Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
An energy-‐aware wireless sensor node for environmental monitoring
Christel Chappuis and Géraldine Mercier
Supervisor : Jorge Soares
1
Plan
• IntroducBon • CalibraBon • Energy-‐saving strategies • Data collecBon • Results • Conclusion • Improvements/Ideas…
2
IntroducBon
• Wireless Sensor Networks • Main issue : managing the power resource • Project aim : save energy AND realisBc data sets • 2 nodes : reference + energy-‐efficient strategy
3
CalibraBon
• Erase the systemaBc error between the nodes • Wide coverage in terms of possible values
• Simultaneously measurements with both nodes at the same rate and in the same environmental condiBons
• Matlab funcBon polyfit()
4
Light
5
Temperature
6
Experimental set-‐up
Light
Experimental set-‐up
Temperature
7
The MICAz
• MICAz sensors implemented in nesC language • TinyOS operaBng system
à Events / tasks triggered by Bmers
• FuncBon get_next_/mer() controls the frequency of the measurements
• Reference node : default Bme interval = 0.5 second
8
Energy-‐efficient strategy implementaBon
9
• Default Bme interval = 1 second • AdapBve behavior: the expected value
Δ(values) < threshold à dt increased Δ(values) not negligible à dt decreased proporBonally to its value Δ(values) very high à dt set to dt_min (0.5 sec)
Light VS temperature
• Time interval depends on light à faster changes than temperature
• BUT! if light condiBons are constant and temperature changes à Bme interval = 2sec
10
Signal Processing
• CalibraBon • ReconstrucBon of the signal by linear interpolaBon
• Conversion of temperature units to °C • Plot the results
11
Data collecBons
• 1-‐hour experiment, at sunset • ArBficial variaBons: to test the strategies implemented
12
• 2 more experiment of 10 minutes each à in constant condiBons à with rapid and wide changes
Experiment 1: Light measurements
13
Event 1
Event 4
Experiment 2 : constant condiBons
14
15
Experiment 3 : variable condiBons
Performance Ma: reconstrucBon of signal accuracy Me: percentage of energy saved compared to the reference node Mo: overall performance, Mo = Ma ×Me
16
Experiment number Ma_t Me_t Mo_t Ma_l Me_l Mo_l Mo_o
1 0.9985 0.9329 0.9315 0.9959 0.9329 0.9296 0.8659
2 0.9990 0.9749 0.9739 1 0.9749 0.9749 0.9495
3 0.9990 0.5900 0.5894 0.9977 0.5900 0.5886 0.3469
Conclusion
• Successful strategies à the energy consumpBon was considerably reduced compared to the reference node AND the accuracy was sBll very good!
• The performances depend on the surroundings condiBons !
17
Improvements, ideas • Suppress messages from sending if similar to the last ones • Test, not conclusive yet…
18
Thank you for your amenBon
QuesBons ?
19