Timeout-based Information Forwarding Protocol for Wireless Sensor ...

International Journal of Distributed Sensor Networks, 3: 331–346, 2007 Copyright © Taylor & Francis Group, LLC ISSN: 1550-1329 print/1550-1477 online DOI: 10.1080/15501320601062148

Timeout-based Information Forwarding Protocol for Wireless Sensor Networks

1550-1477 Journal of Distributed Sensor Networks 1550-1329 UDSN International Networks, Vol. 3, No. 4, September 2007: pp. 1–37

Timeout-based W. Jeong and S.Information Y. Nof Forwarding Protocol

WOOTAE JEONG and SHIMON Y. NOF PRISM Center, School of Industrial Engineering, 315 N. Grant St. Purdue University, W. Lafayette, IN

Recent wireless microsensor network protocols provide more flexible leverage to the applications with dynamically changing topology, but they should be designed to overcome energy constraints, the bandwidth limit, and system latency. Thus, microsensor network protocols should be effective both in energy and in latency. In addition, they should be evaluated through designated tools at each level of networking characteristics. This paper proposes a new Timeout-based Information Forwarding (TIF) protocol for wireless sensor networks. It uses a relatively simple logic to forward the data packet with multi-hop fashion to reduce the overall network energy consumption. The TIF protocol has been implemented into a network evaluation tool, called TIE/MEMS, and provides a design strategy for distributed wireless sensor network systems needed for various emerging applications. The simulated results show that the TIF protocol has low energy consumption and provides design guidelines between energy consumption and latency according to the number of hops by adjusting weight values in the timeout function. Keywords Communication Protocols; Wireless Communication; Microsensor Networks; Network Protocols; Fault Tolerance; Teamwork Integration Evaluation (TIE); Task Administration

1. Introduction Wireless microsensor networks (WSNs) are recently envisioned to provide a seamless link between the physical world and the global information infrastructure. Typically, they are densely deployed with limited power sources and limited capabilities, but networks can produce more widely accessible, reliable, and accurate information than the sum of the individual sensors for decisions about physical environments. However, in order to be used in most realistic applications, distributed sensor network systems should be designed with application-specific communication algorithms and task administration protocols because wireless microsensors are extremely resourceconstrained. Therefore, most research efforts are focused on the network performance evaluation and the trade-off between energy consumption and network parameters such as node density, radio transmission range, network coverage, latency, and distribution. In general, a distributed sensor network (DSN) should provide the following main features: Fault-tolerance: Multiple, redundant sensors increase reliability in case of sensor errors and network link failures. In this case, a DSN enables enough redundancy with the data from different routes or nodes that the system may still produce acceptable quality information. Accuracy Improvement: Redundancy of information can reduce the overall uncertainty and increase the accuracy with which events are perceived. Since nodes located close

Address correspondence to Shimon Y. Nof, PRISM Center, School of Industrial Engineering, 315 N. Grant St., Purdue University, W. Lafayette, IN 47907-2023, USA. Tel.: 765-494-5427, Fax: 765-494-1299. E-mail: [email protected]

331

332

W. Jeong and S. Y. Nof

to each other are combining information about the same event, fused data improve the quality of the event information. Timeliness: DSN can provide the processing parallelism that may be needed to achieve an effective integration process, either at actual speed that a single sensor could provide, or even at a faster operation speed. Network Topology: A high number of nodes deployed throughout the sensory field should be maintained by topology management schemes because any changes in sensor nodes and their deployments affect the overall performance of DSN. Therefore, a flexible and simple topology is usually preferred. Energy Consumption: Since each wireless sensor node is working with a limited power source, the design of power-saving protocols and algorithms is a significant issue for providing longer lifetime of sensor network system. Scalability: A coverage area of a sensor network system depends on the transmission range of each node and the density of the deployed sensors. The density of the deployed nodes should be designed carefully to provide a topology appropriate for the specific application. The network density r can be expressed according to [1], [2] as

r ( R) = ( N ⋅ p ⋅ R 2 ) / A

(1)

where N is the number of scattered sensor nodes in area A, and R is the wireless transmission range with circular propagation. Many factors and communication algorithms in WSNs influence the performance of a wireless communication network. This research proposes simple data forwarding protocol for multi-hop WSNs by applying a timeout-based control and a geometric data forwarding scheme. It includes a coordination scheme of the sensor node activation to save limited battery power without loss of network connectivity. In Section 2, relative research will be examined. The communication model and the data routing scheme of timeout-based information forwarding protocol will be explained in Section 3. In Section 4, the performance of the proposed protocol will be evaluated through analysis and simulations. In Section 5, we will provide a brief discussion and conclusion from the results.

2. Related Work A number of protocols have been developed to prolong the network lifetime and reduce the energy consumption in forwarding data. In order to minimize the energy consumption, these protocols can be broadly classified into two categories; 1. effective data forwarding scheme, 2. sensor node activation scheme. The effective data forwarding scheme provides appropriate routing paths for various network constraints such as energy, fault-tolerance, and latency. The purpose of the sensor node activation scheme is to manage the sensor node activation scheme (sleep schedule). Figure 1 illustrates basic data forwarding/routing architectures for transmitting information to a base station (BS) in a single-hop and multi-hop fashion. As effective data forwarding schemes, clustering or hierarchical techniques can help in reducing the energy consumption for WSNs that require scalability to densely distributed sensor nodes [3]. It consists of a set of sensor nodes, a set of cluster-head (CH) nodes, and a communication network interconnecting the nodes [4], [5]. In general, one sensor

Timeout-based Information Forwarding Protocol

333

BS BS

(a) BS

(b) Cluster BS

(c)

Cluster head

Cluster head

(d)

FIGURE 1 Four basic WSN data forwarding architectures.

node communicates with more than one CH and a set of nodes communicating with a CH is defined as a cluster. Clustering architecture in the context of routing protocols can increase the system load balancing and enable better resource allocation [6]–[9]. However, local partitions, in general, do not reduce the energy consumption because the net communication distance between each sensor and the base station does not change. Therefore, the clustering in energy constrained network considers coordinated node management with localization/self-organization techniques. As an example of clustering protocols, Heinzelman [10] proposed the Low-Energy Adaptive Clustering Hierarchy (LEACH), which is a clustering-based protocol that utilizes a randomized rotation of local cluster base stations to evenly distribute the energy load of sensors in DSN. In order to distribute the energy load among the cluster, LEACH elects a different cluster-head at different time intervals, which depends on the amount of energy left at the node. That is, each node chooses a random number between 0 and 1. If this random number is less than the threshold T(n), the sensor node is a cluster head. T(n) is calculated as

P ⎧ if n ∈ G, ⎪ T (n) = ⎨1 − p[r mod(1 / P )] ⎪ 0 otherwise ⎩

(2)

where P is the desired percentage to become a cluster head; r, the current round; and G, the set of nodes that have not being selected as a cluster head in the last 1/P rounds. A Resource Oriented Protocol (ROP) [11] also uses a hierarchical clustering structure, and its network topology is adaptively formed according to the resources of the cluster member nodes. To achieve a longer system lifetime in heterogeneous networks, the ROP balances the resource among the sensors by maintaining a routing cache. The energy-balance is also addressed in distributed energy-balanced algorithms [12] for a single-hop wireless sensor network. The optimal energy balanced algorithms for a singlehop communication must be extended to multi-hop routing applications for the system

334


level energy optimization, even in the clustering architecture. Therefore, multi-hop routing protocols have also been developed to extend the lifetime of the sensing nodes in a wireless network [13], [14] because a single hop propagation scheme consumes high energy especially for longer distance transmission. For example, a Minimum Transmission Energy (MTE) routing protocol [13] chooses intermediate nodes such that the sum of the squared distances is minimized by assuming a square of distance power loss between two nodes. However, this protocol results in an unbalanced death of nodes with respect to the entire network. In order to route the information in an energy efficient way, the Directed Diffusion routing protocol based on the localized computation model [15], has been discussed for robust communication. The consumer of the data will initiate requests for data with certain attributes. Nodes will then diffuse the requests towards producers via a sequence of local interactions. This process sets up gradients in the network which channel the delivery of the data. Even though the network status is dynamic, the impact of the dynamics can be localized. Probabilistic Forwarding Protocol (PFR) [16] also provides a strong energy efficient scheme by probabilistically favoring certain close to optimal paths towards the sink. The sub-logic of this scheme, such as the calculation of a threshold angle is relatively complicated to be implemented into microsensors operating with a tiny memory and processor. The proposed protocol in our research uses a relatively simple information propagation scheme. In the multi-hop network applications, adaptive random backoff protocols [17] have been proposed to provide trade-offs between fault-tolerance (collision avoidance) and network efficiency (time, energy). These random backoff protocols reduce collisions and the latency introduced by the backoff mechanism, but the transmission energy consumption is, in general, inversely proportional to the end-to-end latency of multi-hop communication. Thus, it should be further extended to deliver a proper trade-off between energy and latency. Besides information routing protocols in WSNs, the proper management of the sensor node activation has been studied to increase the lifetime of the network by turning off the radio to prevent overhearing [18]–[20]. The duration for a node to remain powered off should depend on the residual energy, the transmission range, and the response time of each node. For example, S-MAC, a medium-access control (MAC) protocol [13] enables a traffic-adaptive wake-up in a multi-hop network. It uses in-channel signaling to avoid overhearing unnecessary traffic and reduces the control overhead. Energy efficiency of various sleep scheduling schemes in WSNs has been analyzed in [20]. In general, a distance-based scheduling scheme and a balanced-energy scheduling scheme provide a longer network lifetime and more energy savings than the conventional randomized scheduling scheme. In addition to the coordinated sleeping scheduling, the timeout based node utilization techniques ensure that when any tasks keep the resource idle for too long, their exclusive service by the resource is disabled. That is, the time-based control protocol is intended to provide a rational collaboration rule among the tasks and resources in the networked system [21], [22]. Here, slow sensors will delay the timely response and other sensors may need to consume extra energy. Finally, these coordinated sleeping and timely activated node schemes result in significant energy savings compared with an 802.11 MAC without controlling the sleeping/activating schedule.

3. Timeout-based Information Forwarding Protocol The main goal of our timeout-based information forwarding (TIF) protocol is to reduce the energy consumption in data delivery through minimal overhead communication without


335

centralized control. There are always trade-offs between reducing the energy consumption and the increased latency in the WSNs because the sophisticated information forwarding algorithms to save energy consumption tend to require excessive overhead exchange, and then increase the latency in the network system. Therefore, the TIF algorithm simplifies the overhead exchanges to select proper candidates in forwarding sensory information during the contending medium from sending nodes. 3.1. Assumptions Since WSNs have different characteristics than the traditional ad hoc network, we make some assumptions about the sensor network: a large number of sensors are densely deployed with a redundancy in an ad hoc fashion. When the sensor nodes are uniformly distributed in area A with a uniform density, it can be assumed that these sensor nodes are randomly distributed as a two-dimensional Poisson distribution with density m. Therefore, the expected number of sensors found in the area A becomes mA, and the probability of finding n nodes in a region of area A is equal to

P ( n) =

(m A)n ⋅ e − m A n!

(3)

Randomly distributed nodes, rather than positioned deterministically, are stationary and must be pre-configured. Nodes can be pre-configured with a GPS, localization algorithms, and distance estimation. For example, when the GPS is not feasible, the localization algorithms as addressed in [23] can provide the location of nodes during networking. Note that we only assume that each node knows its position and location of the base station in the network area. Relative locations of neighbor nodes can be identified during the contending medium through RTS and CTS control format of IEEE 802.11 MAC protocol scheme. Therefore, each node can have information of the neighbor nodes, the distance to the sink node, and the location of the final destination of data, i.e., the base station. We assume that a radio signal is transmitted by a line of sight, and then a two degree power transmission model (a = 2) from the Equation (4) can be applied. A relatively long radio communication (a > 2) can be applied in the same manner. Due to the short radio range, all data packets are transmitted over in a multi-hop fashion. In addition, a symmetric communication channel is assumed because all sensor nodes used are homogeneous and they can communicate using the same transmission power. In addition, clocks, i.e., local times of all sensor nodes are synchronized. Otherwise, the clock drifts on each node and can cause synchronization errors. Thus, well-known network time synchronization schemes such as the Network Time Protocol (NTP) [24] can be applied. 3.2. Wireless Sensor Communication Model The wireless microsensor node consists of a sensing unit, a processing unit, a power unit, and communication elements. The sensing unit is an electrical part detecting the physical variable from the environment. The processing unit (a tiny microprocessor) performs signal processing functions, i.e. integrating data and computation required in the processing of information. The communication elements consist of a receiver, a transmitter, and an amplifier if needed. The power unit provides energy source with other units (see Fig. 2). Basically, all individual sensor nodes are operated by a limited battery, but a base station node as a final data collecting center can be modeled with an unlimited energy source.

336


Node

Processing Unit (Processor, Memory)

Receiver

Node

Sensing Unit (Sensor, ADC)

Transmitter

Amplifier

Power Unit

Node

FIGURE 2 A wireless micro-node model. Each node has sensing unit, processing unit, power unit, and communication units.

Power is mostly consumed when nodes are in an active mode. Thus, limited energy in each node can be saved by turning off the radio during idle time. In addition to the wireless micro-node model, it is important to adopt a good communication model to use the energy efficiently in the microsensor network. In this research, the communicational and computational energy models among nodes are based on the models used in [10]. In this model, the power attenuation is dependent on the distance, d, between the transmitter and the receiver in the microsensor nodes. For a relatively short distance, the propagation energy loss is inversely proportional to d2 while for a long distance, it is inversely proportional to d4. Thus, each transmitting microsensor node should amplify the power to ensure the signal at the receiving node. Assuming that node A transmits a k-bit data packet with a distance d to the node B, node A dissipates energy as follows:

Et = max( Emin , k ⋅ (a + b × r a )),

2≤a≤4

(4)

For the receiving node B, it should expend energy as follows:

Er = k × c

(5)

Where the factors, a and c, are coefficients depending on the electronics energy and b is related to the transmit amplifying energy. The electronics energy is power consumption in the digital coding, modulation, and filtering of the signal before it is sent to the transmit amplifier. The amplifying energy factor, b, depends on the required receiver sensitivity and the receiver noise ratio. The Emin is a minimum radio transmission energy and the α is a degree of radio transmission which depends on the transmitting range for each node. From Equations (4) and (5), therefore, the amount of energy required to transmit packets between two sensor nodes can be computed by

E (k,r ) = Et (k,r ) + Er (k ) = max( Emin ,k • (a + b × r α )) + k × c

(6)

The total energy consumption in the entire WSN communication becomes a sum of energy dissipation among the nodes and can be calculated by


Etotal = ∑ E (k,ri ) = ∑ ( Et (k,r ) + Er (k )), i ∈Γ

337

∀ i ∈Γ

i ∈Γ

= ∑ (max( Emin ,k • (a + b × riα )) + k × c)

(7)

i ∈Γ

where the entire number of network links, Γ, can be increased in multiple short communication ranges, i.e., a multi-hop wireless sensor network. The communication energy model adopted here plays an important role not only in generating clusters in the sensor network, but also in deciding the routing path under the failure of the communication link, because the priority of the routing node is a function of response time from nodes in this time-based control protocol. 3.3. Geometrical Information Forwarding The geometrical information forwarding scheme is shown in Fig. 3. Assume that the data packet is forwarded from the source node, S, to the data sink node, or base station (BS). Due to the limited radio transmission range, r, in the node S, data packet should be sent to the base station in a multi-hop fashion, i.e., via intermediate node B or C. Nodes within a virtual circle K with a diameter d, the distance between node S and the BS, become initial forwarding candidates because the nodes in the virtual circle K reduces the transmission energy when a two degree power consumption model is used from Equation (4). For example, when the node S forwards data packets to the BS, node B, C, and E within the circle K can be considered as the first forwarding candidates because they satisfy the following equation 2

2

( N S N A ) + ( N A N BS ) ≤ ( N S N BS )

2

(8)

where ( N S N A ) is the distance between node S and node A. That is, since all nodes contain information about their locations, sending these nodes can decide the first forwarding candidates from the circular region K defined by the equation (8). However, since node S has a limited radio transmission range, r, node B or node C in the overlapping area becomes prior forwarding candidates rather than the node E. Of course, the transmission range of sending node S is equal to or greater than d (r ≥ d), node

Forwarding Zone C F d

S B

E

Base station (BS)

D

Range of S

Circle K

A

FIGURE 3 Geometric data forwarding scheme in the TIF protocol.

338


S forwards data packets directly to the BS. If there exist no nodes in the overlapping area, node S may fail to forward the data packet or need to route the data by forwarding it to other nodes in its transmission range r, e.g., node F. Otherwise, the sender should increase its transmission range, r, until it finds the proper forwarding candidate in the overlapping area. In the high dense and redundant sensor distribution, the forwarding candidates in the overlapping area exist with a high probability, especially, in the two-dimensional Poisson distribution. It would reduce the forwarding failure rate in the WSNs. The network sensitivity with respect to sensor density and the transmission range of the sensors will be analyzed in the simulation. 3.4. Timeout Control Scheme Besides the geometric data forwarding scheme, a timeout-based control scheme in this research is proposed for the MAC scheme in the network scheduling and control. The timeout-based control scheme has been broadly used in the area of distributed collaborated systems for better resource allocation. The basic idea of the timeout-based protocol is that sending nodes stop waiting responds from candidate nodes, and select the next forwarding node with the shortest timeout values (CTS message). That is, the timeout-based control protocol is intended to provide a rational collaboration rule among tasks and resources in the networked system [25]. Here, slow sensors will delay the timely response and other sensors may need to consume extra energy. The patented FTTP1 uses the basic concept of the timeout control scheme effectively in a microsensor communication control [25]. This timeout based control was successfully implemented in a microsensor network system by utilizing sensing nodes with the shortest response time order [22]. The communication contention mechanism in the timeout control scheme is based on the IEEE 802.11, i.e., using RTS (Request-To-Send) and CTS (Clear-To-Send) control packets to solve the hidden node problem and collision. The MAC header format of the timeout control scheme, however, contains the location of the node itself and the base station. This additional location information in the MAC header is exchanged with RTS and CTS packets, and it is also used to calculate relative distances between nodes as shown in Equation (8). When the sender S forwards a data frame to all neighbors within its transmission range, it senses the medium until it is not busy. If it senses a clear medium, it waits an additional amount of time defined by a short inter-frame space (SIFS), and a DFC inter-frame space (DIFS) that is longer than SIFS. After DIFS, if the medium is still clear, the sender S sends RTS. The RTS frame includes a duration value specifying how much time it needs for sending. In this sensor network, the location information of the node itself and BS is also carried in the RTS frame. Therefore, the receiving nodes compute and compare distances by using Equation (8). If they satisfy the condition, the receiving nodes become prior forwarding candidates, i.e., nodes in the overlapping area in Fig. 3. Otherwise, any other nodes within the transmission range of S can be the forwarding candidate, e.g., node F. Once the candidates receive the RTS frame, they set a timer to calculate the amount of time that elapses before sending a CTS packet to the sender. The CTS frame includes this amount of time. We define this amount of time as a timeout threshold which helps to select the most effective forwarding candidate considering the entire networking efficiency. The details of handing the timeout are described in the next section.

1

Fault-Tolerant Timeout Protocol (FTTP) is a patent pending protocol of the PRISM Center at Purdue University.


339

When the sender gets the CTS, it gets the permission to start sending the data packet. Since other nodes will know the duration requested, we have collision avoidance. If no CTS is received by the sender for sometime, it assumes that a collision has happened to it. When the receiver node gets the data packet, it returns the ACK frame to the sender. The process is repeated whenever information needs to be forwarded to the BS and is depicted in Fig. 4. In the information forwarding scheme, sensor nodes within the transmission range of sender S but not selected as a final forwarding node, e.g., sensor F, set their Network Allocation Vector (NAV) when they hear that the sender S has received the CTS from the other candidate. It avoids multiple CTS collision and energy consumption in trials to send out CTS continuously. The updated NAV value maintains a prediction of future traffic on the medium based on the timeout threshold information that is announced in the RTS/CTS frames prior to the actual exchange of data. Data implosion may be another potential problem in the TIF when there exist multiple senders and a forwarding candidate receives a duplicated data packet from different senders because densely distributed nodes may measure the same physical value and deliver it to the BS through multiple intermediate nodes. However, it is difficult to avoid the implosion problem during TIF control because the data forwarding sensors do not control duplicated data packet during transmission. Thus, we assume that forwarding sensors have some capabilities to compress duplicated data packets with their tiny processors. Otherwise, information routing techniques, e.g. Sensor Protocols for Information via Negotiation (SPIN), can be used to solve the implosion problem during information dissemination. 3.5. Handling Timeout As described in our TIF control scheme, a timeout threshold, i.e., the duration of time to reply to RTS, can handle the time duration to reply for the RTS of the sender. The main purpose of using the time threshold is to use sensors that have fast responses and more residual energy with a higher probability. That is, the more remaining energy it has, the shorter responding time it replies with. If there exist many forwarding candidates within the transmission range, it assigns a high priority for the sensor that contains more remaining energy. Otherwise, it selects the sensor located on the path to consume less transmission

Start sensing the medium (idle) DIFS Source S

SIFS RTS

SIFS

DIFS

DATA Time

Tr(B) CTS

Sensor B Tr(C) Sensor C

ACK

NAV

Tr(F) NAV

Sensor F Defers access

FIGURE 4 Timeline of the TIF control scheme.

340


energy. This timeout also can help to tolerate faulty sensors, e.g., dead sensors with exhausted energy and malfunction, which can not respond properly to the RTS packet during the TIF handshake scheme. Therefore, the timeout threshold, Tr, is defined and calculated by

Dr ⎞ ⎛ Er Tr = 1 − ⎜ ⋅ we + ⋅ wd + Fr ⋅ w f ⎟ ⎠ ⎝ E D

(9)

E: Initial energy implemented at each node, Er: Remaining energy at the forwarding node, D: The distance between the current node and the base station, Dr: The distance between the sending node and the forwarding candidate, Fr: The probability of node failure or packet loss We, Wd, Wf : Weighting factors for energy, distance, node failure respectively. (We + Wd + Wf = 1) The timeout threshold must not exceed the DIF Inter Frame Space (DIFS) of IEEE 802.11 because other sensors waiting for sending their data may interfere with the ongoing data transferring schedule.

4. Performance Evaluation and Simulation Results The main goal of simulation described here is to measure the energy and the data forwarding efficiency of the proposed TIF scheme. First of all, network sensitivity with respect to the sensor density and the sensor transmission range is measured on the TIF scheme we proposed. Then, the overall energy consumption, the packet loss rate, and the number of hops during networking are analyzed. Various simulation tools, e.g., a network simulator, NS2 [26], have been used to simulate network protocols, but in this paper, Teamwork Integrated Evaluator (TIE) [21] has been improved by the TIF, which has a concrete timebased communication control and network energy model. The enhanced TIF simulator is programmed by a Matlab tool. 4.1. Simulation Setup A network environment variables and parameters must be specified before executing the network communication protocol. Constants and units used in the network should be checked as well. Randomly distributed microsensor nodes were used for two-dimensional network space which was set to 200m × 200m. The sink node or base station was placed at (200m, 200m) and 5 nodes located at (0,0), (0,100), (0,200), (100,0), (200,0) respectively were selected as source nodes to validate the performance of the TIF scheme effectively. Note that any nodes can be used as source nodes, but the five source nodes used in the simulation were selected to evaluate the performance of the proposed protocol intuitively. That is, the five source nodes require at least several hops to deliver their sensory data to the base station. Each source node transmits data packets (500 byte) to the base station in a multi-hop manner because of their limited transmission range. The basic packet forwarding scheme follows the TIF described in Section 3. It contains the basic characteristics of the simplified IEEE 802.11. For energy consumption between the two nodes, a two degree energy dissipation model was used. The amplifying energy and the transmitting energy


341

TABLE 1 Summary of Parameters in the Microsensor Network Simulator Parameters Network size (M) Number of nodes (N) Radio transmission range Data packet size Packets sent Base station (Sink node) Source nodes Initial energy for each node

Values (Units) 200 × 200 (m2) 100,150,200,250 30 m∼85 m 500 byte 50 per source (200m, 200m) (0,0),(0,100),(0,200),(100,0),(200,0) 2 Joule

was set to 100pJ/bit/m2 and 100nJ/bit respectively. Also, the minimal energy to transmit any signal was set to 200nJ. Parameters and their values used in the time-based network simulation are summarized in Table 1.

4.2. Results and Analysis Based on the parameters and the network constant, the experiments are carried out on the TIF algorithm we have proposed to measure the network performance. The communication algorithm follows the model and procedures described in the previous section; a detailed code has been omitted. We first look at the simulation results on the average energy consumption when the five sources send their sensing data to the base station through the multi-hop manner. Four different node densities are considered for comparison. In each simulation, five source 50 packets of size 500 bytes to the base station which is located at the upper right hand corner of the network area. Figure 5a is the measured average energy consumption and Fig. 5b is the average ratio of the missing packet. When the radio transmission range of each sensor is greater than 35 m, the total energy consumption is reduced as the range increases. Energy consumption becomes very low below 35 m because of the high rate of the missing packet as shown in Fig. 5b. We assumed that the energy consumption from the idle listening sensor is very limited. The total energy consumption of TIF is, in general, lower than that of the 802.11 scheme because it uses a minimal size of the overhead packet, i.e., the TIF control packet does not contain the node address information. We measure the average number of hops to deliver the packet from the sources to the base station as shown in Fig. 6. In general, the multi-hop communication increases the network latency even if it saves some amount of energy through networking. As we expected, the average number of hops from the five source nodes can be reduced by increasing the sensor transmission range. However, the number of hops for each source does not increase over 150 nodes. That is, even if the density of the sensor is increased, the network latency will not increase in the TIF algorithm. This is one of the great features in TIF compared with the 802.11 MAC protocol. The network energy consumption and the number of hops by changing the weighting value in the timeout function of Equation (9) are measured as shown in Fig. 7. The average energy consumption and the average number of hops are also compared with a protocol where a random selection scheme (e.g., random backoff scheme in [17]) is applied. By

342

W. Jeong and S. Y. Nof (a) 60 100 Nodes 150 Nodes 200 Nodes 250 Nodes

Total Energy Consumption (Joule)

55 50 45 40 35 30 25 20 15 30

35

40

45

50

55

60

65

70

75

80

85

Radio Transmission Range (m)

(b) 100 100 Nodes 150 Nodes 200 Nodes 250 Nodes

Total Packet Loss Rate (%)

90 80 70 60 50 40 30 20 10 0 30

35

40

45

50

55

60

65

70

75

80

85

Radio Transmission Range (m)

FIGURE 5 Total energy consumption and packet loss rate by varying the node density.

handling the timeout value to select the forwarding candidate, the TIF algorithm avoids draining energy from specific nodes by distributing forwarding sensors uniformly. Also, more weighted value in the distance term of the TIF timeout function increases the number of hops in delivering packets. The timeout function can adjust the failure rate for each sensor, which happens in real network communication, i.e., failed links and broken nodes. As a result, when the network protocol is controlled by the proposed timeout handling function, the overall energy consumption and the number of hops are reduced by about 20–25% in comparison with non-timeout protocols with a random selection routing scheme.


343

(a) 12 100 Nodes 150 Nodes 200 Nodes 250 Nodes

Average Number of Hops

11 10 9 8 7 6 5 4 30

40

50

60

70

80

90

Radio Transmission Range (m) (b) 10


9

8

7

6

1st Node 2nd Node 3rd Node 4th Node 5th Node

5

4 100

150

200

250

Number of Nodes

FIGURE 6 Average number of hops from sources to the base station; (a) with varying node density, (b) at five different source nodes.

5. Conclusions For the design of the microsensor network protocol, detailed specifications of application should be considered not only because the general architecture of a sensor network cannot meet all the requirements of the application, but also because implemented microsensor nodes have different characteristics. The proposed TIF protocol controlled by applying the timeout function provides speedy and reliable communication. Another interesting property of the protocol is that it has the ability to guide the design consideration between

344

W. Jeong and S. Y. Nof (a)

4

Total Energy Consumption, (Joule)

Without TIF (Random selection) TIF(We = 0.1, Wd = 0.8, Wf = 0.1) TIF(We = 0.45, Wd = 0.45, Wf = 0.1) TIF(We = 0.8, Wd = 0.1, Wf = 0.1)

3.5

3

2.5 30

40

50

60

70

80

90

100

110

120

Radio Transmission Range, Rt (m) (b)

18 Without TIF (Random selection) TIF(We = 0.1, Wd = 0.8, Wf = 0.1)

16


TIF(We = 0.45, Wd = 0.45, Wf = 0.1)

14

TIF(We = 0.8, Wd = 0.1, Wf = 0.1)

12 10 8 6 4 2 30

40

50

60 70 80 90 100 Radio Transmission Range (m)

110

120

FIGURE 7 Energy consumption (a) and number of hops (b) with different weighted value in the TIF timeout function.

energy consumption and latency according to the number of hops. By using a simple control packet and scheme, it also enables seamless connections among wireless microsensors in applications requiring high security and fast response, such as home/office security network systems and transportation/environment surveillance systems. The dynamic and flexible protocol design with the timeout-based control provides the expected high performance under various constraints in wireless network systems.


345

Acknowledgments This work was partially supported by the PRISM Center, Indiana 21st Century Fund for Science & Technology, and the Purdue Research Foundation.

About the Authors Wootae Jeong has earned a Ph.D. in the school of Industrial Engineering from Purdue University, West Lafayette, Indiana, in 2006. He earned his BS and MS in Mechanical Engineering from Yonsei University, Seoul, Korea, in 1998 and 2000 respectively. Since he began his doctoral study in August 2000, he joined the Production, Robotics, and Integration Software for Manufacturing and Management (PRISM) laboratory as a research assistant. He is a student member of several professional organizations including the IEEE, ASME, and IIE. He has received awards during his doctoral program, including an Academic Year Rotary Ambassadorial Scholarship, Rotary International; Purdue Research Foundation, Purdue University. His research interests include communication protocols for wireless micro-sensor network systems, multi-sensor information fusion, and intelligent transportation systems and their integration. Shimon Y. Nof, Professor of Industrial Engineering at Purdue University, has held visiting positions at MIT and universities in Chile, EU, Hong Kong, Israel, Japan, and Mexico. Director of the NSF-industry-supported PRISM Center for Production, Robotics, and Integration Software for Manufacturing and Management; he is a Fellow of IIE, Secretary General of IFPR, and current Chair of IFAC CC – Manufacturing and Logistics Systems. He has published over 300 articles on production engineering and information/ robotics engineering and management, and is the author/editor of nine books in these areas. In 1999 he was elected to the Purdue Book of Great Teachers, and in 2002 he was awarded the Engelberger Medal for Robotics Education. Professor Nof has also had over eight years of experience in industry positions.

References 1. G. Cheng and N. Ansari, “A theoretical Framework for Selecting the Cost Function for Source Routing,” in IEEE International Conference on Communications, vol. 1, 2003, pp. 631–635. 2. N. Bulusu, D. Estrin, L. Girod, J. Heidemann, “Scalable Coordination for Wireless Sensor Networks: self-configuring localization systems,” in the International Symposium on Communication Theory and Applications (ISCTA’01), 2001. 3. W. B. Heinzelman, A. Chandrakasan, and H. Balakrishnam, “An application-Specific protocol Architecture for Wireless Microsensor Networks,” IEEE Transactions on Wireless Communication, vol. 1, no. 4, pp. 660–670, Oct. 2002. 4. S. S. Iyengar, D. N. Jayasimha, D. Nadig, “A versatile architecture for the distributed sensor integration problem,” IEEE Transactions on Computers, vol. 43, no. 2, pp. 175–185, 1994. 5. D. N. Jayasimha, “Fault tolerance in multi sensor networks,” IEEE Transactions on Reliability, vol. 45, no. 2, pp. 308–315, 1996. 6. S. Ghiasi, A. Srivastava, X. Yang, and M. Sarrafzadeh, “Optimal energy aware clustering in sensor networks,” Sensors, vol. 2, pp. 258–269, 2002. 7. C. R. Lin and M. Gerla, “Adaptive clustering for mobile wireless networks,” IEEE Journal on Selected Areas in Communications, vol. 5, no. 7, pp. 1265–1275, 1997. 8. S. Banerjee and S. Khuller, “A clustering Scheme for Hierarchical Control in Multi-hop Wireless Networks,” in Proceedings of IEEE INFOCOM, April 2001. 9. S. Bandyopadhyay and E. Coyle, “An Energy-Efficient Hierarchical Clustering Algorithm for Wireless Sensor Networks,” in Proceedings of IEEE INFOCOM, April 2003.

346


10. W. B. Heinzelman, “Application-Specific Protocol Architectures for Wireless Networks,” Ph.D. Dissertation, Massachusetts Institute of Technology, Boston, MA, 2000. 11. Y. Ma, S. Dala, M. Alwan and J. Aylor, “ROP: A Resource Oriented Protocol for Heterogeneous Sensor Networks,” In The 13th Virginia Tech’s Symposium on Wireless Personal Communications, June 2003, pp. 59–70. 12. M. Singh and V. Prasanna, “Energy-Optimal and Energy-Balanced Sorting in a Single-Hop Wireless Sensor Network,” In The First IEEE International Conference on Pervasive Computing and Communications (PERCOM’03), 2003. 13. M. Ettus, “System capacity, latency, and power consumption in multihop-routed SS-CDMA wireless networks,” in Radio and Wireless Conference (RAWCON’98), 1998, pp. 55–58. 14. Y. Tseng, C. Hsu, and T. Hsieh, “Power-saving protocols for IEEE 802.11-based multi-hop ad hoc networks,” in Proceedings of IEEE INFOCOM, June 2003. 15. C. Intanagonwiwat, R. Govindanand, D. Estrin, “Directed diffusion: A scalable and robust communication paradigm for sensor networks,” In Proceedings of the Sixth Annual International Conference on Mobile Computing and Networking (MobiCOM’00), Boston, Massachusetts, 2000. 16. I. Chatzigiannakis, T. Dimitriou, S. Nikoletseas, and P. Spirakis, “A probabilistic forwarding protocol for efficient data propagation in sensor networks,” In The 5th European Wireless Conference on Mobile and Wireless Systems beyond 3G (EW’04), 2004. 17. I. Chatzigiannakis, A. Kinalis, and S. Nikoletseas, “Wireless Sensor Networks Protocols for Efficient Collision Avoidance in Multi-path Data Propagation,” In The First ACM Workshop on Performance Evaluation of Wireless Ad Hoc, Sensor, and Ubiquitous Networks (PEWASUN’04), 2004. 18. D. Tian and N. D. Georganas, “A coverage-preserving node scheduling scheme for large wireless sensor networks,” in Proceedings of the 1st ACM Int. Workshop on Wireless Sensor Networks and Applications, ACM press, 2002, pp. 32–41. 19. W. Ye, J. Heidemann, and D. Estrin, “Medium access control with coordinated adaptive sleeping for wireless sensor networks,” IEEE/ACM Transaction on Networking, vol. 12, no. 3, pp. 493–506, June 2004. 20. J. Deng, Y. S. Han, W. B. Heinzelman, and P. K. Varshney, “Balanced-energy sleep scheduling scheme for high density cluster-based sensor networks,” in Proceedings of the 4th Workshop on Applications and Services in Wireless Networks (ASWN’04), Boston, 2004. 21. Liu Y., Nof S. Y., “Distributed micro flow-sensor arrays and networks: Design of architecture and communication protocols,” International Journal of Production Research, vol. 42, no.15, pp. 3101–3115, 2004. 22. W. Jeong, S. Y. Nof, “Design of protocols for wireless MEMS sensor network systems in industrial applications,” in The 18th International Conference on Production Research, Italy, July 2005. 23. N. Bulusu, J. Heidemann, and D. Estrin, “GPS-less low cost outdoor localization for very small devices,” IEEE Personal Communication Magazine, Vol. 7, No. 5, pp. 28–34, 2000. 24. D. L. Mills, “Computer network time synchronization: The network time protocol,” CRC Press, 2006. 25. S. Y. Nof, Y. Liu, and W. Jeong, 2003, “Fault-tolerant time-out communication protocol and sensor apparatus for using same,” Patent pending, Purdue University. 26. K. Fall and K. Varadhan, “The NS Manual,” UC Berkeley, 2002, http://www.isi.edu/nsnam/ns. 27. A. Boukerche and S. Nikoletseas, Wireless Communications Systems and Networks: Protocols for Data Propagation in Wireless Sensor Networks: A Survey, Kluwer Academic Publisher, ed. Mohsen Guizani, pp. 23–51, 2004. 28. W. R. Heinzelman, J. Kulik, and H. Balakrishnan., “Adaptive protocols for information dissemination in wireless sensor networks,” In Proceedings of the 5th Annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom’99), pp. 174–185, 1999.