Post on 21-Dec-2015
1
A Novel Mechanism for Flooding Based Route Discovery in Ad hoc Networks
Jian Li and Prasant Mohapatra
Networks Lab, UC Davis
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 2 / 22
Agenda
Introduction
Our Proposal: PANDA
PANDA Algorithms
Performance Evaluation
Conclusion
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 3 / 22
What is a Mobile Ad hoc Network (MANET)? A set of wireless devices that can move around freely Form a dynamic topology via ad hoc wireless links No fixed infrastructure, no central administration Limited bandwidth, battery, etc
MANET applications Military tactical communication Disaster relief Conferencing ...
Intro (1): MANET
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 4 / 22
Intro (2): Routing in MANET
Why routing is different in MANETs? Unpredictable moving pattern
Dynamic link availability, limited bandwidth
Fast changing topology
Other concerns: battery source, security issues, etc
Proposed routing protocols Proactive v.s. Reactive (On-Demand)
Topology-based v.s. Position-based
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 5 / 22
Intro (3): Flooding based Routing Flooding technique is often used by on demand
protocols, such as Dynamic Source Routing(DSR) and Ad hoc On-demand Distance Vector(AODV) routing
S
D Route Request
Route Reply
Broadcast Storm Problem: for example, node C may receive broadcast messages from nodes A and B almost simultaneously, which results in a collision.
CA
B
Proposed Solution: each node applies a random delay before rebroadcasting a message or responding to a broadcast message.
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 6 / 22
Intro(4): Random Rebroadcast Delay
Random Rebroadcast Delay (RRD) approach can solve the problem of “broadcast storm” effectively
But, the “Randomness” also introduces a new problem: “Next-hop Racing” behavior
S
I
J
K
M
L
D
At nodes L and M, node I “wins” over node J,
even if J is better than I in term of link lifetime
Suppose:Node I moves much faster than Node J,so the link S-I will be broken sooner than the link S-J.
It happens that:Node I rebroadcast earlier than Node J.
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 7 / 22
Agenda
Introduction
Our Proposal: PANDA
PANDA Algorithms
Performance Evaluation
Conclusion
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 8 / 22
Our Proposal: PANDA (1)
PANDA Positional Attributes based Next-hop Determination Approach
Basic idea of PANDA Classify neighboring nodes into different classes, each of which
uses a different delay range such that better candidates go first
“Good” or “Bad” candidates Utilize geographical location, velocity, energy, etc., to determine
the rebroadcast delay
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 9 / 22
Our Proposal: PANDA (2)
Greedy approach selecting better link at each hop hopefully leads to better end-to-
end routes
Fully distributed an intermediate node makes local decision without any
communications with its neighboring nodes
Versatile capabilities e.g., search for a route with smallest number of hops, or with
minimal transmission power consumption, etc
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 10 / 22
Agenda
Introduction
Our Proposal: PANDA
PANDA Algorithms
Performance Evaluation
Conclusion
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 11 / 22
PANDA Algorithms
Different Variants PANDA-LO (Location Only)
PANDA-LV (Location & Velocity)
PANDA-TP (Transmission Power)
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 12 / 22
PANDA-LO (Location Only)
Determine rebroadcast delay according to link distance Attempt to make a big jump at each hop
At node A
If |SA| > L1
delay = t1 + uniform(0, t1)
else if |SA| > L2
delay = 2*t1 + uniform(0, t1)
else if |SA| > L3
delay = 3*t1 + uniform(0, t1)
else
delay = 4*t1 + uniform(0, t1)
Note: L1 > L2 > L3 and T1 > T2 > T3.
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 13 / 22
PANDA-LV (Location & Velocity)
Determine rebroadcast delay according to link distance and lifetime Attempt to select more stable link with a BIG jump
At node A
If |SA| > L1 && Lifetime > T1
delay = t1 + uniform(0, t1)
else if |SA| > L2 && Lifetime > T2
delay = 2*t1 + uniform(0, t1)
else if |SA| > L3 && Lifetime > T3
delay = 3*t1 + uniform(0, t1)
else
delay = 4*t1 + uniform(0, t1)
Note: L1 > L2 > L3 and T1 > T2 > T3. This figure shows how to estimate the link lifetime.
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 14 / 22
PANDA-TP (Transmission Power)
Motivation: multiple small hops can save transmission power over a big single hop
Assuming the path loss is a simple function of the transmission distance:
The path energy ratio:
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 15 / 22
Similar to PANDA-LO, only consider link distance Attempt to make a big jump at each hop
PANDA-TP (2)
At node A
If |SA| < L3
delay = t1 + uniform(0, t1)
else if |SA| < L2
delay = 2*t1 + uniform(0, t1)
else if |SA| < L1
delay = 3*t1 + uniform(0, t1)
else
delay = 4*t1 + uniform(0, t1)
Note: L1 > L2 > L3 and T1 > T2 > T3.
smaller
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 16 / 22
Agenda
Introduction
Our Proposal: PANDA
PANDA Algorithms
Performance Evaluation
Conclusion
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 17 / 22
Simulation Setup
NS-2 simulator
simulation area 1500m x 300m
250m transmission range
100 nodes
30 connections
speed (0, 20) m/sec
“random waypoint” mobility model pause time: 0, 30, 60, 150, 300, and 500 sec
simulation time 500 sec
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 18 / 22
Simulation Results (1)
Path optimality ratio = length of actual path / optimal path
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 19 / 22
Simulation Results (2)
End to end delay
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 20 / 22
Simulation Results (3)
Energy conserving route discovery
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 21 / 22
Agenda
Introduction
Our Proposal: PANDA
PANDA Algorithms
Performance Evaluation
Conclusion
Networks Lab @ UC Davis Jian Li, Prasant Mohapatra 22 / 22
Conclusion
PANDA approach outperforms RRD approach
Both PANDA-LO and PANDA-LV can improve path optimality
PANDA-LV can improve end-to-end delay
PANDA-TP can discover routes with much less power consumption than RRD approach