Probabilistic Routing Schemes for Ad-Hoc Opportunistic Networks

parsimoniousknotNetworking and Communications

Feb 16, 2014 (3 years and 5 months ago)

61 views

© The authors



1

Chapter 8:

Probabilistic Routing Schemes

for Ad
-
Hoc Opportunistic Networks

1
Vangelis Angelakis,
2
Elias Tragos

,


3
George Perantinos, and
1
Di Yuan


1
Link
öping
University, Sweden

2
Foundation for Research and Technology

Hellas

3

Forthnet S.A.

Routing in Opportunistic Networks


© The authors



2

Wireless proliferation



Wireless RF Proliferation in the past decades


Bluetooth, 802.11a,b/g, 3/4G




Computing paradigms based on Wireless


Wireless Cloud


Internet of Things


Machine
-
to
-
Machine (ad
-
hoc) communication




Wireless medium backlashes


Range issues


Interference / Communication reliability



© The authors



3

Relaying and forwarding


Transmission range limitations
-
> need for relays



Key decisions in forwarding packets:

1.
What

to send (
my packet or a relayed packet
?)

2.
To

whom

(
to a relay or the destination
?)

3.
When

to do so (
will I suffer collisions, cause interference
?)



Routing deals with 1,2


Scheduling takes care of 3 once 1 and 2 have been decided



Relaying typically assumes:


Some topology knowledge


Collaborating nodes (limited/no selfishness)


Routing needs to work towards these assumptions


© The authors



4

Routing in Opportunistic Networks


The role of
mobility

1. Buffering
taking advantage of transitive transmission

2. Delay
\
Disruption
-
Tolerant Networking



Problems arising from opportunistic communication:

1.
Topology is becoming too variable

2.
Selfishness can arise to conserve resources



Opportunistic Networks’ routing needs to cope with
these two








© The authors



5

Probabilistic Routing


Work
-
around:
Probabilistic routing


Model and take into account the environment (too
complex), or


Randomize on


Whom to send to and


When to send




Cross
-
layer routing approach, taking input from:


Physical layer


Access layer



Trade
-
off: performance / simplicity
-
effectivness






© The authors



6

Probabilistic Routing


Work
-
around:
Probabilistic routing


Model and take into account the environment (too
complex), or


Randomize on


Whom to send to and


When to send




Cross
-
layer routing approach, taking input from:


Physical layer


Access layer



Trade
-
off: performance / simplicity
-
effectivness






© The authors



7

Schemes Overview


1.
Epidemic routing (
Vahdat

& Becker, 2000)


2.
PROPHET (Lindgren, et al. 2003)


3.
MAXPROP (Burgess, et al. 2006)


4.
Parametric Probabilistic Routing (
Barret
, et al.
2005)


5.
PROPICMAN (Nguyen, et al. 2007)



© The authors



8

Epidemic Routing 1/2


Bio
-
inspired: packets are considered to infect nodes








(
Vahdat

& Becker, 2000)



Assumes


Nodes are randomly mobile & have ordered identifiers


Resources sufficiency (battery / buffers)



Forwarding Decision: fixed


flooding



Buffers: FIFO



Buffer (hashed) “index”:
Summary Vector
(
SV
)



Reliability:
ack’s



© The authors



9

Epidemic Routing 2/2


Meeting a newly identified neighbor node


Exchange SVs


Exchange unknown messages

For protocol sake the process is initiated by the node with the
smaller identifier









Per
-
host queuing



New messages given preference over old ones in
terms of buffer availability


1



B



A

3

2

Request: (SV
A
+SV
B
’)

SV
A

Messages unknown to B

© The authors



10

PRoPHET (1/2)


PRoPHET
:

P
robabilistic
Ro
uting
P
rotocol




using
H
istory of
E
ncounters and
T
ransitivity




(
Lindgren, et al. 2003
)



Users move in a “not so random”,
predictable
fashion



Forwarding decision:

by
Delivery Predictability

P
(
M
,
D
)

set up at every node
M

for each known destination
D
.




Epidemic Routing SV’s are used here too to exchange


Delivery Predictability values to updated own P(M,D) as follows:



© The authors



11

PRoPHET (2/2)


When the node
M

encounters another node
N
, the
predictability for

N


increases
as:

P
(
M
,
N
)
new

=
P
(
M
,
N
)
old

+ (1
-

P
(
M
,
N
)
old
) x
L
enc
,

L
enc

is an initialization
constant


The
predictabilities for all destinations

D

other
than

N

suffer
ageing
:

P
(
M
,
D
)
new

=
P
(
M
,
D
)
old

x
γ
K
,

γ

is an aging constant

K

is a time
factor


Transitive
property
updates
the
predictability
of
destination

D

for which

N

has a

P
(
N
,
D
)
value:

P
(
M
,
D
)
new

=
P
(
M
,
D
)
old

+ (1
-

P
(
M
,
D
)
old
) x
P
(
M
,
E
) x
P
(
E
,
D
) x
β

β

is a

scaling factor


The assumption here is that

M

is likely to meet

N

again.

© The authors



12

M
AX
P
ROP

(1/2)


Motivated by pedestrian mobility and city vehicles
(busses)








(Burgess, et al. 2006)


Addressed resources issues considering vehicles


Bulky equipment


energy


Maintains ordered destination based queues


Addresses on top of
PRoPHET


QoS


Stale data


Assumes


Unlimited buffer for own messages per node


Fixed size buffer for relaying messages


No topology knowledge/control




© The authors



13

M
AX
P
ROP

(2/2)


Communication steps (flooding
-
based!):


1. Neighbor Discovery

(no knowledge of when the next opportunity to communicate will be)


2. Data Transfer

a)
Transfer packets destined for neighbor peer,

b)
Transfer routing information,

c)
Acknowledge any delivered data,

d)
prioritize “young” relayed packets,

e)
Send un
-
transmitted packets by
estimated delivery likelihood
,

f)
ensure only new packets are sent.


3. Storage Management

(expunge packets to accommodate the relay buffers)

© The authors



14

PARAMETRIC
PROBABILISTIC
ROUTING
(1/2
)


Developed for Sensor Networks

(
Barret
, et al. 2005)


Based on controlled flooding:


Packet forwarding decision by probability function


Probability function is based
on:


distance to destination,


distance from original source to destination,


number of copies already received, …



Variations:

1. The Destination Attractor


Source
-
Destination
distance and
Current Relay
-
Destination distance

2. Directed transmission


uses also the number of hops packet has already traveled.


© The authors



15

PARAMETRIC
PROBABILISTIC
ROUTING
(2/2
)


Estimating distances to Destination:


Each sensor includes its current estimate of distance to D


receiving such information, each sensor updates its distance
information


A sensor chooses as S
-
D distance the minimum of the currently
received information from neighbors.



Potentially this leads to misinformation



Exponential scheme relaxes the problem, but enables
wider flooding

© The authors



16


Fully context
-
aware routing protocol

(Nguyen, et al. 2007
)


Node Profile: nodes exchanging data must have some
information
about each other.


Selection of best forwarders:


delivery probability based on the profile of the neighbors


For every neighbor a sender calculates 2
-
hop route delivery probability


Forwards only if own delivery probability is less than a potential relay



Security considerations


Assumptions for “community level” security (e.g.
authentication,
signatures)


Messages’ content is secure although the “evidences” of the node profile
can be recovered.






P
ROPICMAN


© The authors



17


Simulation framework for lower layer parameters
inverstigation

(
Gazoni
, et al. 2010)


Forwarding decision:


Probability function based on modular metric


Distance


ETX


Linear or piece wise


selection of shape and slope affects on the number of “certain forwarders”


can be varied upon execution to adapt to losses



Time to send


Back
-
off based scheme implemented (with variable or fixed window size)


Highly probable forwarders get to transmit early.



Passive acknowledgements via overhearing

A F
RAMEWORK

FOR

P
ROBABILISTIC

R
OUTING

© The authors



18

References


A. Vahdat and D. Becker. Epidemic Routing for Partially
-
connected Ad Hoc
Networks. Technical Report: CS
-
200006, Duke University, April 2000.


A. Lindgren, A. Doria, and O. Schelén. Probabilistic Routing in Intermittently
Connected Networks. In proc. of the 2003 ACM MobiHoc.


J. Burgess, et al. MaxProp: Routing for vehicle
-
based disruption
-
tolerant
networks. In proc. of 2006 IEEE INFOCOM.


C. L. Barrett et al. Parametric Probabilistic Routing in Sensor Networks,
Mobile Networks and Applications 10:4, pp 529
-
544, 2005.



H. A. Nguyen, et al. Probabilistic Routing Protocol for Intermittently
Connected Mobile Ad Hoc Networks (PROPICMAN). In proc. of the 2007
IEEE WoWMoM.



Niki Gazoni, et al. A framework for opportunistic routing in multi
-
hop wireless
networks. In proc. of the 2010 ACM PE
-
WASUN.




© The authors



19

Thanks for your
attention!