Sensor Deployment, Self- Organization, And Localization. Model of Sensor Nodes. Model of Sensor Nodes. WiSe

Transcription

1 Sensor Deployment, Self- Organization, And Localization Material taken from Sensor Network Operations by Shashi Phoa, Thomas La Porta and Christopher Griffin, John Wiley, /20/ Smart Wireless Sensor Systems 1 Model of Sensor Nodes A generic model for a wireless sensor node 2 Model of Sensor Nodes each sensor consists of four main components: sensing/adc unit the processing/storage unit the transceiver and the power supply 3 WiSe Lab@WMU; 1

2 Model of Sensor Nodes Each operation Sensing Processing Buffering transmitting costs a certain amount of energy 4 Model of Sensor Nodes The sensing/adc unit and the transceiving unit can support different operating modes, which correspond to different amounts of power consumed and work done. 5 Model of Sensor Nodes Normally we assume that the sensing/adc unit can operate in Active off modes The transceiving unit can operate in Active Idle off modes. 6 WiSe Lab@WMU; 2

3 Introduction Sensor network is to be able to self-form When randomly deployed to be able to organize into an efficient network Capable of gathering data in a useful and efficient manner Gathering data in a useful manner requires exact location 7 Introduction Location Once sensor know their location, and that of their neighbors, redundant sensors can be powered down to save energy Low energy communication paths may be established between nodes Coverage holes in the sensor network may be uncovered 8 A Scalable self-configuration and Adaptive Reconfiguration Scheme for Dense Sensor Networks Harshavardhan Sabbineni and Krishnendu Chakrabarty 9 WiSe Lab@WMU; 3

4 SCARE Algorithm distributes the set of nodes in the network into two subsets Coordinator Nodes Non-Coordinator Nodes Redundancy is Exploited Maintain the coverage Prolong the network lifetime 10 SCARE 11 SCARE When active nodes fail Energy Depletion Wear out SCARE Replaces inactive nodes 12 WiSe 4

5 Background Information Battery-powered sensor nodes have capabilities Sensing Communication processing 13 Background Information Wireless sensor networks are networks of large numbers of sensor nodes. Applications of such sensor networks include the Monitoring Inventory tracking Assembly line monitoring Target tracking in military systems 14 Background Information Deployment in a remote location Sensor nodes might fail Battery power An enemy attack Change in environmental conditions 15 WiSe Lab@WMU; 5

6 Background Information Replacement of each failed sensor node Expensive Infeasible 16 Background Information In such cases large number of redundant sensor nodes are deployed 17 Background Information The self-configuration of a large number of sensor nodes requires a distributed solution. In this section, we present a scalable selfconfiguration and an adaptive reconfiguration (SCARE) algorithm for distributed sensor networks. 18 WiSe Lab@WMU; 6

7 Background Information An effective self-configuration scheme should have the following characteristics. It should be completely distributed localized because a centralized solution is often not scalable 19 Background Information Simple without excessive message over-head because sensor nodes typically have limited energy resources Energy Efficient Require only a small number of nodes to stay awake Perform multihop routing Keep the other nodes in a sleep state 20 Background Information Distributed self-configuration scheme distributes the set of nodes in the sensor network into subsets coordinator nodes noncoordinator nodes 21 WiSe Lab@WMU; 7

8 Background Information Coordinator nodes stay awake provide coverage perform multihop routing in the network Noncoordinator Nodes Sleep 22 Background Information When nodes fail, SCARE adaptively reconfigures the network by selecting appropriate noncoordinator nodes to become coordinators and take over the role of failed coordinators This scheme only needs local topology information and uses simple data structures in its implementation 23 Topology Management A Solution The Geography informed energy conservation for ad hoc routing (GAF scheme [4]) uses geographic location information of the sensor nodes divides the network into fixed-size virtual square grids identifies redundant nodes within each virtual grid and switches off their radios to achieve energy savings. 24 WiSe Lab@WMU; 8

9 WiSe 9

10 WiSe 10

11 WiSe 11

12 34 Topology Management A Solution In contrast, SCARE achieves energy savings by selectively powering down some of the nodes that are within the sensing radius of a coordinator 35 Scheduling Scheme LEACH uses a TDMA based MAC protocol, and in order to maintain a balanced energy consumption, suggests that each node probabilistically become a cluster head. 36 WiSe Lab@WMU; 12

13 Scheduling Scheme LEACH protocol extended by Tian 2002 to design a coverage-preserving node scheduling scheme to achieve energy savings nodes advertise their position information in each round. Each node evaluates its eligibility to switch itself off by calculating its sensing area and comparing it with its neighbors's If a node's sensing area is hold by a union set of its neighbors's, then it turns itself off LEACH: Raghavendra and Singh Scheduling Scheme (Cont.) To prevent blind spots in coverage due to several eligible nodes switching themselves off simultaneously, a back-offbased scheduling is used. After the back-off interval has elapsed, nodes broadcast a status advertisement message to let other nodes know about their on/off status Thus, each node broadcasts two messages in each round 39 WiSe Lab@WMU; 13

14 Scheduling Scheme (Cont.) SCARE needs fewer than two messages per node on average during its operation. The scheme also utilizes location information of the nodes for its operation. SCARE only needs an estimate of the distance between the nodes. 40 Aggressive Sleep and waking The Topology management for energyefficient sensor networks (STEM scheme [6]) trades off latency for energy savings by putting nodes aggressively to sleep and waking them up only when there is data to forward. It uses a lower duty cycle for transmitting periodic beacons to wake up nodes when there is data to forward. STEM: Schurgers, Tsiatsis and Srivastava Aggressive Sleep and waking SCARE does not use a separate paging channel for self-configuration. SCARE can integrate well with STEM to achieve significant energy savings 42 WiSe Lab@WMU; 14

15 Sleep and waking Scheduling In Adaptive energy conservation routing for multihop ad hoc routing (AFECA), nodes listen to the channel for transmissions. AFECA usually tries to keep nodes awake when there are NOT too many neighbors in its radio range. AFECA: Xu and Estrin Sleep and waking Scheduling In order to deduce this information, each node has to listen to transmissions that are not meant for it In SCARE, however, nodes listen at only periodic intervals in order to determine their states AFECA: Xu and Estrin Sleep and waking Scheduling (Cont.) The Power-Aware Multi-Access Protocol with Signaling for Ad Hoc networks (PAMAS multiaccess protocol [10]) saves power by switching off the radio of a node when it is not transmitting or receiving Saves power when idle node just listening PAMAS: Raghavendra and Singh WiSe Lab@WMU; 15

16 More Comprehensive Approach The Span: An energy-efficient coordination algorithm for topology maintenance in ad hoc wireless networks [5] approach appears to be the most closely related to SCARE 46 More Comprehensive Approach Span attempts to save energy by switching off redundant nodes without losing the connectivity of the net-work. Nodes make decisions based on their local topology information. 47 More Comprehensive Approach However, SCARE differs from Span in that it uses distance estimates to determine the state of a node. 48 WiSe Lab@WMU; 16

17 More Comprehensive Approach (Cont.) Span uses a communication mechanism to obtain this information. Span was developed for ad hoc networks, its main focus is on Ensuring network connectivity through energy-efficient topology management. It is not directed toward ensuring the sensing coverage of a given region. 49 More Comprehensive Approach (Cont.) SCARE In addition to ensuring network connectivity and low-energy self-configuration It attempts to provide a high level of sensing coverage 50 More Comprehensive Approach (Cont.) A Performance of a novel self-organization protocol for wireless ad hoc sensor networks [11] uses TDMA-based selforganization scheme for sensor networks TDMA: Sohrabi and Pottie WiSe Lab@WMU; 17

18 More Comprehensive Approach (Cont.) Each node uses a super frame, similar to a TDMA frame, to schedule different time slots for different neighbors However, this scheme does not take advantage of the redundancy inherent in wireless sensor networks to power off some nodes TDMA: Sohrabi and Pottie Localization Work SCARE utilizes a localization scheme for periodic transmission of beacon signals and for the synchronization of the clock signals of sensor nodes. 53 Localization Work A number of such localization schemes have been proposed in the literature for sensor networks. These schemes use a special set of nodes called the reference nodes transmit beacon signals to let the sensor nodes self-estimate their position. 54 WiSe Lab@WMU; 18

19 Relevant Prior Work Traditionally, Global Positioning System (GPS [13]) receivers are used to estimate positions of the nodes in mobile ad hoc networks. GPS: Hoffman, Lichteneger and Collins Relevant Prior Work However, their high cost and the need for more precise location estimates make them unsuitable for sensor networks. It is expensive to add GPS capability to each device in dense sensor networks GPS does not work indoors GPS: Hoffman, Lichteneger and Collins Outline of SCARE SCARE is a decentralized algorithm that distributes all the nodes in the network into subsets Coordinator nodes Noncoordinator nodes 57 WiSe Lab@WMU; 19

20 Outline of SCARE Coordinator nodes stay awake and provide coverage and perform multihop routing in the network Noncoordinator nodes Sleep wake up periodically to check if they should become coordinators to replace failed coordinators. 58 Outline of SCARE SCARE achieves four desirable goals. It saves energy by selecting only a small number of nodes as coordinators and putting other nodes to sleep. It uses only local topology information for coordinator election 59 Outline of SCARE It provides nearly as much sensing coverage compared to the coverage obtained if all the nodes are awake. It preserves network connectivity by using a protocol based on CHECK and CHECK_REPLY messages. 60 WiSe Lab@WMU; 20

21 States of Node A sensor node executing the SCARE procedure can be in one of the following states: coordinator (C) noncoordinator (NC) eligible to be a coordinator (ETC) eligible to be a noncoordinator (ETNC) undecided (U) Basic Scheme In self-configuration based on SCARE, each node starts by generating a random number with uniform probability between 0 and WiSe Lab@WMU; 21

22 Basic Scheme A node becomes eligible to be a coordinator random number > than a threshold small percentage of the nodes become coordinators. 64 Basic Scheme The threshold value can be preset depending on the application. A higher value --> results in a small number of initial coordinator nodes. effect of delaying the convergence of the selfconfiguration algorithm better selection of coordinator nodes. 65 Basic Scheme Low value for the threshold implies that a high number of coordinator nodes are selected. This speedup the convergence of the protocol 66 WiSe Lab@WMU; 22

23 67 68 Basic Scheme A node that is eligible to be a coordinator waits for a random amount of time before declaring itself to be a coordinator by broadcasting a HELLO message. 69 WiSe Lab@WMU; 23

24 Basic Scheme This wait time can be chosen from a uniform distribution of values ε [T, NT] where T is a preset slot time and N is the number of neighbors of the node that are coordinators. Initially, N can be chosen to be a constant, for example Basic Scheme Upon receipt of a HELLO message, a sensor node compares its distance from the sender C of the HELLO message to its sensing ranges (sensing range of C). 72 WiSe Lab@WMU; 24

25 Basic Scheme A node within a distance s from a coordinator immediately becomes a noncoordinator node and stores the ID of the node that sent the HELLO message in its local cache. A node that is at a distance greater than s from C but within transmission range r becomes eligible to be a coordinator node WiSe Lab@WMU; 25

26 76 77 Network Partitioning Problem The basic scheme described above can sometimes result in a partitioning of the network 78 WiSe Lab@WMU; 26

27 79 Network Partitioning Problem Here, coordinator node F makes node A a non-coordinator. However, coordinator node D can communicate with F only through A. 80 Network Partitioning Problem G and K are coordinator nodes and B and C are noncoordinator nodes. This situation again results in network partitioning as nodes G and K cannot reach each other. 81 WiSe Lab@WMU; 27

28 82 Network Partitioning Problem In the basic scheme, if there is a network partition, a node might never receive a HELLO message e.g. node H. This results in the node waiting eternally for a HELLO message, which results in wastage of energy. 83 Network Partitioning Problem time-out value T off after which the nodes that are still undecided about their state can become eligible to become coordinator nodes. The time-out value can be chosen based on the probability threshold A lower value for the threshold means that the procedure converges quickly and needs a lower T off value and vice versa. 84 WiSe Lab@WMU; 28

29 85 86 Network Partitioning Problem To prevent the network partitioning that occurs due to the uncontrolled cases A node that initially receives a HELLO message from a coordinator node does not become a noncoordinator immediately and go to the sleep state. 87 WiSe Lab@WMU; 29

30 Network Partitioning Problem Instead, it continues to listen for messages from other coordinator nodes and remains in the "eligible to be a noncoordinator" (ETNC) state WiSe Lab@WMU; 30

31 Network Partitioning Problem A sensor node that is in the ETNC state can become a coordinator node in two cases: 91 Network Partitioning Problem 1. If it can connect two neighboring coordinator nodes that cannot reach each other in one or two hops. It can deduce this information from the HELLO messages it received earlier. node A, which is in the ETNC state, receives HELLO messages from node F and node D and decides to become a coordinator WiSe Lab@WMU; 31

32 94 Network Partitioning Problem 2. If it can connect two neighboring coordinator nodes that cannot reach each other in one or two hops via a node in the ETNC state. nodes B and C, that are in the ETNC state, receive HELLO messages from nodes G and K, respectively, and decide to become coordinators as there is no match between the node lists of G and K WiSe Lab@WMU; 32

33 97 Network Partitioning Problem To achieve this, each ETNC node sends a CHECK message. This CHECK message contains the neighbor list of the coordinator node that caused this node to be in the ETNC state. 98 Network Partitioning Problem Intuitively, this case is more likely to occur if there are few coordinators in the surrounding area and less likely if there are more coordinator neighbors. 99 WiSe Lab@WMU; 33

34 Network Partitioning Problem Any ETNC node that receives this CHECK message replies with a CHECK_REPLY message and becomes a coordinator if there is no node common to the neighbor lists of both the nodes. Upon receipt of the CHECK_REPLY message, the node that sent the CHECK message also becomes a coordinator. 100 Network Partitioning Problem To prevent oscillations during the selection of coordinators once a node becomes a coordinator, it continues to remain a coordinator until it is unable to provide any service. 101 Network Partitioning Problem As the density of nodes increases, the fraction of noncoordinator nodes increases, and this leads to more energy savings. 102 WiSe Lab@WMU; 34

35 Network Partitioning Problem SCARE has a slightly larger number of coordinators than the minimum number necessary for coverage and connectivity. This also happens due to the randomness involved in the distributed selection of coordinator nodes. 103 Network Partitioning Problem After self-configuration, each coordinator periodically broadcasts a HELLO message along with a list of its one-hop neighbors that are coordinators. Noncoordinator nodes listen to these messages. 104 Network Partitioning Problem Noncoordinator nodes also maintain a timer to keep track of the coordinator node that made them a noncoordinator. 105 WiSe Lab@WMU; 35

36 Network Partitioning Problem If this timer goes off, a noncoordinator node assumes that the corresponding coordinator node has failed and then non coordinator node goes into an undecided state. This results in noncoordinator nodes becoming eligible to become coordinators 106 Network Partitioning Problem SCARE can also be applied to mobile sensor networks. A node that has moved to a new location is treated in the same way as the appearance of a new node at that location. It sets itself to the undecided state and listens to the network until either the timer T off goes off or it receives a HELLO message. 107 Network Partitioning Problem Similarly, when a node moves away from one location, this is treated as a node failure by its neighbors. 108 WiSe Lab@WMU; 36

37 Network Partitioning Problem Failure of noncoordinator nodes does not result in any change in the topology. However, the movement of coordinator nodes is detected by the noncoordinator nodes, and this makes them eligible to subsequently become coordinators 109 Rules of SCARE A set of control rules governs the state of the sensor node, while a set of defer rules decide when a node should postpone its decision. Timeout rules specify the time after which sensor nodes should make a decision. 110 States of Node A sensor node executing the SCARE procedure can be in one of the following states: coordinator (C) noncoordinator (NC) eligible to be a coordinator (ETC) eligible to be a noncoordinator (ETNC) undecided (U) 111 WiSe Lab@WMU; 37

38 112 Time Setup Periods The ETC and ETNC states are temporary and exist only during the T setup period. 113 Time Setup Periods There are seven timeout values in SCARE: 114 WiSe Lab@WMU; 38

39 Time Setup Periods 1.T off Time after which a node that is in the undecided state about its state becomes eligible to be a coordinator and goes into the ETC state. 2.T rand Time for which the sensor node that is in ETC state waits before becoming a coordinator. It then sends a HELLO message along with all its coordinator neighbors that it has identified. 115 Time Setup Periods 3. T runtime After every T runtime units of time, all noncoordinator nodes wakeup and listen. 4.T setup Time interval for which the noncoordinator nodes wake up and listen, after which they go to sleep if they still remain noncoordinators. This is also the period during which beacon messages are sent to synchronize the nodes. 5.T coord Time interval during which only the coordinators send HELLO messages. This occurs at the beginning of the T setup period. 116 Time Setup Periods 6. T noncoord Time interval during which only the noncoordinators send messages. This is the latter part of the T setup period. This period starts immediately after the T coord period ends. 7. T failure A noncoordinator node waits for time T failure for the HELLO messages from the coordinator node that made it the noncoordinator. If no HELLO message is received within this time interval, it decides that the corresponding coordinator node has failed and sets its state to undecided. 117 WiSe Lab@WMU; 39

40 The Type Of Messages In More Detail There are three types of messages in SCARE: HELLO CHECK CHECK-REPLY 118 The Type Of Messages In More Detail HELLO These messages are sent by coordinators. They also contain a list of the one-hop coordinator neighbors of the sender node. 119 The Type Of Messages In More Detail CHECK These messages are periodically sent by the noncoordinator nodes. They are used to remove the potential network partitions. Each CHECK message also contains of list of coordinator neighbors of the node that made it the noncoordinator. 120 WiSe Lab@WMU; 40

41 The Type Of Messages In More Detail CHECK-REPLY Upon receipt of a CHECK message, noncoordinator node compares the coordinator neighbor list included in the CHECK message with the neighbor list of the node that made it a noncoordinator. If there are no common entries in the two lists, it sends a CHECK-REPLY message. 121 The Type Of Messages In More Detail A noncoordinator node becomes a coordinator node if two coordinators at the end of the T coord period cannot reach each other within one or two hops. 122 Rules That Decide The State of the Sensor Node Recall that we used r to denote the transmission radius of a node. Similarly, s the sensing radius of a node. 123 WiSe Lab@WMU; 41

42 Rules That Decide The State of the Sensor Node The control rules that decide the state of the sensor node are as follows: 1.A sensor node that generates a random number between 0 and 1, and greater than a threshold, becomes a coordinator. 2.A sensor node that lies at a distance between s and r of a coordinator node becomes eligible to become a coordinator node and goes into the ETC state. 124 Rules That Decide The State of the Sensor Node 3. A sensor node that lies at a distance at most s from a coordinator node becomes eligible to become a noncoordinator node and goes into the ETNC state. 125 Rules That Decide The State of the Sensor Node 4. A sensor node that is in ETNC state listens to the HELLO messages sent by the coordinator nodes for the T coord period. From this list of coordinator nodes contained in the HELLO messages, if it determines that two coordinator nodes do not have a common neighbor that is a coordinator, this (ETNC) node becomes a coordinator at the end of the T coord period. On the other hand, if there are common neighbors in the node lists, then the node stays in the ETNC state. 126 WiSe Lab@WMU; 42

43 Rules That Decide The State of the Sensor Node 5. A sensor node that is in the ETNC state at the end of T coord period broadcasts a CHECK message. This message contains a list of the coordinator neighbors of the node that caused it to go to the ETNC state. 127 Rules That Decide The State of the Sensor Node 6. A sensor node that receives a CHECK message compares the list of neighbors in the CHECK message with its neighbor list. If there is no match between the two lists, it transmits a CHECK_REPLY message to the sender of the CHECK message. 128 Rules That Decide The State of the Sensor Node 7. Upon receipt of a CHECK_REPLY to its CHECK message, a sender node that is in the ETNC state becomes a coordinator node. The node that sent the CHECK_REPLY also becomes a coordinator. 8. A sensor node that is in the ETNC state and does not satisfy conditions 4 and 5 becomes a noncoordinator node at the end of the setup period. 129 WiSe Lab@WMU; 43

44 Rules That Decide The State of the Sensor Node 9. A sensor node that is in the ETC state becomes a coordinator node after the Tcoord period if it does not become a noncoordinator node due to the selection of some other coordinator node. 10. A sensor node with data to send can opt to become a coordinator for as long as it has data to transmit. 130 The defer rules for SCARE 1. If a node becomes eligible to be a coordinator, it listens for T rand period 2. If a node becomes eligible to be a noncoordinator at the end of the T coord period, it listens for time T noncood period. 131 The timeout rules 1. A sensor node at the end of the T rand period broadcasts a HELLO message. 2. A sensor node at the end of the T setup period becomes a noncoordinator if it is still eligible to be a noncoordinator. 132 WiSe Lab@WMU; 44

45 The timeout rules 3. A sensor node at the end of the T coord becomes a coordinator if it is still eligible to become a coordinator. 4. A sensor node wakes up and listens to the medium after the timer T runtirne expires. 5. After its T off timer expires, a sensor node becomes eligible to become a coordinator if it is still undecided about its state WiSe Lab@WMU; 45

46 WiSe 46

47 WiSe 47

48 Ensuring Network Connectivity We next discuss how SCARE prevents network partitioning. Let S be a set of nodes containing the partial set of coordinators that are connected and the associated nodes in the ETNC state. Each coordinator in set S can reach any other coordinator in set S in a finite number of hops. 142 Ensuring Network Connectivity Let X denote the region enclosing the nodes present in set S. Now consider a node not in set S. Any node not present in S can lead to the following scenarios. We use the notation P A to represent the area within the transmission range of node A. 143 Ensuring Network Connectivity 144 WiSe Lab@WMU; 48

49 Ensuring Network Connectivity Coordinator B outside the region X but within the transmission range of the coordinator A in region X In this case, both the coordinators can reach each other and the set S = S U {B} and the region X expands to include the coordinator B. 145 Ensuring Network Connectivity Coordinator B is outside the transmission range of the coordinator A but is within the transmission range of ETNC node C 146 Ensuring Network Connectivity 147 WiSe Lab@WMU; 49

50 Ensuring Network Connectivity However, as node C listens to the HELLO messages from both coordinator nodes A and B, it becomes a coordinator if there is no other path from A to B by becoming a coordinator. Now this reduces to (case 1) with coordinators C and B within reach of each other. C becomes a coordinator, and the region X expands to include the coordinator B, that is, S = S U {B}. 148 Ensuring Network Connectivity Coordinator B is outside the transmission range of coordinator A. 149 Ensuring Network Connectivity 150 WiSe Lab@WMU; 50

51 Ensuring Network Connectivity However, node C in ETNC state due to node B is within the reach of coordinator A Node C listens to HELLO messages from A and B, and it becomes a coordinator. Now, A and C are within reach of each other, and this reduces to case 2; hence S = S U {C}. By a similar procedure, node B is also included. 151 Ensuring Network Connectivity Coordinator B and coordinator A cannot reach each other 152 Ensuring Network Connectivity Q B P A B A C D X 153 WiSe Lab@WMU; 51

52 Ensuring Network Connectivity However, nodes C and D that are in ETNC state can reach other. Node C and node D send and receive CHECK and CHECK-REPLY messages and become coordinators if there is no other path from node A to B. 154 Ensuring Network Connectivity Once C becomes a coordinator, coordinator C in region X and coordinator D outside region X are within reach of each other. This reduces to case 2 and S = S U {D). Region X expands to include node B and node D. 155 Ensuring Network Connectivity The result of applying SCARE to an example sensor network with 100 randomly deployed node in a 100-m x 100-ni grid. radio range of 25 m 156 WiSe Lab@WMU; 52

53 Ensuring Network Connectivity Timeout values of T falture of 3 S T coord of 3 s T noncoord of 2 s T setup of 5 s T runtime of 95 s. 157 Ensuring Network Connectivity SCARE selects 32 nodes as coordinators and the rest are designated as noncoordinators. 158 SCARE 159 WiSe Lab@WMU; 53

54 160 WiSe 54