Sensor Networks

Chapter 10:
Cross Layer Protocols
Wireless Sensor Networks
Akyildiz/Vuran
1
Traditional Layered Approach
Network Layer
MAC Layer
Physical Layer
Energy Management Plane
Transport Layer
Cross-Layer Management Plane
Application Layer
Each protocol designed
independently
 Limited information is passed
between layers
 Good for abstraction and
development
 Bad for energy efficiency,
overhead, performance
Wireless Sensor Networks
Akyildiz/Vuran
2
Inter-Layer Effects
 Wireless channel (PHY)
 Channel characteristics drastically influence performance
[1,2]
 MAC and Routing
 Significant effect on each other due to interference
[1] M. Zuniga, et.al., ``Analyzing the transitional region in low power wireless links,’’ Proc. IEEE SECON ‘04, Santa
Clara, CA, Oct. 2004.
[2] C. Bettstetter, et.al., “Connectivity of Wireless Multihop Networks in a Shadow Fading Environment,’’
ACM/Springer Wireless Networks, vol. 11, no. 5, pp. 571-579, September 2005.
Wireless Sensor Networks
Akyildiz/Vuran
3
Inter-Layer Effects
 Transport and PHY layer
 Congestion and contention are highly coupled due to
broadcast nature of the wireless channel [3]
 Transmission power control and congestion affect each
other (CDMA scheme) [4]
[3] M. C. Vuran, V. C. Gungor, and O. B. Akan, ``On the interdependency of congestion and contention in
wireless sensor networks,'' Proc. SENMETRICS'05, July 2005.
[4] M. Chiang, ``Balancing transport and physical Layers in wireless multihop networks: jointly optimal
congestion control and power control,’’ IEEE JSAC, vol. 23, no. 1, pp. 104 – 116, Jan. 2005.
Wireless Sensor Networks
Akyildiz/Vuran
4
Our Vision
Application Layer
Energy Management Plane
MAC Layer
Cross-Layer
Melting
Cross-Layer Management Plane
Network Layer
Energy Management Plane
Transport Layer
Cross-Layer Management Plane
Application Layer
PHY Layer
Our View
Traditional Approach
Wireless Sensor Networks
Akyildiz/Vuran
5
Cross-layer Communication
 Receiver-based routing [5] and [6]
 The next hop is determined based on receiver contention
(MAC + Routing)
[5] P. Skraba, et. al., ``Cross-layer optimization for high density sensor networks: Distributed passive routing
Decisions,’’ in Proc. Ad-Hoc Now’04, Vancouver, July 2004.
[6] M. Zorzi, et. al., ``Geographic random forwarding (GeRaF) for ad hoc and sensor networks: multihop
performance,’’ IEEE Trans. Mobile Computing, vol. 2, no. 4, pp. 337- 348, Oct.-Dec. 2003.
Wireless Sensor Networks
Akyildiz/Vuran
6
Cross-layer Communication
 Performance analysis of receiver based routing [6, 7]
 Energy efficiency analysis
 Latency and multihop performance
 Only MAC & routing interaction is considered (no PHY
layer/lossless/simple channel model)
 MAC is based on a two radio node
[6] M. Zorzi, et. al., ``Geographic random forwarding (GeRaF) for ad hoc and
sensor networks: multihop performance,’’ IEEE Trans. Mobile Computing,
vol. 2, no. 4, pp. 337- 348, Oct.-Dec. 2003.
[7] M. Zorzi, et. al., ``Geographic random forwarding (GeRaF) for ad hoc and
sensor networks: energy and latency performance,’’ IEEE Trans. Mobile
Computing, vol. 2, no. 4, pp. 349 – 365, Oct.-Dec. 2003.
Wireless Sensor Networks
Akyildiz/Vuran
7
Cross-layer Communication
 In [8], the scheme in [6,7] is extended for single radio nodes
 Relies purely on geographical information
 Considers perfect channel conditions
[8] M. Zorzi, ``A new contention-based MAC protocol for
geographic forwarding in ad hoc and sensor networks,” in Proc.
IEEE ICC ‘04, vol. 6, pp. 3481 - 3485, June 2004.
Wireless Sensor Networks
Akyildiz/Vuran
8
Cross-layer Communication
 A cross-layer solution with a realistic channel model [9] (includ. fading
channel)
 Receivers contend based on a cost function as well as
geographical location
 Correlation between nodes’ cost functions are considered
 Based on MAC + routing interaction
 Does not consider transport layer and PHY layer issues
 Energy efficiency is not considered
[9] M. Rossi, et. al., ``Cost Efficient Localized Geographical Forwarding Strategies for
Wireless Sensor Networks,’’ in Proc. TIWDC 2005, Sorrento, Italy, 2005.
Wireless Sensor Networks
Akyildiz/Vuran
9
Cross-layer Communication
 Joint routing, MAC, and link layer optimization [10]
 TDMA for MAC and MQAM for modulation is considered
 Analytical results favor single-hop communication instead of multihop
 No communication protocol is proposed
 Transport layer issues such as congestion control are not
considered
[10] S. Cui, et. al., ``Joint routing, MAC, and link layer optimization in sensor
networks with energy constraints,’’ in Proc. IEEE ICC ’05, vol 2, pp. 725 – 729, May
2005.
Wireless Sensor Networks
Akyildiz/Vuran
10
Cross-layer Communications
 Existing work focus on pair wise cross-layering, e.g.,
PHY+MAC, MAC+routing.
 A complete cross-layer suite that replaces each layer is
required
Wireless Sensor Networks
Akyildiz/Vuran
11
Motivation
 Unique characteristics of WSN
 High density
 Limited resources
Energy
Processing
Memory
 Wireless channel
Wireless Sensor Networks
Akyildiz/Vuran
12
Motivation: High Density
 Very low reporting rates
 Duty cycle operation is required
 Traditional routing algorithms require neighborhood information
 Leads to contention and energy consumption
 Duty cycle introduces
 Delays
 Frequent re-routing
 Energy inefficiency
Wireless Sensor Networks
Akyildiz/Vuran
13
Motivation: Wireless Channel
 Unit disk model leads to simple connectivity graphs
R
Wireless Sensor Networks
Akyildiz/Vuran
14
Motivation: Wireless Channel
 Shadow fading model complicates things!
Wireless Sensor Networks
Akyildiz/Vuran
15
Motivation: Wireless Channel
 Shadow fading model complicates things!
 Short links not guaranteed
Wireless Sensor Networks
Akyildiz/Vuran
16
Motivation: Wireless Channel
 Shadow fading model complicates things!
 Short links not guaranteed
 Considerable amount of long links
Wireless Sensor Networks
Akyildiz/Vuran
17
Motivation: Wireless Channel
 Shadow fading model complicates things!
 Short links not guaranteed
 Considerable amount of long links
 Link quality varies with time
Wireless Sensor Networks
Akyildiz/Vuran
18
Motivation: Wireless Channel
 Traditional routing protocols
 Neighborhood information actually varies with time
 Requires frequent updates
 Duty cycle operation
 Nodes sleep for a certain fraction of time
  – duty cycle parameter
  = 0.1  Awake for 10%, sleep for 90% of the time
 Necessitates either frequent re-routing or scheduling
 Channel asymmetry
 Routing decisions made by the source node are not accurate
Wireless Sensor Networks
Akyildiz/Vuran
19
XLP: Cross-Layer Protocol
I. F. Akyildiz, M. C. Vuran, and O. B. Akan, “A Cross-layer Protocol for Wireless Sensor Networks,” in Proc. Conference on
Information Science and Systems (CISS ’06), Princeton, NJ, March 22-24, 2006.
M. C. Vuran, I. F. Akyildiz, ``XLP: A Cross-Layer Protocol for Efficient Communication in Wireless Sensor Networks’’ to appear
in IEEE Trans. Mobile Computing, 2010.
Application Layer
Transport
Network
MAC



Initiative Concept
 Communication incentive is passed to the
receiver
Receiver Contention

Potential receivers contend for packets and become
next-hop
Local XL congestion control

Highly congested nodes do not participate in
communication

Angle-based routing

Channel adaptive operation
PHY



Adaptive to local minima in case of ‘voids’ in the
network
Receivers adapt communication parameters based
on channel conditions
Duty cycle operation

Energy consumption centric operation via duty cycle
Wireless Sensor Networks
Akyildiz/Vuran
20
Initiative Concept
 Core of XLP
 A node participates in communication based on its
initiatives
 When a node has a packet to send, it broadcasts an RTS
packet
 A neighbor node contends for routing of the packet based
on its initiatives
Wireless Sensor Networks
Akyildiz/Vuran
21
Initiative Concept
 Node Initiative

 RTS  Th 


Th 
relay  relay 
1, if 


max
I  
    

min 

E rem  E rem 

0, otherwise


Wireless Sensor Networks
Akyildiz/Vuran
22
Initiative Concept
 Node Initiative depends on






 RTS

Th




Th
 relay  relay 

1, if 


m
ax
I 
 

 

min 

 Erem  Erem 




0
,
otherwise


Received RTS packet’s signal
to noise ratio (SNR) – channel
quality
Input packet rate for Cong. C.
Buffer level for Cong. C.
Remaining energy
 If all the inequalities are satisfied, node participates in communication
Wireless Sensor Networks
Akyildiz/Vuran
23
Initiative Concept
 Node Initiative implications

  RTS   Th 




Th
 relay  relay 

1, if 


m
ax
I 
 



min 

 Erem  Erem 



0, otherwise

Wireless Sensor Networks
Akyildiz/Vuran

Guarantees high channel
quality

Eliminates congested nodes

Eliminates congested nodes

Leverages energy
consumption
24
XLP Mechanisms
Source node
Router/Active node
Passive node
Source
Router
Router
Sink
Receiver-based contention & routing
Local congestion control
Angle-based routing
Duty cycle (d) based operation
Wireless Sensor Networks
Akyildiz/Vuran
25
XLP Mechanisms
 Transmission initiation
 Receiver contention
 Angle-based Routing
 Local XL congestion control
 Duty cycle determination
Wireless Sensor Networks
Akyildiz/Vuran
26
Transmission Initiation
 When a node has packet to send
 Listens to channel
 If channel is occupied, performs backoff with CWRTS
 If channel is idle, broadcasts an RTS packet
 Nodes receiving RTS packet
 Check their location relative to source and destination
 Measure RTS packet SNR, RTS
Wireless Sensor Networks
Akyildiz/Vuran
27
Transmission Initiation
Sink
Wireless Sensor Networks
Akyildiz/Vuran
28
Transmission Initiation
Infeasible nodes
Feasible nodes
Sink
Wireless Sensor Networks
Akyildiz/Vuran
29
Receiver Contention
 Feasible nodes calculate initiative, I
 Nodes with I=1 contend for the packet
 Each node location corresponds to a priority region Ai with backoff


window size CWi
Nodes with longer progress (closer to the sink) have higher priority
over other nodes
Based on the location, each node determines its region Ai and backs
off for
i 1
 CW
j 1
j
 cwi , where cwi  [0, CWi ]
 Contention winner sends a CTS packet
Wireless Sensor Networks
Akyildiz/Vuran
30
Receiver Contention

A3
Closer nodes to the sink are highly
likely to win the contention
A2
A1
Priority regions
Sink
RTS
A1
A2
A3
CTS
CTS
CTS
CW1
Wireless Sensor Networks
Akyildiz/Vuran
CW2
CW3
CW4
31
Receiver Contention
A3
A2
A1
Sink
RTS
CTS
CW1
Wireless Sensor Networks
Akyildiz/Vuran
CW2
32
Receiver Contention
A3
A2
A1
Sink
DATA
RTS
CTS
Wireless Sensor Networks
Akyildiz/Vuran
33
Receiver Contention
A3
A2
A1
Sink
DATA
RTS
CTS
Wireless Sensor Networks
Akyildiz/Vuran
ACK
34
Receiver Contention

A3

A2
A1

I = 1 may not hold for any node (high
congestion)
After CW4 if no CTS is heard, neighbors
send KEEP ALIVE packet
Sender determines congestion and
decreases transmission rate
Sink
A1
A2
A3
KEEP ALIVE
CW1
CW2
Wireless Sensor Networks
Akyildiz/Vuran
CW3
CW4
35
Angle-based Routing (ABR)
 In sparse networks,
‘voids’ exist
 Receiver contention may
lead to local minimum
 Packets need to be
routed around the ‘void’
Wireless Sensor Networks
Akyildiz/Vuran
36
How to determine local minimum
 Node broadcasts an RTS
packet
 If no CTS or KEEP ALIVE

packets received, re-broadcasts
RTS (maybe previous RTS
packet lost)
If no response  local
minimum reached
 Switches to angle-based
routing (ABR)
Wireless Sensor Networks
Akyildiz/Vuran
37
Angle-based Routing (ABR)
 Send RTS with
RTS
1 CW
 Angle-based routing
ABR bit Traverse
direction
(ABR) bit set
 Traverse direction –
clock wise
Wireless Sensor Networks
Akyildiz/Vuran
38
Angle-based Routing (ABR)
 Neighbor nodes send CTS
RTS
1 CW
ABR bit Traverse
direction
based on their locations
Wireless Sensor Networks
Akyildiz/Vuran
39
Angle-based Routing (ABR)


Enlarged view
Neighbor that has the smallest angle
with the source-sink vector reply
earlier
q1,2 < q1,3 < q1,4 < q1,5
2
q1,4
3
q1,3
1
q1,2
2
1
4
3
Source-sink
vector
5
RTS
4
5
Wireless Sensor Networks
Akyildiz/Vuran
CTS
CTS
CTS
CTS
40
Angle-based Routing (ABR)
ABR  basic XLP
 Angle-based routing is
terminated if packet
traverses closer to the
sink than the starting point
Wireless Sensor Networks
Akyildiz/Vuran
41
Angle-based Routing (ABR)
CW  CCW
 If network edge is reached,
perform counter-clock
wise angle-based routing
ABR  basic XLP
Wireless Sensor Networks
Akyildiz/Vuran
42
Angle-based Routing
sink

g
0
Wireless Sensor Networks
Akyildiz/Vuran
Sample route
 0-a: Basic XLP
 a-b: ABR with clock-wise
direction
 b-c: Basic XLP
 c-d: ABR with clock-wise
direction
 d-e: Basic XLP
 e-f: ABR with clock-wise
direction
 f-g: ABR with counter clockwise
 g-sink: Basic XLP
43
Local XL Congestion Control
 Relay nodes
 Participate in communication if
relay  Th
relay
 Source nodes
 Explicitly control the rate of generated packets
 (ii )
Wireless Sensor Networks
Akyildiz/Vuran
44
Local XL Congestion Control
lji
lii
mi
 Input packet rate
i  ii  relay  ii 
 Output packet rate

ji
jΝ iin
i  (1  ei )(ii  relay )
ei packet error rate
Wireless Sensor Networks
Akyildiz/Vuran
45
Local XL Congestion Control:
Relay Nodes
 Node’s transmit & receive times during active period:
(assuming a node is active for average of d fraction of time)
Trx  relayTPKT
Ttx  (1  ei )( ii  relay )TPKT


Tlisten    (1  ei )ii  ( 2  ei )relay  TPKT
where TPKT is the average duration required to successfully transmit
a packet to another node.
Wireless Sensor Networks
Akyildiz/Vuran
46
Local XL Congestion Control:
Relay Nodes
 Tlisten ≥0 (in order to prevent buffer overflow and maintain its
duty cycle), input relay packet rate is bounded by:
relay  
th
relay

th
relay

,

(2  ei )TPKT
(1  ei )

ii
(2  ei )
Wireless Sensor Networks
Akyildiz/Vuran
47
Local XL Congestion Control: Source
Nodes
 Control the rate of generated packets, ( )
 Source node waits for CTS packets
 If no CTS packets are received, performs retransmission
 (assumes there may be RTS packet loss)
 If KEEP ALIVE packet is received, decrease rate
ii
ii  ii 
1

 If data transfer is successfully, increase rate
ii  ii  
Wireless Sensor Networks
Akyildiz/Vuran
48
Local XL Congestion Control: Source
Nodes
  is the transmission rate throttle factor; assumed to be 2.
  is the increase factor and is assumed to be iio/10 where
the iio is the initial value of generated packet rate.
Wireless Sensor Networks
Akyildiz/Vuran
49
Effect of Duty Cycle
 Total energy consumed as a flow-based energy consumption analysis
E flow ( D)  E per  hop E[nhops ( D)]
 Expected number of hops from a source to sink with distance D
E[nhops ( D)] 
D  Rinf
1
E[ d next hop ]
E[dnext-hop] is the expected hop distance. Rinf is the approximated transmission
range.
Wireless Sensor Networks
Akyildiz/Vuran
50
Effect of Duty Cycle
 Consumed energy per hop in one hop for transmitting a
packet
E per hop  ETX  ERX  Eneigh
Eng. consumption of the
transmitter node
Eng. consumption of the
neighbor nodes
Eng. consumption of the
receiver node
 Detailed formula on the paper
Wireless Sensor Networks
Akyildiz/Vuran
51
Energy Consumption (Eflow)
Effect of D on energy consumption
Wireless Sensor Networks
Akyildiz/Vuran
52
Effect of Duty Cycle Parameter ()
 Energy consumption of a flow is minimal for ~0.002
 For small sized networks with <1K nodes, this operating
point may not provide connectivity in the network.
 Energy consumption has a local minima around =0.2,
which is a suitable operating region.
Wireless Sensor Networks
Akyildiz/Vuran
53
Recap: XLP Cross-Layer Interactions
Application Layer
Transport
Network
MAC Layer
PHY
Wireless Sensor Networks
Akyildiz/Vuran
54
Recap: XLP Cross-Layer Interactions
 A node monitors its channel
Application Layer
Transport
Network
MAC
quality using the received
packet.
 It participates in
communication based on the
recent channel state.
 Receiver-based initiative
concept provides accurate
channel-aware operation.
PHY
Wireless Sensor Networks
Akyildiz/Vuran
55
Recap: XLP Cross-Layer Interactions
Application Layer
 Receiver-based contention & routing
 Potential receivers contend for the
Transport
Network
MAC
PHY

transmitted packets and become nexthop
Routing Algorithm
 Feasible receivers are determined
based on geographical information
of source and sink
 Voids are avoided by angle-based
routing
Wireless Sensor Networks
Akyildiz/Vuran
56
Recap: XLP Cross-Layer Interactions
Application Layer
Transport
 Local congestion control
 Nodes monitor their buffer state
 Highly congested nodes do not
Network

MAC

participate in contention & routing
Relay rate is controlled to prevent
congestion
Source rate is controlled based on local
information
PHY
Wireless Sensor Networks
Akyildiz/Vuran
57
Performance Evaluation
 C++ cross-layer simulator (XLS) developed in BWN lab
 300 nodes, 100x100m2 sensor field
 Sink at (80,80), event at (20,20) with event radius 20m
 Throughput, goodput, energy consumption, number of hops
and latency analysis
 Comparative analysis with 5 layered protocol stacks and 1
cross-layer protocol (ALBA-R)
Wireless Sensor Networks
Akyildiz/Vuran
58
Protocol Stacks
 Flooding: At the MAC layer, a simple CSMA type broadcast


mechanism, at the transport layer, constant packet rate with no rate
control.
[GEO]: geographical routing [14], and CC-MAC [15] at transport,
routing, and MAC layers, respectively.
[PRR]: ESRT [13], PRR-based geographical routing [14], and CC-MAC
[15] at transport, routing, and MAC layers, respectively.
[13] Ö. B. Akan and I. F. Akyildiz, ``Event-to-Sink Reliable Transport in Wireless Sensor Networks,'‘ IEEE/ACM Trans.
on Networking, vol. 13, no. 5, pp. 1003-1016, October 2005.
[14] K. Seada, M. Zuniga, A. Helmy, B. Krishnamachari, ``Energy-efficient forwarding strategies for geographic
routing in lossy wireless sensor networks,'' in Proc. ACM Sensys '04, November 2004.
[15] M. C. Vuran, and I. F. Akyildiz, ``Spatial Correlation-based Collaborative Medium Access Control in Wireless
Sensor Networks,'‘ IEEE/ACM Trans. on Networking, August 2006.
Wireless Sensor Networks
Akyildiz/Vuran
59
Protocol Stacks
 [PRR-SMAC]: ESRT [13], PRR-based geographical routing [14], and



SMAC [16]
[DD-RMST]: RMST [17], directed diffusion [18], and CSMA at transport,
routing, and MAC layers, respectively (only for  = 1).
[ALBA-R(x)]: ALBA-R protocol [19] where x represents the traffic rate l
= {3,4,6.25} pkts/sec
XLP: Our proposed cross layer protocol (XLP)
[16] W. Ye, J. Heidemann, and D. Estrin, ``Medium Access Control with Coordinated Adaptive Sleeping for Wireless
Sensor Networks,'‘ IEEE/ACM Transactions on Networking, vol. 12, no. 3, pp. 493-506, June 2004.
[17] F. Stann and J. Heidemann, ``RMST: Reliable data transport in sensor networks,'' in Proc. IEEE SNPA '03, pp.
102-112, Anchorage, Alaska, April, 2003.
[18] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, F. Silva, ``Directed diffusion for wireless sensor
networking,'‘ IEEE/ACM Transactions on Networking, vol. 11, no. 1, pp. 2 - 16, February 2003.
[19] P. Casari, M. Nati, C. Petrioli, and M.Zorzi, “Efficient Non Planar Routing around Dead Ends in Sparse
Topologies using Random Forwarding,” in Proc. ICC’07, Scotland, UK, June 2007.
Wireless Sensor Networks
Akyildiz/Vuran
60
Performance Evaluation
 Effect of angle-based


routing
Up to 70% decrease in
route failure rate
For >0.2, ABR limits route
failure rate to <10%
Wireless Sensor Networks
Akyildiz/Vuran
61
Performance Evaluation
Avg. Throughput
Wireless Sensor Networks
Akyildiz/Vuran
Goodput
62
Performance Evaluation
Avg. Number of Hops
Wireless Sensor Networks
Akyildiz/Vuran
63
Performance Evaluation
End-to-end Latency
Consumed Energy/Packet
Wireless Sensor Networks
Akyildiz/Vuran
64
Conclusions
 Smaller number of hops does not mean faster
routes/minimum energy consumption
 XLP achieves significantly low energy consumption
with low latency
 Cross-layer protocol design outperforms layered
and existing cross-layer protocols
 Lower complexity in design and implementation
Wireless Sensor Networks
Akyildiz/Vuran
65