context aware wireless sensor networks for smart home ... · pdf filebiographical notes: ......

Int. J. Autonomous and Adaptive Communications Systems, Vol. 6, No. 2, 2013 99

Copyright © 2013 Inderscience Enterprises Ltd.

Context aware wireless sensor networks for smart home monitoring

Charles C. Castello* Applied Research Center, Florida International University, 10555 West Flagler Street, EC2100, C4-5, Miami, FL 33174, USA and Department of Electrical and Computer Engineering (ECE), Florida International University, 10555 West Flagler Street, EC3915, Miami, FL 33174, USA E-mail: [email protected] *Corresponding author

Ruei-Xi Chen Computer Science and Information Engineering Department, St. John’s University, 499, Section 4, Tam King Road, Tamsui, Taipei, Taiwan E-mail: [email protected]

Jeffrey Fan Department of ECE, Florida International University, 10555 West Flagler Street, EC3915, Miami, FL 33174, USA E-mail: [email protected]

Asad Davari Department of ECE, West Virginia University Institute of Technology, 405 Fayette Pike, Montgomery, WV 25136, USA E-mail: [email protected]

Abstract: This paper introduces a temperature control framework for smart homes using wireless sensor networks (WSN). A key issue with temperature monitoring and control is standard sampling techniques which take few

100 C.C. Castello et al.

temperature samples into consideration to make heating and cooling decisions in large areas of space. This results in poor controllability of temperature in unmonitored locations with potentially significant temperature variations in comparison with monitored locations. To solve this problem, spatial analysis techniques, namely geostatistical analysis, can be utilised to predict temperature in unmonitored locations to aid in making more informed decisions on how to heat and cool certain parts of a dwelling. Results show independent temperature control in defined areas using the proposed temperature control framework.

Keywords: smart home; intelligent temperature control; WSN; wireless sensor network; context awareness; geostatistical analysis; classical variography; ordinary point kriging.

Reference to this paper should be made as follows: Castello, C.C., Chen, R-X., Fan, J. and Davari, A. (2013) ‘Context aware wireless sensor networks for smart home monitoring’, Int. J. Autonomous and Adaptive Communications Systems, Vol. 6, No. 2, pp.99–114.

Biographical notes: Charles C. Castello received his Bachelor of Science and Master of Science degrees in Computer Engineering from Florida International University (FIU) in the Spring of 2007 and 2009, respectively. Currently, he is pursuing a PhD in Electrical Engineering from FIU and is expected to graduate by Spring 2011. He is also working at Applied Research Center at FIU as a Research Assistant and as part of the Department of Energy (DOE)-FIU Science and Technology Workforce Development Initiative (DOE Fellowship). He is a Student Member of the Institute of Electrical and Electronics Engineers (IEEE), Delta Epsilon Iota Honor Society (DEI), Eta Kappa Nu Honor Society (HKN), Golden Key International Honor Society and Upsilon Pi Epsilon Honor Society (UPE). His research interests include environmental monitoring using wireless sensor networks, in-situ instrumentation for measuring the concentration of methyl-mercury, optimal sensor placement and smart homes.

Ruei-Xi Chen received a Bachelor of Science degree from the Department of Electronic Engineering at Tamkang College, Taipei, Taiwan, in 1975, a Master of Science degree from the Department of Information Engineering at Tamkang University, Taipei, Taiwan, in 1982 and a PhD degree from the Department of Electrical Engineering at National Taiwan University, in 2000. He had been a Lecturer of St. John’s and St. Mary’s Institute of Technology since 1982. Currently, he is an Associate Professor in CSIE at St. John’s University. His group of students obtained over ten awards on the Taiwan National Competition of Specific Application Microcomputer System Design. He has been a consultant of many hi-tech companies, where he assisted in establishing their key technologies in embedded system of DSP applications. His research interests are major in the reconfigurable computing for digital signal processing, including speech and video coding, SOPC architecture design and implementation methodology.

Jeffrey Fan is currently an Assistant Professor in Electrical and Computer Engineering at Florida International University. Prior to his academic career, he served as the Vice President of Vivavr Technology, Inc. and General Manager/Cofounder of Musica Technologies, Inc. From 1988 to 2002, he held various senior technical positions in California at Western Digital, Emulex Corporation, Adaptec Inc. and Toshiba America. His product line of research and development includes virtual reality (VR) 3D animation, MP3 players, hard drives, fibre channel adapters, SCSI/ATAPI adapters, RAID disk array, PCMCIA cards and laser printer controllers. He received his PhD in Electrical

Context aware WSN for smart home monitoring 101

Engineering at University of California, Riverside in 2007 and a Master of Science degree in Electrical Engineering from State University of New York at Buffalo in 1987. He also holds Bachelor of Science degree in Electronics Engineering from National Chiao Tung University in Taiwan. He is a Senior Member of IEEE.

Asad Davari received his Bachelor of Science (1980), Master of Science (1981) and PhD (1985) degrees from the University of Alabama in Huntsville. In 1985, he joined the Department of Electrical Engineering, WVU Tech where he is currently a Full Professor of Electrical Engineering. He teaches and conducts research in the area of control theory, and applications as well as energy-related studies. He is a Senior Member of IEEE, a Member of Eta Kappa Nu and IEEE WV Section Chair 2009. His research was recognised by the West Virginia Governor with a Certificate of Achievement in Scientific Research on 9 February 2004.

This paper was presented in part at the IEEE 10th International Symposium on Pervasive Systems, Algorithms and Networks (I-SPAN09) (Castello et al., 2009).

1 Introduction

Research involving smart homes has flourished in recent years due to advancements in computational resources including processing power, system memory and transmission bandwidth. These improvements allow for state-of-the-art electronic devices to be seamlessly integrated into homes for unparallel comfort, control and energy efficiency. To see these benefits come to fruition, there are three elements that must be considered:

1 internal network

2 intelligent control

3 home automation (King, 2003).

Home automation is the use of products to link services and systems outside the home. These product types include environmental, security, home entertainment, domestic appliances, information and communication, and health (King, 2003). This paper will focus on environmental services, mainly developing a novel framework for intelligent temperature control in smart homes using geostatistical analysis. This work will add to knowledge in context awareness modelling for smart home services.

There has been much quality research work (Baek et al., 2004; Jiang et al., 2004; Pounds-Cornish and Holmes, 2002) done on smart homes, but improvements are needed, particularly in context awareness modelling for smart home services. This paper seeks to accomplish this by developing a novel framework for intelligent temperature control using geostatistical analysis for total spatial awareness. Described in Section 2 of this paper is previous research in temperature control, where decision making is based primarily on temperature data from a handful of sensors, not taking into account locations lacking temperature readings. This gives an inaccurate depiction of the area being monitored, thereby controlling temperature of the home in an inefficient manner. To improve monitoring techniques, geostatistical analysis is used, particularly variography


to determine the spatial dependency of referenced observations and kriging, to interpolate across space for predicted values between observed locations. This information is used in the proposed framework to aid the temperature controller in making more informed decisions for user comfort.

The research discussed in this paper is an extension of previous work (Castello et al., 2009) accomplished for added context awareness in smart homes. The major contribution of this paper is added simulations of the proposed temperature control framework by adding temperature trends for multiple cycles of heating and cooling the independent areas being monitored. This was done by assuming a heating rate of 0.8°C/10 min and a cooling rate from data obtained from the Intel Berkeley Research Lab (Bodik et al., 2004). An extension of previous research on temperature control in general was also added to give a broader understanding, not just in application to smart homes.

The remainder of this research paper is organised as follows: Section 2 reviews previous research pertaining to smart homes, temperature control and context awareness techniques. Section 3 reviews geostatistical analysis methodologies being used in the proposed framework, Section 4 describes the proposed framework and analytical algorithm, Section 5 discusses the experimental dataset, Section 6 examines generated results and Section 7 concludes with final thoughts and future work.

2 Previous research

2.1 Smart homes

Past research in smart homes covers a wide range of topics including control systems, platform and architecture, middleware, security and context awareness. Research in Akhlaghinia et al. (2008) used Laboratory Virtual Instrument Engineering Workbench (LabVIEWTM) and Personal Digital Assistants (PDA) to control and monitor smart home services including elevator positioning, power, gas, light, water temperatures and ambient temperatures. A threshold is set by the user for ambient temperature using a PDA where temperature is monitored with a single sensor. Work accomplished in Hodá(2005) developed a cost-effective home automation system having the capability to adjust security, cooling, heating and lighting services for added comfort, energy conservation and user convenience. The SmartHome unit developed was based on the Rabbit 3000 microprocessor, particularly, the RMC3700 RabbitCore designed for internet applications. The temperature control of this unit utilised sensors with user-defined programs. Work in Baillie and Schatz (2006) described a lightweight, user-controlled smart home system geared towards the elderly living for longer periods of time at home. This was done by concentrating on three key topics: privacy, control of personal space and enjoyment within the home. The developed system was composed primarily of a multimodal robot, for use as a personal assistant, a mobile assistant (PDA or a smart phone) and an interactive TV, which were both utilised for control and monitoring. The multimodal robot has the ability to travel throughout the home to fulfil various tasks, such as measuring temperature and checking whether lights are on or off. Control of the multimodal robot is either through voice commands, the mobile assistant or TV. Cooling and heating temperatures can also be monitored or changed through the mobile assistant or TV. A recurring characteristic of research just mentioned is the lack of intelligent temperature control, where the majority base decisions primarily on a single data sample,


giving an inaccurate depiction of the environment in question. To improve system awareness, geostatistical methods are used in the proposed temperature control framework described in Section 4. To the best knowledge of the authors, there are no other temperature control schemes in smart home design using geostatistical analysis methods for improved spatial awareness.

2.2 Temperature control

Temperature control techniques in general will now be described from Moreira et al. (2004), Du et al. (2008a,b,c), Liceaga-Castro et al. (2006), Guo et al. (2009), Danmei et al. (2008) and Du et al. (2008). A predictive control model for air conditioning systems is proposed in Moreira et al. (2004) which is entitled mixed logical dynamical, based on the orthonormal series function (MLDOS). This approach is described by interacting physical laws, logic rules and operational requirements which are integrated using continuous and binary variables in the model as linear inequality constraints. Predictive control is obtained by optimising a performance index which is subject to model constraints. The novelty behind this technique is the use of an orthonormal series for modelling the process dynamics.

There has also been much research dealing with variable air volume (VAV) air conditioning control systems (Du et al., 2008a,b,c). Du et al. (2008a,b,c) focuses on using VAV in combination with the designing and connecting method of LonWorks intelligent control network. Advantages of using this system are added reference value and energy savings in practical applications with the use of variable frequency speed regulation technology. The effect of adjusting the air volume is observed using real-time acquisition data. Another research paper by Du et al. (2008a,b,c) also utilises VAV and LonWorks as the intelligent control network. A practical application is introduced where the control strategy and work principle of the system are analysed. A LonWorks network is then designed, programming using a project software tool for intelligent control networks is then described and integration of an intelligent control network by showing a common central integration platform is explained.

Intelligent control for air conditioning systems is described in Liceaga-Castro et al. (2006) and Guo et al. (2009). Liceaga-Castro et al. (2006) researches a multivariable control system for air condition processing. The generalised predictive control (GPC) scheme is utilised for the controller design. Individual channel design (ICD) is used to analyse the resulting stability, robustness and performance properties. The goal of this control system is to maintain temperature and humidity values within certain ranges. The Gunt Hamburg ET605 Recirculating Air Conditioning Trainer was used to test the proposed control system. Demonstration of excellent performance has been achieved through real-time results. Guo et al. (2009) studied temperature and humidity control in industrial workshops by introducing an intelligent control method which is based on improved fuzzy techniques. This methodology applies a partly serialising control face based on the basic fuzzy control algorithm. This results in rapid response and robustness to disturbances in the control system.

Fuzzy logic controllers are utilised in Danmei et al. (2008) and Du et al. (2008a,b,c) for air conditioning systems. Danmei et al. (2008) uses a fuzzy discrete event system (FDES) for energy-saving control in air conditioning control. Modelling for air condition


energy-saving control and effectiveness optimisation is accomplished with fuzzy automation modelling. Priorities were given to major control steps which are based on logical reasoning and forward-looking tree modelling to facilitate energy savings. Lastly, Du et al. (2008a,b,c) focuses on an improved control algorithm for air conditioning by utilising an expert fuzzy logic controller. The idea behind the control system is that boundaries of a membership function are decided by expert experience and then, a flexible control rule set is designed. Control precision and flexibility are ensured by using adjustable parameters for the Fuzzy-PD algorithm. Temperature control techniques previously reviewed lack intelligent predictive ability using spatial analysis techniques. The use of geostatistical analysis will allow for the prediction of temperature in non-monitored locations. The added data will aid the temperature controller in making more informed decisions for user comfort.

3 Geostatistical analysis

There have been many studies in environmental sciences that use geostatistical analysis to help to understand a substance or a parameter’s concentrations on a spatial plane. This type of analysis could provide valuable information for improved scientific understanding, risk assessment of contaminants or pollutants and decision support. Examples of such work include using geostatistics to assess mercury in soils around a coal-fired power plant in Baoji, China (Yang and Wang, 2008), analysing spatial correlation of nitrogen dioxide (NO2) concentrations in Milan, Italy (Dubois et al., 2002), studying spatial distribution of soil lead in the mining site of Silvermines, Ireland (McGrath et al., 2004) and development of a spectrophotometer in generating detailed soil maps for rice paddy fields using geostatistical analysis (Shibusawa, 2003).

The geostatistical methods used in this paper will now be described which include variography and kriging, briefly summarised from Trauth (2006). The theoretical basis of both techniques can be found in greater detail in Olea (1999), Isaaks and Srivastava (1989) and Matheron (1963).

3.1 Classical variogram

In geostatistical analysis, the variogram is used to describe the spatial dependency between referenced observations within the analysed plane where the true variogram is generally unknown. Therefore, an estimation of the variogram can be calculated through known observations. This is accomplished by first determining the experimental variogram, given by the semivariance:

( ) 0.5 x x hh z z (1)

where zx is the observed value at point x and zx + h is an observed value at another point within distance h. (h) is also known as a semivariogram or variogram and the distance between observations is known as lag distance.

The next step is to summarise the experimental variogram with a variogram estimator, which determines the central tendency and is similar to descriptive statistics derived from univariate observations (Trauth, 2006). The variogram estimator is


calculated by Equation (2) where N(h) represents the amount of pairs within lag interval h.

( )2

1

1( )2 ( )

N h

E xi xi hi

h z zN h

(2)

Lastly, a variogram model or parametric curve is fitted to the variogram estimator. This is similar to frequency distribution fitting where the frequency distribution is modelled by a distribution type and its parameters (Olea, 1999). The three commonly used models in variography are: spherical, exponential and linear.

3.2 Ordinary point kriging

To interpolate observations on a regular grid, ordinary point kriging is used, which is considered the most commonly utilised method. Ordinary point kriging uses weighted averages and neighbouring observations to predict unobserved points:

0ˆN

x i xii

z z (3)

where i are estimated weight values. To guarantee that estimates are unbiased, the sum should equal to one.

1N

ii

(4)

Equation (5) shows the average estimation error must equal zero where zx0 is the unknown value.

0 0ˆ 0x xE z z (5)

Calculating the mean-square error using Equations (3)–(5) in terms of the variogram and using a Lagrange multiplier v for optimisation yields a linear kriging system of N + 1 equations and N + 1 unknowns which is calculated by:

01

, ,N

i i j ii

x x v x x (6)

where i is the weight for the ith data point, (xi,xj) is the variogram between data points xi and xj and (xi,x0)is the variogram between the data point and unobserved point. The kriging variance is calculated by:

20 0 0

1

,N

i ii

x x x v x (7)


4 Temperature control framework

The temperature control framework will now be described, which is shown in Figure 1. The process starts with data sampling throughout the WSN nodes at a user-defined rate (e.g. 1 min). This data must first undergo an initial training sequence where the appropriate distribution model is determined for the variogram model. It is assumed the selected distribution model along with sill and range values are acceptable for subsequent sets of data samples being analysed. Once the variogram model has been determined, kriging analysis is executed in the spatial analysis step to interpolate observations for the monitored area. Results from kriging are then divided into user-defined sub-regions where each sub-region must possess a heating or cooling element with the ability to be controlled separately from other elements in the system. The mean is then calculated for each sub-region. The calculated mean for each is then compared with user-defined thresholds. If the calculated mean is less than the lower threshold, Tlow,or greater than the upper threshold, Thigh, a signal is sent to the heating or cooling system, respectively, to engage the sub-region in question. This process minus the initial training step repeats at a user-defined basis (e.g. 1 min). If the calculated mean is between the temperature thresholds, Tlow and Thigh, the heating or cooling system is disengaged. A user-defined time delay is then put into effect. Pseudo-code of the proposed temperature control framework is shown in Figure 2.

Figure 1 Temperature control framework


Figure 2 Framework algorithm (see online version for colours)

5 Experimental dataset

The experimental dataset i.e. used to test the proposed temperature control framework for smart homes is from Intel Berkeley Research Lab (Bodik et al., 2004). Collected data consists of temperature, humidity, light and voltage for 54 sensor nodes in a WSN. Samples were collected every 31 sec between 28 February and 5 April 2004, totalling 2.3 million readings. Locations are also given for each node relative to the upper right corner of the lab, which is shown in Figure 3. Temperature data will be focused upon in this research paper.


Figure 3 Intel Berkeley Research Lab [6] (see online version for colours)

6 Results

The first dataset used to analyse the proposed framework consists of a data sample from each of the 54 sensors minus samples from nodes 5 and 20 due to irregular readings. Samples were taken approximately at 01:10:00 am on 28 February 2004. Due to the use of linear kriging techniques, temperature samples should be Gaussian in nature, which is confirmed in the histogram shown in Figure 4.

The experimental variogram and lag values are then calculated for the initial dataset and plotted in Figure 5. Analysis of the variogram cloud conveys the dispersion of variogram values at different lags. A variogram model is fitted to the variogram estimator in Figure 6 where linear, exponential and spherical models are fitted. By visual inspection, the spherical model was chosen for kriging analysis where the sill is set to 0.73 and range to 13. Kriging results for the initial dataset from Intel Berkeley Research Lab are shown in Figure 7 where predictions are calculated for locations without sensor nodes. Figure 8 displays the sensor node locations along with variance values for the entire grid. It is shown that variance tends to be lower or near sensor node locations rather than predicted locations. This demonstrates variance being a measure of information density (Wackernagel, 2003). It should be noted the nugget affect was not taken into consideration during this study. Temperature trends are plotted in Figure 9 for data from Intel Berkeley Research Lab every 10 min from 01:10:00 am to 04:20:00 am on 4 February 2004 minus samples from nodes 5 and 20. The mean and standard deviation values were taken from the initial dataset analysis where – = 17.5749 is defined as Tlow, and + = 19.2633 is defined as Thigh. and values were used to define Tlow and Thigh values in this study, although other user-defined amounts are acceptable.


Figure 4 Histogram of initial dataset (see online version for colours)

Figure 5 Variogram cloud of initial dataset (see online version for colours)


Figure 6 Variogram model of initial dataset (see online version for colours)

Figure 7 Kriging results of initial dataset (see online version for colours)


Figure 8 Variance results of initial dataset (see online version for colours)

Figure 9 Sub-regional temperature trends without control (see online version for colours)


Temperature trends for each sub-region, defined in Figure 7, are shown in the graph. Following the temperature control framework of Figure 1, the binary control variable eh4,shown in the algorithm of Figure 2, would change from a ‘0’ to ‘1’ value to produce heat for sub-region 4 at 01:20:00 am. The same would occur for sub-region 1 at 02:40:00 am, sub-regions 2 and 8 at 02:50:00 am, sub-regions 6 and 5 at 03:00:00 am and sub-regions 3 and 7 at 03:50:00 am. Heating would continue for each sub-region until the averaged temperature reaches = 18.4191 which is shown in Figure 10. Data from Intel Berkeley Research Lab was utilised to the point of passing – . Afterwards, the heating mechanism for each sub-region was assumed to follow an increase in temperature with a slope of 0.8°C/10 min. Once the temperature reaches above the value, heating is disengaged and the decrease in temperature is assumed to follow the original trend from Intel Berkeley Research Lab’s data. Sampling occurs at a user-defined rate of 10 min where the temperature control framework is simulated from 01:10:00 am to 07:40:00 am.

Figure 10 Sub-regional temperature trends with control (see online version for colours)

7 Conclusion

A novel temperature framework for smart homes has been designed and developed using geostatistical analysis tools, particularly variography and kriging. This allows the monitored area to be described in greater detail for improved temperature control, thereby improving user comfort. This work adds to knowledge in context awareness modelling for smart home services. To the best knowledge of the authors, there are no other temperature control schemes in smart home design using geostatistical analysis methods for improved spatial awareness.


An important observation in the discussed temperature framework that should be mentioned involves the user-defined time for sampling. This is an extremely important parameter because if set too high, temperature can travel from below Tlow to above Thighvalues or vice versa in less than a system cycle. If the sampling time is set too low, unnecessary analysis and system resources are wasted. Although this issue is beyond the scope of this paper, further analysis would be beneficial in determining the most optimal values for sampling times.

Future work involving the discussed temperature control framework includes further research and analysis of user-defined sampling rates. Other research efforts can be invested in the possible affect on power consumption and accuracy analysis of kriging methods. Log-normal distributions using log-normal kriging can also be studied, which is not uncommon in spatial analysis. Lastly, further research is possible in improving context awareness for various aspects in smart homes using spatial analysis besides temperature control.

References Akhlaghinia, M.J., Lotfi, A., Langensiepen, C. and Sherkat, N. (2008) ‘Occupant behaviour

prediction in ambient intelligence computing environment’, Journal of Uncertain Systems,Vol. 2, No. 2, pp.85–100.

Baek, S-M., Chang, K-B., Shim, I-J. and Park, G-T. (2004) ‘Implementation of smart home control using LabVIEW and PDA’, IEEE International Symposium on Consumer Electronics, 1–3 September, pp.558–562.

Baillie, L. and Schatz, R. (2006) ‘A lightweight, user-controlled system for the home’, Human Technology Journal, Vol. 2, No. 1, pp.84–102.

Bodik, P., Hong, W., Guestrin, C., Madden, S., Paskin, M. and Thibaux, R. (2004) ‘Intel lab data’, Available at: http: db.csail.mit.edu/labdata/labdata.html.

Castello, C.C., Fan, J., Davari, A. and Chen, R. (2009) ‘Temperature control framework using wireless sensor networks and geostatistical analysis for total spatial awareness’, IEEE 10th International Symposium on Pervasive Systems, Algorithms and Networks (I-SPAN09), Kaohsiung, Taiwan, 14–16 December, pp.717–721.

Danmei, L., Weichun, L., Hui, Z. and Shihuangi, S. (2008) ‘A fuzzy discrete event system for air conditioning system energy-saving control and optimization’, The 2nd International Conference on Genetic and Evolutionary Computing.

Du, M., Fan, T. and Huang, N. (2008a) ‘Research and application for VAV air conditioning system with LonWorks fieldbus intelligent network’, The 2nd International Symposium on Intelligent Information Technology Application.

Du, M., Fan, T., Su, W. and Huang, N. (2008b) ‘Design of VAV energy saving air conditioning system based on intelligent control network’, IEEE Pacific-Asia Workshop on Computational Intelligence and Industrial Application.

Du, M., Fan, T., Su, W. and Li, H. (2008c) ‘Design of a new practical expert fuzzy controller in central air conditioning control system’, IEEE Pacific-Asia Workshop on Computational Intelligence and Industrial Application.

Dubois, G., Saisana, M., Chaloulakou, A. and Spyrellis, N. (2002) ‘Spatial correlation analysis of nitrogen dioxide concentrations in the area of Milan, Italy’, Proceedings of the 1st Biennial Meeting of the International Modeling and Software Society, June, pp.176–183.

Guo, Y., Cao, D. and Zheng, G. (2009) ‘Application of intelligent control techniques for temperature-humidity control in industrial workshop’, The 2nd International Symposium on Intelligent Information Technology and Security Informatics.


Hodá , I.F. (2005) ‘SmartHome, cost-effective home automation system’, Available at: http: www.feec.vutbr.cz/EEICT/2005/sbornik/03-Doktorske_projekty/03-Kybernetika_a_ automatizace/01-xhodan00.pdf.

Isaaks, E.H. and Srivastava, R.M. (1989) An Introduction to Applied Geostatistics. New York: Oxford University Press.

Jiang, L., Liu, D-Y. and Yang, B. (2004) ‘Smart home research’, Proceedings of the 3rd International Conference on Machine Learning and Cybernetics, Shanghai, China, 26–29 August.

King, N. (2003) ‘Smart home – a definition’, Available at: http: www.housingcare.org/downloads/ kbase/2545.pdf.

Liceaga-Castro, J., Ramirez-Espana, C. and Liceaga-Castro, E. (2006) ‘GPC control design for a temperature and humidity prototype using ICD analysis’, Proceedings of the 2006 IEEE International Conference on Control Applications, October.

Matheron, G. (1963) ‘Principles of geostatistics’, Journal of Economic Geology, Vol. 58, pp.1246–1266.

McGrath, D., Zhang, C. and Carton, O.T. (2004) ‘Geostatistical analyses and hazard assessment on soil lead in Silvermines area, Ireland’, Journal of Environmental Pollution, Vol. 127, pp.239–248.

Moreira, V.D., do Amaral, W.C., Augusto, P. and Ferreira, V. (2004) ‘A new approach for robust model predictive control with an application to an air-conditioning system’, Proceedings of the IEEE International Conference on Control Applications, Taipei, Taiwan, September.

Olea, R.A. (1999) Geostatistics for Engineers and Earth Scientists. London, UK: Kluwer. Pounds-Cornish, A. and Holmes, A. (2002) ‘The iDorm – a practical deployment of grid

technology’, Proceedings of the 2nd IEEE/ACM International Symposium on Cluster Computing and the Grid (CCGRID’02), Berlin, Germany, 21–24 May.

Shibusawa, S. (2003) ‘On-line real time soil sensor’, Proceedings of IEEE/ASME International Conference on Advanced Intelligent Mechatronics, pp.1061–1066.

Trauth, M.H. (2006) MATLAB® Recipes for Earth Scientists. Heidelberg, Germany: Springer-Verlag.

Wackernagel, H. (2003) Multivariate Geostatistics: an Introduction with Applications (3rd ed.). Berlin: Springer-Verlag.

Yang, X. and Wang, L. (2008) ‘Spatial analysis and hazard assessment of mercury in soil around the coal-fired power plant: a case study from the city of Baoji, China’, Journal of Environmental Geology, Vol. 53, pp.1381–1388.

context aware wireless sensor networks for smart home ... · pdf filebiographical notes: ......

Documents