Designing a smart home environment using a wireless sensor networking of everyday objects

Transcription

1 Designing a smart home environment using a wireless sensor networking of everyday objects LAGUIONIE Olivier November 27, 2008 Master s Thesis in Computing Science, 30 ECTS credits Supervisor at CS-UmU: SURIE Dipak Examiner: Per Lindström Umeå University Department of Computing Science SE UMEÅ SWEDEN

2

3 Abstract This thesis report describes the development of a smart home environment based on an accurate wireless sensors network using ZigBee[26] communication protocol. In order to recognize the states/state changes of everyday objects, we used 81 sensors to sense the 42 everyday objects we selected in a realistic home environment. Knowing the states of the everyday objects is essential to perform further advanced computation for instance activity recognition. The promising results we obtained consequently provide a good base for such developments. The system has been evaluated in a realistic home environment with background noise (ambient light variations, WI-FI network with same frequency, etc). The sensing nodes have shown some interesting results with a precision value of 91.2% and a recall value of 98.8% concerning the recognition of the everyday objects state changes. The experimentation has been accomplished by four individual users during one week. They were all wearing the same equipment for collecting data: a wearable camera and digital video recorder in order to obtain the ground truth, and an ultra portable laptop (the personal server) connected to the receiving node. The ground truth obtained from the video recording is compared with the state changes information recorded in the personal server to evaluate the performance of the system.

4 ii

5 Contents 1 Introduction Outline Presentation Problem Description Goals Related works Communication protocols Infrared WI-FI Bluetooth ZigBee Existing wireless sensor networking platforms MITes Berkeley motes The BTnodes Intel mote The U Smart-its Conclusion Conceptual design Everyday objects in a home environment Sensor nodes Everyday object: fridge Everyday object: freezer Everyday object: microwave oven Everyday objects: ovens and stove Everyday object: cupboards and milk box Everyday object: coffee machine iii

6 iv CONTENTS Everyday object: dustbin Everyday object: drawer Everyday object: dish washer Everyday object: lamps Everyday object: soap provider and flush Possible future scenarios System implementation Requirements for the wireless sensor network Hardware implementation Description of the board The main components Sensors Operating mode Software implementation The class board Classes board 1 to board Global functions Evaluation The sensor network Scenario of evaluation The experimental setup Energy consumption Range Accuracy of the system Discussion and limitations 45 8 Future works 47 9 Acknowledgments 49 References 51 A Appendix 55 A.1 Maxstream XBee s data sheet A.2 Atmal ATMEGA88 data sheet s front page A.3 TAOS light to voltage converter data sheet s front page A.4 Dallas digital temperature sensor TO92 data sheet s front page A.5 Omron microswitch data sheet s front page A.6 Microchip low drop voltage regulator data sheet s front page

7 List of Figures 2.1 Integration of the sensing board The first version of the MITes An example of a berkeley mote(the sensing part) The BTnode The intel mote The U 3 node The 2 modules of the Smart-its Light and temperature sensors in the fridge Light and temperature sensors in the freezer The sensors of the microwave Switches around the knob Sensors on the stove/oven panel A light sensor in a cupboard (a) and in the milk box (b) Press buttons on the coffee machine (a) and light sensor in the coffee container (b) Light sensor in the dustbin Light sensor in a drawer Light sensors on the dish washer panel Light sensor on a lamp Press button on a soap provider (a) and on the toilet s flush (b) Photo of the sensing board Description of the sensing board The XBee transceiver Antenna connector (a) and antenna (b) Microcontroller The voltage regulator (a) and on the battery (b) The light sensor (a) and the temperature sensor (b) The switch (a) and the press button (b) Example of a data frame v

8 vi LIST OF FIGURES 5.10 The classes diagram A live-in laboratory home environment with nodes placement and signal strength measures A user wearing our system during the experimentation session A.1 XBee datasheet page A.2 XBee datasheet page A.3 Microcontroller datasheet page A.4 Light sensor datasheet page A.5 Temperature sensor datasheet page A.6 Switch datasheet page A.7 Voltage regulator datasheet page

9 List of Tables 3.1 Comparative statement of different communication protocols List of selected everyday objects Evaluation of signal strength at 8 different locations vii

10 viii LIST OF TABLES

11 Chapter 1 Introduction 1.1 Outline This report is composed of the following parts: Chapter 1 is a short introduction to this report, describing the goal of this project. Chapter 2 explains in detail the problems addressed in this thesis. It also summarizes all the important reflections of the undertaken work and draws the conclusions concerning the development of the system. Chapter 3 references several related works which have been studied during this thesis. Chapter 4 describes in detail the conceptual design of the system. Chapter 5 explains the implementation of both the hardware and the software application. Chapter 6 concerns the evaluation of the system. Chapter 7 is a discussion about the results of the evaluation of the system and details its limitations. Chapter 8 makes a list of some possible future works on the system. Chapter 9 is the acknowledgments part. Appendix: some extracts of the important data sheets. 1

12 2 Chapter 1. Introduction 1.2 Presentation This report describes the development of a wireless sensor network[4] of everyday objects in a smart home environment[30]. It will present all the different aspects of this project, from the related works to the final evaluation. This thesis is related to the field of ubiquitous computing. Ubiquitous computing (computing everywhere) is a vision in using technology (wireless networking, sensing, actuation, computation, etc) in everyday environments like home, airports, etc. Ubiquitous computing technologies can be used for situation awareness[35], context awareness [2], activity recognition[34], etc. The system to be described in this thesis could be regarded as a human-computer interface (implicit interface) between a user and his/her home environment based on wireless sensor networking of everyday objects[28]. There are many related works that exist concerning the design of smart environments using wireless sensor networks[29][31][3]. However, considering the special requirements for this research work in obtaining the states/state changes of everyday objects, the optimal sensors required, the modularity and upgradeability, we have chosen to develop our own sensing and communication module. We have used ZigBee communication considering its low power consumption requirements. The choice of the ZigBee communication protocol can appear to be pretty obvious nowadays for the development of smart objects especially due to the low power consumption of these transceivers. When we started to discuss the goals of this project two years ago, the ZigBee technology was not that widespread compared to other wireless communication protocol like for instance Bluetooth[27]. However, considering its efficiency and the promise shown by this technology, we have decided to use ZigBeeZigBee[5]. As it can be seen from this thesis, ZigBee communication can be used efficiently for smart home environments. The use of simple and cheap sensors has also played an important role concerning our decision in developing our own sensor module. Many wireless sensor projects use complex sensors like camera, microphone which need higher level computation for feature extraction. We preferred to use simple sensors like switches or light sensor, which are cheap, easy to implement, available off-the-shelf, resistant to hostile environmental conditions and can be used in many different situations. Furthermore, we also decided to make our sensing board as much generalized as possible. As mentioned earlier, the wireless sensor network described in this thesis is part of a larger system[8]. Other modules of the larger system will be added in the near future using the data collected by the wireless sensor network to enable activity recognition[6][9], situation awareness, interaction management, etc. Consequently, the choice of all the components of the electronic board representing the sensing node have been very important. This board had to be modular and easily upgradeable to make it possible to use other kind of sensors (RFID reader[32], accelerometer, etc). That is why all the components from the microcontroller to the ZigBee transceiver and the way the sensors are connected to the sensing node have been throughly investigated. Such a generalized board can be also be used for other applications beyond the scope of this thesis. The size of this board can be minimized by removing some sensor connectors depending on the application requirements. One such application requirement is to attach accelerometers to the sensor module presented in this thesis in a wearable form

13 1.2. Presentation 3 (for the left-hand and right-hand) for gesture recognition[1] and gesture based interaction. However the sensor board presented in this thesis is a first version of the prototype used mainly for activity recognition. Wireless sensor networking could be used for elderly care in a home environment. For instance in such a home environment, an efficient activity recognition system could assist the elderly person in performing his/her daily activities in an appropriate manner without forgetting important actions.

14 4 Chapter 1. Introduction

15 Chapter 2 Problem Description Many related works can be found concerning the development of a smart environment or a wireless sensor networking. A literature survey is presented to compare the different technologies used, the components used, etc., to understand the advantages and limitations of the existing works and to see the improvements that are required for our current design. As mentioned earlier, this thesis aims at developing an efficient hardware prototype for sensing the state changes of everyday objects in a home environment as it is shown on this sketch 2.1. The sensing board should resist certain hostile environmental conditions in order to be efficient, robust and scalable. The prototype to be developed should be easy for future developers to be able to modify the board and to extend it. The sensing modules should allow for background noise, real environment constraints, etc, while providing state/state changes information with reliable accuracy. Cost is an important factor to consider and the sensors to be used should be cheap, durable and easy to use. Some sensors have to be set up inside the fridge, inside the freezer, on a microwave oven, on a stove, on a dish washer, etc., which produce some perturbation. Figure 2.1: Integration of the sensing board 5

16 6 Chapter 2. Problem Description It is also important to investigate the environment in which the system is supposed to be deployed. This includes investigating on the objects present in the environment and their states/state changes based on the user s interaction with them while performing everyday activities and to analyse the states/state changes that might be useful for sensing. All the objects do not have the same possible states/state changes and hence cannot be sensed in the same way. This investigation is important to decide the objects to be sensed, their states, and the sensors that are required for sensing thse states. A market survey is also required to investigate on the available sensors, their cost, technology used, compatibility, etc.. The data that is received from the sensor network is collected in a personal server currently under development. The personal server runs an activity centric middleware for facilitating personal activity support in smart environments[8]. The operating system and programming language were chosen according to the projects standards (Microsoft Windows XP and C++). The middleware module developed as part of this thesis should be modular so that other developers can modify and extend it easily. It also has to be stable and efficient. 2.1 Goals 1. To develop an efficient sensing module that provides an accurate state/state change information concerning everyday objects. Such information is important to develop an efficient activity recognizer. 2. To select the appropriate objects in a home environment based on studying the users activity patterns in a home environment. 3. To select the states that are interesting and important for knowing the activity performed by the user. 4. To select the appropriate sensors available in the market for sensing the objects states. 5. To build a complex sensing mechanism for states that are not directly senseable with the existing sensors. 6. To basically build the sensing module such that it is easy to modify or upgrade it. 7. To design the system in a power efficient manner since the system is supposed to be deployed for everyday use in a home environment. 8. Comparison of different communication protocols and to choose an appropriate one.

17 2.1. Goals 7 9. To design the sensing module such that it is not confined to any specific application and has a broader usage. 10. To deploy the system built as part of the thesis in a living laboratory home environment for collecting data, evaluating the system and to analyse the results obtained.

18 8 Chapter 2. Problem Description

19 Chapter 3 Related works 3.1 Communication protocols Concerning the communication protocols, several possibilities are available. Each of them have some advantages and disadvantages. To choose the most appropriate one, the pros and cons have to be weighted for each one. The different protocols are compared considering the application requirements including data transmission rate, range, battery life and cost, an it is presented in this comparative table Infrared The infrared light has a wide bandwidth and a high rate transmission (10 Mb/s), but the range is short (30 meters). Finally, the infrared technology is relatively old, consequently the infrared transceivers are cheap compared to the other technologies. The infrared transmissions should also be intense in order to not be confused with other light sources like television, window, light bulbs, etc. Finally the infrared technology is a well established technology resulting in infrared transceivers which are cheap compared to other technologies. Considering the risk of light confusion, this communication protocol has not been retained. Advantages: High rate transmission Cheap components Disadvantages: Short range Light confusion WI-FI WI-FI is certainly the technology which has the highest data transmission rate with a range of 300 meters with the b norm. The data transmission rate differs according to the norms, but concerning the b which is the most standard one today, it can reach 11 Mb/s. However the battery life of a WI-FI transceiver is very short ranging from half a day to five days and does not suit our requirements. 9

20 10 Chapter 3. Related works Advantages: High rate transmission Very important range Disadvantages: Short battery life Interaction risk Bluetooth Bluetooth is a short range communication technology (< 10 meters), but this range can easily be extended with a power booster. This communication protocol is considered as low power consumption, but the battery life of a Bluetooth transceiver is not much longer than the battery life of WI-FI transceiver (1 day to 7 days), and is not according to our requirements. The data transmission rate is lesser that of WI-FI, but it is still very high for our application (720 Kb/s). Advantages: Sufficient range High data rate Disadvantages: Short life battery ZigBee ZigBee is a new standard uniquely designed for low rate wireless personal area networks. It targets low data rate, low power consumption and low cost wireless networking. The range of the transceivers reaches 100 meters and can be risen until more than 1000 meters with a new kind of transceivers. The data rate is certainly the smallest of this comparative statement, as it reaches only 250 Kb/s. However, the battery life is much bigger than the other technologies. A ZigBee transceiver can work from 100 to days, which is the highest battery life of this comparative statement. Advantages: High range Very low power consumption Sufficient data rate? Low cost Disadvantages: Sufficient data rate?

21 3.2. Existing wireless sensor networking platforms 11 Communication protocols Advantages Disadvantages Infrared High rate transmission Short range Cheap components Light confusion WI-FI High rate transmission Short battery life Very important range Interaction risk Bluetooth Sufficient range Short life battery High data rate High range Sufficient data rate? ZigBee Very low power consumption Sufficient data rate? Low cost Table 3.1: Comparative statement of different communication protocols 3.2 Existing wireless sensor networking platforms MITes MITes: MIT Environmental Sensors, is a portable wireless sensor platform that can be used to collect data for activity recognition in non-laboratory settings such as homes[33]. A photography of the first version of these sensor modules can be found on this figure 3.1. The MITes platform includes six environmental sensor types, five of them being among the most typically needed in ubiquitous and pervasive computing applications: movement using ball, mercury, and reed switches movement tuned for object-usage detection (using acceleration) light temperature proximity current consumption. The MITes platform also includes five wearable sensors: accelerometers to acquire body motion information heart rate ultra violet radiation exposure an RFID reader in a wristband form factor location beacons. All of these sensors can be used simultaneously, and a single receiver acquires the data, which is sent to a PC or mobile computing device for real-time processing.

22 12 Chapter 3. Related works Figure 3.1: The first version of the MITes Berkeley motes The Berkeley motes are small sensors devices developed at UC Berkeley, which use a radio frequency transmission[13]. The motes contain a microcontroller, a radio frequency transceiver and many other components. All the sensors are placed on a separate board. The following figure 3.2 is a photography of the sensing board. The sensor boards can be composed of different sensors considering the need and the function of the wireless sensing device. Some examples of sensors include light sensor, temperature sensor, accelerometer, magnetometer and humidity sensor. The UC Berkeley also developed TinyOS, an operating system which is especially suited to run on the sensor devices. Figure 3.2: An example of a berkeley mote(the sensing part)

23 3.2. Existing wireless sensor networking platforms The BTnodes The BTnode is an autonomous wireless communication and computing platform based on a Bluetooth radio and a microcontroller. The BTnode has been jointly developed at ETH Zurich by the Computer Engineering and Networks Laboratory (TIK) and the Research Group for Distributed Systems[14]. The figure 3.3 is a photography of a BTnode. Figure 3.3: The BTnode Intel mote The Intel mote is built around an integrated microcontroller which is composed of an ARM 7 core, a Bluetooth radio transceiver, RAM, Flash memory and several I/O options[31]. The Intel motes use a tree topology in which the root node is connected to a pc for the data collection. The motes are connected to an accelerometer board as you can notice in figure 3.4. Figure 3.4: The intel mote

24 14 Chapter 3. Related works The U 3 The U 3 is a sensor network that has been designed by the university of Tokyo, Japan[29]. The sensor node is represented by a 50mm cube that contains 4 separate modules. Each module consists in an independent functional board. The 4 different board of each node are placed on top of each other and interconnected inside a cube. A power module, a CPU module, a RF communication module and a sensor module are composing each module. The power module consists in 3 AAA rechargeable batteries. The CPU module is based on a PIC18F452. The RF communication module is a 300 MHz RF transceiver, and the range transmission reaches 30m. The sensor module is composed of a temperature sensor, a brightness sensor a,d a motion sensor. The concept of this sensor node is very interesting especially concerning the design. The fact of having all the different modules totally independent makes it easily modular. However, this concept makes the sensor node pretty big, and each module should not exceed a done sized to not increase the size of the whole module. Besides, the fact of having all the sensors plugged on the sensor module was not suitable for our application. A photography of a U 3 node can be found on the figure 3.5. Figure 3.5: The U 3 node

25 3.3. Conclusion Smart-its The Smart-its are small devices which are composed of two hardware modules[3]. The first one is the communication module which contains a microcontroller PIC 16F87x and a RFM transceiver TR 1001 at MHz. The second module is the sensing board. Both modules are interconnected by an I2C data bus and a power bus. These modules are shown on the figure 3.6. One or several sensing boards can be connected to the communication board. The sensing module is composed of a microphone, a light sensor, an accelerometer, a pressure sensor and a temperature sensor. Figure 3.6: The 2 modules of the Smart-its 3.3 Conclusion There have been many more related works including IRIS[25], MICAz[23], Imote2[24], etc. Even though many of such systems have their own advantages, they do not meet our requirements in sensing the state changes to everyday objects including home appliances, furniture, simple objects, walls, etc, in a home environment.

26 16 Chapter 3. Related works

27 Chapter 4 Conceptual design 4.1 Everyday objects in a home environment A typical home environment is a complex one with many everyday objects that an user can interact with while performing some activities. To recognize the activity of a person in such an environment, the system needs to know what object is currently used but more precisely its state/state change information[6]. The states of objects are of two types. We make a difference between the internal states and the external states. Internal states for instance is a fridge that is open or closed while external states represent the relationship of an object with reference to other objects. For instance a fridge might contain other objects like milk packet, vegetables packet, etc. The work to be presented in this thesis will deal with the internal state/state changes alone. External state changes are left for future work. However the wireless sensor networking module developed as part of this thesis could be used for sensing the external state changes as well. For instance a RFID reader could be easily integrated to the system presented in this thesis to sense the external states/ state changes of everyday objects. Many internal states are common to several objects, but some are specific for individual objects. To know if an object is open or closed; switched on or off, are common states. However to sense these states can be different depending on the object. For instance, to know if a coffee-maker is switched on can be sensed using a light sensor which detects the feedback light signals from the LED integrated onto the coffee-maker. Instead a vibration sensor can be used to detect that the blender is switched on. Furthermore some state like for instance the temperature, the timer or the power specifications are specific to individual devices. Therefore each individual object that needs to be sensed has to be studied in detail in order to choose the appropriate states and the best way to sense them. Refer to table 4.1 for a list of selected objects with their state information implemented as part of this thesis. Note that V represents the state that are sensed and X the state that are not sensed. 4.2 Sensor nodes The overall goal for this thesis is to develop sensor and communication nodes for individual objects such that they communicate wirelessly with the user s personal server 17

28 Table 4.1: List of selected everyday objects Object Open/closed On/off Temperature Full/not full Pressed/released Fridge V X V X X Freezer V X V X X Microwave oven V V X X V Up oven X V X X V Down oven X V X X V Stove X V X X V Cupboards V X X X X Coffee machine V X X V V Lamps X V X X X Dustbins X X X V X Drawers V X X X X Dish washer X V X X V Exhaust fan X V X X X Milk box V X X X X Toilet s flush X X X X V Soap provider X X X X V 18 Chapter 4. Conceptual design

29 4.2. Sensor nodes 19 and provide information related to objects states/state changes. However to reduce the amount of boards which are required we have chosen to use common sensor boards for several objects. Our system is composed of 10 boards. Mobility is an important requirement since the user normally is mobile when interacting with everyday objects hence we have decided to go with wireless approach instead of a wired approach. The sensing board is the same for all the objects that includes a microcontroller, a transceiver and a few other electronic components. However the sensors that are embedded in different objects are different. A detailed account of the sensors used for sensing the states/state changes to everyday objects is provided below Everyday object: fridge State type - open/closed: light sensor Several sensors can be used to detect if the fridge is open or closed. The simplest sensors which can be used are certainly the switch and the light sensor. The light sensor has the advantage that there is no mechanical interaction, so it has a longer average lifetime. By placing the light sensor inside the fridge it is easy to detect if the door is open or closed as all the fridges have a light inside. Of course, there can be a problem if the light breaks down. The photography 4.1 is an example of the integration of the sensor in the fridge. State type - temperature: temperature sensor To sense the temperature, there is no other choice than to use a temperature sensor. However this sensor has to be used in a low temperature environment (around 0 C). Figure 4.1: Light and temperature sensors in the fridge Everyday object: freezer State type - open/closed: light sensor This is similar to the fridge but in this case the sensor has to resist very low temperature (from -6 C to -30 C following the quality of the freezer). Another problem is that in some freezer there is no light inside when the door is open, but

30 20 Chapter 4. Conceptual design this can be probably solved by placing the sensor in such a position that it can detect the outside light (for instance on the door). State type - temperature: temperature sensor Similar to the fridge but the sensor has to be able to work in a very low temperature environment. The photography 4.2 shows an example of the integration of the sensor in the freezer. Figure 4.2: Light and temperature sensors in the freezer Everyday object: microwave oven State type - open/closed and on/off: light sensor We placed a light sensor on the glass door of the microwave. This light sensor can either detect if this door is open by receiving the ambient light or it can also detect if the microwave is switched on by receiving the inside light of the microwave. An example of the integration of the sensors can be found in figure 4.3. State type - start,stop and defrost: press buttons A press button has been placed over the start button of the microwave. This allows our system to detect when the user is pressing the start button. Stop and defrost buttons are sensed in a similar way Everyday objects: ovens and stove State type - on/off: light sensor Some light sensors have been placed over the LEDs to detect when the devices are switched on. State type - knobs: switches Some switches have been placed all around the knobs, at each of their different positions in order to detect the state of each of them. The sketch 4.4 presents this way of sensing. It has been used for the experimentation as you can see on the figure 4.5.

31 4.2. Sensor nodes 21 (a) (b) Figure 4.3: The sensors of the microwave Figure 4.4: Switches around the knob (a) (b) Figure 4.5: Sensors on the stove/oven panel

32 22 Chapter 4. Conceptual design Everyday object: cupboards and milk box State type - open/closed: light sensor The open/closed state of the cupboard is quite easy to sense with a switch or a light sensor. The light sensor maybe the most efficient as the door of the cupboard is not obliged to be totally closed to be sensed as closed. A light sensor has also been used for the milk box as you can notice on the figure 4.6. (a) (b) Figure 4.6: A light sensor in a cupboard (a) and in the milk box (b) Everyday object: coffee machine State type - open/closed: light sensor The state open/closed of the coffee machine is sensed by a light sensor placed inside the coffee machine to detect if this one is opened or closed. State type - full/not full: light sensor A light sensor has been placed in the coffee container of the coffee machine to detect if this one needs more coffee. State type - selection buttons: press buttons Some press buttons have been placed on the coffee machine to detect which selection the user is doing as you can see on the figure Everyday object: dustbin State type - full/not full: light sensor A light sensor has been placed in the dustbin to detect if it is full or not as it can be seen on the photography 4.8. Indeed, if the dustbin is full then the light sensor is hidden from the ambient light.

33 4.2. Sensor nodes 23 (a) (b) Figure 4.7: Press buttons on the coffee machine (a) and light sensor in the coffee container (b) Figure 4.8: Light sensor in the dustbin Everyday object: drawer State type - open/closed: light sensor The situation of the drawer is very similar to the cupboard. A switch can also be used to detect its state. However the use of a light sensor appeared to be simpler, and has been used as it can be seen in figure 4.9.

34 24 Chapter 4. Conceptual design Figure 4.9: Light sensor in a drawer Everyday object: dish washer State type - on/off: light sensor A light sensor has been placed on the LED corresponding to the on/off state of the dish washer. Another way could be to place an accelerometer on the door of the dish washer to detect the vibrations generated during its functioning. State type - program selection: light sensors Some light sensors have been place on the LEDs corresponding to the different programs of the dish washer that the user can select, as it can be seen on the photography Figure 4.10: Light sensors on the dish washer panel Everyday object: lamps State type - on/off: light sensor A light sensor seems to be the easiest way to sense the state on/off of a lamp by placing it very close to the bulb, as it can be seen on the photography However it has to be isolated from the ambient light Everyday object: soap provider and flush State type - pressed/released: press button Several possibilities can be found to sense this state. A touch sensor, a press button or other sensors can suit for that.

35 4.2. Sensor nodes 25 Figure 4.11: Light sensor on a lamp The press button seems to be the simplest and cheapest way to sense it. A press button has also been used for the toilet s flush as it is shown on the photography (a) (b) Figure 4.12: Press button on a soap provider (a) and on the toilet s flush (b)

36 26 Chapter 4. Conceptual design 4.3 Possible future scenarios Several future scenarios can be imagined concerning the application of our system. Knowing the state of the objects provide some very useful information for higher level applications. First, these data can be used for an activity recognition based system which could analyses all the habits of a user, the time schedule, etc. In case of an elderly person, such a system could be used as a permanent assistant which could detect if something abnormal is happening. Some automates could also be added in order to either help the user in case he forgets to perform something, or to simply improve the comfort of a normal user (switch on the TV and select the channel following the taste, etc). The states of the objects could also be used by a tracking location system to detect where the user is situated in the environment.

37 Chapter 5 System implementation The device that has been developed during this thesis consists in an electronic board (refer to figure 5.1) on which sensors are connected. This board is powered with a battery and uses ZigBee wireless technology to send the state of the sensors. The data is sent to a receiving board connected to the personal server. We will explain in details in the following parts which components are implanted on the sensing board and how the sensing data is collected. Figure 5.1: Photo of the sensing board 27

38 28 Chapter 5. System implementation 5.1 Requirements for the wireless sensor network A workshop has been conducted with several users and members of the projects in order to constitute some requirements for our wireless sensor network[7]: Usability/Availability/Installation: The users prefer cheap products, available offthe-shelf and easy to set up. They do not want to wear to many devices, and do not want to change their habits in the environment. Consequently, the wireless sensor network has to use a simple network topology which does not involve any complex installation. The size of the sensor nodes has to be as minimized as possible, as the personal server which should remain the only wearable device carried by the user. Performance/Reliability: The reliability of the system has been emphasized by the users compared to its performance. The transmission range also has to be high enough to obtain a stable system. However, as the user is constantly wearing the mobile receiver, the transmission of the state changes should not be a problem with the everyday objects the user is interacting with. Finally, the battery life is an important criteria to avoid any repetitive recharge cycles of the batteries. 5.2 Hardware implementation Description of the board The two main components of the board are the ZigBee transceiver and the microcontroller. Both communicate through the USART serial communication device, which concretely consists in linking the Rx port of the transceiver to the Tx port of the microcontroller and vice versa. A programming interface which consists in a six pins header has been had to the board to allow to reprogram the microcontroller without removing it. One I/O microcontroller pin, port D4, is connected to the pin of the transceiver called sleep control line and is used either to wake up the transceiver or to put it in sleep mode. All the other I/O pins are connected to a three pins header which is also linked to the ground and +3V. This allows to connect the sensors to the microcontroller and also to power them. Three of the I/O pins of the transceiver can be used as indicators and consequently have been connected to LEDs. All the other I/O pins of the transceiver are connected to a pin header so that they can be used in a future evolution. The sketch 5.2 presents the different parts of the board The main components The ZigBee transceiver is a XBee module from Maxstream (figure 5.3), series 1 and standard version with a U.FL antenna connector[15] This transceiver operates at ISM 2.4 GHz frequency and allows data rates up to 250 Kbps. This rate is more than enough for our application. The standard version of this module can be replaced by the professional one which has a higher transmitting power and consequently a bigger range. We chose to use an external antenna instead of an antenna integrated to the transceiver. This allows a better transmission and reception and also a larger range. A new series (series 2) of these XBee modules has been developed whereas this thesis was going to its end. This

39 5.2. Hardware implementation 29 Figure 5.2: Description of the sensing board Figure 5.3: The XBee transceiver new series uses Mesh networking instead of the point to multi point network architecture of the series 1. It could be interesting to use this new series for a more expanded wireless sensors network. The U.FL antenna connector of the XBee is connected to a MMCX (U.FL) to SMA reverse converter[20]:

40 30 Chapter 5. System implementation This converter is used to connect a 2.4 GHz antenna for SMA reverse connector[21]. This antenna (figure 5.4) is a 2.4 GHz omnidirectional antenna with half a wavelength and a gain of 2.90 dbi. (a) (b) Figure 5.4: Antenna connector (a) and antenna (b) The microcontroller (figure 5.5) is an Atmel ATMEGA88[10]: Figure 5.5: Microcontroller This microcontroller has been chosen for several reasons. First, this microcontroller is pretty cheap and available off-the-shelf. It also provides a very interesting connection set with for instance five analog-to-digital converters (here used for light sensors, but which can also be used for accelerometers or others), USART serial communication device (used for the communication with the XBee transceiver) and SPI serial interface (used as a programming interface). Finally, the firmware can be coded in C language with the free environment AVR studio. This microcontroller is running at 8 MHz. The power is regulated through a low drop voltage regulator (figure 5.6) which delivers 3.0V[22]. The whole board is powered with 3.0V. When operating under this voltage, the XBee module consumes much less power when it is in sleep mode. The microcontroller can also work with this voltage as the sensors used so far. The power is provided by 3 NIMH batteries of 1.2V each and 2600mA[17]:

41 5.2. Hardware implementation 31 (a) (b) Figure 5.6: The voltage regulator (a) and on the battery (b) Sensors The sensors used are basic ones. It was important for this thesis to provide a system easy to install, to use, and affordable. Some more suitable sensors for proper state changes certainly exist and we probably be used in a future evolution of our system. However we tried to select some basic sensors for this first version that could be used for the most common state changes. The board has been developed in such a way that it is easy to use other sensors. Consequently, some new sensors can easily be added, and some new state changes can be sensed. Light sensor(figure 5.7): TAOS light to voltage converter[18] Temperature sensor(figure 5.7): Dallas digital temperature sensor TO92[19] (a) (b) Figure 5.7: The light sensor (a) and the temperature sensor (b) Switch(figure 5.8): Omron microswitch[16] Press button(figure 5.8): Push button sw. NO

42 32 Chapter 5. System implementation (a) (b) Figure 5.8: The switch (a) and the press button (b) Operating mode Global description The sensors are plugged into the 3 pins headers corresponding to the I/O ports of the microcontroller. Only the light sensors need a special port as it is explained in the following part. The microcontroller is periodically checking the values of the sensors and in case a new value differs from a previous one more than a certain threshold, it wakes up the transceiver, sends it the information, and then put the transceiver back in sleep mode to save some power. The sampling rate is low, around 100 Hz or less, depending on the amount of sensors. When the transceiver receives some information from the microcontroller, it automatically sends it to the destination address corresponding to the receiving node. The sensors The temperature sensor The temperature sensor is the most particular sensor used here. This is a digital temperature sensor which uses the one wire protocol from Dallas. That means this sensor only uses one wire for data communication with the microcontroller. This sensor needs to be initialized at the beginning of the processing part of the microcontroller. Consequently, the temperature sensors need to be plugged to the board before the power is switched on. This sensor is the only one among the other sensors that we are using that can be called passive. In fact, even if the temperature is changing, it will not be seen as an event. Consequently, the temperature will be updated in the personal server when another state change will occur. An application note with source code examples is provided by Atmel explaining how to communicate with sensors using this protocol[11] The light sensor The light sensor used is a light to voltage converter. Consequently, it needs to be connected to an Analog-to-Digital converter. The Atmel ATmega88 provides 6 Analog-to-Digital converters represented by the ports C.

43 5.2. Hardware implementation 33 The microcontroller The microcontroller is constantly checking the values of all the sensors connected to it through an endless loop. At each iteration of the loop, the values of all the sensors are compared to their previous value stored in a variable. If at least one of these values has changed, then the microcontroller sends the new values of all the sensors to the transceiver. First, it wakes up the transceiver using the port D2 which is reserved for this use. This only consists in clearing the corresponding bit followed by a certain delay to let enough time to the transceiver to wake up. Then, a data frame is sent through the UART. This data frame depends on each sensing board and on the sensors used. Indeed, this data frame first contains an identification number unique for each board. Then it contains the values of all the sensors. The number of bytes used depends on the sensors. Finally, a character is added at the end of the frame as you can notice on the sketch 5.9. Figure 5.9: Example of a data frame The transceiver The transceiver works in transparent mode (See the data sheet for more information). When working in this mode, the transceiver immediately sends the data it receives on its RX port of the UART serial communication device. These data are sent to the transceiver whose address is the destination address of the previous transceiver. In this thesis, all the transceivers of the sensing boards send the data to the transceiver of the receiving board plugged to the computer. Consequently, all these transceivers have the same destination address. Processing part The processing part of the microcontroller is more or less the same for all the different boards. First, it is composed of a delay of a few milliseconds to let all the sensors and transceivers to be ready, an initialization part which includes the initialization of the uart, initialization of the A2D converter, configuration of the ports (input, output, pull up), configuration of the port dedicated to wake up the transceiver before sending data, and then put the transceiver in sleep mode. If the microcontroller also deals with a temperature sensor, it also has to initialize the one wire protocol and search for the temperature sensors on the bus. For each sensor, the processing part attributes 2 values for

44 34 Chapter 5. System implementation each sensor. One represents the previous value of the sensor and another one represents the new value. This is used to determine if the value of the sensor has changed or not. After the previous initialization part, all the variables representing the previous values of the sensors are initialized with the current values of the sensors. The microcontroller wakes up the transceiver, sends the states of the sensors (4 times) and then puts the transceiver back in sleep mode. When the microcontroller is sending the data to the transceiver, it always sends it 4 times with a delay of a few milliseconds in between. We did that to avoid the loss of data. Indeed, we noticed when the signal strength is pretty low that the data frame sometimes reaches the personal server with some data missing and consequently cannot be used. (no crc or other error detection algorithm has been implemented, but has been recognized as a good future improvement). Then the microcontroller enters in an endless loop. This endless loop consists in reading all the new values of the sensors (the reading process depending on the sensors), comparing these new values with the previous ones, and then, with some specific thresholds and filtering functions, deciding to send or not these new values. If the new values are sent, then the previous values are replaced with these new values. The light sensors The light sensors have been the most complicated sensors to calibrate. In fact, the ambient light plays an important role concerning the accuracy of these sensors. To obtain the best results we could, we introduced several filtering functions with different thresholds either in the processing part of the microcontroller and in the application of the personal server itself. First of all, to read the value of a light sensor, we read 10 sample values with a delay of 1 millisecond in between, and then deduce the value by taking an average of these 10 sample values. In fact, these light sensors are very sensitive and their value can change a lot even if the state of the object did not change at all. This can be due to the shadow of the user, shadow of an object, light reflected by an object, etc..., or other kind of perturbation on the sensor. When this new value is calculated, we compare it with the previous value of this sensor. The difference between the new value and the previous one has to be equal or bigger to a certain threshold to be interpreted as a possible state change, and sent to the transceiver. This threshold is specific to each sensor. Indeed, depending on the place where they are situated, a light sensor is not exposed to the same light conditions. For example, a light sensor placed in a cupboard will detect a pretty low light level when the cupboard will be closed. In the contrary, a light sensor placed near the bulb of a lamp will always detect a pretty high level of light because the ambient light will always be stronger than the one in a closed cupboard. Consequently, the range of the light values detected by the light sensor placed in the cupboard will be larger than the one detected by the light sensor placed near the bulb. That is why they need different thresholds. If we had applied the same thresholds for these 2 sensors, the microcontroller would have constantly detected the new value of the light sensor in the cupboard as a possible new state change because of its too high sensitivity. If a light sensor is not well calibrated, and is too sensitive due to a too low threshold, it can happen that the corresponding node will constantly send the new values of this sensor and consequently monopolize the receptor preventing all the other nodes from sending data. Furthermore, another threshold is used in the application of the server to interpret the new value of a sensor as a new state or not. This threshold is specific for each sensor. Indeed, this threshold depends on the object the sensor is sensing, the place of this sensor in the object, the place of this object in the environment and the ambient light of the environment. Consequently, to obtain the best results, this threshold has to be evaluated in different environment light conditions.

45 5.3. Software implementation 35 Switches ad press buttons The switches and press buttons are the most simple sensors used in this system. There is no threshold or filtering done concerning those sensors. Their state is directly interpreted in the microcontroller as the state pressed or released, and is represented as a simple character in the data frame. 5.3 Software implementation The application developed for this thesis is responsible for the interaction between the hardware prototype and the laptop (middleware application). It has been written in C++ to respect the standards of coding of the project (object oriented). It has several tasks. First, it has to read the data received by the reception board. Then, the application treats these data and updates the state of the objects in consequence. Finally, these states are displayed in order to inform the developer of a possible problem, and are also saved in a.log file to keep track of all the state changes. This application is a multi-threaded application developed with Microsoft visual studio 6.0. One thread is responsible for the reception of the data from the reception board. Another one treats these data and update the states of the objects. A last one displays the information. The application is composed of several classes as you can notice on the class diagram default classes deal with the application and the dialog boxes. One class is dedicated to the communication with the serial port: CCom[12]. Indeed, even if the receiving board used is plugged to the personal server through a usb connection, the driver provided is a usb to serial converter driver. Consequently, this board is recognized as a com port. Then the application is composed of one class per board: board 1, board 2,..., board 9. All these classes inherit from a common class board The class board Attributes This class is composed of 2 attributes. First a string id, unique for each board, used to identify it, and then an integer state change which notifies if a state change occurred in one of the objects sensed by the board. Both of these attributes are private ones. Public methods string get id() Access function which just returns the id of the board. string identify(string data) This function reads the id (first 3 bytes) of a data frame given as a parameter and returns it. string read id(string data) This functions reads the id of a data frame given as a parameter, store this id in the id attributes of the board, removes the id from the data frame and returns this string. int has changed() Access function returns the attribute state change of the board. void event occurred() This function gives the value 1 to the state change attribute. void event treated() This function gives the value 0 to the state change attribute.

46 36 Chapter 5. System implementation Figure 5.10: The classes diagram Classes board 1 to board 10 The classes board 1,..., board 10 all inherit from the class board.

47 5.3. Software implementation 37 Attributes Each board class has as attributes all the possible states of the objects they are sensing. For each state, the board has 3 attributes: one for the current value of a sensor, one for the current corresponding state, and one for the previous state. For example, the board class responsible for the fridge will have the following attributes corresponding to the light sensor of the fridge: string fridge light value string fridge light state string fridge light previous state All these attributes are private. Public methods Attributes accessors First, all the board classes have functions in order to get the values of all attributes. int data(string data) This function consists in reading the data frame given as a parameter. It reads in the appropriate order first the id of the board, and then all the values of all the sensors. This function depends on the structure of the data frame. CString display() This function consists in displaying all the values of the sensors of the board in the debugging window. string write file() This function consists in writing the states of all the states of the sensed objects in the.log file. Private methods These functions are the ones which are composing the previous data method. They consist in reading one of the sensor value included in the data frame, storing the previous value in the appropriate attribute, and interpret this value as a new state or not Global functions The global functions consist in all the calibration functions. These functions interpret a sensor value as a state, following a certain threshold value. These functions are global ones in order to be used by different classes.

48 38 Chapter 5. System implementation

49 Chapter 6 Evaluation 6.1 The sensor network For this experimentation, we used 10 boards, one board containing one transceiver. Each board represents a node in the wireless sensors network. The 10 nodes represented by these 10 boards are sending the data to an 11th node which represents the center of the star topology of this wireless sensor network. The number of the nodes can easily be increased as the XBee transceivers can take one of the 65,000 addresses available. 6.2 Scenario of evaluation Our system has to be evaluated in a real living home environment to precisely know its accuracy and stability. A real environment always offers lots of perturbations that cannot be thought in advance during the developing state of a project. This step is essential for the evaluation, but also in a first time to test the system and correct the possible problems. This system has to be able to face several perturbations. First, a user that does not know the system and how it works, does not behave in the same way than the person who designed it. For instance, if the user knows how the sensor of a specific object works or is fragile, he will differently interact with this object than a user who absolutely does not know anything about it. Besides, the sensors themselves have to face a lot of constraints. For example, the sensors placed in a fridge or a freezer have to face some humidity and temperature conditions very specific which cannot be simulated. The shadow of the user or the ambient light conditions can also represent some constraints that only exist in a real environment. Finally, it is important to see how the wireless sensor network behaves when the user is interacting with several objects at the same time, and if the personal server is not missing any event. 6.3 The experimental setup The experiments were performed in the fika room at the 4th floor of MIT huset. The configuration of this environment is described in this sketch

50 40 Chapter 6. Evaluation Figure 6.1: A live-in laboratory home environment with nodes placement and signal strength measures Four users were recruited for the experimentation. The experiments were performed individually by the users during a week. In addition to the system components (wearable personal server + receiving node), the users were also carrying a wearable camera connected to a mobile digital video recorder (6.2) to obtain the ground truth. 6.4 Energy consumption We conducted the experimentation of our wireless sensors network in real conditions during 4 days. The batteries of the sensing boards were fully charged before we started and they were switched on during 27 hours. The sensing boards do not function in the same way following the sensors they are using and the devices they are sensing. Consequently, the batteries do not need to provide the same amount of energy following the board they are powering. However, during these 4 consecutive days, we never encountered any problem due to a low battery level, and the batteries could have certainly been used during a larger amount of time. To reduce the power consumption of the sensing boards, we put the XBee transceiver in sleep mode when the board is not sending any data. But first of all, the whole board is powered in 3V which consequently reduces the power consumption when the transceiver is in sleep mode. Then the XBee transceiver consumes less than 3µA when it is in sleep mode. If it had been powered in 3.3V, it would have consumed 101µA when it is in sleep mode. 6.5 Range The signal strength measure above the acceptable limit of >10 db to obtain reliable data communication was found to be 97.5% checked at 8 different locations in a home environment. The results have been summarized in this table 6.1. Finally the transmissionreception range was evaluated to be 33 m with a single wall obstruction and 19 m with multiple wall obstruction.

51 Table 6.1: Evaluation of signal strength at 8 different locations Signal Living Dining Kitchen Bedroom Bedroom Toilet Toilet Office strength space hall (door open) corridor (door closed) (door closed) corridor (door open) (%) (%) (%) (%) (%) (%) (%) (%) Best (>30 db) Good (>20 db) Medium (>10 db) Low (<10 db) Range 41

52 42 Chapter 6. Evaluation Figure 6.2: A user wearing our system during the experimentation session 6.6 Accuracy of the system The accuracy of the system is essential for such an application. Indeed, if the sensing data is not well received by the personal server or if some data is missing, then any application built on top of this wireless sensor network cannot be reliable. To perform the evaluation, the users were performing a set of activities of daily living (Preparing lunch, baking, etc). The characteristic values concerning the evaluation of the system, precision and recall values, are defined as follows:

53 6.6. Accuracy of the system 43 P recision = T ruep ositives (T ruep ositives+f alsep ositives Recall = T ruep ositives (T ruep ositives+f alsenegatives There is a True Positive when a state change happens and the system recognizes it. There is a False Positive when the system recognizes a state change whereas it did not happen in reality. There is a False Negative when the system does not recognize a state change whereas it happened in reality. The system got a global precision value of 91.2% and a global recall value of 98.8%.For more information, please refer to the related paper[7]. These values are promising taking into consideration that some perturbations due to the environment itself (interferences, light conditions, etc) influence the performance of the network. Besides, some state changes were really difficult to sense, and consequently some more appropriate sensors would significally improve the results.

54 44 Chapter 6. Evaluation

55 Chapter 7 Discussion and limitations Even though the results obtained after this evaluation are very encouraging, some limitations of the system have been revealed and should be reduced in a next version. First of all, some important problems have been encountered concerning the use of the light sensors. Those sensors are very sensitive, and their accuracy depend a lot about the ambient light conditions. The experimentation has been performed at different moments of the day, with different daylight conditions, which caused some troubles for the light sensors. To solve this problem, we recalibrated the sensors, more precisely, so that they could work in all cases. However, this represents a lot of work, and could be even more difficult to realize with a larger amount of sensors. That is why, the use of the light sensors should be improved. A possibility could be to use several of these sensors to sense a same state (2 or 3) instead of one, and also to sense the ambient light intensity which could allow some kind of dynamic sensor calibration. Another limitation concerns the battery life of our sensor modules. In fact, we did not really tested for how long a sensor node can work without any power supply problem. However, as the state of the sensors are periodically checked by the microcontroller, a non negligeable amount of energy is wasted when a state change does not occur. This should be improved to increase the battery life of our system. Finally, one of the constraint of the system is the wearable equipment that the user is carrying. Even if its size and weight are not that important, it still represents a limitation for the user. An easy way of avoiding this constraint could be to use a device the user is wearing by himself. Actually, nowadays almost everybody has a mobile phone and carries it in any situation. Besides, the mobile phones becoming more and more powerful (smartphone, PDAphone, etc), it could be possible (or maybe in a near future) to embed our application in such a device. 45

56 46 Chapter 7. Discussion and limitations

57 Chapter 8 Future works Several future works can be planned concerning this thesis. First, some solutions can be found in reference to the limitations mentioned above. The hard-coding of the objects states should be removed. It would certainly be more efficient to create a database with all the everyday objects and their state changes. Like this, it would be much faster and easier to treat a larger amount of devices. This would also allow other applications to access the sensor data which can then be used for other purposes. Another improvement concerns the communication protocol. In fact, even if nowadays ZigBee seems to be the most efficient technology for this kind of work, a reception board is needed to receive the data from the sensor nodes to the computer. A technology like Bluetooth is already embedded in the laptops which avoids to use any additional board. It would not be surprising to see in a near future a laptop with an embedded ZigBee transceiver which would then reduce the wearable equipment. Besides, as it has been explained before, lots of the objects which are sensed also have a containment property (fridge, freezer, cupboard, etc) of which the state would be interesting to know. This future work had already been planned since the beginning of this thesis. Some work about the use of an RFID reader and some id tags has been performed, and should start to be tested very soon. The electronic board of the sensor nodes has been designed to be as modular and upgradeable as possible. However, if this board has to take part of a wearable equipment, for instance connected to several accelerometers on the user s body, the design should be modified. Indeed, for such an application, the size of the pcb could be significally reduced. Finally, a CRC or other kind of error detection algorithm should be implemented to avoid the loss of data during the transmission. 47

58 48 Chapter 8. Future works

59 Chapter 9 Acknowledgments I would like to thank all the following persons who helpt me a lot to write this master s thesis: Thomas Pederson, manager of the easyadl project in computer science department at Umeå university, for accepting me in his team and for advising me during this long period. Dipak Surie, my supervisor, Phd student in computer science and team member of the easyadl project, for supervising me, orientating me and supporting me. Nils-Erik Eriksson, electronic teacher at the TFE department, for bringing me lot of help concerning the electronic part of this thesis. All the other members of the project, master thesis students and research assistants, for integrating me in the team and helping me in my work. I also want to thank all the university personal, teachers, international office, and others for integrating me so well and supporting me during my stay. 49

60 50 Chapter 9. Acknowledgments

61 References [1] Pentland A. Looking at people: sensing for ubiquitous and wearable computing. Pattern Analysis and Machine Intelligence, IEEE, Volume 22, Issue 1, [2] Gregory D. Abowd, Anind Dey, Robert Orr, and Jason Brotherton. Contextawareness in wearable and ubiquitous computing. GVU Center, College of Computing, Georgia Institute of Technology, Atlanta, [3] Michael Beigl and Hans Gellersen. Smart-its: An embedded platform for smart objects. Smart Objects Conference, [4] Jan Beutel, Oliver Kasten, Friedemann Mattern, Kay Römer, Frank Siegemund, and Lothar Thiele. Prototyping wireless sensor network applications with btnodes. ETH Zurich, [5] Dennis Cox, Emil Jovanov, and Aleksandar Milenkovic. Time synchronization for zigbee networks. IEE, Proceedings of the Thirty-Seventh Southeastern Symposium on System Theory, [6] Surie D., Lagriffoul F., Pederson T., and Sjölie D. Activity recognition based on intra and extra manipulation of everyday objects. In Proceedings of IFIP UCS 2007 Conference on Ubiquitous Computing Systems, Springer LNCS 4836, pp [7] Surie D., Laguionie O., and Pederson T. Wireless sensor networking of everyday objects in a smart home environment. ISSNIP 08, Australia, Sydney, [8] Surie D. and Pederson T. An activity-centered wearable computing infrastructure for intelligent environment applications. In Proceedings of IFIP EUC 2007 Conference on Embedded and Ubiquitous Computing, Springer LNCS 4808, pp [9] Surie D., Pederson T., Lagriffoul F., Janlert L.-E., and Sjölie D. Activity recognition using an egocnetric perspective of everyday objects. In Proceedings of IFIP UIC 2007 Conference on Ubiquitous Intelligence and Computing, Springer LNCS 4611, pp [10] Keyword: ATMEGA88 (Visited the 21/10/2008). Atmel AT- MEGA88. [11] documents/doc2579.pdf (Visited the 21/10/2008). Atmel application note about Dallas one wire protocol. [12] (Visited the 21/10/2008). Serial port communication class. 51

62 52 REFERENCES [13] (Visited the 21/10/2008). Berkeley motes. [14] (Accessed the 21/10/2008). The BTnodes. [15] (Visited the 21/10/2008). XBee transceiver from Maxstream. [16] Stock n : (Visited the 21/10/2008). Omron microswitch. [17] Stock n : (Visited the 21/10/2008). NIMH battery. [18] Stock n : (Visited the 21/10/2008). Light sensor. [19] Stock n : (Visited the 21/10/2008). Temperature sensor. [20] (Visited the 21/10/2008). U.FL antenna connector. [21] (Visited the 21/10/2008). 2.4 GHz antenna. [22] Part n : mcp e/to (Visited the 21/10/2008). Voltage regulator. [23] (Visited the 21/10/2008). MICAZ. [24] (Visited the 21/10/2008). Imote2. [25] (Visited the 21/10/2008). IRIS. [26] (Visited ). ZigBee Alliance. [27] Emil Jovanov, Amanda O Donnell Lords, Dejan Raskovic, Paul G. Cox, Reza Adhami, and Frank Andrasik. Stress monitoring using a distributed wireless intelligent sensor system. IEEE Engineering in Medicine and Biology Magazine, May/June [28] Achilles Kameas, Stephen Bellis, Irene Mavrommati, Kieran Delaney, Martin Colley, and Anthony Pounds-Cornish. An architecture that treats everyday objects as communication tangible components. PerCom, [29] Yoshihiro KAWAHARA, Masateru MINAMI, Hiroyuki MORIKAWA, and Tomonori AOYAMA. Design and implementation of a sensor network node for ubiquitous computing environment. Vehicular Technology Conference, [30] Cory D. Kidd, Robert Orr, Gregory D. Abowd, Christopher G. Atkeson, Irfan A. Essa, Blair MacIntyre, Elizabeth Mynatt, Thad E. Starner, and Wendy Newstetter. The aware home: A living laboratory for ubiquitous computing research. College of Computing and GVU Center, Georgia Institute of Technology, Atlanta.

63 REFERENCES 53 [31] Lama Nachman, Ralph King, Robert Adler, Jonathan Huang, and Vincent Hummel. The intel mote platform: a bluetooth-based sensor network for industrial monitoring. IEEE, Information Processing In Sensor Networks. Proceedings of the 4th international symposium on information processing in sensor networks, [32] Kay Römer, Thomas Schoch, Friedemann Mattern, and Thomas Dübendorfer. Smart identification frameworks for ubiquitous computing applications. Wireless Networks, Springer Netherlands, [33] E. Munguia Tapia, S. S. Intille, L. Lopez,, and K. Larson. The design of a portable kit of wireless sensors for naturalistic data collection. in Proceedings of PERVA- SIVE, [34] Emmanuel Munguia Tapia. Activity recognition in the home setting using simple and ubiquitous sensors. Massachusetts Institute of Technology, September [35] Stephen S. Yau, Yu Wang, and Fariaz Karim. Development of situation-aware application software for ubiquitous computing environments. COMPSAC, 2002.

64 54 REFERENCES

65 Appendix A Appendix A.1 Maxstream XBee s data sheet A.2 Atmal ATMEGA88 data sheet s front page A.3 TAOS light to voltage converter data sheet s front page A.4 Dallas digital temperature sensor TO92 data sheet s front page A.5 Omron microswitch data sheet s front page A.6 Microchip low drop voltage regulator data sheet s front page 55

66 Figure A.1: XBee datasheet page 1

67 Figure A.2: XBee datasheet page 2

68 Figure A.3: Microcontroller datasheet page 1

69 Figure A.4: Light sensor datasheet page 1

70 Figure A.5: Temperature sensor datasheet page 1

71 Figure A.6: Switch datasheet page 1

72 Figure A.7: Voltage regulator datasheet page 1