Comprehensive Study of Routing Management in

Wireless Sensor Networks- Part-1

Vasu Jolly, Shahram latifi
{jolly,latifi}@egr.unlv.edu

Abstract
Increasing computing and wireless communication capabilities will expand the role of the sensors from
mere information dissemination to more demanding tasks as sensor fusion, classification, collaborative
target tracking. Wireless sensor nodes can be mobile within a chosen area and communicate with
neighboring nodes in the bounds of protocol limits. Typically they are densely deployed in large numbers
and are prone to failures, and frequent topology changes. Among many concerns about design of sensor
networks are growing bandwidth demands, speed of information retrieval and transporting bytes over the
wireless networks to provide a quality service for the diverse requirements of the users, such as signal
processing or multimedia applications. This Paper has been split in two parts to discuss past and ongoing
research in routing schemes in WSNs. Current work a comprehensive investigation of different routing
schemes used in wireless sensor networks. Depending on the sensor applications, the design challenge,
advantage and performance concern for each routing protocols is also revealed.

1. Introduction
We review the state-of-the-art routing techniques for wireless sensor networks. The routing
scheme for wireless sensor networks has to be straightforward and simple which does not expend much
computation power and memory, and eventually minimize communication among nodes to save its power.
Routing protocols may be classified into one of the ensuing three models [1] (a) single hop model, (b)
multi-hop model and (c) Cluster-based hierarchical model. We will discuss each model briefly and further
classify the protocol based on network structure and protocol operation in the subsequent sections. Single
hop model is the simplest model and act as a direct communication model. In this model, fig 1 (a), all the
nodes travel one hop to reach to a base station or the sink node. This kind of single hop transmission is
highly unrealistic in the real world. The transmission range of each node and the energy consumed plays a
crucial role in defining the sensor network. The multi-hop model supports the collaborative effort of several
nodes within the sensor cloud, fig 1 (b). Each sensor node has a radio range, which is referred to as the
distance at which the signal strength remains above the minimum usable level for that particular node to
transmit and receive. If two nodes cannot communicate directly, the nodes positioned between those two
nodes, transmit an information packet from the source node to the destination node. Information is received
only by nodes within the radio range of the forwarding node in a wireless medium. In view of efficient
energy consumption, this model follows more practically feasible approach and is employed by [2], [3], [4],
[5], [6]. The multi hop model uses the data aggregation techniques.

Figure 1 (a) Single hop routing model (b) Multi-hop routing model
 Within a sensor cloud, variation in the rate of consuming power by each node depends on factors
such as event sensing rate, distance from sink node, and location of each node relative to other nodes. This
disparity in energy consumption in wireless sensor network causes an imbalance of node power status.,
Figure 1(c) resulting in diminishing overall network lifetime. If the sink node is at one fixed location,
information packets gather from the entire network to one fixed sink. This result in denser information
traffic around the nodes in vicinity of the sink, as compared with the nodes placed farther from the sink.
Hence, the nodes close to the sink will exhaust energy at the faster pace. If the nodes around the sink drain
their energy, the sink is isolated from the entire sensor network, thereby making the data collection
impossible. The segregation of the sink node from entire network is called self induced black hole effect
[1]. To avoid isolation of sink node from the network, it is necessary to adopt an energy conservation
heuristic on nodes located around the sink.

Figure 1(c) Disparities in power spending

In the cluster based model, fig 2, the network is divided into clusters comprising of “X”
amount of nodes. Cluster head, which is master node, within each respective cluster is responsible for
routing the information to the other cluster head. Data is first aggregated within the cluster and then from
cluster to cluster. As the data packets moves from one cluster to another, it covers larger distances. This
results in very low data latency as compared to multi hop model and single hop model respectively.
However this model has a drawback. As the distance between clustering levels increases, the energy spent
grows proportional by the square of distance. This definitely increases the energy consumption of the
sensor network.

Figure 2 Cluster based hierarchical routing model

2. Protocol Assessment
In traditional networks, the focus is on Quality of Service (QoS). In wireless sensor networks QoS
requirements can be relaxed to preserve energy and the network lifetime. At each layer of protocol stack,
steps must be taken to (a) save energy, (b) allow sensor nodes to reconfigure, and (c) update their
respective tasks according to the resources available. The simulation test bed should be as simple as
possible. Diverse environmental conditions need to be implemented in analyses and simulations [1]. Some
of the parameters used to evaluate the routing performance are (a) Energy consumption, both with respect
to an individual node and the whole network (b) Simulation time and latency, (c) success rate of the data
packets reaching the sink, (d) network size, and finally (e) fault tolerance capability of the entire network.
In addition, the routing protocol should incorporate some kind of security to evade vulnerability from
adversaries.
3. Routing Methodology in Wireless Sensor Networks
In WSNs, discovering the routes and then sustaining them is practically insignificant because of the
energy constraints and sudden node failures. First of all, there is no control on topology of the nodes within
a sensor network cloud. Secondly the unpredictable topological change makes it impossible to stick to a
fixed routing strategy. Some well known routing plans such as data-centric methods, in-network
processing, clustering, data diffusion, data aggregation and energy aware methods are proposed in the
literature to cater the requirements for wireless sensor networks. Figure 3 illustrates the lineage of the
routing protocols in wireless sensor networks. Broad classification is on the basis of
• Network structure and
• Protocol operation
Based on the network structure, the protocols are further classified as flat-network routing,
hierarchical-based routing, and location-based routing. Flat routing protocols distribute information as
needed to any reachable sensor node within the sensor cloud. No effort is made to organize the network or
its traffic, only to discover the best route hop by hop to a destination by any path. All nodes are assigned
uniform functionality. In hierarchical-network routing, nodes play different roles in the network. These
protocols often group sensor nodes together by functionality and merge them into a hierarchy. Location-
based routing uses the physical position of a sensor node in the network to route packets to that node. If a
node changes location, the connection to that node will be broken and another route is required to establish
its new location. Based on their operation, the protocols are further classified into Query based, negotiation
based, multipath based, QoS based and coherent routing. The protocols categorized here often overlap on
top of each other. We will explain each relevant protocol on the basis of network structure and protocol
operation respectively.

Figure 3 Classification of Routing Protocols [7]

4. Flat Routing
In flat network architecture all the sensor nodes are equal and connections between nodes are set
up in short distance to establish the radio communication. Route discovery can be carried out in sensor
networks using flooding or broadcasting which do not involve topology maintenance. In this section, we
will provide an in-depth knowledge of the flat routing based protocols.
4.1 Flooding
Flooding is an example of simple flat routing scheme. When a sensor node receives a data packet, it stores
the data and broadcasts it to its neighboring nodes. This process repeats until the information reaches all the
sensor nodes in an entire network. To perform flooding, figure 4 (a), sensor nodes do not need any
knowledge of the network configuration. Sensor nodes distinguish each data packet, while receiving or
transmitting a data packet. This will save the limited memory space of each node. Since flooding does not
require any complicated routing algorithms, it can be easily implemented for sensor networks. However,
there are some shortcomings in this scheme which dissipate the limited resources of the sensor nodes [2].
One such problem in classic flooding is implosion. Implosion occurs, when a data receiving node
broadcast the data packet to its all neighboring nodes, irrespective of whether the neighboring node already
has the same data or not. Figure 4 (b) illustrates implosion; here node D floods the information to its
neighboring nodes E and F respectively. Node H being the neighbor to both E and F, gets the same copy of
the information from both E and F. Due to indiscriminate transmission of data, sensor nodes in this scheme
expend limited transmission energy and bandwidth. Another problem associated with flooding is overlap.
This situation occurs when multiple nodes observe the same sensor region, they generate the overlapping
data. The sensor nodes within its neighborhood receive the multiple copies of the same packet containing
the same information. Overlap, like implosion expends transmission power and bandwidth. Figure 4 (c)
shows that J and K collect overlapped information and flood to their respective common neighbor/s, here
node L. In flooding, the overlap problem is more difficult to solve than the implosion problem, because
implosion is a function of network topology, whereas overlap is a function of both topology and observed
data. Additionally, in classic flooding, the sensor nodes are not resource aware, i.e. sensor nodes do not
update their activity status according to the energy constraints at any given time. These shortcomings
reduce the battery life of sensor node and therefore shorten the entire network life span.

Figure 4 (c) The Overlap Problem

Figure 4(a) Flooding Techniques Figure 4 (b) Implosion Problem associated with Flooding

Nonetheless, Flooding, due to its easier implementation and simple design has been investigated
intensively to overcome the abovementioned shortcomings. In [9] and [10], each sensor node only needs to
know a small portion of entire network configuration, which is the location information of its neighboring
nodes, instead of the information of entire network topology. Gossiping is another by-product of flooding
[8].When a node receives the data packet it selects a subset of its all neighboring nodes and transmits the
data packet to the subset, instead of all neighboring nodes. This reduces the consumption of transmission
energy.

4.2 SPIN- Sensor Protocols for Information via Negotiation

Another example of flat routing is negotiation based protocols. Sensor Protocols for Information via
Negotiation- SPIN [2] intend to disseminate data towards the sink using negotiations. It is assumed that the
source has an observed data meant to be transported to sink node. The source node advertises its data over
the sensor network. Those nodes desiring the sensor data, request it from the source. For the negotiations,
the information descriptors called “meta-data” are used. Upon sensing the information packet, figure 5, the
observer or source node transmits a small advertisement packet (ADV) to its all neighboring nodes except
the one from which the node receives the data packet. The ADV contains the information of actual data.
Upon receiving the ADV, a neighboring node checks its local cache whether the node already has the same
data or not. If the neighboring node already has the data, the ADV is rejected. If the node does not have the
desired data, it sends a request message (REQ) to the receiving node. Then, the receiving node transmits
the data packet (DATA) to the neighboring nodes, which request the data by sending the REQ message.
The corresponding neighboring node then replicates this procedure with its neighbors. As a result, the
entire sensor network will acquire a copy of the data. This guarantees that there is no redundant information
sent throughout the network. The SPIN family of protocols includes several schemes with minor
modifications on the actual proposal [2].
SPIN-1 includes negotiation before transmitting information to guarantee that only useful
information will be transferred. It is a three-way handshake protocol, as mentioned above. SPIN-2 is a
modification to SPIN-1 which in addition to a three- way handshake includes a resource awareness
mechanism [7]. SPIN-2 works under resource constraint environment. Each sensor node has its own
resource manager, which keeps account of the expended and remaining power. Before each transmission,
the nodes examine their resource manager and curb on other energy expending activities to increase the
lifetime of the node.

Figure 5 (a) Observer / Source node acquiring Data, (b) Receiver sending ADV message, (c) Desired
nodes sending REQ message and (d) Source sending the Data to selective neighbors

SPIN-BC is developed for broadcast networks in which the sensor nodes use a single shared
channel to communicate with each other. In SPIN-BC, the sensor node on receiving the ADV message does
not send a REQ message instantaneously. Instead, it waits for a certain time before sending the REQ
message. This is to avoid the redundant request for the same message. In SPIN-RL, each node keeps track
of all the nodes from where it hears the advertisement. If it does not receive any requested data within a
certain period of time, it sends out the request again. Similarly after transmitting the data message, sensor
node waits for a certain period of time before responding to other requests for the same data message.
Mainly SPIN-RL is used for the lossy channels. SPIN-PP has been developed to work with point-to-point
communication. In SPIN-PP, two nodes can have a direct communication, without the need for
intermediate nodes. It is a simple 3-way handshake scheme in which energy is not considered to be a
constraint. [11]. In SPIN-EC the sensor nodes follow the 3-way handshake like SPIN-PP but there is an
energy-conservation heuristic added to it. Sensor node contributes actively in the protocol only if it is above
a certain energy threshold and believes it can complete all the other stages of the protocol. Performance
evaluation of SPIN [2] demonstrates that SPIN is more energy-efficient than flooding or gossiping while
distributing data at the same rate or faster than these protocols. However the SPIN suffers from the
weakness [12] of transmitting all the data packets at the same Energy level and not using the distance to a
neighbor to adjust the energy level. Besides a large overhead in broadcasting the data, energy consumption
is a concern in SPIN.

4.3 Directed Diffusion

Directed Diffusion, [13], [14] is a data-centric and application-aware routing scheme where all the
information generated by sensor nodes is named by attribute-value pairs. This is a Sink-initiated reactive
routing paradigm in which routes are established as they are requested. In data centric routing the data
originating from different sources are combined with in-network aggregation by eliminating redundancy,
minimizing the number of transmissions; thus saving network energy and prolonging its lifetime. Unlike
conventional networks, maintaining a node addressing table, results in a large overhead. Instead the data
queries are exchanged. For example in the task of monitoring a building’s entrance, the request to gather
the “The total number of people going out” is more appropriate than the request to “gather the readings
from the nodes W,X,Y,Z”. The sink node requests data by broadcasting “interests” or sensing task. Interest
specifies the sensing task; include type of sensing event, sensing area, duration of sensing task, and event
transmission frequency. Figure 6 illustrates the operation of directed diffusion. The interest is disseminated
throughout the network in a hop-by-hop manner. The query is initiated by the sink node and it broadcasts
its interest message periodically to all of its neighbors. An interest cache is maintained by each node. When
a node receives an interest, it stores the interest and also sets up a gradient toward the node, from which it
received the interest. This process continues until gradients are established from the sources back to the
sink. If a respective node has the requested data, which matches the received interest, the node sends back
the data packet to the sink in multiple paths according to the gradients. On receiving the data packet at the
sink, the reinforcement of the optimal path is initiated by the sink. The criterion for the selecting the
optimal path highly depends on the application. It may be the shortest path or minimum energy consuming
path, whichever suits the application. The best path is reinforced by the sink sending a new interest to the
path.

Figure 6 (a) Directed diffusion operation - Diffusing interests

Figure 6 (b) Directed diffusion operation- Setting up gradients

Figure 6 (c) Directed diffusion operation - Path reinforcement

In flooding and SPIN, data collection is initiated by source nodes. In other words, source nodes
start transmitting data whenever an event is observed. On the contrary, in directed diffusion data collection
is initiated by sink nodes. Because of sink initiated data collection, directed diffusion can limit data flow.
By doing so, it will reduce unnecessary data transmissions and thus energy consumption of sensor nodes
will be reduced. Such type of information retrieval is appropriate for continual queries where requesting
nodes do not anticipate data that fulfill a query for duration of time. Possibility of transmission overhead
created by interests creates another disadvantage of using this scheme. When a sink broadcasts an interest,
the sink does not know whether the data, which will match the interest, is obtainable or not. If the data is
not available at that time, sink node can not collect any data at all. Furthermore, this makes it unsuitable for
one-time queries, as it is not worth setting up gradients, which use the path only once. For instance,
directed diffusion is not applicable data dissemination scheme for surveillance purpose because sensor
nodes have to transmit data as soon as they detect abnormality.
4.4 Rumor Routing
Rumor routing [15] is a data centric scheme proposed for applications where geographic routing is
not practically possible. Rumor routing directs the queries to the sensor nodes that have observed a specific
event rather than flooding the entire sensor network to extract information about the occurring events. The
rumor routing utilizes a set of long-lived agents which create paths that are directed towards the events they
encounter. On detecting an event, sensor node adds it to its events table, and generates an agent. Agents
travel within the sensor network to disseminate information about local events to distant nodes. When a
node generates a query for an event, the nodes that know the route, may reply to the query by inspecting its
respective event table. The agent has a lifetime of a specific amount of hops after which it dies. Any node
creating a query will transmit the query if it has a route to the event, otherwise it will transmit it in a
random direction. If the node discovers that the query did not reach the destination, then it will flood the
network. The fewer the number of queries which flood, the less is the energy consumed. Unlike directed
diffusion where data can be routed through multiple paths at low rates, rumor routing keeps only one path
between source and destination. Simulation results [15] show that rumor routing can handle the node
failure and achieve major energy savings when compared to flooding. Rumor routing performs well as long
as the number of events is small. For the large number of events, the cost of generating and maintaining the
event tables and agents results in a large overhead [7].

4.5 COUGAR
COUGAR [16] is a data- centric protocol and follows the directed diffusion model along the database
approach. COUGAR utilizes in-network data aggregation for power saving. There exists an additional
query layer between network and application layer. It abstracts the information generated by network in an
update-only relational table. The attribute in this relational table is (a) details about the sensor node, for
example its location, ID and (b) information collected from respective node (e.g., temperature, light).
Sensor applications are often interested in summarized and consolidated data that are produced by
aggregated queries rather than detailed data. In addition, sensor nodes choose a leader node to initiate the
aggregation and transmit the information to the sink. Sink is liable for generating a query, which specifies
the details about the data flow and in-network computation for the incoming query and sends it to the
appropriate nodes. In COUGAR, sensor readings are treated like “virtual” relational database tables and a
query language like SQL may be used to issue tasks to the WSN. COUGAR has some drawbacks [7]. First,
inclusion of query layer on each sensor node puts an extra overhead on the nodes and the entire network in
terms of power and memory storage. Second, a high level of synchronization among the sensor nodes is
required to achieve the in-network data computation.

4.6 ACQUIRE
ACQUIRE [17] stands for active query forwarding in sensor networks. It is a data-centric,
application specific scheme for querying wireless sensor networks. In ACQUIRE an active query is passed
through the sensor network. The intermediate sensor nodes use cached local information (within the look
ahead of “d” hops) to partially resolve the query. When the query is resolved an entire response is sent
directly back to the querying node. For the complex queries, directed diffusion may not be the right choice,
because it uses flooding based query mechanism, which would expend energy. ACQUIRE can adjust the
look ahead parameter “d” to offer an efficient querying [7]. When d is equal to the network diameter,
ACQUIRE performs similar to flooding. If “d” is too small, the query has to travel more hops. ACQUIRE
performs better than directed diffusion and the optimal ACQUIRE can lessen the energy consumption by
more than 60% as compared to expanding the ring search.

5.0 Conclusions
The sensor networks, jointly with sensing devices, embedded processors, and communication
components, uses an appropriate energy-efficient and fast routing strategy to deliver the data to the desired
node.To make sensor networks truly advantageous for common applications, they must be reliable, robust,
energy efficient and resistive to topology changes. Although the commercially available sensor nodes are
very cheap, but designing the infrastructure and application usage cost should be minimal. Collecting data
and routing appropriate and needed data to the end user is a challenging issue in such a wireless battery
operated small sensor networks. Sensor information is data centric and using traditional network protocols
are not always appropriate or sufficient. This paper provides a comprehensive investigation of different
routing schemes used in wireless sensor networks. Depending on the sensor applications, the design
challenge, advantage and performance concern for each routing protocols is also revealed.

Reference:
[1] Jamil Ibriq, Imad Mahgoub,"Cluster-Based Routing in Wireless Sensor Networks:: Issues and
Challenges", Proceedings of the 2004 Symposium on Performance Evaluation of Computer and
Telecommunication Systems- SPECTS, July 25-29 2004
[2] W. R. Heinzelman, J. Kulik, and H. Balakrishnan, “Adaptive Protocols for Information
Dissemination in Wireless Sensor Networks,” In Proceedings of 5th ACM/IEEE Mobicom Conference,
August 1999.
[3] Narayanan Sadagopan, Bhaskar Krishnamachari, and Ahmed Helmy, "Active Query Forwarding in
Sensor Networks (ACQUIRE),"Journal of Ad Hoc Networks, Elsevier, Vol 3, Issue 1, pp. 91-113, January
2005.
[4] Stephanie Lindsey, C. S. Raghavendra, "PEGASIS: Power Efficient Gathering in Sensor Information
Systems," 2002 IEEE Aerospace Conference, March 2002.
[5] Chalermek Intanagonwiwat, Ramesh Govindan, Deborah Estrin, John Heidemann and Fabio Silva,
Directed Diffusion for Wireless Sensor Networking IEEE/ACM transactions on networking, vol. 11, no.
1, February 2003
[6] Koustuv Dasgupta, Konstantinos Kalpakis, Parag Namjoshi "An Efficient Clustering--based Heuristic
for Data Gathering and Aggregation in Sensor Networks". IEEE Wireless Communications and
Networking Conference (WCNC), New Orleans, Louisiana, USA, March 16-20, 2003.
[7]Jamal N. Al-Karaki Ahmed E. Kamal, Routing Techniques in Wireless Sensor Networks: A Survey",
IEEE Wireless Communications, Volume: 11, Issue: 6 ,26- 28, Dec. 2004
[8] B. Krishnamachari, Y. Mourtada, and S. Wicker, The Energy-robustness Tradeoff for Routing in
Wireless Sensor Networks, Autonomous Networks Research Group Technical Report ANRG TR 02-001,
September 2002.
[9] H. Lim and C. Kim, Multicast Tree Construction and Flooding in Wireless Ad Hoc Networks, In
Proceedings of ACM Modeling, Analysis, and Simulation of Wireless and Mobile Systems, pp. 61-68,
2000.
[10] H. Sabbineni and K. Chakrabarty, “Location-Aided Flooding: An Energy-Efficient Data
Dissemination Protocol for Wireless Sensor Networks,” IEEE Transactions on Computers, Volume 54,
Issue 1, pp. 36-46, January 2005.
[11] Archana Bharathidasan and Vijay Anand Sai Ponduru,"Sensor networks: An overview" Department of
Computer Science University of California, Davis, UCDAVIS 2003.
[12] Gunjan Khanna, Saurabh Bagchi, Yu-Sung Wu,” Fault Tolerant Energy Aware Data Dissemination
Protocol in Sensor Networks ” 2004 International Conference on Dependable Systems and Networks
(DSN'04) 2004 Florence, Italy, pp 795
[13] Chalermek Intanagonwiwat, Ramesh Govindan, Deborah Estrin, John Heidemann and Fabio Silva,
Directed Diffusion for Wireless Sensor Networking IEEE/ACM transactions on networking, vol. 11, no.
1, February 2003
[14] C. Intanagonwiwat, R. Govindan, and D. Estrin, "Directed diffusion: a scalable and robust
communication paradigm for sensor networks," Proceedings of ACM MobiCom '00, Boston, MA, 2000,
pp. 56-67.
[15] D. Braginsky and D. Estrin, "Rumor Routing Algorithm for Sensor Networks," in the Proceedings of
the First Workshop on Sensor Networks and Applications (WSNA), Atlanta, GA, October 2002.
[16] Yong Yao and J. E. Gehrke, The Cougar Approach to in-Network Query Processing in Sensor
Networks. Sigmod Record, Volume 31, Number 3, September 2002.
[17] N. Sadagopan et al., The ACQUIRE mechanism for efficient querying in sensor networks, in the
Proceedings of the First International Workshop on Sensor Network Protocol and Applications, Anchorage,
Alaska, May 2003.