CS541 Lecture7

CS-541
Wireless Sensor Networks

Lecture 7: Routing for Wireless Sensor Networks.
Spring Semester 2016-2017
Prof Panagiotis Tsakalides, Dr Athanasia Panousopoulou, Dr Gregory Tsagkatakis
CS-541 Wireless Sensor Networks

Spring Semester 2016-2017 1
University of Crete, Computer Science Department
Today’s objectives
• Routing Objectives for Wireless Sensor Networks
• Routing Techniques
• Routing in practice & associated services at the transport layer

Routing for Wireless Sensor Networks
Energy
Scalability
Application
& Density
Addressing Topology
Robustness
“Wireless Sensor Networks”, Ian F. Akyildiz, Mehmet Can Vuran, 2010, Willey, Ch 7
• Neighborhood discovery & local information exchange
• Network is more important than the sensor (WSN are
application-specific - information from a set of sensors
could be more important than the information of a
Energy single sensor (clustering or selecting or compressing)
Scalability
Application
& Density
Addressing Topology
Robustness
Scalability • Knowledge of the entire

Application topology is not realistic (also
& Density
because of its changes, due to,
e.g. nodes failures or RDC
policy)
• Autonomic behavior (self-
organizing and optimizing) –
from local decisions to converge
to a global optimum
• Topology is not network-driven
Addressing Topology but application driven!
• In-network processing for
combining & extracting the
important information
Robustness

& Density
policy)
organizing and optimizing) –
from local decisions to converge
to a global optimum
• Topology is not network-driven
Addressing Topology but application driven!
• Topology is not static
• In-network processing for
combining & extracting the
important information
Robustness
• Short-range and/or multi-hop
• Cheap hardware – prone to failures
• Unattended and hostile environments (RF harsh)
“Wireless Sensor Networks”, Ian F. Akyildiz, Mehmet Can Vuran, 2010, Willey, Ch 7 • Resilient to changes (e.g. the routing paths)

• Network is more important than the sensor (WSN are application
specific - information from a set of sensors could be more
important than the information of a single sensor (clustering or
Energy selecting or compressing)

& Density
policy)
• Conventional addressing (e.g. a organizing and optimizing) –
unique MAC address) is not met from local decisions to converge
in practice to a global optimum
• Addressing is overhead – • Topology is not network-driven
routing-based on addressing Addressing Topology but application driven!
may not work (dynamic • In-network processing for
topologies and failures) combining & extracting the
• Routing that does not rely on important information
unique IDs (position or context)
Robustness

• Periodic monitoring
• Event-driven
• Sometimes also combined, e.g. Scalability • Knowledge of the entire
Cyber-physical Systems Application topology is not realistic (also
& Density
• Different applications need different because of its changes, due to,
routing protocols e.g. nodes failures or RDC
policy)
Robustness

• Periodic monitoring
• Event-driven
• Sometimes also combined, e.g. Scalability • Knowledge of the entire
Cyber-physical Systems Application topology is not realistic (also
& Density
• Different applications need different because of its changes, due to,
routing protocols e.g. nodes failures or RDC
policy)
Robustness

Classification of routing techniques
Geographical-
Data-centric Topology-driven
based
• Attribute-based • Information available with • Routing decisions are

addressing respect to the network taken with respect to the
• Sensor nodes are topology locations of the sensors
homogeneous (in terms of • Associated network metrics • Localization in the
sensing modality) OR are (e.g. QoS) background
clever enough to extract • Proactive or reactive (similar • Destination is a
some attributes from to ad-hoc routing protocols) geographical location
different modalities • Address-based routing for • Virtual coordinates to
• The physical phenomenon constructing a specific routing route messages in the
can be described by more topology (e.g. tree) network
than one sensors..
• Flat architectures
• Fault-tolerant
+Hierarchical approaches (clustering +routing)

Flooding & Gossiping

Data-centric
SPIN (Sensor Protocols for Information via
Negotiation)
• Attribute-based
addressing
• Sensor nodes are
homogeneous (in terms of Directed Diffusion
sensing modality) OR are
clever enough to extract
some attributes from
different modalities ACQUIRE
• The physical phenomenon
can be described by more
than one sensors..
• Fault tolerant

Flooding & Gossiping
flooding
Data-centric Flooding:
• Blind broadcast of every
packet received
• Attribute-based
addressing • Limiting rebroadcast (time-
• Sensor nodes are to-live policy)
homogeneous (in terms of
Gossiping:
different modalities • Random selection of a
• The physical phenomenon gossiping
can be described by more single neighbor for packet
than one sensors.. relaying
• Fault tolerant • Route trace to avoid having
rooting loops…

Flooding & Gossiping - Problems
Implosion flooding
Data-centric • Always sends data to a neighbor,
even it has already received the
data from another node
• Attribute-based • Density of the network
addressing
homogeneous (in terms of
Overlap
sensing modality) OR are • Nodes often cover overlapping
clever enough to extract areas (e.g. temperature distr.)
different modalities
• Density and mapping of observed
data
• The physical phenomenon gossiping
than one sensors.. Resource blindness
• Fault tolerant • Amount of energy available is not
taken into account

SPIN (information via negotiation)
Data-centric Controlled flooding
A family of routing protocols

• Attribute-based
addressing
• Sensor nodes are Broadcast information’s meta-data (e.g. ID,
homogeneous (in terms of location, etc) to negotiate before starting
sensing modality) OR are transmitting their data. -> ADV
different modalities Interested Parties respond with a REQ
than one sensors..
DATA packet is sent only to interested Heinzelmann, W. R.; Kulik, J.; and
• Flat architectures Balakrishnan, H. Adaptive Protocols for
parties (multicasting)
• Fault tolerant Information Dissemination in Wireless
Information propagates within the network Sensor Networks. In Fifth ACM/IEEE
MOBICOM Conference (August 1999).
No redundant data sent over the network.

SPIN (information via negotiation) + Energy
conservation (resource blindness)
Data-centric
Reduce participation of node when
approaching low-energy-threshold
• Attribute-based
addressing
• Sensor nodes are When node receives data, it only initiates
homogeneous (in terms of protocol if it can participate in all three stages
sensing modality) OR are with all neighbor nodes
Nodes with energy below a threshold:
can be described by more • When node receives advertisement, it does
than one sensors.. not request the data Heinzelmann, W. R.; Kulik, J.; and
• Flat architectures Balakrishnan, H. Adaptive Protocols for
• Fault tolerant • Node still exhausts energy below threshold Information Dissemination in Wireless
by receiving ADV or REQ messages Sensor Networks. In Fifth ACM/IEEE
MOBICOM Conference (August 1999).

Gossiping has the slowest convergence, flooding the fastest.
Gossiping: 85% of data in a small amount of time, most spent for the remaining 15% (random choice of neighbor – can be
optimized if with the route tracing mechanism)
SPIN: 80ms longer to converge than flooding
Results from a 25-nodes WSN

SpringW.
Heinzelmann, Semester 2016-2017
R.; Kulik, J.; and Balakrishnan, H. Adaptive Protocols for Computer
Information Dissemination 16
University of Crete, Science Department in
Wireless Sensor Networks. In Fifth ACM/IEEE MOBICOM Conference (August 1999).
Gossiping does not converge!

Gossiping is the most costly protocol
SPIN 3.5 times less energy than flooding
Total energy in the system: 1.6 Joule

None cannot deliver 100% of the requested
data volume
<0.2 J: none cannot distribute any data
@0.2J: gossiping distributes a small amount of data SPIN-2: SPIN-EC
@0.5 J: SPIN begins to distribute data
@1.1 J: flooding begins to distribute the data.
Directed Diffusion
Data flow is initiated from the sink (requests specific information from the sensors)
Data-centric
(a) Interest propagation: exploring for matching data.
• Attribute-based (b) Gradient setup: caching the interest message and the trace of the route from which
addressing it has been received (forming a reserved path towards the sink). Forwarding based
on flooding (or optimized, to compensate implosion and/or resource blindness)
homogeneous (in terms of (c) Source node: the one that has data / task matching the interest
sensing modality) OR are (d) When data become available, source sends data to sink using the gradient route.
some attributes from Routing becomes proactive
than one sensors..
• Fault tolerant
Intanagonwiwat, Chalermek, et al. "Directed diffusion for wireless sensor networking." Networking, IEEE/ACM Transactions on 11.1 (2003): 2-16.

Directed Diffusion
Data-centric Source node may have more than one gradients -> sink can reinforce
the use of a specific path, based on network metrics, e.g.:
• Link quality (e.g. SNIR, LQI)
• Delay
• Attribute-based
addressing
• Sensor nodes are Update the gradients w.r.t. to topology changes & apply negative
homogeneous (in terms of reinforcement to suppress a routing path OR establish multi-path
sensing modality) OR are delivery (e.g. for robustness)
• The physical phenomenon • Flooding: critical overhead
than one sensors.. • Variation: Push diffusion initiated from source nodes (SPIN) + For
• Flat architectures applications where data need to be initiated from the sensors
• Fault tolerant
• Upon receiving the advertisements, the sink sends reinforcement
packets to establish routes between the sources and the sink.


ACQUIRE
• Reduce the gradients overhead of directed diffusion
Data-centric
• No distinct query/response phase: complex query (i.e. sub-queries) that is
forwarded within the network on a random or predetermined path.
• Sensor nodes partially resolve queries over a d-hop neighborhood (on
• Attribute-based demand by restricted flooding) & then forward the query further (randomly
addressing or not) – performing updates within their d-hop neighborhood
• Sensor nodes are • The initial query is progressively resolved into smaller and smaller
homogeneous (in terms of components until it is completely solved and is returned back sink as a
sensing modality) OR are completed response.
clever enough to extract • Shortest path or reserved path is used to route the response back to the sink
can be described by more N modalities
than one sensors.. M sub-queries
• Fault tolerant
Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed Helmy. "Active query

forwarding in sensor networks." Ad Hoc Networks 3.1 (2005): 91-113.

ACQUIRE
Data-centric
• d: latency vs. energy-efficiency -> the larger the d the faster the
network resolves the query (at the expense of update costs…)
• Attribute-based • Amortization factor c: the frequency of updating w.r.t. queries,
addressing e.g.:
homogeneous (in terms of If c = 0.01 =>
Then an update has
some attributes from to be made every 100 queries
than one sensors..
• Fault tolerant
Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed

Helmy. "Active query forwarding in sensor networks." Ad Hoc
Networks 3.1 (2005): 91-113.
Stable network:
With a larger look-ahead,
the sensor can get more
ACQUIRE information that will remain
stable for a longer period of
Data-centric time which will help it to
answer subsequent queries
Increased dynamics in the network, better

to keep a small look-ahead parameters
Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed

Helmy. "Active query forwarding in sensor networks." Ad Hoc
Networks 3.1 (2005): 91-113.
• SAR
Topology-driven
• SPEED
• Information available with respect to the

network topology
• Associated network metrics (e.g. QoS)
• Proactive or reactive (similar to ad-hoc
routing protocols)
• Address-based routing for constructing a
specific routing topology (e.g. tree)

Sequential Assignment Routing (SAR)
• Quality-of-Service
• Table-driven approach
Topology-driven
• Multiple trees for the 1-hop neighborhood of the sink
(QoS-metrics)
• Each node specifies two parameters for each path to the
• Information available with
respect to the network
sink:
topology (a) energy resources
• Associated network (b) Additive QoS metric: Each path is also associated with an
metrics (e.g. QoS) additive QoS metric (energy and delay)
• Proactive or reactive
(similar to ad-hoc routing
(c) weighted QoS: product of QoS metric and the packet’s
protocols) priority => paths with higher QoS are used for packets with
• Address-based routing for higher priority
constructing a specific (which would be an application area for
routing topology (e.g. this metric?)
tree)
Establishment of paths based on

minimum cost forwarding.
Aggregating cost from hop to hop.
SPEED
• To take into account the spatial correlations

Topology-driven
• Packet guaranteed advancement at a given time.
• Information available with • End-to-end delay experienced by the packets is proportional

respect to the network to the distance between the source and the destination.
topology
• Associated network
metrics (e.g. QoS) • Components of SPEED:
• Proactive or reactive • neighbor beacon exchange: location information between neighbors
(similar to ad-hoc routing (NeighborID, Position, SendToDelay, and ExpireTime)
protocols)
• Address-based routing for • Delay estimation: time it takes between packet and ACK reception
(or moving average)
constructing a specific
routing topology (e.g.
tree)
• SendToDelay: forwarding using the stateless nondeterministic
geographic forwarding (SNGF) algorithm.
• SNGF forward packets to neighbors that can optimize delay

SPEED
A is the sender, D is the destination
Topology-driven
Nodes E,F,G are the
candidates for
• Information available with forwarding the
respect to the network
topology packet
• Associated network Selection is made
metrics (e.g. QoS)
• Proactive or reactive w.r.t.:
(similar to ad-hoc routing Advancing in
protocols)
• Address-based routing for distance
constructing a specific
routing topology (e.g.
tree)

A single speed guarantee can be supported in the network. Heterogeneous traffic?
• Geographical forwarding techniques

Geographical-
based
• Planar Graph Techniques
• Routing decisions are
taken with respect to the
locations of the sensors • With virtual coordinates
• Localization in the
background
• Destination is a
geographical location
• Virtual coordinates to
route messages in the
network

Geographical forwarding techniques
• Localized: 1-hop decisions (?)

• Next hop selection depends on the
forwarding technique.
• Partitioning the transmission range to
feasible and infeasible areas Feasible area
• Distance-based forwarding or Infeasible
reception-based forwarding i area
(sink – static)
S

• Distance-based:
A. Greedy Routing Scheme

Feasible area
Closest to the destination
Infeasible
i area
S
(sink – static)

• Distance-based:
B. Most Forward within Range: the one

that allows the most forward progress Feasible area
towards destination Infeasible
i area
S
(sink – static)
Progress of a 1-hop neighbor (X) within the

CS-541 feasible region: the orthogonal projection of the
32
line
connecting X and i onto the line connecting S and i.
• Distance-based:
C. Nearest Forward Progress:

Feasible area
The nearest neighbor with forward
progress Infeasible
i area
S
(sink – static)
(adjust transmission power & minimize
the possibility of collisions
• Distance-based:
D. Compass-based routing
Feasible area
Choose next hop that is angularly
closest to destination Infeasible
i area
S
(sink – static)

• Distance - Blacklisted: Feasible area

1-hop located at distance > threshold
are blacklisted from the selection
process
Why: to increase the chances of Infeasible
delivery – the higher the distance the i area
higher the attenuation… S
(sink – static)

• Reception-based:
0.7
• Best-reception within the feasible area
• Blacklisted based on Packet Reception
Ratio (i.e. the most reliable) within the
feasible area. Feasible area
0.9 0.4
Infeasible
i area
S 0.6
1
(sink – static) 0.2
Blacklisting can be absolute (i.e. predefined threshold) or

relative (i.e. percentage)
e.g. PRR threshold: 0.5
• Reception + Distance:
PRR x Distance: select from the 1-hop

neighbors closer to the destination the Feasible area
one that maximizes: Infeasible
PRR x distance improvement i area
S
𝑑(𝑥, 𝑆)
1- (sink – static)
𝑑(𝑖, 𝑆)
X

• Comparison over noisy and faulty channels (Marco Zúñiga, et al. , 2008)
PRR x distance improvement towards is highly suitable metric for geographic
forwarding in realistic environments (i.e. considering a realistic PHY model –
based on empirical measurements)
Zamalloa, Marco Zúñiga, et al. "Efficient geographic routing over

lossy links Spring
in wireless sensor networks." ACM Transactions on
Semester 2016-2017
Sensor
38
Networks (TOSN) 4.3 (2008): 12. University of Crete, Computer Science Department
Succeeds in dense networks
Dead-end: when the packet reaches a node that has no neighbor closer to the destination
𝑑(𝐵, 𝐷)
A A: Concave node
𝑑(𝐴, 𝐷)
“Guide to Wireless Sensor Networks”, S. Misra, I. Woungang, S. C. Misra, 2009, Springer, Ch 4

GEDIR (Geographical distance routing) – applies a greedy forwarding policy

(e.g., distance, progress, direction)
The concave node will not forward the message
Handling concave routing decision:

𝑑(𝐵, 𝐷) • Drop the packet (packet failure)
A
• Concave node disconnects itself from the network
𝑑(𝐴, 𝐷) w.r.t. the specific route and/or message (not-
A: Concave node stateless)
• No trace route
“Guide to Wireless Sensor Networks”, S. Misra, I. Woungang, S. C. Misra, 2009, Springer, Ch 4

• Geographic forwarding techniques

Alternative GEDIR:
Generating as many copies of the same message as D

the number of 1st hop neighbors.
Each intermediate node forwards i-th received copy

of the same message to the i-th best closest to
destination neighbor.
A node becomes concave if it has fewer neighbors

GEDIR: A-B-J (pck dropped)
than the number of received copies.
Temporary loops may be created, but the exit is
guaranteed with maximum node degree.
Xu Lin, Ivan Stojmenovic, Location-based localized alternate, disjoint and multi-path routing algorithms for wireless
networks, Journal of Parallel and Distributed Computing, Volume 63, Issue 1, January 2003, Pages 22-32


Alternative GEDIR: Which are the silent assumptions made?
D
Each intermediate node forwards i-th received copy
of the same message to the i-th best closest to
destination neighbor.
A node becomes concave if it has fewer neighbors

than the number of received copies.
GEDIR: A-B-J (pck dropped)
Temporary loops may be created.
Altr. GEDIR: A-B-J-B-C-J-C-F-G-H-D


Is J concave node?
Disjoint GEDIR:
D
Each interm. node will forward to the best neighbor
among those who never received the same message
before. Then it becomes inactive w.r.t. to that
message (drops duplicates).
A node becomes concave if there is no neighbors left GEDIR: A-B-J (pck dropped)
to forward the message. Altr. GEDIR: A-B-J-B-C-J-C-F-G-H-D
Disj. GEDIR: A-B-J-C-F-G-H-D

• Planar graph routing techniques

An alternative to geographic forward routing for alleviating the local-minima problem..
Abstracting the network as a graph
Applying locally the left-hand or the right-hand rule. How: find a successor node in clockwise or counter-
clockwise order respectively
S D S D
Non-planar planar
How to make a graph planar:

initial planar

How to make a graph planar (examples):

Initial Gabriel graph (GG)
GG: If max(d(u,w ),
d(v,w)) < d(u,v) -> uv
is not part of the GG
46
How to make a graph planar (examples cont’):

Relative Neighbors graph
Initial Gabriel graph (RNG)
GG: If max(d(u,w ), RNG: If w is in the

d(v,w)) < d(u,v) -> uv intersection of the u
is not part of the GG – v range -> uv is not
part
University of Crete, Computer Scienceof the RNG
Department
47
How to make a graph planar (examples cont’):

Delaunay graph

• Compass Routing II (FACE):
Traverses a sequence of adjacent faces until reaching the destination
Each face is traversed completely in order to determine the edge that intersects the s-t line
and is closest to the target.
The message is passed to that endpoint & face changes.
Ruhrup, Stephan. "Theory and practice of

geographic routing." Ad Hoc and Sensor Wireless
Networks: Architectures, Algorithms and Protocols CS-541 Wireless Sensor Networks
(2009): 69. University of Crete, Computer Science Department
• Face 2: Avoids the complete traversals

and performs the face change before
crossing the s-t line.
On each face traversal, a node u checks

whether the edge to the next node
intersects the s-t-line. If this is the case,
then u changes the face and continues
traversing the next face.
++guarantees delivery on planar graphs

++local position-based rules (stateless)
Ruhrup, Stephan. "Theory and practice of
geographic routing." Ad Hoc and Sensor Wireless
Networks: Architectures, Algorithms and Protocols CS-541 Wireless Sensor Networks
(2009): 69. University of Crete, Computer Science Department

GPSR: Greedy perimeter stateless routing
Combines greedy + face routing
Greedy & recovery phase: if a local minimum (concave node) is

reached, the algorithm switches to recovery mode (FACE routing
is applied)
greedy
recovery
Face: moving along the planar graph edges
Use position where face was entered and destination position to determine
when face can be left again, switch back to greedy routing
Ruhrup, Stephan. "Theory and practice of geographic routing." Ad Hoc and

Sensor Wireless Networks: Architectures, Algorithms and Protocols (2009): 69.

• Planar graph routing

techniques
In recovery mode you need to
know:
• Distance from destination
when entering
• First node in recovery mode

when entering
• Previous hop
Ruhrup, Stephan. "Theory and practice of geographic routing." Ad Hoc and

Sensor Wireless Networks: Architectures, Algorithms and Protocols (2009): 69.

i+1 i+1 Traverse order update: w.r.t. the cross

cw product between the projection of the
intersecting edge and the reference line
ccw
i i

GPSR: Greedy perimeter stateless routing
--calculation of the planar subgraph (planar graph could increase the

hop count)
--no PRR / link quality (considers deterministic propagation range)
Detour..
--calculation of the optimal direction is not possible with only local
information – solution. (the face change when crossing the s-t –line)
Adaptive Face Routing
Message @ ellipse on the traversal path: travel towards opposite direction.

Termination of face traversal: ellipse is reached for the second time/ traversal has reached the start point.
After the traversal, the message is sent to the node closest to the destination on the current face.
This node changes the face and continues the traversal of the next face.
AFR @ recovery mode: Ruhrup, Stephan. "Theory and practice of geographic

routing." Ad Hoc and Sensor Wireless Networks:
GreedySpring
Other
Semester Adaptive
2016-2017 Face Routing (GOAFR)
Architectures, Algorithms and Protocols (2009):5469.
• To sum up:
• Geometric properties and assumptions regarding the communication graph
• Theoretical analysis of the efficiency vs. real-life constraints (delivery cannot

always be guaranteed)
• Greedy forwarding is still an efficient and robust method for geographic

routing in dense networks.
• To think about: PRRxdistance + energy over greedy w. recovery phase

• Virtual Coordinates
Why?
(a) Location is not always accurate

(b) Distance is not the optimal metric for routing WSN
Generate virtual coordinates for the sensor nodes – e.g. using the connectivity graph
And then apply (standard) geographical routing on these coordinates.

Virtual Polar Coordinate Routing:
Spanning tree of the sensor nodes and assign to each

node a polar coordinate w.r.t its position within the
tree
Relaying to the next node, closer to destination

(minimum distance between two angles @the same
level)
Packets can also be routed towards the root of the tree
“Guide to Wireless Sensor Networks”, S. Misra, I. Woungang, S. C. Misra,

2009, Springer, Ch 5

NoGeo (Geographic Routing without Location Information)
Perimeter nodes: nodes lying on the outer boundary of WSN deployment
Assumptions to gradually generalize:

• Perimeter nodes know their location
• Perimeter nodes know that they are perimeter nodes,
but don’t know their location
• Nodes know neither their location, nor whether they
are on perimeter
Rao, Ananth, et al. "Geographic routing without location

information." Proceedings of the 9th annual international
conference on Mobile computing and networking. ACM, 2003.
NoGeo
• Perimeter nodes know their location
Non-perimeter nodes can recursively calculate their virtual

coordinates based on the barycenter of the polygon
defined by their 1st hop neighbors
Imply that non-perimeter nodes that have a perimeter node

among its neighbors will “move” towards that perimeter
node.
Rao, Ananth, et al. "Geographic routing without location information."

Proceedings of the 9th annual international conference on Mobile computing and
networking. ACM, 2003. University of Crete, Computer Science Department
NoGeo
• Perimeter nodes know that they are perimeter nodes, but don’t know their location
Flooding to discover the distances (in hops) between all perimeter nodes
Create a globally-known distance vector between all perimeters.
Coordinates: minimize hop distance and Euclidean distance of Virtual Coordinates
Triangulation algorithm that computes perimeter positions
(𝑚𝑒𝑎𝑠𝑑𝑖𝑠𝑡 𝑖,𝑗 − 𝑑(𝑖, 𝑗))2

𝑖,𝑗 𝑖𝑛 𝑝𝑒𝑟𝑖𝑚𝑒𝑡𝑒𝑟

2009, Springer, Ch 4-5
NoGeo
• Nodes know neither their location, nor whether they are on perimeter
Explore phase for determining if it is on the perimeter: if a node is the farthest away,
among all its two-hop neighbors from the first bootstrap node, then the node decides
that it is on the perimeter.
Randomly selected
2009, Springer, Ch 4-5
True positions Virtual coordinates
Rao, Ananth, et al. "Geographic routing without location information." Proceedings of the 9th annual international
conference on Mobile
Spring Semester computing and networking.
2016-2017
UniversityACM,
of Crete,2003.
Computer Science Department
62
Landmark Routing:
A small subset of nodes: landmarks.
Landmarks flood the network & every node records its hop count distances to these
landmarks.
The landmark distances are used to generate virtual coordinates for routing between the
nodes.
++simple to implement
++do not depend on the scale

“Guide to Wireless Sensor Networks”, S. Misra, I.
Woungang, S. C. Misra, 2009, Springer, Ch 4-5
Gradient Landmark based Routing (GLIDER)
With holes and/or irregular shapes
Landmark nodes partition the space in Voronoi cells
Voronoi cell: each node in cell

has the smallest distance from
its landmark
Packets are routed through

CS-541 Wireless Sensor Networks the tiles 64
Gradient Landmark based Routing (GLIDER)

“Guide to Wireless Sensor Networks”, S. Misra, I.
Woungang, S. C. Misra, 2009, Springer, Ch 4-5
2-level hierarchy:
global routing (how to get around holes)
local routing (to deliver a packet to the destination):
from tile to tile
Global topology is likely to remain stable: nodes may come and go, but large-scale topology
remains the same features.
Landmarks should be selected to capture the global topology of the sensor field.

So far…
Geographical-
Data-centric Topology-driven
based
• Flooding • SAR • Greedy forwarding

• Gossiping • SPEED • GEDIR
• SPIN • FACE
• Directed diffusion • GPSR
• Virtual Coordinates /
Landmark-based routing

Energy
Scalability
Application
& Density
Geographical-
Data-centric based
Addressing Topology
Geographical- Robustness
based Topology-driven
Energy
Scalability
Application
& Density
Geographical-
Data-centric
Standard? based
Addressing Topology
Geographical- Robustness
based
Topology-driven
Routing Standards: IEFT Routing protocol for Low-power Lossy Networks (LLN)
• Internet Engineering Task Force: the working groups

behind AODV, OLSR, etc.
Rationale: to prevent fragmentation in the sensor

networking market by providing an IP-based routing
standard
• Broad industrial support

• IPv6 routing over networks with significantly

higher packet loss rates than traditional IP
routing protocols.
• Extra scalable solutions
• Interoperable w.r.t. to PHY and MAC
• Many-to-one traffic pattern (e.g. sensor

nodes to sink node)
Routing in Wireless Sensors
Uses IP layer for constructing and

maintaining a routing tree-> 6LoWPAN RPL
A set of rules for supporting the implementation of

the μIP stack in WSN
μIP
6LoWPAN is not a network protocol!
• Runs on each node – not responsible for MAC
functions.
• Compressing the IP headers 6LoPWAN
• Packet (de)fragmentation Link Layer
e.g. IPv6: 1280 bytes @ link layer
Missing out on heavy functions &
Maximum capacity of IEEE 802.15.4: 128 bytes stripping packet headers to
Address resolving & Route discovery absolute minimum
• Routing tree: Destination Oriented Directed Acyclic Graph (DODAG)

Each DODAG is uniquely identified Tsvetkov, Tsvetko. "RPL: IPv6 Routing Protocol for Low Power and Lossy Networks."
Sensor Nodes–Operation, Network and Application (SN) 59 (2011): 2.
by:
(a) RPL instance ID –independent

set of DODAGs for a given
scenario (using the same routing
criteria)
(b) DODAG ID – root’s IPv6 address
(c) DODAG version number –
incremented as the DODAG is
reconstructed.
WithinSpring
the same network different instances
Semester 2016-2017
ofNetworks
CS-541 Wireless Sensor RPL (labeled with different IDs) 72
Logical sets of communication:
Candidate neighbors: all reachable nodes (same/different DODAG Versions)
Candidate parents: It includes only neighbor nodes that belong to the same DODAG
Version.
When a node stores a neighbor into the
parent set, it becomes attached to the given
DODAG.
The most preferred next hop taken from the

parent set.
A node may belong to more than one RPL
Instance. In this case, it must join a DODAG for
each RPL Instance.
Tsvetkov,
CS-541 Tsvetko.
Wireless "RPL: IPv6 Routing Protocol for Low Power and Lossy Networks." Sensor
Sensor Networks
University Nodes–Operation, Network
of Crete, Computer Science and Application (SN) 59 (2011): 2.
Department
4 types of control messages (ICMPv6) for topology maintenance and information

exchange
• DODAG Information Object (DIO): Routing control information (Rank of a node,

the current RPL Instance, IPv6 address of the root, etc.) – within 1hop
neighborhood for topology construction and maintenance.
• Destination Advertisement Object (DAO): Downwards traffic - propagates
destination information.
• DODAG Information Solicitation (DIS): on demand requests for DIO messages
from a reachable neighbor.
• DAO-ACK: in response to a DAO message.

Distinction between the levels of the tree -> Rank:
A scalar value that represents the relative position of the node w.r.t. other nodes
and the DODAG root.
Rank of Root: minimum (0) & increases as we move away from the root.
MinRankIncrease: the minimum difference in ranking between different levels of

the tree.
MaxRankIncrease: the maximum allowable difference in ranking between

different levels of the tree.
root
Tree construction:
Root node: sends DIO message. (Rank = 0)
Each node within range: Standby for DIO messages.
Selects parent based on the Objective Function and computes its

own Rank. (Rank is not the same as the Objective Function)
If it is a parent (i.e. there are nodes with a higher rank than its root
own), it updates the Rank in the DIO message and sends it to all
neighboring peers.
Process repeated until reaching a leaf, or when no more nodes are

left in range.
How to cope with routing loops
Loop example
DIOs are sent in the 1st hop neighborhood.
It may happen that child nodes are selected as

best next hops.
ETX: Expected transmissions count

Each node does not process the DIO messages
from nodes with higher Rank than itself
Within a DODAG a node cannot advertise a
If link fails, then the node in trouble should rank L>L+MaxRankIncrease
declare an invalid state, poison its routes, and
join the DODAG Version again.
Tsvetkov,
CS-541 Tsvetko.
Wireless "RPL: IPv6 Routing Protocol for Low Power and Lossy Networks." Sensor
Sensor Networks
of Crete, Computer Science and Application (SN) 59 (2011): 2.
Department
How to detect problems
Within data packet: RPL Packet Information field
1. Direction (upwards or downwards) Root Repairing
2. Rank mismatch (between sender and receiver) –

packet is not immediately dropped but if it
happens again RPL enters the tree route repair
mechanism.
Global repair or local repair
3. Forwarding error field (child node): there is no
valid route (Rank of the sender and the RPL
Instance ID)

Root Repairing / Topology Maintenance
DIO messages are periodically sent – how periodically?
Trickle Timer
JP Vasseur , IoT
Workshop “RPL
Tutorial” April 2011

Downwards Communication
To support controlling mechanisms (or specific queries) from

the root towards the tree members
2 operational modes:
Non-storing mode: Each node propagates its parent up to

the root (DAO message). The root calculates proactively the
path for each destination
Storing mode: Parents keep a routing table for possible

destinations (remember that in uplink traffic the parent
simply checks if the rank of the sender is consistent to its
own rank & does not check the objective function of the
sender) Tsvetkov, Tsvetko. "RPL: IPv6 Routing Protocol for Low Power and Lossy Networks." Sensor
of Crete, Computer Science and Application (SN)
Department 59 (2011): 2.
Sensor-to-sensor routes:
Non-storing mode: the data will reach the root -> destination sensor.
Storing mode: up to the closest common ancestor & then to the destination

RPL is flexible w.r.t. objective function for selecting the parent.
RPL Objective Function: combination of end-to-end metrics & application

constraints.
Max/Min Metric s.t. Constraint
Defined by the standard:
Node Metrics Link Metrics

Node state & attributes object: reflecting Throughput
the node workload
Node energy object Latency (maximum or over entire route)
Hop count Link reliability
RPL is flexible w.r.t. objective function for selecting the parent.
RPL Objective Function: combination of end-to-end metrics & application constraints.

Max/Min Metric s.t. Constraint
Default: hop distance from the root. (OF0)

Popular: ETX: The expected number of transmissions between two neighbors (or between
source and destination – additive over path). ETX = 1/PRR
Alternatives: Plenty! E.g:
R-metric: based on the probability that a packet is correctly received in each link of the
route (considering backoff & MAC retransmissions) OR
Q-metric: forward load balancing and power consumption during TX & RX
Di Marco, P.; Fischione, C.; Athanasiou, G.; Mekikis, P.-V., "Harmonizing MAC and routing in low power
and lossy networks," Global Communications
Conference
CS-541 (GLOBECOM), 2013 IEEE , vol., no., pp.231,236,
83
9-13 Dec. 2013
Oliveira, Jaudelice de, and

Joydeep Tripathi. "Performance
Evaluation of the Routing
Protocol for Low-Power and
CS-541 Wireless Sensor Networks Lossy Networks (RPL)."
Spring Semester 2016-2017 84 (2012).
Performance Evaluation
ETX
Q-metric
References and Material for Reading
“Guide to Wireless Sensor Networks”, S. Misra, I. Woungang, S. C. Misra, 2009, Springer, Ch 4-5
Intanagonwiwat, Chalermek, et al. "Directed diffusion for wireless sensor networking." Networking,
IEEE/ACM Transactions on 11.1 (2003): 2-16.
Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed Helmy. "Active query forwarding in
sensor networks." Ad Hoc Networks 3.1 (2005): 91-113.
Heinzelmann, W. R.; Kulik, J.; and Balakrishnan, H. Adaptive Protocols for Information Dissemination
in Wireless Sensor Networks. In Fifth ACM/IEEE MOBICOM Conference (August 1999).
"Wireless Sensor Networks: An Information Processing Approach”, F. Zhao, L. Guibas, 2004, Elsevier
/ Morgan Kaufmann – Ch 3

Kemal Akkaya, Mohamed Younis, A survey on routing protocols for wireless sensor networks, Ad
Hoc Networks, Volume 3, Issue 3, May 2005, Pages 325-349, ISSN 1570-8705,
http://dx.doi.org/10.1016/j.adhoc.2003.09.010.
Rodoplu, Volkan, and Teresa H. Meng. "Minimum energy mobile wireless networks." Selected
Areas in Communications, IEEE Journal on 17.8 (1999): 1333-1344.
Marco Zúñiga, et al. "Efficient geographic routing over lossy links in wireless sensor networks."
ACM Transactions on Sensor Networks (TOSN) 4.3 (2008): 12.
“Guide to Wireless Sensor Networks”, S. Misra, I. Woungang, S. C. Misra, 2009, Springer, Ch 4-5
"Wireless Sensor Networks: An Information Processing Approach”, F. Zhao, L. Guibas, 2004,
Elsevier / Morgan Kaufmann – Ch 3
Xu Lin, Ivan Stojmenovic, Location-based localized alternate, disjoint and multi-path routing
algorithms for wireless networks, Journal of Parallel and Distributed Computing, Volume 63, Issue
1, January 2003, Pages 22-32
Ruhrup, Stephan. "Theory and practice of geographic routing." Ad Hoc and Sensor Wireless
Networks: Architectures, Algorithms and Protocols (2009): 69.
Karp, Brad, and Hsiang-Tsung Kung. "GPSR: Greedy perimeter stateless routing for wireless
networks." Proceedings of the 6th annual international conference on Mobile computing and
networking. ACM, 2000.

Rao, Ananth, et al. "Geographic routing without location information." Proceedings of the 9th
annual international conference on Mobile computing and networking. ACM, 2003.
Tsvetkov, Tsvetko. "RPL: IPv6 Routing Protocol for Low Power and Lossy Networks." Sensor Nodes–
Operation, Network and Application (SN) 59 (2011): 2.
Di Marco, P.; Fischione, C.; Athanasiou, G.; Mekikis, P.-V., "Harmonizing MAC and routing in low
power and lossy networks," Global Communications Conference (GLOBECOM), 2013 IEEE , vol., no.,
pp.231,236, 9-13 Dec. 2013
Routing Over Low power and Lossy networks (ROLL). RFC6550 Description of Working Group.
http://datatracker.ietf.org/wg/roll/charter/, 2011.
Routing Metrics Used for Path Calculation in Low-Power and Lossy Networks. RFC6551 Description
of Working Group. http://datatracker.ietf.org/wg/roll/charter/, 2012.
Oliveira, Jaudelice de, and Joydeep Tripathi. "Performance Evaluation of the Routing Protocol for
Low-Power and Lossy Networks (RPL)." Performance Evaluation (2012).
JP Vasseur , IoT Workshop “RPL Tutorial” April 2011


CS541 Lecture7

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CS541 Lecture7

Uploaded by

Copyright:

Available Formats

CS-541

Wireless Sensor Networks

Spring Semester 2016-2017

Prof Panagiotis Tsakalides, Dr Athanasia Panousopoulou, Dr Gregory Tsagkatakis

CS-541 Wireless Sensor Networks

• Routing Objectives for Wireless Sensor Networks

• Routing in practice & associated services at the transport layer

CS-541 Wireless Sensor Networks

Scalability • Knowledge of the entire

Scalability • Knowledge of the entire

CS-541 Wireless Sensor Networks

Scalability • Knowledge of the entire

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

• Attribute-based • Information available with • Routing decisions are

CS-541 Wireless Sensor Networks

Flooding & Gossiping

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

Data-centric Controlled flooding

A family of routing protocols

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

Results from a 25-nodes WSN

CS-541 Wireless Sensor Networks

Gossiping does not converge!

Total energy in the system: 1.6 Joule

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed Helmy. "Active query

CS-541 Wireless Sensor Networks

Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed

Increased dynamics in the network, better

Sadagopan, Narayanan, Bhaskar Krishnamachari, and Ahmed

• Information available with respect to the

CS-541 Wireless Sensor Networks

Establishment of paths based on

• To take into account the spatial correlations

• Information available with • End-to-end delay experienced by the packets is proportional

• SNGF forward packets to neighbors that can optimize delay

CS-541 Wireless Sensor Networks

CS-541 Wireless Sensor Networks

• Geographical forwarding techniques

CS-541 Wireless Sensor Networks

Geographical forwarding techniques

• Localized: 1-hop decisions (?)

CS-541 Wireless Sensor Networks

Geographical forwarding techniques

A. Greedy Routing Scheme

CS-541 Wireless Sensor Networks

Geographical forwarding techniques

B. Most Forward within Range: the one

Progress of a 1-hop neighbor (X) within the

Geographical forwarding techniques

C. Nearest Forward Progress:

Geographical forwarding techniques

CS-541 Wireless Sensor Networks

Geographical forwarding techniques

• Distance - Blacklisted: Feasible area

CS-541 Wireless Sensor Networks