Title: Driving%20the%20Power%20Out%20of%20Wireless%20Sensor%20Networks
1Driving the Power Out of Wireless Sensor Networks
2Overall Mote Vision
- Integrate computation, communication and sensing
into a single unit - Advanced behavior emerges from collection of
simplistic devices - Exploit peer-to-peer architectures that scale to
thousands of nodes - Allow seamless integration with physical world
- Robust networking achieved through redundancy
- Short range communication reduces power
requirements and conserves bandwidth
3Example uses
- Env. Monitoring, Conservation biology, ...
- Precision agriculture, land conservation, ...
- built environment comfort efficiency ...
- alarms, security, surveillance, treaty
verification ... - Civil Engineering structures response
- condition-based maintenance
- disaster management
- urban terrain mapping monitoring
- Interactive Environments
- context aware computing, non-verbal communication
- handicap assistance
- home/elder care
- asset tracking
CENS.ucla.edu
4Outline
- Hardware generations
- Mote Success Stories
- Power is King Discussion of power saving
techniques - Specialized hardware
- Spec Node
- Specialized protocols
- Low power listening
- RF Wake-up
- Time Synchronized Comm.
- Current Status
5Design Lineage
- COTS dust prototypes (Kris Pister et al.)
- weC Mote (30 produced)
- Rene Mote (850 produced)
- Dot (1000 produced)
- Mica node (current, 5000 produced)
- Spec
?
6Mote Success Stories
729 Palms Military Vehicle Tracking
8Intel Developers Forum
- 800 Nodes
- Live demo
- Displayed networktopology in lights
9Great Duck Island
10Power Saving Techniques
- Improve performance of communication operations
(hardware) - Radio power consumption
- Transmission bit rate
- Efficiency of protocol implementation
- Streamline protocols for power savings (software)
- Flexible, application specific protocols
- Trade latency and bandwidth for decreased power
consumption
Must do both!!!
11Spec Requirements
- Drastically reduce power consumption, cost and
size - Maintain a tight integration between processing,
communication, and sensing that allows
cross-layer optimizations - Allows for rich interfaces to hardware
accelerators and flexible resource pools - Provides
- Efficiency
- Through specialized concurrency mechanisms and
optimal hardware accelerators - Flexibility
- By using software to compose basic protocol
building blocks into application specific
protocols with rich interfaces
12The Spec Architecture
- Single CPU for Base band, OS and Application
- Shared system resources can be divided between
system components dynamically - High bandwidth, flexible interfaces can be
exposed across system components - Allows applications access to fine-grained system
control - Hardware accelerators to support key sensor
network challenges - Communication, synchronization, power management,
concurrency - Shared memory interface model
13General Architecture Diagram
14Spec Primitives
- Integrated Transmitter
- 800-gt1100 MHz transmitter
- FM/AM
- Inductor and Reference crystal are only external
components - Consumes 1 mW _at_ 2.7 V minimum voltage
- .5 mW TX power
- Targeting 300 uA receive mode
- -95 dBm receiver sensitivity
- Same sensitivity as RFM at 1/10 power
- Tunable IF (Low)
- lt100 us turn-on time
- Designed by Al Molnar
15Spec Primitives (cont.)
- Integrated ADC
- Ultra low power 8-bit ADC
- 27 pJ per sample (1000x improvement over industry
standard) - Designed my Mike Scott
- Register Windows
- Allows for fast interrupt support
- With early versions of TinyOS, 50 of CPU energy
consumption went to saving register sets - 28x reduction in interrupt overhead vs. Mica
16Spec Primitives (cont.)
- Encryption Support
- Stream based, LFSR based cipher
- Automatic generation of random pad
- Encryption automatically performed during
transmission and reception - Allows for efficient secure communication and
authentication - Encrypted MAC can serve as both CRC check and
authentication signature - Communication Accelerators
- Hardware Start symbol detection
- Automatic timing extraction and synchronization
- DMA memory engine
17Spec Layout
2.5mm
- IO Pads
- RAM blocks
- MMU logic
- Debug logic
- ADC
- AVR CPU Core
- RF Frequency Synthesizer
- Transmitter
.25 um CMOS
Core Area only 50 full
18Spec
19Communication Interface
- Hardware provides message-level AM interface
- Same functionality only implemented in hardware
- gt 5000 x cost reduction
- Hardware handles
- Message send command with TOSMsgPtr
- Hardware signals
- Message arrival event with TOSMsgPtr
- CPU communication overhead dropped from approx.
2MIPS down to 0.
20Spec Demonstration
21(No Transcript)
22Power savings through adaptive protocols
23Trade latency for power
- Low power receiving
- Reduce receiver power consumption while
introducing minimal additional transmitter
overhead - Total Power Power of Receiver Send Rate
Energy of Transmitter - Sample briefly to detect activity
ms
ms
2429-Palm low-power listening
- Frequent messages expected
- 2-3 per second possible
- 10 duty cycle selected for receiver
- Channel Check in 30 us
- 300 us extra transmission required
- 10x radio power consumption for receiver
- Bonus 50 reduction in CPU usage
25IDF RF Wake-up
- Expected message frequency near zero
- Ultra-low power network sleep state
- Network can be woken from uA sleep state within
seconds - Flexible radio use reduces check time from 100
ms to 50 us over packet based protocols - 334x transmission cost increase for 30 byte
message
ms
ms
26Time Synchronized Communication
- Global time used to schedule communication
- Caps transmitter energy requirements
- Accuracy of timing determines worst-case overhead
- Precise packet timing leads to lt ms
synchronization accuracy - Still trading latency for power
ms
27GDI GSK Power
- GDI uses low-power listening varrient
- Periodic checks every 500 ms for 10 ms -- 5 Duty
Cycle - Message preamble 1200 bytes long prior to
sending actual data - Expected message interval 10 mintues
- Expected 3 month lifetime
- GSK uses time windows for communication
- Periods of active communication followed by sleep
- Time intervals in seconds
- Resynchronization required during active periods
Both can used unmodified routing algorithms.
28Dust Inc.s HDK
- Combines multiple power saving techniques
- Low-power listening used for infrequent
configuration and emergency messages - TDMA based message scheduling used for regular
data collection messages - Central server determines TDMA schedules and
routing topologies
29Putting it all together
- Apply combined optimizations into typical
application - Sample application that
- Check sensor every 4 seconds
- Collects data every 5 minutes
- Aggregates data from child nodes
- Checks for Alarm conditions every 4 seconds
- 2 AA battery power supply
30Sample Application Performance
Samples 21600
Reports 288
Communication Events 1728
Alarm Checks 21600
ADC Conversion 2.4 mJ/Day
Communication 298.1 mJ/Day
Synchronization 103.7 mJ/Day
Alarm Checks 254.0 mJ/Day
Idle cost 518.4 mJ/Day
Total 1176.6 mJ/Day
Pair of AA Batteries 1.17E07 mJ
Expected Lifetime 27.2 Years
x
31Questions?