Computer Science Technical Report Archive

USC Computer Science Technical Reports, no. 806 (2003)

(USC DC Other)

USC Computer Science Technical Reports, no. 806 (2003)

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content RUGGED: RoUting on finGerprint Gradients in sEnsor Networks Jabed Faruque and Ahmed Helmy Department of Computer Engineering University of Southern California, Los Angeles, CA 90089, USA, {faruque,helmy}@usc.edu Abstract. Every physical event produces a fingerprint in the environ- ment resulting in a natural information gradient in the proximity of the phenomenon. Moreover, many physical phenomena follow diffusion laws. In this paper, we propose a novel scheme to effectively exploit the natural information gradient to route a query in a sensor network. Our scheme uses multiple path exploration, and controls the instantiation of paths by simulated annealing. Unlike other information-driven protocols, our scheme eliminates the overhead of preparing and maintaining the gradi- ent information repository. We apply our scheme to study three different problems: (1) single-value query, (2) global maxima search, and (3) mul- tiple events detection. Simulation results have demonstrated that the routing protocol, based on our proposed scheme, is highly energy effi- cient and achieves over 98% success rate in routing around sensor holes, even in the presence of environmental noise and malfunctioning sensor nodes. We also illustrate that our scheme is well suited for a broad-range of applications; e.g., time gradient based target tracking, boundary de- tection. 1 Introduction A distributed sensor network (DSN) consists of sensor nodes with limited en- ergy source, sensor devices, short-range radio and on-board processing capability. Sensing capability of the attached sensing devices and their small size, make these sensor nodes highly suitable for monitoring physical phenomena. Sensor net- works are most widely used for habitat and environmental monitoring[9][10][11]. More specifically, advances in the MEMS technology makes it possible to de- velop sensors to detect and/or measure most of the usual physical phenomena like temperature, light, sound, radiation, humidity, chemical contamination, ni- trate level in water etc. Every physical event leaves some fingerprints in the envi- ronment in terms of the event’s effect; e.g., fire increases temperature, chemical spilling increases contamination, nuclear leakage increases radiation. Moreover, most of the physical phenomena follow diffusion property[18][19] with distance, This research is supported by Pratt&Whitney UTC Institute for Collaborative En- gineering(PWICE), Intel and NSF CAREER awards. II i.e., f(d)∝ 1 d α ,where d is the distance from the point having maximum effect of the event, f(d) is magnitude of the event’s effect and α is the diffusion pa- rameter depending on the type of effect; e.g., for light α=2, heat α 1. As an example, if some location’s temperature is 100 o C then, nearby locations temper- ature should be correlated with that based on the distance. Thus the property of physical phenomena implicity creates a distributed information repository about the event’s effect. Furthermore, the information gradient concept is not limited to physical phenomena. For example, by recording the time stamp about a mov- ing object when it passed by a sensor node, DSN can establish time gradient towards the object’s current location. So, routing protocols for DSN can exploit this natural, freely available information gradient as an important attribute to forward the query efficiently towards the source. Characteristics of the sensor nodes; e.g., limited battery life, energy expen- sive wireless communication, high probability of failure or malfunction and un- structured nature of the DSN, make routing in the DSN a challenging problem. Traditional routing protocols of the DSN are based mostly on flooding (Directed Diffusion[1]) or random-walk (Rumor routing[4], ACQUIRE[5], etc.). These ap- proaches, however, do not utilize the domain-specific knowledge, i.e., the event’s fingerprint gradient about the monitored phenomenon. In this paper, we effec- tively exploit the fingerprint gradient to design a scheme for efficient routing in sensor networks. Previous data-centric routing protocols that are based on information gra- dient[6][7][12][14], use a proactive phase to prepare distributed or cluster based gradient information repository towards a target or an event. To adapt to dy- namic behavior of DSN, periodic update of information repository is required. To route a query from sink to source, most of these protocols use greedy routing algorithms based on information gradient. For preparing gradient information repository and routing query, protocols do not take advantage of using well established law of physical events. Moreover, the query proceeds towards the source through a single path, which usually get trapped at the local maxima or minima, or reach a dead-end due to imperfect sensor device and environmental noise. On the other hand, creating unlimited multiple paths resembles flooding. In this paper, we propose a scheme for using the fingerprint gradient of the event’s effect to avoid the proactive phase of preparing distributed gradient in- formation repository and a novel energy efficient, fully distributed and reactive routing protocol based on that information gradient. Our protocol effectively exploits the laws of physical events for routing decision. Also it overcomes the limitations of usual information-driven routing protocols due to local maxima or minima using simulated annealing concepts in distributed manner and estab- lishes an effective balance between single path and multiple path exploration to discover routes. To design and test the protocol, we consider a somewhat realistic model of the environment consisting of flat information region and en- vironmental noise in addition to the region having noisy information gradient about the event’s effect. We also provide several potential applications for uti- lizing information gradients including (but not limited to) target tracking us- III ing time gradients, event boundary detection, predicting spread of phenomenon (e.g., chemical contamination), determining earth quark patters, among others. An overview of such applications is discussed in this document. On-going and future research will investigate application details. Here we keep our focus on static sensor networks which are used mainly for environmental monitoring. The rest of the paper is organized as follows. Section 2 discusses related work on routing in sensor networks. Section 3 discusses the environment and noise models. Our protocol is described in section 4. Section 5 describes the sim- ulation model. Approach to solve three different problems, using of the proposed protocol, and results are presented in section 6. Other potential applications are discussed in section 7. Finally section 8, concludes the paper and discusses future work. 2 Related work Several approaches have been proposed for routing in sensor networks. Directed Diffusion[1][2], is one of the first data-centric query dissemination protocol that is particularly useful for long lived data flow or continuous queries. In this scheme, a node’s interest for named data is initially distributed through the network via flooding to find the sources of relevant data. Diffusion results in high quality paths and well-suited for long continuous queries. Initial flooding overhead is amortized by the duration of long flows. The work in [3] attempts to adapt directed diffusion according to specific applications. Several protocols[8][5][4][13][17] have been designed based on Random walk. Asymptotically, random-walk shows good performance. But in practice, it causes high latency and without directionality and/or proper value of TTL, it may fail to discover resource. In [13], Servetto and Barrenechea have shown that multiple random-walks help to improve load balancing and to minimize critical point of failure and latency with increased communication cost. One of the major differences between our information gradient based ap- proach and flooding and/or random-walk based approaches is that our scheme uses sensor’s measurement, about the event’s effect, for routing decision. In this context our protocol is information driven, utilizing natural gradients created signature by the mentioned phenomena. Chu, Hausseker and Zhao propose Information Driven Sensor Querying(IDSQ) and Constrained Anisotropic Routing (CADR) mechanisms[6], especially for lo- calization and target tracking. IDSQ is a proactive sensor selection strategy for correlated information based on a criterion which combines information gain and communication cost. Dynamic environment and low query rate may cause extra overhead due to frequent exchange of information between neighbors and cluster leader. CADR is a greedy algorithm which routes a query to its optimal destination by following the local gradients to maximize the information gain through the sensor network. It may suffer from getting trapped at local maxima or minima. IV The recent work by Liu, Zhao and Petrovic [7] proposes min-hop routing al- gorithm to overcome the limitation of CADR about handling local maxima and minima. It uses a multiple step look-ahead approach in which initial network discovery phase determines the minimum look-ahead horizon so that the path planning phase can avoid network irregularity. The algorithm improves the suc- cess rate of routing the message with additional search cost. Also the increase in the neighborhood size causes more communications between cluster the leaders and their neighbors. Another information driven routing protocol is GRAdient Broadcast (GRAB) [14] which introduces virtual gradient by building a cost field towards a particular node and then routes queries across a limited size mesh towards that node. Initial overhead for flooding the network can be amortized by routing queries reliably along the shortest path. In [12], a navigation protocol is proposed to guide along the safest path using a distributed repository of information about the area covered by the sensor network. The network can adapt to sensor failure or addition of new nodes by continuous updates to the distributed information content. Both building and updating of the information repository causes significant communication overhead. Also our work has some similarities to techniques proposed in [15] and [16] for routing in ad-hoc networks using mobility diffusion to disseminate the desti- nation location information. However, cost-model and mobility issue of ad-hoc networks make it inapplicable for static sensor networks. Our scheme routes the query using a fully distributed decision making proce- dure by effectively exploiting the natural gradient information repository, which is the consequence of the fingerprint gradient of physical phenomenon being monitored and follows well established physical laws. Multiple path exploration to discover the route or the event and controlling instantiation of multiple paths, using a probabilistic function based on simulated annealing concept, is another key difference with existing information-driven protocols. The ability to query for multiple sources (Sec.6.3) and to use time gradient for target tracking (Sec.7) are important features of our scheme. We target applications in which queries are generally triggered to identify the origin of an event after its occurrence and its effect follows diffusion laws. For example, fire event, earth quake, chemi- cal contamination etc. We shall discuss more specific applications in Sec.7. Our scheme works without the location information. However, location information, when available, can make our protocol more energy efficient and robust. 3 Environment Model We now describe the environment model used in this paper. The following three components are considered in attempt to properly model the environmental ef- fects, and to design and analyze the performance of our protocol. – Area covered by event’s effect: When an event occurs, diffusion of its effect is a function of distance, d and time, t, i.e., f(d, t) ∝ t d α .Now,con- V sidering sensors reading at particular time instance, say t 1 , diffusion can be expressed as a function of distance only, i.e., f(d) ∝ 1 d α . Theoretically, the tail of the diffusion of the event’s effect is infinite; but in real life, sensors are unable to detect or measure the effect of an event below a certain threshold. So, after a certain distance from the event’s location, it is not possible to measure the event’s effect using the small sensors of DSN. Zero reading of the sensors of these region creates a flat information region where information gradient is unavailable. – Erroneous reading of malfunctioning sensors: Real life sensors are not perfect and subject to malfunction due to obstacle or sensor/node failures. So, some of the sensors erroneous reading may cause irregular pattern, i.e., local maxima or minima in the information gradient of the event’s effect. To model the environment for our protocol, we consider malfunctioning sen- sors are uniformly distributed in the DSN and each malfunctioning sensor is assigned a random reading. An analytical technique is presented in the Appendix.9.1 to filter such arbitrary readings of isolated malfunctioning sen- sor nodes. Also simulation results of Sec.6.2 demonstrates the effectiveness of the filter. To analyze the performance of our protocol in the presence of sensor error, we vary the percentage of malfunctioning sensors between 0 to 20%. – Environmental noise: Condition of the surrounding environment, such as direction of airflow or fluid, humidity, etc., of sensors and the event is re- sponsible for this type of noise. Although its effect is less where gradient information level about the event’s effect is high, it gradually increases to- wards the low gradient zone. Due to this noise, sensor’s reading can increase or decrease by a certain amount. We model this noise as follows, f(d i )= f ∗ (d i )± f EN (f ∗ (d i )), f EN (f ∗ (d i ))∝ (f max − f ∗ (d i )) where, d i = distance of the location from peak information point (i.e., the event) f(d i ) = gradient information of the location with environmental noise, f max = peak information, f ∗ (d i ) = gradient information of the location without environmental noise. The proportional constant is considered 0.03 to model the environmental for our protocol, i.e., 3% environmental noise is consider. Thus our environmental model (Fig.1) consists of flat (i.e. zero) informa- tion region and gradient information region. Environmental noise is present only in the gradient information region, while malfunctioning sensors are uniformly distributed in the both regions. 4Protocol With the intuition of natural information gradient discussed in Sec.1 and the environment model presented in Sec.3, our basic information-driven routing pro- VI (a) (b) Fig.1. Environment model: (a) Events are at the peaks and its effect reduces with dis- tance. (b) Event is located at “E”. Radial gradient of color represents the event’s effect. Black dots denote good sensor nodes and gray dots (e.g., “M”) denote malfunctioning sensor nodes. Nodes in the white region are flat information region nodes. tocol is designed. It is assumed that to prevent looping, each querier generates unique sequence number for the query it sends. Based on the environment model, each query can have two different modes - (1) flat region mode and (2) gradient region mode. Initially, a query starts with flat region mode. It switches to gradi- ent region mode as soon as it finds gradient information about the event’s effect. Thus, the query packet needs fields for query ID and query mode in addition to other information. A query may be initiated from any arbitrary node. Upon receiving the query, the querier sets the query mode to flat region mode and forwards the query to its neighbors with its gradient information level about the queried event. Then, each neighbor independently decides whether to forward the query based on the algorithm described below. – In the flat information region, if the query mode is flat, node uses flooding to forward the query towards the gradient information region (Fig.2(a)). Otherwise, a node uses probabilistic forwarding described next. The query does not switch to the gradient mode unless gradient information is found. Hence, in the absence of event(s), gradient information will be zero (in ideal condition) and the protocol will only use this flooding approach. – In the gradient information region, a node uses greedy forwarding approach. If a node is able to improve the information level, it forwards the query to its neighbors for further improvement (Fig.2(b) and 2(d)). Otherwise, the node performs probabilistic forwarding, described next. Note that, this greedy forwarding approach is different from classical greedy forwarding algorithms VII which either choose the best neighbor or a set of best neighbors based on collected information of neighbors like information level, close to destination etc. In our greedy forwarding approach underlying concept is, if a node’s reading is more than that of its parent node along the forwarding path, then the forwarding path through the node may reach to nodes having higher readings. – The type of irregular patterns created in gradient information region, due to erroneous sensors reading as discussed in Sec.3, can be sharp drop or rise of information level about the queried event. To overcome such local and isolated maxima or information hole, the protocol uses a probabilistic forwarding (Fig.2(c)) which is a function based on simulated annealing con- cept. As a parameter, the function takes the hop-count(x) in the gradient information region. That is, the probabilistic function, f p (x)= 1 x a where, a depends on the diffusion parameter, α and controls the reachability of the protocol. As we shall discuss, the performance is a function of the interplay between ‘α’and‘a’. This will be discussed in Sec.6. Nodes use the reverse path as a basic mechanism to send the reply of the resolved query to the querier. However, depending on the type of query, the reply mech- anism may be optimized to suppress unpromising responses; more discussed in Sec.6.2. Initially the protocol instantiates multiple paths to discover source(s); but, in the absence of multiple sources most of the paths will terminate after few hops. Note that our algorithm does “not” require neighbor “Hello” messages (i.e., a node processing the query is not assumed to know all its neighbors readings). This has proven to save significant overhead over using “Hello” message. 5 Simulation Model Simulations were carried out to validate and characterize the performance of the proposed information gradient based routing protocol. The objective of the simulation model is not only to analyze the performance of our routing protocol, but also to study how to exploit the natural information gradient effectively. In the simulation, two different sensor layouts are used. The first layout is a regular 100× 100 grid of 10000 sensor nodes and each node has eight neighbors. The second layout is a uniform random grid (Fig.3(a)) used in [7] to simulate a sensor field of dimension 225× 375 m 2 with sensors having 50m communication radius. It is generated from a regular 15×6 grid of 90 sensor nodes by perturbing the grid points with independent Gaussian noise(0,25). For single-value query (Sec.6.1), to test the success rate of our routing protocol in the presence of sensor holes, we remove grid points from row 5-6 and columns 2-5 before perturbing the grid points (Fig.3(b)). In addition to sensor layouts, the environment is modelled using its three components described in Sec.3. For all simulations, parameter α, of the phenomenon diffusion function is set to 0.8. Evaluation of the routing algorithm is done in terms of average energy dis- sipation and success rate, i.e., reachability. To compute energy dissipation, we consider total number of transmissions required for both query forward and reply. VIII (a) ‘Q’ forwards query to its neighbors. All ‘f’nodes are in the flat information re- gion, so they use flooding again. (b) All ‘g’nodes arein gra- dient information region, so they switch the query mode to gradient mode and must forward again. (c) All neighbors(p)of Mx have less information, so they will probabilistically forward the query to their neighbors. (d) All neighbors(g)of Mn have more information, so they will forward the query to their neighbors. Fig.2. Routing protocol: Event is at ‘E’ and querier ‘Q’ at flat information region. Effect of ‘E’ follows diffusion law.Mx is local maxima andMn is local minima. IX E 350 300 250 200 150 100 50 0 0 50 100 150 200 (a) 0 0 E 350 300 250 200 150 100 50 50 100 150 200 (b) Fig.3. Sensor layout: (a) uniform random grid. Sensors within dotted rectangle are removed to create sensor hole, (b) uniform random grid with sensor hole. “E” denotes the location of the event. Performance of our routing algorithm is compared with flooding and expanding ring search (ERS). Effectiveness of the information gradient is analyzed by vary- ing the first two components of the environment model, i.e., the percentage of flat information region nodes and percentage of uniformly distributed malfunction- ing nodes. Also we pay attention to tune the parameter ‘a’ of the probabilistic function described in Sec.4, to find optimal trade-off between energy dissipation and improving reachability of our proposed routing protocol. 6 Performance Evaluation Three different problems are studied and analyzed using the information gradient based scheme. These are (1) Single-value query, (2) global maxima search, and (3) multiple event detection. 6.1 Single-value Query This type of query searches for a specific value and have a simple response. Here, we assume that only one node has the response. And the response is about an event, the effect of follows a diffusion law and creates information gradient. Also the event source is assumed to be stationary. For example, search for a source of chemical leakage, where information gradient is a fingerprint of the chemical contaminant. So, the routing algorithm needs to find the source. Here the source node replies to the querier using the reverse path. X 0 2 4 6 8 10 12 0 10 20 30 40 50 60 70 Percentage of failure (to find source) Percentage of nodes in flat region a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (a) Failure rate 0 1000 2000 3000 4000 5000 0 10 20 30 40 50 60 70 Avg. energy dissipation Percentage of nodes in flat region a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (b) Average energy dissipation Fig.4. Effect of flat information region nodes, while environmental noise is 3% and malfunctioning nodes are 15%. For one-shot query, the event is simulated at location (74,49) of the first sensor layout and the querier can be any of the remaining nodes. In Figure 4(a) and 4(b), for 15% malfunctioning nodes, we vary the number of nodes in the flat information region from 0-66% to show change of query failure rate and average energy dissipation respectively for different values of ‘a’ of probabilistic for- warding function (Sec.4). With the increase of the flat information region nodes, flooding overhead becomes dominant. The protocol creates multiple paths, so failure rate decreases but at the same time energy dissipation increases. But malfunctioning nodes (Sec.3) cause the protocol to switch to gradient mode er- roneously; so, failure rate increases for more than 57% flat information region nodes. In Figure 5(a) and 5(b), for 36% flat information region nodes, the number of malfunctioning nodes are varied. With increase of malfunctioning nodes, the protocol switches from flat region mode to gradient region mode rapidly, which reduces flooding overhead, but increases failure rates especially for higher value of ‘a’. It is important to notice from Figures.4 and 5 that for higher value of ‘a’, probabilistic forwarding function, f p (x)= 1 x a (Sec.2) drops sharply and the protocol explores less number of nodes. As a result, failure rate increases and average energy dissipation decreases. For a<α, probabilistic function drops slowly and allows to follow diffusion pattern due to α by more probabilistic for- warding. Thus, values of a< α but close to α give optimal trade-off between energy dissipation and reachability. In simulations, though α is 0.8, but due to simulated environmental noise, it is found from Figures.4 and 5 that a=0.65 is optimal for the simulated scenario. Further, in Figure.6, average energy dissipa- tion of our algorithm is compared with that of flooding-based querying (FBQ) and expanding ring search (ERS) algorithms using the same configuration of the layout. Our routing protocol reduces energy dissipation by 47-80% over FBQ, XI 0 2 4 6 8 10 12 0 5 10 15 20 Percentage of failure (to find source) Percentage of nodes in malfunctioning nodes a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (a) Failure rate 0 500 1000 1500 2000 2500 3000 0 5 10 15 20 Average energy dissipation Percentage of nodes in malfunctioning nodes a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (b) Average energy dissipation Fig.5. Effect of malfunctioning nodes, while environmental noise is 3% and flat infor- mation region nodes are 36%. while the flat information region nodes of 66% or less. Also in the presence of 47% or less flat information region nodes, our protocol reduces energy dissipation by 18-50% over ERS. The second sensor network layout is used to test reachability in the presence of a deployment gap or hole. Target is simulated at location ‘E’ and queriers can be any node below the sensors hole of the sensor network layout as shown in Fig.3(b). In Figure 7, for 20% malfunctioning nodes, flat information region nodes are varied from 20-94%. For lower values of ‘a’, the success rates of our protocol to route the query around the sensors hole are above 98%, even at the presence of 55% flat region nodes. 6.2 Global Maxima Search The query finds the maximum value of the event’s effect within the sensor mon- itored region. This important statistic gives the current critical status about the observed phenomenon. For FBQ and ERS algorithms, to decide about the maximum value, need to explore all nodes of the DSN. However, using gradi- ent information, our protocol determines the global maxima by exploring only a limited number of nodes. Due to distributed nature of sensor network, any node with some gradient information about the observed phenomenon, can become a potential responder to the query; so reply overhead for this type of query may become significant. Hence, we propose a reply suppression scheme in which intermediate nodes sup- press non-promising replies by caching and comparing the maximum value of the responses passing through the node for the same query ID, as shown in Figure.8. In the suppression scheme, a node forwards the reply unless it receives some higher gradient information from its neighbors’ broadcast replies. To make XII 10000 3804 3942 4093 5138 5265 3340 3401 1906 2460 Flood ERS-3 ERS-5 ERS-7 GBR-s1a1 GBR-s1a2 GRB-s2a1 GRB-s2a2 GRB-s3a1 GRB-s3a2 Fig.6. Comparison with FBQ, ERS, and our gradient based routing (GBR). Here ERS ring sizes are 3, 5 and 7. For GBR, s1,s2 and s3 indicate 66%,47%, and 36% flat information region nodes respectively. And a1 and a2 indicate ‘a’ is 0.7 and 0.5 respectively. 0 5 10 15 20 20 30 40 50 60 70 80 90 100 Percentage of failure (to find source) Percentage of nodes in flat region a=0.90 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 a=0.45 a=0.40 Fig.7. Query failure rate to route a query around sensors hole of the second sensor layout. this scheme even more effective a node may use a timer (per reply) that is set before a reply is sent or forwarded, while the timer is running the node listens to other broadcast replies and suppresses unnecessary reports. The timeout is a function of the network size. For this type of query, we use the first sensor network layout with the same configuration used for single-value query (Sec.6.1). The failure rates to find the global maxima are similar to those of single-value query (on Figure.4(a)). In Figure.9(a) and 9(b), for 15% malfunctioning nodes, we vary the percentage of flat information region nodes. As discussed in Sec.6.1, erroneous reading of malfunctioning nodes in the flat information region causes a query to switch to gradient region mode. Due to absence of further gradient information and sharp drop of probabilistic function for higher values of ‘a’, along that path, informa- tion improvement halts after the malfunctioning node. So, the node initiates a reply towards the querier after timeout of the query. But, for lower value of ‘a’, the probabilistic function drops slowly which causes more probabilistic forward- ing of the query. It increases the probability to reach the gradient information region and the reply comes from actual source node. For these reasons, in both Figure.9(a) and 9(b), for 44% to 65% of flat information region nodes, higher value of ‘a’ shows more energy dissipation. For more than 65% flat information region nodes, gradient information region size reduces more and average energy dissipations for all values of ‘a’ become identical. However, after using the filter to detect malfunctioning nodes described in the Appendix.9.1, number of replies from malfunctioning is reduced dramatically. As a result, the average energy dissipation reduces significantly, as shown in Figure.9(b). We notice that for this type of query effects of malfunctioning nodes are similar to those of single-value query. Also, it is found that a=0.65 is optimal for the simulated scenario. XIII Q E Li = 7 Li = 2 Li = 10 Li = 4 Li = 3 (a) Query forwarding ends at the shaded nodes. Q E Li = 7 Li = 2 Li = 10 Li = 4 Li = 3 (b) Intermediate nodes suppress the non- promising replies. Fig.8. Optimized reply mechanism. Here, event and querier are at ‘E’ and ‘Q’ respec- tively.Li is the information about the event’s effect. 6.3 Multiple Events Detection This type of query searches for multiple events of the same type. So sensors, within the gradient information region of multiple sources, record resultant gra- dient information. For example, fire incidents cause multiple sources as it spreads with time. So, a sensor monitoring temperature, actually records resultant tem- perature at its location due to these multiple fire sources. Using multiple path exploration and the gradient information, our protocol attempts to find all the multiple sources. In our simulation, sources are uniformly distributed in the first sensor network layout and the querier can be any of the remaining nodes. In Figure.10(a) and 10(b), for 15% malfunctioning nodes, we vary the number of sources from 1 to 18. We notice that for less than five sources, a =0.4gives optimal trade-off between number of sources found and average energy dissipation. But, with the increase of number of sources in the sensor network layout, resultant gradient information from multiple sources creates some plateaux regions and requires more probabilistic forwarding to forward query towards sources through that region. So lower value of ‘a’ is required so that probabilistic function drops slowly and allows more probabilistic forwarding. For five or more number of sources, a=0.35 is found optimal in the simulated scenario. XIV 0 2000 4000 6000 8000 10000 12000 14000 0 10 20 30 40 50 60 70 Average energy dissipation Percentage of nodes in flat region a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (a) Average energy dissipation with- out filter to avoid malfunctioning nodes. 0 1000 2000 3000 4000 5000 6000 7000 8000 0 10 20 30 40 50 60 70 Average energy dissipation Percentage of nodes in flat region a=0.90 a=0.85 a=0.80 a=0.75 a=0.70 a=0.65 a=0.60 a=0.55 a=0.50 (b) Average energy dissipation with filter to avoid malfunctioning nodes. Fig.9. For global maxima search, effect of flat information region nodes, while envi- ronmental noise is 3% and malfunctioning nodes are 15%. Notice that y-scale of (a) and (b) are 0-14000 and 0-8000 respectively. 7 Other Potential Applications In Sec.6, three generic types of query routing are discussed and solved using the natural information gradient based approaches. These approaches can be combined to solve several essential real life applications. Due to space limitation, here we only illustrate two problems: (1) Event boundary detection, and (2) Target tracking. 7.1 Event Boundary Detection Boundary detection is the process of estimating contour between two homoge- neous regions. Therefore, the locations of the sensor nodes along the contour, determine the boundary of an event. Here, each of the sensor nodes along the contour is a potential source to reply the query to detect the boundary. So, a query for boundary detection is essentially similar to a query for multiple events detection (Sec.6.3). Upon receiving the query, each source node replies the querier with its location information. The querier estimates the boundary us- ing the location information of the sources. Accuracy of the estimated boundary depends on the success rate of finding the sources along the contour. 7.2 Target Tracking In a tracking scenario, consider targets as point sources of signals and the signal amplitude attenuates with distance according to diffusion law. Here, each sensor XV 0 20 40 60 80 100 0 5 10 15 20 Percentage of sources found Number of sources a=0.65 a=0.60 a=0.55 a=0.50 a=0.45 a=0.40 a=0.35 a=0.30 (a) Percentage of sources found vs number of sources. 0 1000 2000 3000 4000 5000 6000 7000 0 5 10 15 20 Average energy dissipation Number of sources a=0.65 a=0.60 a=0.55 a=0.50 a=0.45 a=0.40 a=0.35 a=0.30 (b) Average energy dissipation. Fig.10. Multiple events detection, while environmental noise is 3% and malfunctioning nodes are 15%. node requires its location information to report the target’s current position. This is not required for routing decision. Now, as the target enters and moves, sensors within the signal range of the target record the local time stamp when the target has been sensed. At each time instance, multiple sensors may detect the same target. To precisely determine the location, each of these sensor nodes initiates query for maximum signal strength similar of global maxima search (Sec.6.2), within the small gradient region of signal, created due to the point source, i.e., the target. Here a higher value of ‘a’ is required so that the probabilistic function (Sec.4) drops sharply and limits unnecessary forwarding. Also the query alerts the nodes immediately outside the gradient region, about the possible movement of the target. Furthermore, from the recorded local time stamp for the target, sensor nodes can create time gradients in terms of number of clock ticks has been passed after sensing the target which decreases along the direction of the moving target. Now routing a query for the finding the minimum value of such clock ticks, is essentially following the target. Since, the sensor node close to current position of the target has the minimum value of clock ticks after sensing the target at that instance. 8 Conclusion and Future Work In this paper, we presented a scheme to route on fingerprint gradients in sensor networks. The main contributions of this paper are 1. The proposed novel scheme to exploit natural information gradient reposi- tory, which is a consequence of the fingerprint gradients of the event’s effect. 2. The novel reactive, fully distributed routing protocol for sensor network, based on that information gradient repository. XVI Unlike other information-driven protocols for sensor network, our scheme elimi- nates the overhead of preparing and maintaining information gradient repository. Three different problems were studied using our scheme and performance of the routing protocol for each problem, was demonstrated by simulations. Overall energy dissipation of the protocol was found significantly low compared to FBQ and ERS. Also its success rate to route around sensors hole, was found to be over 98%. Multiple path exploration and controlling instantiation of paths by simulated annealing, makes our protocol well suited for broad range of applications includ- ing time gradient based target tracking, event boundary detection. One possible future research direction is to develop protocols for target tracking and target counting using our proposed scheme. Also we have found that the parameter ‘a’ of probabilistic function depends on diffusion parameter α. So, another impor- tant future work will be to establish analytical relationship between ‘a’and α to further reduce the energy dissipation. References 1. C. Intanagonwiwat, R. Govindan and D. Estrin, “Directed Diffusion: A Scalable and Robust Communication Paradigm for Sensor Networks, MobiCom 2000. 2. C. Intanagonwiwat, D. Estrin, R. Govindan, and J. Heidemann, “Impact of Net- work Density on Data Aggregation inWireless Sensor Networks , In Proceedings of the 22nd International Conference on Distributed Computing Systems (ICDCS02), Vienna, Austria. July, 2002. 3. J. Heidemann, F. Silva and D. Estrin, “Matching Data Dissemination Algorithms to Application Requirements”, SenSys 2003. 4. D. Braginsky and D. Estrin, “Rumor Routing Algorithm for Sensor Networks”, WSNA ’02, September 28, 2002. 5. N. Sadagopan, B. Krishnamachari, and A. Helmy, “Active Query Forwarding in Sensor Networks (ACQUIRE)”, SNPA ’03, May 2003. 6. M. Chu, H. Haussecker, and F. Zhao, “Scalable Information-Driven Sensor Querying and Routing for ad hoc Heterogeneous Sensor Networks”, Int’l J. High Performance Computing Applications, 16(3):90-110, Fall 2002. 7. J. Liu, F. Zhao, and D. Petrovic, “Information-Directed Routing in Ad Hoc Sensor Networks”, WSNA ’03, September 19, 2003. 8. H. Matsuo and K. Mori, “Accelerated Ants Routing in Dynamic Networks”, in Proc. Intl. Conf. On Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, pp.333-339, Aug. 2001. 9. J. Warrior, “Smart Sensor Networks of the Future”, Sensors Magazine, March 1997. 10. G.J. Pottie, W.J. Kaiser, “Wireless Integrated Network Sensors”, Communications of the ACM, vol. 43, no. 5, pp. 551-8, May 2000. 11. A. Cerpa et al., “Habitat monitoring: Application driver for wireless communica- tions technology”, 2001 ACM SIGCOMM Workshop on Data Communications in Latin America and the Caribbean, Costa Rica, April 2001. 12. Q. Li, M.D. Rosa and D. Rus, “Distributed Algorithms for Guiding Navigation across a Sensor Network”, MobiCom 2003. 13. S. D. Servetto and G. Barrenechea, “Constrained Random Walks on Random Graphs: Routing Algorithms for Large Scale Wireless Sensor Networks”, WSNA ’02, September 28, 2002. XVII 14. Fan Ye, Gary Zhong, Songwu Lu and Lixia Zhang, “GRAdient Broadcast: A Ro- bust Data Delivery Protocol for Large Scale Sensor Networks”, accepted for publi- cation in ACM WINET (Wireless Networks). 15. H. Dubois-Ferriere, M. Grossglauser and M. Vetterli, “Age Matters: Efficient Route Discovery in Mobile Ad Hoc Networks Using Encounter Ages”, MobiHoc 2003. 16. M. Grossglauser and M. Vetterli, “Locating nodes with EASE: Last Encounter Routing for Ad Hoc Networks through Mobility Diffusion”, INFOCOM 2003. 17. Q. Lv, P. Cao, E. Cohen, K. Li, and S. Shenker. “Search and replication in un- structured peer-to-peer networks”. In ICS02, New York, USA, June 2002. 18. D.R. Askeland, The Science and Engineering of Materials, PWS Publishing Co., 1994. 19. J.F. Shackelford, Intro to Materials Science For Engineers, 5th Ed., Prentice Hall, 2000. 9 Appendix 9.1 Filtering Erroneous Readings of Malfunctioning Sensors: Consider two sensor nodes S 1 and S 2 and they are neighbor of each other. Let, distance of S 1 and S 2 from the event’s location are d and d+1 hops respec- tively, as grid topology is used. Due to event’s effect, reading of S 1 and S 2 are R 1 and R 2 respectively, and we get, R 1 = C d α ,and R 2 = C (d+1) α , respectively. Here, C is proportional constant. Hence, R1 R2 = (d+1) α d α =(1+ 1 d ) α ≈1+ α d ,where| 1 d | < 1 Therefore, if sensor reading follows diffusion law, then R1 R2 ≈ 1+ α d In our simulation study, α=0.8. Hence, R1 R2 ≈ 1+ α d < 2.0 is used as a reasonable simple filter to detect malfunc- tioning nodes. 9.2 Routing Algorithm for Each Node This section presents a combined form of the three parts of the our routing algorithm described in Sec.4. XVIII d : current hop-count of the query from the querier. x : current hop-count of the query in the gradient information region. Initialized to 0 at querier. q id :queryID I d : gradient information at current node I d−1 : gradient information at parent node along the forwarding path q mode : query mode. Initialized to “flat region mode” at querier. S : Set of already seen query IDs step 0: wait to receive a query routing request (q id ,I d−1 ,d,x,q mode ) step 1: if q id ∈ S, discard the request and goto step 0. step 2: S← S∪{q id } [stores query ID to avoid looping] if node has reply, sends the reply to querier and goto step 0. [reply decision and mechanism depends on the query type.] step 3: compute information gain, I gain = I d − I d−1 step 4: if I gain > 0 goto step 5 [as gradient information is available] if q mode = flat [flat information region and query is in flat mode also] broadcast the query routing request among neighbors having parameters (q id ,I d ,d+1,x,q mode ) else [Unsmooth gradient information region] [No information gain and query is in gradient mode] [Uses probabilistic forward to handle local maxima or minima] broadcast the query routing request among neighbors having parameters (q id ,I d ,d+1,x+1,q mode ) with probability f p (x), where f p (x)= 1 x a goto step 0 step 5: [gradient information region follows greedy approach] if q mode = flat, [information gradient is found, so switching the query mode] q mode ← gradient x← 1 [information gain is positive and query is in gradient mode] broadcast the query routing request among neighbors having parameters (q id ,I d ,d+1,x+1,q mode ) and goto step 0

Linked assets

Computer Science Technical Report Archive

Conceptually similar

PDF

USC Computer Science Technical Reports, no. 887 (2007)

PDF

USC Computer Science Technical Reports, no. 812 (2003)

PDF

USC Computer Science Technical Reports, no. 797 (2003)

PDF

USC Computer Science Technical Reports, no. 814 (2004)

PDF

USC Computer Science Technical Reports, no. 837 (2004)

PDF

USC Computer Science Technical Reports, no. 789 (2003)

PDF

USC Computer Science Technical Reports, no. 781 (2002)

PDF

USC Computer Science Technical Reports, no. 788 (2003)

PDF

USC Computer Science Technical Reports, no. 765 (2002)

PDF

USC Computer Science Technical Reports, no. 803 (2003)

PDF

USC Computer Science Technical Reports, no. 804 (2003)

PDF

USC Computer Science Technical Reports, no. 673 (1998)

PDF

USC Computer Science Technical Reports, no. 884 (2006)

PDF

USC Computer Science Technical Reports, no. 674 (1998)

PDF

USC Computer Science Technical Reports, no. 780 (2002)

PDF

USC Computer Science Technical Reports, no. 654 (1997)

PDF

USC Computer Science Technical Reports, no. 770 (2002)

PDF

USC Computer Science Technical Reports, no. 663 (1998)

PDF

USC Computer Science Technical Reports, no. 775 (2002)

PDF

USC Computer Science Technical Reports, no. 657 (1997)

Description

Jabed Faruque, Ahmed Helmy. "RUGGED: Routing on fingerprint gradients in sensor networks." Computer Science Technical Reports (Los Angeles, California, USA: University of Southern California. Department of Computer Science) no. 806 (2003).

Asset Metadata

Creator Faruque, Jabed (author), Helmy, Ahmed (author)

Core Title USC Computer Science Technical Reports, no. 806 (2003)

Alternative Title RUGGED: Routing on fingerprint gradients in sensor networks (title)

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)

Tag OAI-PMH Harvest

Format 18 pages (extent), technical reports (aat)

Language English

Unique identifier UC16269150

Identifier 03-806 RUGGED RoUting on finGerprint Gradients in sEnsor Networks (filename)

Legacy Identifier usc-cstr-03-806

Format 18 pages (extent),technical reports (aat)

Rights Department of Computer Science (University of Southern California) and the author(s).

Internet Media Type application/pdf

Source 20180426-rozan-cstechreports-shoaf (batch), Computer Science Technical Report Archive (collection), University of Southern California. Department of Computer Science. Technical Reports (series)

Access Conditions The author(s) retain rights to their work according to U.S. copyright law. Electronic access is being provided by the USC Libraries, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Repository Email csdept@usc.edu

Inherited Values

Title Computer Science Technical Report Archive

Description Archive of computer science technical reports published by the USC Department of Computer Science from 1991 - 2017.

Coverage Temporal 1991/2017

Repository Email csdept@usc.edu

Repository Name USC Viterbi School of Engineering Department of Computer Science

Repository Location Department of Computer Science. USC Viterbi School of Engineering. Los Angeles\, CA\, 90089

Publisher Department of Computer Science,USC Viterbi School of Engineering, University of Southern California, 3650 McClintock Avenue, Los Angeles, California, 90089, USA (publisher)