52
Implementation of a self-triggered control strategy over IEEE 802.15.4 DAN PANTZAR Master’s Degree Project Stockholm, Sweden 2011 XR-EE-RT 2011:004

Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Implementation of a self-triggered controlstrategy over IEEE 802.15.4

DAN PANTZAR

Master’s Degree ProjectStockholm, Sweden 2011

XR-EE-RT 2011:004

Page 2: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 3: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

i

AbstractA self-triggered control strategy implementation on a IEEE 802.15.4 based networkcan prove to be quite difficult. In this master thesis project a study of the IEEE802.15.4 protocol is conducted to understand what constraints it poses to an imple-mentation of a self-triggered controller and the possibility to use it as a feedbacklink in a control process. A self-triggered control strategy recently introduced byTiberi et al. is implemented on a wireless sensor network. An experimental test toprove that it is possible to use a self-trigger control is presented. A PI-controller isdesigned and used on a system where the temperature is measured with a wirelesssensor located on a Mote(iv) Tmote Sky. The feedback link in the control processis over a IEEE 802.15.4 network.It is shown that a self-triggered control strategy delivers an equal control resultcompared to a periodic control strategy while reducing the energy consumption ofthe network.The results from the conducted experiments show that the hardware in use is notyet ready for the self-triggered control strategy. Even though the same controlperformance as a periodic controller was achieved, the hardware has some flaws.Problems with beacon adaptivity as well as computation constraints were discov-ered. It is concluded that for a system with low demands on precision and witha slow dynamic the self-triggered control performs well on Mote(iv) Tmote Skyunits. However, a fast system with a high demand on precision is not ready for theself-triggered control implementation while using Tmote Sky platforms.

Page 4: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

ii

SammanfattningDetta examensarbete handlar om en studie av IEEE 802.15.4 protokollet och vilkabegränsningar det medför på en implementation av en sjäv-triggningsalgoritm förtrådlösa sensorer. En analys av protokollet genomförs och visar att protokollet ärfullt anpassningsbart för att fungera som en återkoppling i ett system där själv-triggningen används, samt att energiförbrukningen sjunker när den används.För att bevisa detta genomförs ett experiment med en PI regulator. Som återkop-pling i systemet används en temperatursensor som sitter inbyggd på en Mote(iv)Tmote Sky plattform. Återkopplingen styrs av själv-triggningsalogritmen och re-sultatet visar att ett likvärdigt resultat uppnås som med en periodisk strategi föratt sköta återkopplingen.Slutsaterna som tas är att även om ett likvärdigt resultat uppnåddes av själv-triggningsalgoritmen som den periodiska algoritmen så visar dem endast detta förett långsamt system som inte har någon ordentlig precision i sig. Det visade sig ävenatt hårdvaran som användes inte riktigt fungerade enligt förväntan vilket medfördebland annat förluster av packet och tidsfördröjningar vid exempelvis byte av ”Bea-con Order”. Själv-triggningesalgoritmen fungerar men hårdvaran är inte anpassadtill den, varför ett snabbt system med krav på hög precision inte skulle fungera medutrustningen som anvönts i detta examensarbete.

Page 5: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

iii

AcknowledgementI would like to thank my examiner Carlo Fischione for all his help and guidancethroughout this thesis. I would also like to extend my gratitude to PiergiuseppeDi Marco for all his help with Linux. Pangun Park also needs to be mentioned forall his help with TinyOS and nesC. I am extremely thankful for all the code andtechnical support provided by Aitor Hernandez (sorry for the spam).A special thanks to Ubaldo Tiberi for spending a lot of time helping me by explaininghow the self-trigger control works.I also need to mention all the help I received with the lab equipment provided tome by the help of Chithrupa Ramesh, José Araujo and Burak Demirel.Finally I also need to mention all the people around the world providing supporton TinyOS forums, without you I would be lost.Thanks y’all!

Page 6: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Contents

Abstract i

Sammanfattning ii

Acknowledgement iii

Contents iv

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.3 Past work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.4 Assumptions of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Wireless Sensor Networks 52.1 General description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1.1 Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.1 Limiting factors . . . . . . . . . . . . . . . . . . . . . . . . . . 72.3 TinyOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 IEEE 802.15.4 93.1 General information . . . . . . . . . . . . . . . . . . . . . . . . . . . 93.2 OSI layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.2.1 PHY layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.2.2 MAC layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3.3 Constraints imposed by the protocol . . . . . . . . . . . . . . . . . . 153.4 Zigbee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.4.1 General information . . . . . . . . . . . . . . . . . . . . . . . 153.4.2 Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.5 IEEE 802.15.4e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

iv

Page 7: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

v

4 Wireless Networked Control Systems 194.1 General description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.2 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.3 Control strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4.3.1 Periodic Control . . . . . . . . . . . . . . . . . . . . . . . . . 204.3.2 Self-triggered Control . . . . . . . . . . . . . . . . . . . . . . 204.3.3 Control strategy used in this master thesis . . . . . . . . . . . 22

5 System testing 255.1 The main experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . 255.2 Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255.3 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5.3.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265.3.2 Labview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5.4 Experiment setup and implementation . . . . . . . . . . . . . . . . . 275.4.1 Self-triggered control . . . . . . . . . . . . . . . . . . . . . . . 295.4.2 Periodic control . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

6 Implementation Issues 37

7 Conclusion and future work 397.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Bibliography 43

Page 8: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 9: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 1

Introduction

1.1 Background

A Wireless Sensor Network (WSN) is a wide technology appearing everywhere fromsimple home automation to large and complex industry automation. Since thesenetworks are wireless, they need to follow a networking standard to be accepted inthe industry. A popular standard is based on IEEE 802.15.4. The main goal withthis protocol is to be used with wireless networks working with a low data transmis-sion rate. Due to the low transmission rate and the low demand on computationalcapacity on devices used in these networks, they have a low energy consumption,hence their main life support can be batteries.The next step in wireless systems is Network Controlled Systems (NCS). On themarket today there are several products that use information from WSNs to con-trol, in example heating, energy consumption or lights etc. As the demand for morewireless information increase, more and more data is handled by devices on thesenetworks which results in an increased energy consumption. To counterbalance thiseffect, devices ”sleep” when not in use and wake up in different time intervals tocheck if there is any information to send or receive. This is quite an effective strat-egy but in most systems these devices wake up and detect no changes and go backto sleep, resulting in a useless power consumption. The next step in the quest forlow energy consumption is to try to estimate the behaviour of a system and onlyuse the devices when they are needed. This type of strategy is called Self-triggeredcontrol and is one of the parts covered in this master thesis. The IEEE 802.15.4protocol will be analysed in this master thesis to determine if it is possible to useit on a feedback channel in a self-triggered control system.

1.2 Objective

The main goal of this thesis is to investigate if it is possible to design a feedbackcontrol over an IEEE 802.15.4 network running on Telos Mote(iv) devices. To be

1

Page 10: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

2 CHAPTER 1. INTRODUCTION

more precise, the main task is to use a controller with a wireless feedback controlledby a self-triggering algorithm to sample the temperature in a system.The following items are the stated objectives for this master thesis:

• Investigate what constraints the IEEE 802.15.4 protocol imposes to the feed-back channel.

• Investigate the energy consumption for a wireless sensors network with feed-back.

• Construct a self-triggering control algorithm to reduce network traffic at thesame time as controller performance is kept at an acceptable level.

• Is it possible to do an effective feedback controller with limited power resourcesover an IEEE 802.15.4 network?

1.3 Past work

In the past there has been a lot of projects, scientific articles and theses basedsolemnly on the IEEE 802.15.4 protocol or on wireless sensor networks. Hernandezwrote a master thesis project [8] focusing on how to control a process with a wirelessfeedback. The code produced in that thesis covers some of the work that this masterthesis is bringing up. Hence, a lot of the code used in this master thesis project isprovided by Hernandez.Rodrigues da Silva Severino presented the thesis ”On the use of IEEE 802.15.4/ZigBeefor Time-Sensitive Wireless Sensor Network Applications”[5] where a similar taskwas discussed. However, it only covered the wireless protocol and no self-triggeringcontrol as mentioned before. Another thesis presented by Maciel called ”Devel-opment of a Detection System Using a Wireless Sensor Network” [14] covers thearea of wireless sensor networks, but with no constraints on power consumption.There are also a another couple of theses in the same area. To summarize thesearticles and theses, there has already been a lot of work on investigating the IEEE802.15.4 protocol and on analysing WSNs. However, due to endless amount of waysto evolve WSNs and the protocol, it is far from finished and a lot of more work canbe conducted.In the recent article ”Adaptive Self-triggered Control of IEEE 802.15.4 Networks”[20] of Tiberi, Fischione et al. the authors introduce the idea of using a self-triggeredcontrol for IEEE 802.15.4 networks. The article determines that the stability of asystem is not always guaranteed if the protocol parameters are not properly tuned.Throughout the article there are numerous equations that have been adapted intothis thesis as well as a pseudo-code explaining exactly how each iteration of theself-triggered control should work.

Page 11: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

1.4. ASSUMPTIONS OF THE THESIS 3

1.4 Assumptions of the ThesisSince the area of wireless communication is quite large the decision was made thatthe plant would be connected by wire to the controller and only the feedback linkwould be wireless.Only one sensor will be used during the experiment phase. This is mostly due tothe difficulty of a multiple sensor network. When using more than one sensor an ad-vanced communication algorithm needs to be designed that can handle everythingfrom simple packet losses, collisions and ad-hoc communications. To investigate themultiple sensor network is easily one or two more master theses.A PI-controller will be used in this master thesis. The reason is that a PI-controlleris fairly simple to design, it is universally adopted, and more focus will be devotedto the wireless protocol and the self-triggering control.

Page 12: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 13: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 2

Wireless Sensor Networks

2.1 General descriptionA Wireless Sensor Network (WSN) is the next step in system monitoring and con-trol. By introducing cheap battery powered sensors based on the IEEE 802.15.4protocol, the lifespan can vary from a couple of months to years on a pare ofAA-batteries. By installing wireless sensors instead of expensive wired sensors com-panies can reduce cost of maintenance and get rid of cables that prevent certaintypes of motion on robots. There are however a couple of drawbacks with wirelesssystems since they are not as reliable as wired systems. Since energy consumptionstill play one of the key roles in WSNs, they will probably not replace the entirewired industry until there is a solution to it. Hence, industries will probably staymainly wired over the next couple of years.But what areas are WSNs limited to? At the moment WSNs appear practically ev-erywhere where monitoring is in process. As an example there are already a lot ofmonitoring within health care and process industries. The next step is to integratesmall sensors into the human body that connects to an access point. By doing thispeople can continuously have there vitals monitored while at home, at work or evenon vacation.In process industries it would be possible to replace all wired sensors with wirelesssensors on robots . However, due to the fact that these robots have a high demandon precision, they demand a low packet loss and since the power consumption ofsensors is not satisfied yet, a wireless solution is not yet applicable.

2.1.1 Topology

Since WSNs usually consists of numerous amount of devices linked together, thereare a couple of different topologies that can be used. There are three main differenttypes of topologies called star, peer-to-peer and cluster tree [6].The star topology is the most common one where all devices are connected to acentral device acting as a coordinator. This is the same type of network used in

5

Page 14: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

6 CHAPTER 2. WIRELESS SENSOR NETWORKS

COORDINATOR REDUCED FUNCTIONAL DEVICEREGULAR DEVICE

STAR PEER-TO-PEER CLUSTER TREE

Figure 2.1: Different network topologies

regular home computer networks. All communications go through the coordinator,hence a limiting factor. This type of network can be connected to other networksthrough the coordinator.The peer-to-peer network also includes a coordinator but it does not need to be inthe center of the network. Communication works as an ad-hoc network meaningthat all devices can talk to each other as long as they are in reach of the coordinator.The main feature with the ad-hoc strategy is that devices can be connected anddisconnected to a live network. This is one of the key features making it possibleto implement monitoring equipment into motion based units, since they can jumpfrom network to network.The cluster tree topology is also a peer-to-peer network, but different type of devicescan act as different parts of a network. Fully Functional Devices can take the roleas coordinators, which can shift when devices come and go, and Reduced FunctionDevices can connect up to more advanced devices and only send or receive data.To get a better understanding of what these topologies look like see Figure 2.1. Thenetwork used in this master thesis project is of a peer-to-peer character.

2.2 HardwareThe market for monitoring equipment such as wireless sensors is rapidly expanding.There are sensors monitoring everything from heating, speed, electricity consump-tion, heartbeats to lights. In this thesis the main focus is on sensors detectingchanges in temperature. To be more precise a small devices called Mote(iv) TmoteSky will be used.Mote(iv) Tmote Sky is a wireless platform designed at the University of CaliforniaBerkley, based on a Texas Instruments MSP430 microcontroller and also referred toas Telos B [15]. The Tmote is a clever wireless module preconfigured with sensorstracking humidity, temperature and light. It is operating in the 2.4 GHz frequency

Page 15: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

2.2. HARDWARE 7

Current Consumption Nom Max UnitMCU on, Radio RX 21.8 23 mAMCU on, Radio TX 19.5 21 mAMCU on, Radio off 1800 2400 µAMCU idle, Radio off 54.5 1200 µAMCU standby 5.1 21.0 µA

Table 2.1: Typical operating conditions for the Tmote Sky platform. Data obtainedfrom [15]

range and is able to transmit data up to speeds of 250 kbps, all under the super-vision of the IEEE 802.15.4 protocol. One of the key advantages with the Tmoteis that it has one of the fastest ”wake up from sleep” time (<6 µs) compared tosimilar products on the market today [15].Another key feature with these motes are that they consume very little energy.They can be powered either by two regular AA-batteries or through the USB-port.In table 2.1 the energy consumption is listed for different operating modes. As thetable shows, the highest energy consumption occurs when the radio unit is sendingand receiving. This clearly shows that there is room for improvement on the radiousage, hence the self-triggered control discussed in subsection 4.3.2 might lead tothat. The main advantage with Tmote Sky despite the easy usage is that theycan be programmed through the USB-port. These motes are running an operationsystem called TinyOS, which will be discussed in section 2.3.There are however a couple of drawbacks with these devices, which will be discussedin section 2.2.1.

2.2.1 Limiting factors

Since these Tmote Sky devices are cheap and easy to use due to their simplicity thereare some drawbacks. If only small amounts of information retrieval is of importance,they preform with excellent result. But if something more than only informationretrieval is needed, in example heavier computation, problems will arise. The mem-ory unit on the circuit boards are limited to 48k Flash + 1024k serial storage [15]which means that it is not possible to store large amounts of data. This means thatif you combine a simple computation of an exponential together with an advancedscheduling strategy for sending and receiving data the memory unit is not sufficient.To counter this problem look-up tables can be created, devices can be connectedto a computer that can compute everything and send the right values back to thedevice. This will however lead to a loss of precision in the system as well as addedtime-delays due to the external computation.Since these motes follow the IEEE 802.15.4 protocol they are also limited to certain

Page 16: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

8 CHAPTER 2. WIRELESS SENSOR NETWORKS

sending/receiving times stated by the protocol, which will be explained in Subsec-tion 3.3.

2.3 TinyOSWhen working with low-powered wireless sensors all sensors need to be intelligentenough to understand how communications are handled and what to do with allinformation received. To make this possible, all sensors are equipped with a mi-croprocessor containing an operating system, OS, which tells the hardware what todo. TinyOS is an open-source OS suitable for this type of application due to itsdesign for low energy consumption[2]. One of the key factors to the widely spreaduse of TinyOS is because of the open-source feature. By opening it up to the publica lot of people all over the world frequently update the TinyOS CVS library withnew and sometimes unique features. There are also a lot of web pages coveringeverything from regular information to frequently asked questions and support.The hardware in the devices (as mentioned in subsection 2.2) using this type of OSoften contains a small memory unit which TinyOS has been able to adapt to. Thesedevices are not suited for any heavy computation, often resulting in a network wherea controller not using TinyOS, does all the heavy computation which is one of themajor weaknesses of TinyOS [1].TinyOS uses a programming language called nesC which is a dialect to C. Due tothe different environment compared to traditional C it is considered by many to bea bit difficult to master in the beginning.

Page 17: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 3

IEEE 802.15.4

3.1 General information

To make sure that certain standards are kept when constructing wireless products,there are specific protocols to follow. The Institute of Electrical and ElectronicsEngineers (IEEE) is a widely spread association focused on establishing industrystandards within all different parts of the electrical market. IEEE 802.15.4 is aprotocol designed mainly for use with wireless sensor networks demanding a low-rate data transfer. Due to the constraints on data transfer in these networks, theprotocol standardizes the physical layer (PHY) and a Medium Access Control layer(MAC), originating from the Open System Interconnection (OSI) model [6]. Themain design of these layers make it possible to communicate over a network with alow power consumption at the same time as the delivery rate is kept at an acceptablerate.This protocol is favourable to use due to its simplicity and strict rules on howcommunication should be handled. By only standardizing the PHY and MAC andnot interacting with the upper layers, focus is kept on low transmission rates.This chapter will cover the protocol design, technical theory on how it works andsome basic information about how it is implemented throughout the industry.

3.2 OSI layers

In 1984 the International Organization for Standardization (ISO) developed theseven layered OSI model [23]. This model shows how a standard computer networkshould be configured and is typically used in all networked systems today. SinceIEEE 802.15.4 only standardizes the MAC and PHY layers, they will be discussedand analysed in the following parts. To get a better understanding of how thetypical OSI model differs from the one defined in the IEEE 802.15.4 protocol, seefigure 3.1.

9

Page 18: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

10 CHAPTER 3. IEEE 802.15.4

Figure 3.1: The left image shows the original OSI model. To the right the IEEE802.15.4 adapted OSI model. The original pictures can be found in [18]

3.2.1 PHY layer

The Physical layer is the lowest layer in the OSI model. Here all interpretationbetween the network data and the hardware is made. Hence, this layer is often re-ferred to as the most complex layer. Since it interprets information, it is responsiblefor modulating and demodulating the data sent over the network [12]. In the IEEE802.15.4 protocol devices can for example sample data through sensors, which isthen relayed over the network as symbols using a modulation called Offset Quater-nary Phase Shift Keying (O-QPSK). By using a standard sinusoidal wave, a twobit symbol is sent with π/4 transition. In wireless communication there is alwaysa probability of packet loss, interference or phase-shifted signals. This is often thecase with this type of modulation as well. However, due to the transition rate inO-QPSK, symbols can only get a maximum phase shift of 900, contrary to QPSKwhere it can shift up to 1800. That is one of the main reason for choosing theO-QPSK over QPSK as modulation strategy.The symbols that are sent over the network make up a 32-bit message. This mes-sage contains information about the intended destination, payload and informationabout the payload etc. The data is of ”little endian” character which basicallymeans that all data sent in the payload area of the message can be found in the endof the message.

Page 19: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

3.2. OSI LAYERS 11

Even though the PHY layer is quite competent by its own, it is not able to solemnlysend and receive data. In other words, it needs to interact with the next OSI Layer,MAC, to be able to send and receive data.

3.2.2 MAC layer

The MAC layer is the next OSI layer located above the PHY layer and is in chargeof regulating the access to a shared medium for a number of nodes [12]. In WirelessSensor Networks (WSN) it is the MAC layer that is struggling with energy conser-vation and is usually using a multi-hop communication scheme to relay frames asfast as possible to different nodes, that is if a device is connected to a cluster treenetwork.The MAC layer can be designed in different ways depending on what type of hard-ware (device) it is connected to. As mentioned in subsection 2.1.1, there aretwo types of devices, Full Function Device (FFD) and Reduced Function Device(RFD)[12]. A FFD is capable of working as all types of devices in a network. Inother words it is able to take the role as a PAN Coordinator, a coordinator or asa device. The difference between a PAN coordinator and a coordinator is that thePAN coordinator controls everything in a large network containing smaller networks(also known as a Cluster tree network) and a coordinator can only control a smallernetwork (Star or Peer-topeer networks). A device can only listen to a PAN or ordi-nary coordinator and obey orders, see subsection 2.1.1. The RFD can however onlywork as simple devices, thus needing at least one FFD in the same network to beable to relay data.The MAC layer in the IEEE 802.15.4 protocol supports two types of operationalmodes, nonbeaconed and beaconed mode.

Nonbeaconed mode

The nonbeaconed mode is a non schedule based transmission mode based on anunslotted Carrier Sense Multiple Action Collision Avoidance (CSMA-CA) protocol,see figure 3.2. This type of transmission is used in a network containing a numberof nodes connected to a network coordinator. At first contact each node maps upthe surrounding nodes in a table. This is done to let the newly connected nodediscover the layout of the network at the same time as it gives away its position tothe other nodes and coordinator. When connection is established each node livesits own life and sends data according to a predefined rule.If a node needs to send data over the network, it follows the flow diagram seen infigure 3.2. In short it senses if the channel is idle. If not, it backs off and waits arandomly generated time and tries again. If this procedure continues a couple oftimes without the success of getting a window to send the data, the package willbe discarded. If it however finds the channel to be idle, it occupies the channeland sends the data. The main advantage with this transmission mode is that alive network can expand continuously. Nodes can be set up to sleep during long

Page 20: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

12 CHAPTER 3. IEEE 802.15.4

NB = 0,BE = macMinBE

Delay forrandom(2BE - 1) unit

backoff periods

Perform CCA

Channel idle?

NB = NB+1,BE = min(BE+1, aMaxBE)

NB>macMaxCSMABackoffs

?

Failure Success

Un-slotted CSMA

Y

Y

N

N

Figure 3.2: A flow diagram showing each step in the process that a device need tofollow if transmitting data in an unslotted CSMA-CA mode. The original picturecan be found in [16]

time intervals, wake up and send their information and then go back to sleep again.This is an effective way of conserving energy, but it has a major drawback. Sincea device is not constantly connected to a coordinator it needs to follow the stepsshown in figure 3.2 to establish a connection. If there is a lot of network trafficin the area, a lot of energy might be spent on transmission failure and retries. Ifthere are lots of nodes in a network there will be a lot of collisions when trying tosend data. Hence important information might be lost. It might even lead to the”hidden node” problem, where a distant node might believe that the channel is idleand try to send information to the network coordinator at the same time as anothernode is active. This will lead to a collision and important data might be lost.Some nodes spend a lot of time in sleep, hence only doing a time synchronizationwhen transmitting data [12]. As mentioned before, nodes used in WSNs are oftenconstructed by cheap electronics, hence their internal clocks tend to drift. With alonger interval between time synchronisation it might lead to an early or late action

Page 21: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

3.2. OSI LAYERS 13

B.I.

S.D.

GTS

CFPCAP

ACTIVE INACTIVE

Figure 3.3: The superframe structure. The large black lines represent the beaconssent by the coordinator.

from one part of a system. This can have a vital effect on the system.

Beaconed mode

When using the beaconed mode strategy, a ”superframe” is initiated by a networkcoordinator when it transmits a beacon to all nodes in a network. The beaconcontains information about the frame. The information sent indicates the lengthof the superframe, if there are any Guaranteed Time Slots (GTS) for any nodesetc. The superframe contains two major periods, an active period and an inactiveperiod. In the active period there are two parts, Contention Access Period (CAP)and Contention-Free Period (CFP), as can be seen in figure (3.3)[6]. In the inactiveperiod sensors can be put to sleep to conserve energy, which is extremely usefulwhen working with WSNs and battery powered devices. By changing the lengthof a superframe it is possible to change the energy consumption of a node (will bediscussed in chapter 5).The active period is divided into 16 time slots where the first slot is dedicated tothe beacon. In the CAP area a slotted CSMA-CA strategy is used to let all nodescompete for a connection. The slotted CSMA-CA procedure is almost the same asthe unslotted version used in the Nonbeaconed mode, but with a couple of moresteps to go through, see Figure 3.4.The rest of the timeslots (a maximum of seven) located in the CFP are calledGuaranteed Time Slots. In these time slots selected devices have the privilege tosend information without having to compete with other devices, also known as TimeDivision Multiple Access (TDMA)[12]. By competing in the CAP-period nodes canget a timeslot in the CFP-period. This timeslot can then be used only once or itis possible to dedicate it to the same device during the coming superframes. Bydedicating a slot, devices can save even more energy since they do not need to

Page 22: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

14 CHAPTER 3. IEEE 802.15.4

NB = 0, CW = 0

Battery lifeextension?

BE = macMinBE

BE = lesser of(2, macMinBE)

Locate backoffperiod boundary

Delay forrandom(2BE - 1) unit

backoff periods

Perform CCA onbackoff period

boundary

Channel idle?

CW = 2, NB = NB+1,BE = min(BE+1, aMaxBE)

CW = CW - 1

CW = 0?NB>

macMaxCSMABackoffs?

Failure Success

Slotted CSMA

Y

Y Y

Y

N

N

N

N

Figure 3.4: A flow diagram showing each step in the process that a device need tofollow if transmitting data in a slotted CSMA-CA mode. The original picture canbe found in [16]

compete for a time slot every time a new superframe begins.The length of a superframe is calculated by using the following equation:

aBaseSlotDuration× aNumSuperframeSlots (3.1)

The minimum length of a superframe is the results of when aBaseSlotDuration isat its lowest value of 60 symbols and when the aNumSuperframeSlots consist of 16slots [10]. With the modulation and configurations of the IEEE 802.15.4 protocol inthe 2.4 GHz frequency band, devices are able to send 62.5ksymbols/s. This resultsin a minimum superframe duration of 15.36 ms, also known asaBaseSuperFrameDuration.By doing the calculation shown in table 3.1, we see that a maximum superframeduration is 252.3136 s. The superframe can be altered in different ways. By changingparamters like the Beacon Order (B.O.), Beacon Interval (B.I.), Superframe Order(S.O.) or Superframe Duration (S.D.) it is possible to create a unique superframefor a network [6]. Table 3.1 shows how they are connected to the superframe andin what way they are limited.Since one of the goals of this master thesis is to conserve energy, the beaconedmode configuration with a dedicated GTS will be used in the experimental phase(as mentioned in [4]).

Page 23: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

3.3. CONSTRAINTS IMPOSED BY THE PROTOCOL 15

S.D. = aBaseSuperFrameDuration2SO , 0≤ SO ≤ 14B.I. = aBaseSuperFrameDuration2BO , 0≤ BO ≤ 14

Table 3.1: How the different constants from the superframe are limited and con-nected to each other. By combining this table together with figure 3.3, it is possibleto see how the superframe changes with different values. The data was obtainedfrom [6].

3.3 Constraints imposed by the protocolThe IEEE 802.15.4 protocol imposes some major constraints on products using it.One major constraint is that all users need to follow a certain pattern when it comesto sending data, also known as frames. This is good if you want the product to beof a plug and play character. But if there is only one way communication betweentwo motes it will force users to send a lot of unuseful data.The size of a superframe also imposes a constraint. As was discussed in subsection3.2.2, the longest duration of a superframe is aBaseSuperFrameDuration214 =252.3136 s. If the wanted sleep time is longer than that it will result in an unwantedwake up for each device. This will result in an unwanted energy consumption. Theminimum superframe length also imposes a constraint since it needs to be taken into account for when calculating start and stop times for the equipment.There are also a limited amount of GTS available. This can pose a problem if thereare more nodes, than available slots, in need of sending important packages.Then there is the traditional drawback with packet losses as well as delays causedby interference from surrounding equipment.

3.4 Zigbee

3.4.1 General information

Zigbee Alliance is a well known association providing a wide series of products withinwireless sensor networks (WSN). All products are based on the IEEE 802.15.4 pro-tocol and appear everywhere from home automation to industry automation.The Zigbee Alliance have contrary to their main competitor, Dust Networks, chosento use the superframe strategy (see subsection 3.2.2) to control the network flow.Dust Networks selected a strategy called Time Division Multiple Access, TDMA,which in a way resembles the superframe strategy by dividing the total transmissiontime into transmission slots for each device. By doing that all sensors are guaran-teed transmission time hence a more reliable network. However, since all sensorsneed to be active during their time slot the energy consumption rises [13]. In theactive part of a superframe there is however also a TDMA strategy used, henceZigbee and Dust are quite close competitors. Since not all sensors are forced to

Page 24: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

16 CHAPTER 3. IEEE 802.15.4

have a time slot in the superframe structure followed by Zigbee, it manages to keepthe energy consumption down. Other drawbacks with TDMA is that it indulgesan increase of signal traffic during network setup and maintenance, as well as morefrequent time synchronising due to clocks drifting on cheap sensors [12].Zigbee Alliance main focus is on products working in mesh networks. The wholedesign principle of these mesh networks are that a number of different sensors workwith an ad-hoc network strategy making it possible to relay information from onesensor through a couple of Fully Functional Device (FFD) to reach the designateddevice. The ad-hoc strategy makes it possible to introduce new sensors anywhere ina network without taking the network layout in to account. As soon as a new sensoris connected to the network all Zigbee products issues a set up protocol where theytalk to each other and map out the entire network. At the same time an FFD ischosen to work as the network PAN coordinator.

3.4.2 Products

Out on the market today there are a lot of Zigbee products within the WSN area.It is almost vital for a company to be certified with the Zigbee logo if they wanttheir products to sell on the market.Most of the products are simple sensor products which are all connected to a PANcoordinator and all controllers are based on event triggering. The only company onthe market that seems to be focusing on PID-control over WSNs is called JohnsonControl [9]. Worth noticing is that they only work with periodic triggering whichdemands more energy than a self-triggering strategy.There are even products focusing on home improvement, for example smart lightswitches, energy meters and so on.The equipment used in the experiment phase in this master thesis all followed theZigbee standard, since Tmote Sky is classified as a Zigbee product.

3.5 IEEE 802.15.4eAt the moment of writing this thesis a new version of the IEEE 802.15.4 protocolis being released. A task group called 4e is currently working on enhancing theMAC layer to make it more flexible. This version of the protocol is known as the802.15.4e and the main focus is set on industry automation [3].One of the major implementations this group is working on is the ability to extendthe superframe time. In other words, make it possible for products to sleep longer.By increasing the sleep time they are aiming at increasing the battery life up tomore than five years [11]. An interesting problem with the increased sleep time willbe to see if the clock drift on the devices will impose a larger problem when sleepinglonger.They are also working on adding more Guaranteed Time Slots in a superframeand implementing a technique called channel hopping. By introducing a channel

Page 25: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

3.5. IEEE 802.15.4E 17

hopping strategy it is possible to send more information at the same time sincemultiple channels are in use.To summarize the work of the task group, everything they are working on is actuallysomething that is vital when it comes to the competition against wired products,energy consumption and data flow. The IEEE 802.15.4 protocol is today limited tothe use of beaconing mode strategy when it comes to saving energy. By introducingthese new features the MAC layer will be working much more efficiently, hence theenergy consumption will go down. It will be interesting to see if they manage toimplement all of these new features on the current hardware.

Page 26: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 27: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 4

Wireless Networked ControlSystems

4.1 General description

As mentioned earlier, wireless sensor networks (WSNs) are starting to appear ev-erywhere. These networks are now evolving from monitoring systems to networkedcontrol systems, where a control loop is closed by a WSN. When a wireless net-work is used to interconnect several processes and controllers, we have a WirelessNetwork Controlled System (WNCS). By incorporating WSN into control systemsnew opportunities occur and in the future, most control systems will probably beconverted into WNCS.In this master thesis, a WSN will be used to provide a feedback link to a controller.Therefore, we will give some information about WNCS. In the following chapterconstraints on a WNCS will be discussed and the control strategy which will beused in this master thesis will be analysed.

4.2 Constraints

Since the information exchange between the process and the controller travels on awireless channel, there will be some kind of interference or disturbance, for examplepacket loss, congestion, noisy environments etc [7]. This will result in a systemuncertainty which can’t be estimated to 100%.Congestion is the result of too much traffic on a channel, hence it poses a threatto WNCSs. When there is congestion in a network, information does not arrive atthe controller, hence the closed loop performance may deteriorate. Depending onthe amount of data flowing in the feedback link, the maximum amount of nodes ina network changes. In other words, a lower amount of nodes will allow each nodea longer connection time to send its data. However, since this master thesis onlyfocuses on a single device providing the feedback, this will not be a problem.The noisy environment is something that can not be prevented. But with the devel-

19

Page 28: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

20 CHAPTER 4. WIRELESS NETWORKED CONTROL SYSTEMS

opment of the IEEE 802.15.4e protocol (Subsection 3.5) with its channel hoppingstrategy it will in the future be possible to use more and more wireless feedback innoisy environments.Then there is also the security issue. In a wireless system it is easier for an externalsource to gain access than in a wired system, hence unwanted visitors can accessdata and introduce disturbances. However, since security and hacking is not a partof this master thesis, this issue will not be analysed further.

4.3 Control strategy

4.3.1 Periodic Control

In traditional implementations of control systems, sensors wake up, sample dataand send information to the control unit periodically (with packet losses in somecases and thus with random reception), even if there is no change in the system.Then they return to sleep. Data transmission in a WSN is energy demanding, whichmakes a periodic implementation energy inefficient. A method to reduce the numberof transmissions is to send data only when relevant change in the system is detected.This method is called ”Event-triggered control”. This control strategy works quitewell, but it has its drawbacks as well. To wake up and sample a system that has notchanged is a power-consuming operation without useful results. As an example, aGuaranteed Time Slot strategy is implemented on a wireless network. Beacons aresent from the PAN coordinator with a fixed interval, forcing the receiving mote towake up and sample the system periodically. If change is detected it is reported,otherwise it ignores this part and goes back to sleep.The next step in WNCS is to apply new strategies to get the same results asan periodic controller, but with major improvement on the power consumption.One strategy fulfilling that requirement is the "Self-triggered control" and will bediscussed in the coming subsection.

4.3.2 Self-triggered Control

A self-triggered control strategy is an attempt to lower the energy consumption inWNCSs. By forcing a device to send feedback over the wireless channel accordingto a strategic pattern, it will result in a lower energy consumption.In self-triggered control, there are a couple parameters that needs to be taken into account. The next transmission time is of the feedback information based onthe current measured state of the system closed loop performance [21]. In short,a prediction of the next needed sampling time is done by looking at how a systemchanges during a couple of samples. Each time a sample of the system is taken,an estimation of what the next sampled value will be is computed. Judging byhow far off this estimation is from the reference signal, the next beacon interval ofthe superframe is adjusted so that feedback information can be sent at the desiredpredicted time. By doing this the energy consumption for transmission will go down,

Page 29: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

4.3. CONTROL STRATEGY 21

Tk Tk Tk Tk+1 +2 +3BO BO BOk+2k k+1

Tk Tk Tk+

k

aBaseSlotDuration aBaseSlotDuration

(x(T ), x(T ))

k

k+1Beacon

Node

PAN

Figure 4.1: The figure illustrates how the different parameters are connected to eachother. The picture is inspired by [20]

because feedback information is transmitted only when needed. It follows that thenetwork traffic decreases. Due to the adaptivity of this strategy the robustness ofthe system should not be affected in a negative way. But exactly how should aneffective Self-triggering controller be designed?This master thesis has followed a new self-triggered control described in [22] and[20]. In particular, the control system is described in the following equation:

x = A(η)x+B(η)uu(t) = Kx(tk), t ∈ [tk + τk, tk+1 + τk+1] (4.1)

The η in the matrices represents the parameter uncertainty in the system , tk is thetime at which the system is sampled and τk represents the time delay caused by thehardware (e.g. transmission delay, sampling delay).The adaptive algorithm roughly works as follows; first a beacon is sent from a PANcoordinator to a device at time Tk. Then, when the device receives this beacon,a measurement of the system is taken at time Tk, see figure (4.1). By using thisvalue, the device then calculates the next estimated state by using equation (4.2),as proposed in [22].

||x(Tk+1)|| = ||eA(Tk+1−Tk−τk) +BK

∫ Tk+1

Tk+τk

eA(Tk+1−s)ds||||x(Tk)|| (4.2)

Once the next estimated value is calculated, the device sends this data together with

Page 30: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

22 CHAPTER 4. WIRELESS NETWORKED CONTROL SYSTEMS

the sampled value to the PAN coordinator. With those values, it is now possible forthe PAN coordinator to calculate the next Beacon Order. By using equation (4.3),provided by [22], the second next time is calculated.

tk+2 = Tk+1 + 1a

ln(1 + aε

b||x(Tk+1)||+ d− b||x(Tk+1)||+ d

b||x(Tk)||+ d(eaτk − 1)

)− τmax (4.3)

The d is the process disturbance, ε is the output error bound guarantee highestvalue that provides stability according to the Lyapunov stability theorem, a is thenorm of A together with a uncertainty, b is the same as a, but with the B matrixinstead. If the estimated value reaches zero, which could happen, equation (4.3)will go to infinity. Hence it is important to set a maximum gain close to zero, inexample 0.0001, forcing the system to use it instead of doing computation resultingin a divide with zero.The next step for the PAN coordinator is to use tk+2 to calculated the next BeaconOrder, which is denoted by B.O.k+1. This is shown in this equation:

B.O.k+1 =⌊log2

( tk+2 − T0,k+1aBaseSuperFrameDuration

)⌋(4.4)

The newly calculated B.O.k+1 is then transmitted with the next beacon to thereceiving mote, which will adapt to it in the next superframe. These steps are runover and over while the system is runningTo do all calculations previously mentioned is fairly straight forward, but there aresome constraints that we need to keep in mind. The IEEE 802.15.4 protocol clearlystates that the maximum Beacon Interval is achieved by letting the B.O. = 14.Hence there will be a problem if a system demands a higher value. However, asdiscussed in subsection 3.5, the new ”4e” implementation will increase the maximumlength of a superframe, hence this problem will disappear. Meanwhile, a way ofsolving this problem is to simply tell the sensor to ignore sampling the system atthe time of the first beacon. This ”small fix” will decrease the energy consumptionon the device since it will not sample and thereby spend less time awake. The mainpower drainer however is the radio unit on a device which means that the problemwith the power consumption is still there.There is also a lower limit to the choice of Beacon Order. As mentioned in Subsection3.2.2, the S.O.k ≤ B.O.k.

4.3.3 Control strategy used in this master thesis

Control theory is rapidly growing and covers a large amount of interesting strate-gies. Since one of the main objectives of this thesis is the self-triggered control,the decision was made to use a simple controller. At first the thought of using atraditional PID-controller was formed. The reason for this is mostly because it isfairly simple to construct and preforms quite well. However, after realising that thesystem in use in the experimental phase of this master thesis is of a slow dynamic,

Page 31: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

4.3. CONTROL STRATEGY 23

the decision was made to construct a PI-controller instead. The main advantagewith a PI-controller is that it is fast and the deriving part of a PID indulges un-wanted noise on the control signal and is pretty much useless if you have a processwith a slow dynamic.A PI-controller consists of two parts, a Proportional part and an Integrating part.The KP , also known as the proportional term, makes changes to the output signalthat is proportional to the error input value. When changes are made to it, it showsup fast in the output signal.The KI , also known as the integral term, eliminates the steady-state error thatoccurs when a KP is applied. Equation (4.5) shows the controller.

u(t) = KP e(t) +KI

∫ t

0e(τ)dτ (4.5)

There are a couple of different techniques on how to tune a PI-controller. By farthe most correct way of doing this is to calculate a cut-off frequency and use itto calculate a couple of equations that finally delivers tailored constants for thecontroller.However, in this master thesis project a slightly less proper technique is used. To bemore specific, all parameters were tuned by implementing a PI control into Simulinkin Matlab and simulating a test with a model of the system. By submitting a unitstep as an input signal, the different parameters were calibrated until an outputsignal that satisfied the system was achieved.

Page 32: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 33: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 5

System testing

This chapter covers the experimental test phase of what has been discussed in theprevious chapters. The main goal is to see how the self-triggered control works on awireless sensor network (WSN). The layout of this chapter is as follows. First a shortsubsection regarding the experiment setup. Then the system modelling is reportedin subsection 5.2 followed by a subsection discussing the equipment used and howall parameters are calculated. The chapter concludes with the experimental resultsfrom all tests.

5.1 The experiments

The performed experiments consist in controlling the temperature in a room byusing the embedded temperature sensor on a wireless device (Mote(iv) Tmote Sky).The temperature samples are then sent as feedback over a 802.15.4 network whencontrolling a heat source. A room with a radiator is rigged with three motes and acomputer. The computer is connected to the actuator controlling the warm watersupply to the radiator. One mote is connected to the computer, hence making itthe PAN coordinator in the network. The other device is placed close to the heaterand programmed to work as the sensor checking the temperature on demand bythe PAN-coordinator. At first the room temperature is 15 ◦C and when the systemstarts it will control the heater to go to the reference value of 21 ◦C. Once thesystem is stable at that temperature the test is stopped. Then a test is conductedto see how the controller handles external disturbances. When the system is stableat the reference value, a window is opened to see how the controller reacts.Meantime, while the experiments are in progress, the third mote will be used in thebackground to monitor the amount of traffic flowing in the network.

5.2 Modelling

For the sake of simple yet insightful experimental implementation, a complicatemodelling of hydro mechanics and temperature dynamics, the modelling was kept

25

Page 34: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

26 CHAPTER 5. SYSTEM TESTING

simple. Since heaters in common use are quite slow, the model of the system needsto be very slow. The following model was derived:

xP = −0.002xP + 0.003uxC = −xP + r

(5.1)

Since the system is controlled by a PI-controller, the model needs to take both thesampled values of the process and the control in to account. Hence the variable xPrepresents the process and xC the controller. The symbol r represents the referencevalue in the control part. Since the control signal, u, is based on values regardingboth the process and the controller, it was derived as:

u = −KP (xP − r) +KIxC (5.2)

When dealing with a Self-triggered controller, the estimation of the next samplevalue depends on the current sample. Since the system derived in equation (5.1)covers both the control and the process, the sampled value of the system needs tocover both as well. There also need to be an error value added to the sampled value[19]. Hence the sampled value x(Tk) should be according to equation (5.3), wherea and b are the scalars from equation (5.1).

||x(Tk)|| =∣∣∣∣∣∣[xP − r xC −

ar

bKI

]∣∣∣∣∣∣ (5.3)

According to this model, the system will lose temperature if no control signal isapplied and gain in temperature if the controller is applied.The system described in equation (5.1) gives an open loop system with the matrices,which are later used when tuning the controller parameters, expressed as:

A =[ −0.002 0−1 0

]B =

[ 0.0030

]C =

[ 1 00 1

](5.4)

5.3 Equipment

5.3.1 Hardware

As previously mentioned, the main experiment covers three Tmote Sky devices (seesection 2.2). The first mote acts as a device which, upon request, samples thetemperature sensor. The sampled value of the process together with the value fromthe control is then used to compute equation (5.3). It is then used to calculate thethe next estimated value, ||x(Tk+1)||. However, these motes are extremely limitedwhen it comes to heavier computations, hence the next estimated value needs to becomputed in advance and saved in a look-up table.The estimated value and the sampled value are then sent to the second mote, whichacts as a PAN coordinator. This mote coordinate the communication as well ascalculate the upcoming beacon order. The beacon order parameter is also computed

Page 35: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

5.4. EXPERIMENT SETUP AND IMPLEMENTATION 27

in advance and programmed into a look-up table. Of course, the utilisation of look-up tables limits the precision of the control system, but still delivers a reasonableperformance.Another goal of this experiment is to keep the power consumption down on thenetwork devices. Since transmission over radio is the biggest energy consumer inthe entire experiment it is also important to monitor the traffic. The third mote, asmentioned before, is set up in the background to measure the network traffic. Byusing a ”packetsniffer” provided by the TinyOS CVS library, all messages sent overthe network are monitored. A program called Z-monitor is then used to collect thedata and provide statistics.The heating in the room is controlled by changing the valve-opening from 3−5 mmon a radiator. The valve is controlled by an actuator supplied by a company calledSauter. By applying a control signal ranging from 0−10 VDC it is possible to controlthe heat [17]. By using a NI PCI-6221 card (also known as a data acquisition, DAQ)provided by National Instruments, a control signal can be constructed. To get theDAQ-card working, a program call NI Labview was used, see subsection 5.3.2

5.3.2 Labview

National Instruments Labview is an intuitive program used to design measuringsoftware connected to a DAQ-card which relays all signals to a shielded I/O-blockcalled ”SCB-68”. In this experiment the DAQ-card will handle all communicationbetween the software and the actuator on the heater.In Labview a loop is designed with a PI-controller, tuned according to Subsection4.3.3 and connected according to figure 5.2. The controller-block has a referencesignal summed with the feedback channel coming from the sensor. The outputsignal is then relayed to the DAQ-card which then relays the signal to the actuator.The entire loop is updated every time the sensor provides a newly sampled value.All data is saved while running the test. There is however a drawback with theutilization of Labview. Since all data from Labview is saved into a single file, ittends to grow very large in just a couple of minutes. Since each experiment coversroughly 10− 25 minutes, a way of making sure that the computer does not run outof memory is to set the time interval between every time data is saved from Labviewto ones every second. This will make the plots shown in the results part look veryedgy and simple, but the control result will not be affected. Worth mentioning isalso that by reading the sampled value into Labview will add a small time delay tothe controller. But since the dynamic of the system is quite slow, this time delayhas no significant effect.

5.4 Experiment setup and implementationThe main objective of this master thesis is to adapt the 802.15.4 network parametersaccordingly to a self-trigger control strategy. To see the efficiency of the controlstrategy, a test with an periodically triggered control implementation and a test

Page 36: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

28 CHAPTER 5. SYSTEM TESTING

Figure 5.1: The layout of the program-loop designed in Labview. Information fromthe PAN coordinator is continuously saved in a file that is read over and over (seethe yellow box to the left of the image). The output signal is relayed through theDAQ-card to the I/O-card that is connected to the actuator on the heater.

Page 37: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

5.4. EXPERIMENT SETUP AND IMPLEMENTATION 29

with the self-triggered control implementation is compared. Since only one mote isused and to keep the energy consumption down a beacon mode with a GuaranteedTime Slot strategy is used in both experiments.

5.4.1 Self-triggered control

As previously mentioned, the look-up tables for the next estimated value, ||x(Tk+1)||,and the upcoming beacon order, B.O.k+1, needs to be calculated. Since the systemin use is slow and since we lose precision in the system by using look-up tables all theconstants containing anything from external distortion to time delays in equation(4.2) are removed. Equation (5.5) is the result when simplifying the equations.

||x(Tk+1)|| = ||eA(Tk+1−Tk) +BK

∫ Tk+1

Tk

eA(Tk+1−s)ds||︸ ︷︷ ︸ ||x(Tk)||

ξ

(5.5)

By assuming that at each measurement Tk = 0, we have that Tk+1 − Tk is theequivalent to the Beacon Interval. By using that assumption, a look-up table isconstructed (see table 5.1). The value of ξ at the current B.O. and its multipli-cation with x(Tk), the estimated value is obtained. Worth mentioning is that thesensing mote only samples the process and does not get any information aboutthe control, hence a substitute is calculated by looking at how the system evolvesbetween each beacon interval and summarized the changes, just as an simple inte-grator.The second look-up table covering the upcoming B.O. is a bit more tricky to cal-culate. First we need to calculate tk+2, which is done by using equation (4.3). Yetagain all constants containing delays, distortion or equivalent are removed with thesame reason as mentioned earlier. The new equation can be seen in equation (5.6)below. The ε is only a constant and can be picked randomly since it is related tostability based on delays and distortion described in [20]. In this experiment a valueof 1 is selected.

tk+2 = Tk+1 + 1a

ln(1 + aε

b||x(Tk+1)||)

(5.6)

Finally the next Beacon Order is calculated by using equation (4.4). By runninga test where a timer is started when a beacon frame starts to transmit and stopswhen the value was sampled on the receiving device it is possible to extract the timeit takes for a device to react. That value then tells us that Tk+1 − T0,k+1 = 0.96ms. Hence we can rewrite equation (4.4) to the following:

B.O.k+1 =

log2(0.96 + 1

aln(1 + aε

b||x(Tk+1)||)

15.36) (5.7)

Due to the Tmote sky platforms inability to handle decimals, the resulting B.O.k+1are rounded off. All values can be found in the right table in table 5.1. The motes

Page 38: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

30 CHAPTER 5. SYSTEM TESTING

B.O. ξ

3 1.947354 1.903705 1.836876 1.755407 1.686958 1.654719 1.6464810 1.6393111 1.6258612 1.6016013 1.5621714 1.50978

||x(Tk+1)|| B.O.k+1 Round value≤0.1 7.2868 7≤0.5 6.1613 6≤1.0 5.4975 5≤1.5 5.0578 5≤2.0 4.7253 5≤2.5 4.4569 4≤3.0 4.2315 4≤3.5 4.0372 4≤4.0 3.8664 4≤4.5 3.7140 4>4.5 3.5763 4

Table 5.1: The left table is the look-up table used when calculating the next esti-mated value x(Tk+1) (done on the receiving mote). By multiplying the measuredx(Tk) with ξ at the current B.O., x(Tk+1) can be extracted. The right table repre-sents the look-up table in the PAN coordinator. By looking at how close the x(Tk+1)is, the next B.O can be extracted. Due to the wireless platform’s inability to handledecimals, the B.O. value needs to be rounded (as seen in the right column).

KP KI

10 .15

Table 5.2: The PID paramters tuned in Simulink.

are then programmed with the newly constructed look-up tables.Finally the PI-controller is tuned in Simulink. By introducing a unit step to thecontroller connected to the previously discussed model, all parameters are tuned,see table 5.2 and figure 5.2. These parameters were then inserted into the controllerdesigned in Labview.When all parameters are tuned and programmed, the experimental testing phasecan start. Worth mentioning is that to make sure that the motes can communicatewith each other the S.O. is set to a value of 3. To get a clearer picture of how theexperiment is constructed see figure 5.3.The system is tested in two different stages, undisturbed and disturbed. The firsttest covers a temperature rise from 15 ◦C to 21 ◦C. The next test shows how thesystem reacts when a window is opened. These two tests should show how thesampling rate changes over time and that the controller works with this type offeedback. All results can be seen in subsection 5.5.Finally a test is conducted to see what the traffic rate is during one hour with a

Page 39: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

5.4. EXPERIMENT SETUP AND IMPLEMENTATION 31

Figure 5.2: The PI-tuning process. Since the main system is slow it will take a longwhile until it is completely stable. However, since it is within the 7% margin afterthe 200 s mark, it is acceptable.

fairly stable temperature.After these tests it is time to test the periodically triggered controller.

5.4.2 Periodic control

The main difference with this setup compared to the self-triggered controller test isthat now the motes are running with a fixed B.O. Since a heating system is rela-tively slow, the B.O. was set to 5. The main reason for not setting it to the highestpossible value is that even if the system is slow it might get exposed to externaldisturbances. If that happens it needs to be able to react to it, which is easier if ashorter superframe duration is selected. The shortest possible B.O. was not selectedsince it would be obvious that the energy consumption would clearly be too highwhen compared to the self-triggered control. Hence, the selected B.O. should be abalance between energy efficiency and control performance.Worth pointing out is that the PI-control parameters stay the same in this experi-

Page 40: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

32 CHAPTER 5. SYSTEM TESTING

PLANT

HEATER

Control

COMPUTERDAQ-cardACTUATOR PAN

SENSOR

x(t )k k+1B.O.

xuv

v v v

v

gap

v

Figure 5.3: The experiment setup.

ments as well. The same type of tests are conducted on this system setup as in theprevious setup. All results are shown in subsection 5.5.

Page 41: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

5.5. RESULTS 33

5.5 ResultsIn figure 5.4 the result of the self-trigger control is displayed. The Temperaturein the system changed from 15 ◦C to 21 ◦C and the controller performed with anacceptable result. As mentioned before, Labview’s inability to save data in a shortertime interval than 1 s is proven in the results from this test (used an interval of 0.1ms) since a lot of disturbance is added to the data, see figure 5.4. Hence a timeinterval of 1 s is used in the periodic control strategy. However, the result showthat the system actually stabilizes at 21 ◦C.Figure 5.5 shows the result of the equivalent test for the periodic controller. Theresult shows that the system reaches 21 ◦C and stabilizes after a while. Worthmentioning is that since the system is not in an optimal closed cell with the sameprerequisites every time, the results can vary a bit. This is clearly shown since thesettling time for the self-triggered controller is much longer than in the case of theevent triggering controller, as well as the overshot is much higher. However, sincethe system stabilized in both cases it shows that the controller works. Also worthmentioning is that the motes were switched between the two different strategy tests,which might be the reason why the self-triggered controller reacted worse in the firsttest.The second test covering the external disturbance to the system shows that bothsystems can handle disturbance. Figures 5.6 and 5.7 show the resluts from bothcontrollers. To get a better understanding of how the beacon adapts during thedifferent temperature states, see figure 5.8. In table 5.3 the statistics from thisexperiment recorded by the monitoring mote is shown.When letting both systems drop in temperature and then regain the reference valuethe network traffic was measured by the third mote. Judging by the results a slightimprovement was achieved by the self-triggered controller compared to the periodiccontroller. However, when letting both systems run undisturbed for one hour, theresult was better. Table 5.4 shows that a 15% lower amount of traffic was achievedby the self-triggered control.Worth mentioning is that the hardware used during these experiments seems tobe unreliable. Sometimes the motes lose connection or simply just stop sendinginformation. This could indicate that some beacons were missed by the surveyingdevice and vice versa.

Page 42: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

34 CHAPTER 5. SYSTEM TESTING

0 2000 4000 6000 8000 100000

5

10

15

20

25

Time [0.1 s]

Uni

ts

Temperature [C]Control Signal [V]

Figure 5.4: Result of the self-triggered control trying to increase the temperaturein the system. Starting from 15 ◦C and working its way up to 21 ◦C. The dottedline represents the temperature and the solid line is the control signal

0 200 400 600 800 1000 1200

0

5

10

15

20

Time [s]

Uni

ts

Temperature [C]Control signal [V]

Figure 5.5: A periodic control is used. The beacon order is fixed and the systemis trying to control a temperature gain from 15 ◦C to 21 ◦C. System is stable after1180 s. The dotted line represents the temperature and the solid line is the controlsignal.

Page 43: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

5.5. RESULTS 35

0 1000 2000 3000 4000 5000 6000 7000 80000

5

10

15

20

Time [0.1 s]

Uni

ts

Temperature [C]Control Signal [V]

Figure 5.6: The self-triggered control reacting to a disturbance in the system. Thedotted red line represents the sampled values and the blue line represents the controlsignal.

0 50 100 150 200 250 300 350 400 450 500

0

5

10

15

20

Time [s]

Uni

ts

Temperature [C]Control signal [V]

Figure 5.7: The periodic control reacting to a disturbance in the system. The dottedred line represents the sampled values and the blue line represents the control signal.

Page 44: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

36 CHAPTER 5. SYSTEM TESTING

Strategy Packages sent Duration[s] Ratio of packages/sSelf-trigger control 997 859.975 1.159Periodic control 672 576.014 1.167

Table 5.3: Traffic flowing in the network during the different strategies when runningthe disturbance test.

Strategy Samples sent Duration[s] Average samples sent/sSelf-trigger control 1156 3597 0.321379Periodic control 1475 3900 0.378205

Table 5.4: The amount of samples sent when using different control strategies.

0 1000 2000 3000 4000 5000 6000 7000 80003

3.5

4

4.5

5

5.5

6

6.5

7

Time[0.1 s]

Bea

con

Ord

er

Figure 5.8: This figure illustrates how the adaptation of the beacon order changesduring a temperature drop. The graph begins when a drop is registered and theB.O. is at its lowest value. When the estimated value is within certain limits itstarts to increase the the B.O.

Page 45: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 6

Implementation Issues

During this master thesis a lot of problems occurred. At first there was a problemwith the compatibility between the computer OS (Windows 7) and the Tmote Skydevices. This was solved by downgrading to Windows XP.Since the devices are programmed through a terminal, Windows is not capable ofhandling it. To solve this Cygwin was installed, but not without any problems.TinyOS needs a lot of different libraries and support of programming languagesnot included in Cygwin. Hence a lot of time was spent on downloading the rightpackages containing all libraries and programs. In the end, a virtual machine withUbuntu, Linux, was installed making it much easier to do all communication withthe motes and programming.The code developed in [8] was reused in this master thesis However, there wereproblems with the encoding and decoding messages sent between the motes. It wassolved after several weeks of programming and error searching. Information need tobe arranged in a specific structure before it can be added to the payload part of theframe. The solution was to define a structure, containing two eight bit variables,in the profile-file used by all devices. By calling this structure when adding thepayload to the frame and when extracting the payload from the frame, the rightdata is relayed.Since an ordinary computer is not able to relay a control signal to the actuator, inthe experimental test phase, another problem occurred. By reflecting on previouscourses at KTH, an idea of using the same equipment used in the lab in the basiccontrol course was thought of. After further investigation it was decided to use theDAQ-card and the I/O-block from the water tank process station, which met thespecifications demanded by the actuator. A problem provided by the DAQ-cardwas that it demanded the use of National Instruments Labview instead of Matlab.Before the experimental test phase could begin there was a problem with the wirelessmotes and the heavier calculations demanded by the self-trigger control algorithm.After discussing the problem with the examiner it was decided that all computationwould be done in advanced and look-up tables would be used on the devices.At first, the equations described in [20] was used to calculate all parameters in the

37

Page 46: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

38 CHAPTER 6. IMPLEMENTATION ISSUES

self-triggered control. These equations were big and some what hard to calculate,hence posing minor problems. After several conversations with the authors of thearticle, some of the equations were simplified, hence all problems were solved.The biggest problem during this master thesis project occurred when it was discov-ered that the actuator on the air-condition unit, that was intended to be used inthe experiment phase, was of a totally different type than expected. It meant thata new unit had to be found or the entire experiment would have to be redesigned.The solution came when another heater with a controllable actuator was found atanother building.To conclude this chapter, it can be said that it is not possible to foresee all problems.They will randomly arise and the only way to solve them is to try to be dynamicand solve them as they occur. All problems can not always be solved and sometimesit means that a lot of work needs to be scrapped.

Page 47: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Chapter 7

Conclusion and future work

7.1 Conclusion

This master thesis project covers the investigation of the system-level design of theIEEE 802.15.4 protocol and a possible self-triggered control with a wireless feed-back link. It delivers some facts indicating that a wireless feedback link is actuallypossible, but with minor trade-off.At first the IEEE 802.15.4 protocol was analysed to see what constraints it im-posed on a wireless network system. The investigation showed that the protocolis designed to permit users to change only a selected number of factors with veryhard constraints. This is mainly because of the industry standard, which meansthat even if a user has a unique setup of a system, it should still be recognizedby other devices as a IEEE 802.15.4 network. When running a Beaconed modeconfiguration, the only way of altering the protocol is to change the Beacon Orderand/or Superframe Order. As mentioned in subsection 3.2.2, the maximum lengthof a superframe is approximately 252 s. This means that if a slow system is in useand nothing will happen during a time cycle a device is forced to wake up and useits radio without a cause. Hence this protocol is not properly adjusted to systemswith a slow dynamic.The experiment phase showed that a controller using a self-triggering strategy run-ning on an IEEE 802.15.4 network performed a result equal to a periodic controller.This clearly shows that a self-triggered control is capable of producing a good result.The traffic rate did go down when an external disturbance was introduced, but onlywith a small rate. This was expected since the Beacon Order should go down toa lower value compared to the periodic controller when the estimated value growslarger. Judging by the results shown in subsection 5.5, the traffic rate did go downeven more over time when the system settled at its reference value, hence a lowerenergy consumption is achieved. With a better model of a system, the look-uptables can be adapted with better precision, conducing a smoother Beacon Orderconversion which will lead to an even more improved traffic rate.The hardware in use, Mote(iv) Tmote Sky, is not the best equipment to use with the

39

Page 48: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

40 CHAPTER 7. CONCLUSION AND FUTURE WORK

self-triggered controller. Its simplicity provides an easy to use interface and energyefficient schedule. However, the simplicity also provides a limit to the self-triggeredcontroller due to its inability to compute advanced algorithms. The solution withlookup tables works, but with a cost of lost precision. There are however newerTmote Sky platforms out on the market today with bigger and better microproces-sors. By introducing a self-triggered control strategy to a platform like that couldimprove the end result.A couple of times the Mote(iv) Tmote Sky devices showed some strange behaviourwhen in use. When running an adaptive Beacon Order they seem to lose connec-tion if the jump between two consecutive beacons is too large. The phenomena wasobserved when going from B.O. 10 to 6. This can be the result of that the hardwareis not adapted to a frequent update of the B.O., hence missing a frame or two whenadapting to a new B.O. with a much smaller time interval than previous B.O.The experiments conducted do however show that it is possible to produce a wire-less feedback link running a self-triggered control strategy over a IEEE 802.15.4network. Keep in mind that this result is only for a system with a slow dynamic,which means that the Beacon Order adaptation might not work on a system with afast dynamic, with the same equipment used in this master thesis. For a fast systemwith a high demand on precision, the look-up tables and the problem with a missedframe during an large B.O. transition is disastrous. Hence, this thesis only provesthat the theory works on a slow system with a low demand on precision (when usingthese Mote(iv) Tmote Sky platforms).

7.2 Future work

In this master thesis it has been shown that it is possible to construct a feedbacklink over a wireless network. Some of the questions/problems that occurred alongthe way could be addressed in future work. Since the Mote(iv) Tmote Sky device’sinability to handle heavier computation and since newer versions have arrived, onepossibility would be to try a faster system with that equipment. With more ad-vanced equipment it would also be possible to get rid of the computer doing allcalculations and control. This could lead to the possibility to implement a self-triggered control on a SMART-grid with battery powered devices.The hardware’s inability to change its beacon order from a high value to a low valueis an area of future work. Since self-triggered control with its beacon order adap-tation is beginning to be acknowledged as a potential strategy in wireless sensornetworks, work groups are currently looking into different ways for the hardware toadapt the beacon order in smoother ways. However, since there is no solution yet,it could be an interesting area to investigate further.The IEEE 802.15.4 protocol presented some problems with its maximum BeaconOrder of 14. As mentioned in subsection 3.5, a new extension of the protocol isbeing developed. With it, new possibilities arises which changes the constraintsimposed on the self-trigger control. Hence an investigation of that protocol could

Page 49: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

7.2. FUTURE WORK 41

be of interest.The experiment phase only covered a single mote and a PAN coordinator. Thenext step is to implement a multiple hop network. To implement multiple motesconnected to a PAN coordinator forcing packages to do multiple hops to reach thePAN coordinator could be interesting. This would lead to more complex equationswhen computing the self-trigger control since all the time delays imposed by thenumber of hops will inflict on the result.As mentioned in section 3.4, DUST networks are using a TDMA scheduling totransfer data. To investigate the possibility to introduce a self-trigger control intoa TDMA scheduling could prove to be a better solution than the implementationon the IEEE 802.15.4 protocol.

Page 50: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today
Page 51: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

Bibliography

[1] Tinyos faq http://docs.tinyos.net/index.php/FAQ, 2010.

[2] Tinyos http://www.tinyos.net/, 2010.

[3] Task group 4e http://www.ieee802.org/15/pub/TG4e.html, 2011.

[4] A. Bonivento, C. Fischione, L. Necchi, F. Pianegiani, and A. Sangiovanni-Vincentelli. ”system level design for clustered wireless sensor networks”. InIEEE Transactions on Industrial Informatics, volume 3, pages 202–214. August2007.

[5] Ricardo Augusto Rodrigues da Silva Severino. On the use of ieee 802.15.4/zig-bee for time-sensitive wireless sensor network applications. Master’s thesis,Polytechnic Institute of Porto School of Engineering, October 2008.

[6] Sinem Coleri Ergen. Zigbee/ieee 802.15.4 summary. Technical report, 2004.

[7] C. Fischione, P. Park, P. Di Marco, and K. H. Johansson. ”Wireless BasedNetwork Control”, chapter ”Design Principles of Wireless Sensor Networks Pro-tocols for Control Applications”. Springer, 2010.

[8] Aitor Hernandez Herranz. Wireless process control using 802.15.4 protocol.Master’s thesis, Royal Institute of Technology, November 2010.

[9] Johnson Controls. Supervisory Controllers, 2009.

[10] Joonyoung Jung, Kiryong Ha, Jeonwoo Lee, Youngsung Kim, and DaeyoungKim. Wireless body area network in a ubiquitous healthcare system for physio-logical signal monitoring and health consulting. International Journal of SignalProcessing, Image Processing and Pattern Recognition, 2008.

[11] Chol Su Kang and Kuor Hsin Chang. Enhancement to ieee 802.15.4-2006 forindustrial markets. In Project: IEEE P802.15 Working Group for WirelessPersonal Area Networks (WPANs).

[12] Holger Karl and Andreas Willig. Protocols and Architectures for Wireless Sen-sor Networks. Wiley, sixth edition, 2007.

43

Page 52: Implementation of a self-triggered control strategy over IEEE 802.15472496/FULLTEXT01.pdf · The next step in wireless systems is Network Controlled Systems (NCS). On the market today

44 BIBLIOGRAPHY

[13] Gary Legg. "zigbee: Wireless technology for low-power sensor networks"http://www.eetimes.com/design/communications-design/4017853/ZigBee-Wireless-Technology-for-Low-Power-Sensor-Networks, 2004.

[14] Bernardo Maciel. Development of a detection system using a wireless sensornetwork. Master’s thesis, Royal Institute of Technology, August 2008.

[15] Moteiv Corporation. Tmote Sky: Datasheet.

[16] Marco Naeve. Ieee 802.15.4 mac overview.https://mentor.ieee.org/802.15/dcn/04/15-04-0218-01-004a-ieee802-15-4-mac-overview.ppt, 5 2004.

[17] Sauter. 55.014 AXS 111S: Actuator for unit valves, continuous.

[18] Freescale Semiconductor. Freescale ieee 802.15.4 / zigbee, software selectorguide, 6 2010. AN3403.

[19] A. Speranzon, C.Fischione, and K.H. Johansson. ”distributed and collaborativeestimation over wireless sensor networks”. In IEEE 45th Conference on Decisionand Control 2006 (IEEE CDC 06), San Diego, December 2006.

[20] U. Tiberi, C. Fischione, K.H. Johansson, and M.D. Di Benedetto. ”adaptiveself-triggered control over ieee 802.15.4 networks”. In IEEE 49th Conference onDecision and Control 2010 (IEEE CDC 10), Atlanta, Georgia, USA, December2010.

[21] U. Tiberi, C. Fischione, K.H. Johansson, and M.D. Di Benedetto. "systemlevel design of self-triggered ieee 802.15.4 networked control loops increases thenetwork capacity". In IFAC World Congress, Milan, Italy, August 2011.

[22] Ubaldo Tiberi. Analysis and design of IEEE 802.15.4 networked control sys-tems. PhD thesis, Dept. of Electrical and Information Engineering, Universityof L’Aquila, 2011.

[23] Jeff Tyson. How osi works http://computer.howstuffworks.com/osi.htm. HowStuff Works, 2011.