation. International Journal of Distributed Sensor Networks 2005;1(2):14962. [5] Chatzigiannakis Ioannis, Mylonas George, Nikoletseas Sotiris E. jwebdust : A java-based generic application environment for wireless sensor networks. In: Proceedings of the IEEE international conference on distributed computing in sensor networks (DCOSS) 2005; p. 37686. [6] Chen JC, Yip L, Elson J, Wang H, Maniezzo D, Hudson RE, Yao K, Estrin D. Coherent acoustic array processing and localization on wireless sensor networks. Proceedings of the IEEE, vol. 91. 2003. p. 115462. [7] Crossbow Technology Inc. Website at: http://www.xbow.com. [8] Mills DL. Internet time synchronization: The network time protocol. In: Global states and time in distributed systems. IEEE Computer Society Press; 1994. [9] Dubois-Ferri` ere Henri, Estrin Deborah, Vetterli Martin. Packet combining in sensor networks. In: SenSys 05: Proceedings of the 3rd international conference on embedded networked sensor systems. New York (NY, USA): ACM; 2005. p. 10215. [10] Coleri Ergen Sinem, Varaiya Pravin. Pedamacs: Power efcient and delay aware medium access protocol for sensor networks. IEEE Transactions on Mobile Computing 2006;5(7):92030. [11] In-Situ sensor networks for habitat monitoring. http://www.coa.edu/html/ motes.htm. [12] Li Mo, Liu Yunhao. Underground structure monitoring with wireless sensor networks. In: IPSN 07: Proceedings of the 6th international conference on information processing in sensor networks. New York (NY, USA): ACM; 2007. p. 6978. [13] Lofar Agro project. http://www.lofar.org/p/Agriculture.htm. [14] Madden Samuel, Franklin Michael J, Hellerstein Joseph M, Hong Wei.

255

[15]

[16] [17] [18] [19] [20]

[21]

[22] [23] [24]

[25]

[26]

Tag: A tiny aggregation service for ad-hoc sensor networks. In: OSDI 02: Proceedings of the 5th symposium on operating systems design and implementation. 2002. p. 13146. Mainwaring A, Polastre J, Szewczyk R, Culler D, Anderson J. Wireless sensor networks for habitat monitoring. In: ACM international workshop on wireless sensor networks and applications. 2002. Matlab. http://www.mathworks.com. Xu N. A survey of sensor network applications. IEEE Communications Magazine 2002;40(8):10214. Ozern Hasan. Wireless sensor networks for crack displacement measurement. Masters thesis. Northwest University; 2005. Prowler. http://www.isis.vanderbilt.edu/projects/nest/prowler/. Shin Kee-Young, Lee Kwangyong, Kim Haeyong, Soo Mah Pyeong, Park Seungmin, Lim Chaedeok. Heung-nam kim: A exible, high-precise time synchronization for multi-hop sensor networks. In: Proceedings of the ninth IEEE international symposium on object and component-oriented real-time distributed computing. 2006. p. 16973. Steere DC, Baptista A, McNamee D, Pu C, Walpole J. Research challenges in environmental observation and forecasting systems. In: Proceedings of the 6th annual international conference on mobile computing and networking (MobiCom00). p. 29299. Tiny DB. A Declarative Database for Sensor Networks. http://telegraph. cs.berkeley.edu/tinydb/. Tiny OS. http://www.tinyos.net. Xu Ning, Rangwala Sumit, Chintalapudi KrishnaKant, Ganesan Deepak, Broad Alan, Govindan Ramesh, Estrin Deborah. A wireless sensor network for structural monitoring. In: SenSys 04: Proceedings of the 2nd international conference on embedded networked sensor systems. New York (NY, USA): ACM; 2004. p. 1324. Zhang Jingbin, Zhou Gang, Son Sang H, Stankovic John A. Ears on the ground: An acoustic streaming service in wireless sensor networks. In: Fifth IEEE/ACM international conference on information processing in sensor networks (IEEE/ACM IPSN06, Demo Abstract). 2006. Zhi-Jun XIE, Lei WANG, Ya-Ping LIN, Hong CHEN, Yong-He LIU. An algorithm of data aggregation based on data compression for sensor networks. Chinese Journal of Software 2006;17(4):8607.