Optimizing the Topology of Bluetooth Wireless Personal Area Networks

1
Optimizing the Topology of Bluetooth
Wireless Personal Area Networks

• Marco Ajmone Marsan, Carla F. Chiasserini,
• Antonio Nucci , Giuliana Carello , Luigi De
  Giovanni

??????? ??? R91006_at_im.ntu.edu.tw 2003/02/24
2
Outline

• Introduction
• Problem statement
• ILP Formulation
• Numerical result
• Conclusions and future work

3
Introduction

• Universal short-range wireless capability
• Uses 2.4-GHz band (ISM)
• Cable replacement
• Use frequency hopping spread spectrum

4
International Bluetooth Frequency Allocations
From wireless communications and networks
5
Piconet

• The Bluetooth units share a common hopping
  sequence can form a piconet
• A Piconet Composed of one master and up to seven
  active salves

6
Scatternet
7
Problem Statement
8
Problem Statement(cont)

• Leader Election Distributed Topology
  Construction of Bluetooth Personal Area Networks
  (IEEE Infocom 2001)

9
Link establishment in Bluetooth

• Inquiry Procedure Collected neighborhood
  information.
• Paging procedure To establish the connections
  between neighboring devices.
• Asymmetric processes they involve two types of
  nodes (which we call senders and receivers) each
  performing differentactions.

10
Link establishment (cont)
11
A symmetric protocol for link formation

• Forcing the two nodes to alternate independently
  between the sender (INQUIRY state) and receiver
  (INQUIRY SCAN state) roles

12
BTCP Bluetooth Topology Construction Protocol

• Phase I Coordinator Election
• Phase II Role Determination
• Phase III The actual connection establishment

13
Phase I Coordinator Election

• Purpose The coordinator node will know the
  count, identities and clocks of all the nodes
  (FHS packets)
• Coordinator election procedure
• Each node x has a variable called VOTE
• One-to-one confrontation compare their VOTE
  variable
• The loser y sends all the device FHS packets of
  the nodes it has won so far to the winner x.
• The winner x increase its VOTE variable by
  VOTE(y) and continues on the leader election.

14
Phase II Role Determination

• The coordinator check the number of nodes that it
  has discovered during phase I
• If the number lt 8 connects to all of the nodes
  and one piconet is formed.
• If the number ? 8 calculates the number of
  piconets P.

From university of Maryland Technical Report.
15
Phase II Role Determination(cont)

• After calculating P
• the coordinator selects itself and (P-1) nodes to
  be the designated masters
• other nodes to be the scatternet
  bridges (assume two piconet share only one
  bridge)
• The remaining nodes to be the pure slave.

16
Phase II Role Determination(cont)

• After the role assignment
• For each master x , the coordinator has a
  connectivity list set (SLAVELIST(x),
  BRIDGELIST(x)) consisting of the masters
  assigned slaves and bridges.
• Then the coordinator connects to the designated
  masters. Thus a temporary piconet is formed.
  (according to equation p , p is always less than
  8. Thus the temporary piconet can always be
  formed.)

17
Phase III the actual connection establishment

• Each master x pages and connects to the slaves
  and bridges defined in its SLAVELIST(x) and
  BRIDGELIST(x).
• As soon as a node is notified by its master that
  it is a bridge , it waits to be paged by its
  second master. When this happens, the bridge node
  send a CONNECTED notification to both masters.
• When a master receives a CONNECTED notification
  from all its assigned bridges, a fully connected
  scatternet of P piconets is guaranteed to be
  formed and the protocol terminates.

18
Example the connection establishment protocol
19
Problem Statement(cont)

• The major factor influencing the nodes energy
  consumption is the amount of transmitted,
  received, and processed traffic rather than the
  distance between transmitter and receiver.

20
Problem Statement(cont)

• Define a performance metric to be minimized,
  the traffic load of the most congested node, or
  equivalently its energy consumption
• And require that the optimal BT-WPAN topology
  meets the following constraints.
• Full network connectivity.
• System specification.
• System complexity.
• Traffic requirement.
• Constraints on the nodes role.

21
Problem Statement(cont)

• Full network connectivity
• There must be at least one path between any two
  nodes in the network.
• System specification
• The number of nodes participating in a piconet
  cannot be greater than a given value
• The distance between each master-slave pair must
  be less than the maximum piconet radius.
• Also, a node can be master in one piconet only.
• System complexity
• the number of formed piconets is limited to a
  fixed value.
• Traffic requirement
• The network must support the desired
  source-destination connections.
• Constraints on the nodes role
• For instance, nodes that are gateways to the
  fixed network should be chosen as masters.

22
Problem Statement(cont)

• Master as a node that is assigned to p? 1
  piconets and is the master in one of them.
• Slave as a node that is assigned to one piconet
  only
• Bridge as a node that is assigned to p? 2
  overlapping piconets and is a slave in all of
  them.

23
3. ILP Formulation

• A. Notation and Definitions
• The BT-WPAN is represented as (N , Z )
• where
• N the set of network nodes
• Z the matrix containing the values of
• the distances between any two nodes (
  i , j )
• i , j N

24
Notation and Definitions (cont)

• S the set of traffic sources
• D the set of destination nodes
• N the numbers of nodes
• S the numbers of sources
• D the numbers of destinations
• ZMAX the maximum radius of a piconet

25
Notation and Definitions (cont)

• C (s , d ) s S , d D the set
  of source-destination connections
• C C the total number of connections that
  have to be routed through the network.
• Assume just only one route is used for each
  source-destination pair.

26
Notation and Definitions (cont)

• T tij the traffic matrix that indicating
  the information rate on each source-destination
  connection , normalized to the network capacity.
• ?s the total average traffic rate for each
  traffic source s , s S

27
Notation and Definitions (cont)
For each node i , i N , defined three
binary variables µ i ß i s i
1 , if node i is a master 0 , otherwise
1 , if node i is a bridge 0 , otherwise
1 , if node i is a slave 0 , otherwise
28
Notation and Definitions (cont)
MMAX the maximum number of piconets XMAX the
maximum number of active nodes that can be
assigned to a piconet . M the set of nodes
that are forced to be masters , M ? MMAX V
the set of nodes that are forced to be slaves
, V ? N - MMAX
29
Notation and Definitions (cont)
For each pair of nodes ( i , j ) i , j N ,
define the set of assignment variables X xij
, as follows The set of flow variables , F
fij, as
30
Notation and Definitions (cont)
For each source-destination pair (s , d ) C,
and for each pair of nodes ( i , j ) i , j
N , define the following routing variables
so that the set R defines the
connection path through the network for any
connection in C.
31
B. Optimization Problem
Given the set of routing variables R , the
traffic load of the generic node i , i N , is
defined as the sum of the Incoming and outgoing
traffic that i has to handle. For each
connection (s , d ) C , we have that the load
of node i as follows
32
B. Optimization Problem(cont)
By adding over all the connections in C
, we obtain the load of node i as
We define as the network bottleneck the node that
experiences the highest traffic load, i.e., whose
traffic load, B , is
33
B. Optimization Problem(cont)
Our objective is to select the network topology
so as to obtain the optimal BT-WPAN topology,
which minimizes the traffic load of the most
congested node. This is defined to be the
optimization problem identified as
34
B. Optimization Problem(cont)
Constraints on assignment variables xij are as
follows.
35
B. Optimization Problem(cont)
Constraints on assignment variables xij are as
follows. (cont)
36
B. Optimization Problem(cont)
Constraints on assignment variables xij are as
follows. (cont)
37
B. Optimization Problem(cont)
Next, in order to meet requirement 1 in Section
II, we consider a graph connecting all the
masters in the network.
38
B. Optimization Problem(cont)
Finally, given the source-destination connections
that the network has to support, the constraints
on the routing variables are as follows
39
B. Optimization Problem(cont)
The constraints on the routing variables are as
follows (cont)
40
D. Example
41
D. Example(cont)
42
D. Example(cont)
43
(No Transcript)
44
D. Example(cont)
45
4. Numerical Results
46
Numerical Results(cont)
47
5.Conclusions and Future work

• Determining an optimal topology for Bluetooth
  Wireless Personal Area Networks.
• Results show that optimized topologies can be
  quite robust to changes in the traffic pattern.
• The main limitation of the approach lies in the
  centralized nature of the optimization algorithm.
  
• Future work the identification of heuristic
  approaches for the construction of good
  topologies in a distributed fashion.

48
The End