Vous êtes sur la page 1sur 19

BY

G.PADMAVATHI
06-530
 Off-the-shelf radio  Multihop capabilities
 Representative of spread-  Scatternet support has been
spectrum radios (frequency announced for years but was
hopping) not supported at the time of
 Free 2.4 GHz band
our study
 How to build multihop
 Promise to be cheap Bluetooth-based networks?
 The Bluetooth stack is  Bluetooth is connection-
complex based
 Stripped down version of  How to define network self-
Bluetooth adapted to assembly based on Bluetooth
constraints of sensor nodes? device discovery. What is the
impact on performance?
 A Bluetooth module  Bluetooth implements Time
embeds front-end radio, Division Multiplexing (TDM)
baseband and MAC layer at the radio level
 Are standard Bluetooth  Can applications leverage
physical layer and MAC layer radio-driven TDM?
adapted to the sensor
network regime?
Pragmatic Approach  Port of TinyOS to BTNodes
 Development of
BTNodes from ETH Zurich TinyBluetooth
 Atmega 128 7.32 MHz  Self-Assembly Procedure
 128 KiB flash  Application using Radio-
 Dual-radio level TDM
 Ericsson’s Bluetooth  UC Berkeley’s TinyDB on top
of TinyBluetooth
module ROK 101 007
 Performance Evaluation
 Intrinsic properties
 Prototype properties
1.Physical bus driver
2.Hcl driver
 Physical bus driver:This layer abstracts
the characteristics of the physical bus
 HCL driver:This layer maps the interface
of the underlaying HCL layer into the
programming model used for
implementing the higher layers
Applications

Profiles

L2Cap

Physical Bus Hardware

HCI

Baseband

RF
TinyOS Application
TinyBluetooth
Physical Bus Hardware

HCI

Baseband

RF
 Asynchronous
Programming Model
 HCI mapped onto tinyOS
events and commands
 UART events decoupled
from HCI events
 Buffer Trading
 Buffers swapped between
modules
 Generic Packet type casted
into specific packet
depending on
event/command
 Interesting information
encapsulated inside
Bluetooth module
 Each node is equipped
with 2 radios
 For each node
 To which node to
connect?
 Connect as master or
slave?
 3 node configurations:
 S-S
M-M
M-S, S-M
M
 Building a connection tree as
a baseline (BlueTree [Petrioli, M
Basagni 2002])
 Each node has a radio set up as S
M M
a master, the other as a slave
 Recursive connection
establishment S M
 First slave radio is turned on.
S
 One node is chosen as the root of S
the connection tree. M
 Master radio turned on once a
connection is established on slave S
radio. M M
 Rely on Bluetooth device
discovery and connection S
establishment
M
S
S

S
M
 Connection tree
supports hierarchical
routing tree.
 Radio drives TDM M
 Bluetooth radio in Sniff
mode: Master and
Slaves agree on
synchronization points S
(ideally once per
epoch). Rest of time M
sensor node sleeps or M
senses. Microcontroller
waken up on radio
signal. M
 Pipelined aggregation S
S
along the routing tree.
 Separated Channels
 No unplanned collisions S
M

 Problem # 1: The sniff


period is not longer than 40 M
secs.

 Problem # 2: When a S
connection is in sniff mode,
the microcontroller sleeps M M
in idle mode (which is less
efficient than the power
save mode according to the M
Atmel specs). S
S

S
73,2 59,73 90,4
50 - Point-to-point
45 20 bytes payload throughput is
40
668 bytes payload (max) high!
Theoretical max
- The performance we
Throughput (kb/sec)

35
achieve is far from the
30
theoretical max
25  UART limit is 45
20 Kib/sec
15
 Junk sent by
Bluetooth module
10
- Slave-to-master and
5 master-to-slave
0 throughput
DM1 DH1 DM3 DH3 DM5 DH5 are similar
Bluetooth Encoding - Throughput degrades
for
• DM and DH are two encoding schemes. Multipoint connections
DM offers a lower error rate. 1 2 3
• 1, 3 and 5 corresponds to the number of
Aggregate 38.1 25.4 19.3
consecutive slots during which slaves and masters
Per Slave 38.1 12.7 6.4
communicate.
 50mW when idle and
Maintaining connections is very expensive 250 mW when
communicating
 Berkeley’s mica
motes: 10 mW when
idle and 160 mW
when communicating
Different sleep modes!
a. Node is turned on.
b. Connection on slave radio
c. 1st Connection on master
radio
d. 2nd Connection on master
radio
e. Master radio is
discoverable
f. Data packets are
transmittted
g. Disconnections on master
radio
h. Disconnection on slave
radio
 Throughput is high
 Best suited for applications that transmit
lots of data
 Energy consumption is high (in
particular connections)
 Life time of applicatoin must be short
(days)
 Short periods of connections
 Suited for Asynchronous In-Network
Processing with radio driven TDM.
Bluetooth-Based Sensor Network well suited for short lived deployments
With unplanned burst of data with high throughput (images, video).
 Intrinsic Bluetooth Properties
 It is feasible to develop a Bluetooth stack for TinyOS
devices
 Encapsulation within Bluetooth module hurts
 Frequency hopping hurts (40 sec period for sniff mode)
 Inquiry, connection establishment is slow
 Better engineering might improve
 Scatternet support
 Cost of connection maintenance
 Throughput
 max
 decrease on point-to-multipoint
 Code available in TinyOS contrib
directory
 More info on our project home page:
http://www.distlab.dk/manatee
 This study is a baseline for:
 Intel motes
 802.15.4 radios
 Tailored radios relying on Bluetooth front-end
(Pico Radio)
QUESTIONS

Vous aimerez peut-être aussi