Wearable Communication Part 2
Jan Beutel, Thomas von Büren, Holger Junker, Matthias Stäger
Computer Engineering and Networks Lab - Electronics LabMay 3, 2002
Computer EngineeringComputer Engineeringand Networks Laboratoryand Networks Laboratory
2
ETH
Zuri
ch
May 3, 2002May 3, 2002
The Wearable Perspective
displaycontext sensor array: camera, light, microphone, GPS
distributed reconfigurable computer
body area network:wireless
communication:WLAN, GSM,
3
ETH
Zuri
ch
May 3, 2002May 3, 2002
Wireless Access Systems
Why Wireless: MobilityTarget: IP Connectivity, Compatibility
• Multi Rate/Multi System Mobile Wireless Access– UMTS– GPRS– IEEE 802.11a and IEEE 802.11b– Bluetooth
• Frontends are self-contained subsystems that are operated by commands from a host system
• Status of our work: Each system operational on IpaQ 3870, joint operation will be available in June
4
ETH
Zuri
ch
May 3, 2002May 3, 2002
Overlay Access Networking
Area Scenario
Range System Bandwidth Resources
Global 1000 km UMTS Satellite + + +
WAN 20 km GPRS/EDGE/… 473.6 kbit/sec + +
Campus 500 m IEEE 802.11 1 Mbit/sec + +
Building 100 m IEEE 802.b 11 Mbit/sec + + +
Floor 12 m IEEE 802.a 54 Mbit/sec + + + +
Room 10 m Bluetooth 768 kbit/sec + / -
PAN 1 m
Sensor 10 cm RFID/Passive bits/sec - - -
5
ETH
Zuri
ch
May 3, 2002May 3, 2002
Low Duty Cycle Concurrent Operation
Host System running Linux OS Communication Frontend B
Communication Frontend A
BAN Communication Frontend CCost/Perf TableCommand Mapping
All Interfaces disabled
Application requesting Data Transfer at Host System
Evaluation of optimal cost/performance
Init of selected Interface; Channel setup
Data Transfer
Shutdown of selected Interface
Update of Cost/Performance Table
6
ETH
Zuri
ch
May 3, 2002May 3, 2002
Benefits
• Operation in best performance mode• Scalable from 2.4 kbit/sec to n54 Mbit/sec• Quasi seamless handover• Reduction of the ON/TX duty cycles
– Best energy performance for non power optimized systems– Operation adaptive to the applications– Optimal usage of the available TX channel– Accounting for burstiness of data transfers
• Use of common protocols and applications
Example: TX of 1 Mbyte data with ACK from 10 kbit/sec data source
1. Constant transmitting of data1 sec init 71 mW + 2 sec setup 162 mW + 800 sec 151 mW = 121195 mWsec
2. 768 kbit/sec Bursts10 1 sec init 71 mW + 10 2 sec setup 162 mW + 10 1 sec 151 mW = 5460 mWsec
7
ETH
Zuri
ch
May 3, 2002May 3, 2002
Local Ad-Hoc Networking
Why Wireless: Easy Configuration, ScalabilityTarget: Better Utilization of Resources, Flexible Usage
• Very short range (sub meter)
• Medium range (meter)
Status of our work: • Prototype on Bluetooth will be used to test
setups/configurations but will not be optimal in terms of energy consumption, size, setup latency.
• First Motion Sensor Network based on I2C and serial communication in test phase.
Todo: • Research in low power, very short range data
communication
8
ETH
Zuri
ch
May 3, 2002May 3, 2002
Topology Issues
• Star topology does not work– Too many frontends– Asymmetric communication demands– Not everything will need central coordination
• Fully distributed does not work– Coordination too difficult
• Idea: Hierarchical Subsystemswith sub-controllers– Linking Units
geographically and logically
– Outsourcing processing tasks
Host System
SubSystem
SubSystem
SubSystem
Leg Motion Sensors Head Up Display
Audio System
9
ETH
Zuri
ch
May 3, 2002May 3, 2002
Demo Application 1: Audio Recording
• Wired transmission from microphone to codec• Audio preprocessing with codec logic
• Reduction to 10 kbit/sec• Buffering to allow burst transfers
• Wireless transmit with 10 packets of 64 kbits each every 65 seconds
• Temporary storage of audio data in cache file system on mobile device
• Move to fileserver storage if not accessed on mobile device
wireless link
reconfigurable
logic, inside watch
wired
microphone, attached to watch
main computing
unit, in jacket
10
ETH
Zuri
ch
May 3, 2002May 3, 2002
Demo Application 2: Motion Sensor (hj)
• Use multiple motion sensors for context awareness
• Idea: Many sensors reveal „more context“
Architecture required to combine those sensors.
Hierarchical approach makes senseI. Information content of sensors is weighted
differently
II. Reduce overall data load (minimize bandwidth requirements)
III. ...
• Map hierarchical topology to human body
11
ETH
Zuri
ch
May 3, 2002May 3, 2002
Current Development: BTnode rev2
• 8-Bit RISC CPU, 1 MIPS/MHz @ 3.6 (or 5.2) MHz
• 128 k Flash, 68 k SRAM• IO, Timer, Analog, UART, I2C
• Idle @3.6 MHz, 3.3V 6 mW
• Active @3.6 MHz, 3.3V 15 mW
• Will be used as subsystem controller
Status:• Testing of the design now• SW Kit/Drivers and Volume
production in June
61 mm
40
mm
12
ETH
Zuri
ch
May 3, 2002May 3, 2002
Parameters Influencing Communication
• Please identify the parameters important for you component/application until mid May:
– Radio Range– Frequency– Modulation– Cell Capacity– Raw Bandwidth– Data Types– Power Consumption– Duty Cycle– Setup Times– Latency– Burst Behavior– Quality of Service Guarantees