Adaptive Clustering

for Wireless Mobile Networks

2
Abstract
This paper describes a self-organizing, multihop, mobile radio
network which relies on a code-division access scheme for
multimedia support.

This network architecture has three main advantages:
It provides spatial reuse of the bandwidth.
Bandwidth can be shared or reserved in each cluster.
The cluster algorithm is robust in the face of topological changes caused
by node motion, node failure, and node insertion/removal.

Simulation shows that this architecture provides an efficient, stable
infrastructure for the integration of different types of traffic in a
dynamic radio network.
3
Contents
Introduction
The Multicluster Architecture
Transport Protocols
QoS Routing
System Performance
Conclusions
4
Introduction(1/2)
Packet radio network (PRNET), developed in the 1970s to address
the battlefield and disaster recovery communication requirements
[16], [17].


Fig. 1. Conventional cellular networks (single hop).
5
Introduction(2/2)






We develop an architecture and networking algorithms which support
a rapidly deployable radio communications infrastructure.

The network provides guaranteed quality of service (QoS) to real-
time multimedia traffic among mobile users without requiring a fixed
infrastructure (e.g., no base station).
Fig. 2. A multihop situation occurs when base station B fails.
6
The Multicluster Architecture(1/9)
The Clustering Algorithm
Cluster Maintenance in The Presence of Mobility
Code Assignment
Network Initialization
7
The Multicluster Architecture(2/9)
The Clustering Algorithm
We make the following operational assumptions underlying the
construction of the algorithm in a radio network.

These assumptions are common to most radio data link protocols [3],
[4], [6], [7].

Every node has a unique ID and knows the IDs of its one-hop neighbors.
A message sent by a node is received correctly within a finite time by all
of its one-hop neighbors.
Network topology does not change during the algorithm execution.
8
The Multicluster Architecture(3/9)
The Clustering Algorithm
Fig. 3. Distributed clustering algorithm.
Fig. 4. System topology.
Fig. 5. Clustering
9
The Multicluster Architecture(4/9)
The Clustering Algorithm
We begin studying the impact of transmission range on connectivity.
Fig. 6. Connectivity property. Fig. 7. Average order of repeaters.
10
The Multicluster Architecture(5/9)
The Clustering Algorithm
The existence of at least one path between a pair of nodes is
required.

The number of repeaters will affect the number of paths.
Fig. 8. Number of repeaters.
11
The Multicluster Architecture(6/9)
Cluster Maintenance in The Presence of Mobility
In the dynamic radio network
Nodes can change location
Nodes can be removed
Nodes can be added.

The cluster maintenance scheme was designed to minimize the
number of node transitions from one cluster to another.
Fig. 9. Reclustering.
12
The Multicluster Architecture(7/9)
Cluster Maintenance in The Presence of Mobility
Fig. 10. Stability of the multicluster architecture.
13
The Multicluster Architecture(8/9)
Code Assignment
Each node has a transceiver which can either transmit or receive at
any given time.

In the spread-spectrum code-division system, the receiver should be
set to the same code as the designated transmitter.

There are three options for using the dedicated code within a cluster.
Receiver-based code assignment
Transmitter-based code assignment
To assign a common codes to all transmitterreceiver pairs within a
cluster.

Based on transmitter-based code assignment, when a node is not in
transmitting mode, it randomly selects and listens to one of the codes
used by its neighbors.
14
The Multicluster Architecture(9/9)
Network Initialization

This basic communication facility allows nodes to organize
themselves in clusters following the algorithm just described.

Once a cluster is formed, the cluster leader communicates with the
neighbors (using the control code) to select the codes.

Only when the code assignment is completed (i.e., each cluster has
been assigned its code) can user data be accepted by the nodes and
transmitted in the network.
15
Transport Protocols(1/4)
The aim of our design is to support integrated traffic (i.e., datagram
and real time) efficiently.

Channel Access Scheme
Acknowledgment for Datagram
Mobility
16
Transport Protocols(2/4)
Within each cluster, the MAC layer is implemented using a TDMA
scheme.

Time is divided into slots which are grouped into frames.
Channel Access Scheme
Fig. 11. Channel access frame within a cluster.
17
Transport Protocols(3/4)
Acknowledgment for Datagram
Fig. 12. Implicit acknowledgment scheme.
18
Transport Protocols(4/4)
Node changes its own cluster ID, using the free slot to transmit
packets in the new code.
Mobility
the nodes in a cluster
will recompute the new
TDMA frame format.

In the same way, if a
node is removed from a
cluster, the frame is
reduced.
Fig. 14. Average number of links between an adjacent cluster pair.
19
QoS Routing (1/4)
Multimedia applications such as digital audio and video have much
more stringent QoS requirements than traditional datagram
applications.

For a network to deliver QoS guarantees, it must reserve and control
resources.

Bandwidth in The Cluster Infrastructure
QoS Routing Scheme
20
QoS Routing (2/4)
The key resource for multimedia QoS support is bandwidth.

A node can at most transmit one packet per frame, the bandwidth of
a node is given by
Bandwidth in The Cluster Infrastructure
(

=
time frame
time cycle
bandwith
Fig. 15. Node bandwidth.
21
QoS Routing (3/4)
Bandwidth in The Cluster Infrastructure
Fig. 16. Bandwidth of node C in cluster C1.
22
QoS Routing (4/4)
The goal of the bandwidth routing algorithm is to find the shortest
path.
QoS Routing Scheme
Fig. 17. Standby routing.
Fig. 18. The primary route fails and the
standby route becomes the primary route.
Another standby route is constructed.
23
System Performance(1/6)
The default CYCLE time is 100 ms.
The offered traffic consists of two components:
Real-time sessions
Datagrams

Weighted end-to-end Throughput
Real-time and Datagram Traffic Mix
Standby Routing
Scheme Comparison Synopsis
24
System Performance(2/6)
The link throughput is defined as the sum of the throughputs on the
links which are simultaneously active in the network.

We measure the end-to-end throughput accounting for possible
network disconnection:
Weighted end-to-end Throughput

=
=
DC
i
i
i
i
L
LT
f
1
throughput
i
L
i
LT
i
f
DC total number of disconnected components;
fraction of node pairs in component ( ) ( ) ( ) ( ) 2 / 1 / 2 / 1 = N N n n i
i i
total link throughput of component i ;
average path length in component i ;
25
System Performance(3/6)
Weighted end-to-end Throughput
Fig. 19. End-to-end throughput.
26
System Performance(4/6)
Real-time and Datagram Traffic Mix
TABLE I
SYSTEM TOPOLOGY (N = 20)
Fig. 20. Throughput of mix traffic.
27
System Performance(5/6)
Standby Routing
TABLE II
PERFORMANCE OF STANDBY ROUTING (MAXIMUM SPEED = 2 FT/S)
TABLE III
THE PERFORMANCE OF STANDBY ROUTING (MAXIMUM SPEED = 8 FT/S)
28
System Performance(6/6)
Scheme Comparison Synopsis
TABLE IV
OVERALL PERFORMANCE COMPARISON (2 FT/S)
TABLE V
OVERALL PERFORMANCE COMPARISON (8 FT/S)
29
Conclusions
In order to reduce control overhead and to overcome the limitation of
the number of orthogonal codes, we use only one code within each
cluster.

Packet transmissions by data sources and real-time sources are
interwoven, with top priority given to real-time sources.

The performance of the proposed cluster scheme is similar to that of
cluster TDMA [7], with less implementation complexity.
Thanks for Your Attention !