MuleSoft Online Meetup Group - B2B Crash Course: Release SparkNotes
Presentation03 27 03
1. MEMS ‘Smart Dust Motes’ for
Designing, Monitoring & Enabling
Efficient Lighting
MICRO Project Industry Sponsor
General Electric Company; Global
Research Center
2. Professor Alice Agogino, Faculty Advisor
Jessica Granderson, Ph.D. Student
Johnnie Kim, B.S. Student
Yao-Jung Wen, Ph.D. Student
Rebekah Yozell-Epstein, M.S. Student
3. Commercial Lighting
• Electrical Consumption and Savings
Potential
• Advanced Commercial Control
Technologies
- Up to 45% energy savings possible with
occupant and light sensors
- Limited adoption in commercial
building sector
4. Commercial Lighting
• Problems With Advanced Control
Technologies
– Uncertainty is not considered --> sensor
signals, estimation, target maintenance
– Time is not considered, lost savings
through demand reduction
– All occupants are treated the same
– Wires, retro-fit and commissioning
5. Intelligent Decision-Making
with Motes
• An intelligent decision algorithm allows:
validation of sensor signals
uncertainty in illuminance estimation
differences in preference and perception
peak load reduction/demand response
• Smart dust motes potentially offer:
wireless sensing at the work surface, increased
sensing density, simpler retro-fitting and
commissioning, wireless actuation, and an
increased number of control points
6. BEST Lab Energy Research
• Characterization, validation, and fusion of
mote signals
• Modeling the decision space for automatic
dimming in large commercial office spaces
(cubicles)
• Benchmarking a specific decision space for
switching and occupancy patterns,
proposed smart lighting design
• Determination of occupant preferences and
perceptions for a specific decision space
7. Modeling the Decision Space
• Goal is a model that can balance
occupant preferences and perceptions
with real-time electricity prices in
daylighting decisions
• Hierarchical problem breakdown
– Local validation of sensor signals
– Regional fusion of sensed data, actuation
– Global optimization of regional decisions
9. Immediate Work
• Regional Decision-Making
– Balance occupant preferences
– Empirical occupant testing without
windows to control for the effects of
natural light
– Incorporation of electricity prices for
demand-responsive load shedding
11. Future Work
• Daylighting decisions
– Glare, blinds
– Natural/artificial light contributions
– Contrast
• Design of a global value function
– Optimal combination of regional
decisions
12. Features of Sensor Validation
and Fusion for Sensor Networks
• Purpose
– Provide reliable information of current
environment for decision-making
– Feed appropriate value back to the control
system
• Main Idea
– Fuse sensor of the same kind into one or
more reliable virtual sensor
– Fuse disparate sensors
13. Research Goals
• Characterize mote sensors
• Find and construct the most suitable
sensor validation and fusion algorithm
for sensor networks
• Build algorithm for sensor locating
based on the result of sensor validation
and fusion.
14. Purpose of Sensor Validation
• Noise rejection
• Fault detection
– Sensor failure
– Process failure
– System failure
• Ultimate purpose
To provide the most reliable data for fusing
15. Methodology for Sensor
Validation
1. Signal check Sensed data
2. Absolute limits Signal output check
check Sensor
Absolute limits check
3. System feature
performance limits Previous Performance limits check
value
check
Expect behavior check
4. Expected behavior
check Correlation check
5. Empirical Fusion procedure
correlation check
16. Possible Methodology for
Sensor Fusion
• Fuzzy Approach
• Kalman filter
• Bayesian network
• Neural network
17. Sensor Fusion and Validation
Sensor readings
Supervisory controller
Diagnosis Calculate fused value using old
predicted value for validation
gate and incoming readings
Sensor Fusion
Fused value
Calculate new predicted
Sensor Validation value using fused value
Controller
Machine Level
Controller
Decision-making system
Sensor Readings
Algorithm for sensor
Architecture for Sensor Validation
validation and fusion
and Sensor Fusion
18. The Mote
Processor and Radio Platform
• Atmega 128L processor (4MHz)
• 916MHz transceiver
• 100 feet maximum radio range
• 40Kbits/sec data rate
21. The Mote
Other Accessories
• Basic Sensorboard
This board has two
sensors:
temperature
photo
and is capable of
integrating other kinds of
sensors on it.
• Interface Board
Programming each mote
platform via parallel port.
Aggregation of sensor
network data onto a PC via
serial port.
22. Example I
Analyzing of Old Cory Hall Data
Mote node_id 6174
Mote Location and
Environment
23. Example I
Analyzing of Old Cory Hall Data
Mote node_id 6174
Mote Location and
Environment
24. Example I (contd.)
Analyzing of Old Cory Hall Data
Mote node_id 6174
Light Readings and
Temperature readings
5/24/01~5/31/01
25. Example I (contd.)
Analyzing of Old Cory Hall Data
Mote node_id 6174
Possible
failure of
light sensor
Possible failure of
both light and
temperature sensor
Light Readings and
Temperature readings
5/24/01~5/31/01
26. Example II
Analyzing of Old Cory Hall Data
Mote node_id 6190 & 6191 in Room 490
Sensor Readings in
Cory Hall 490
5/17/01~5/22/01
27. Example II (Contd.)
Analyzing of Old Cory Hall Data
Mote node_id 6190 & 6191 in Room 490
Fusion of Light
Reading of 5/17
Using Dr. Goebel’s
FUSVAF Algorithm
28. Potential Difficulties:
Validation and Fusion
• There is not a specific sensor on the
sensor board for sensing occupancy
• Error of mapping sensor signals to
physical readings due to the non-linearity
and sensitivity of each sensor element
• The sampled data for the same time
stamps might be received at different
time due to wireless communication
• Only one sensor per board functions at
any given time
29. Plans for the Next
Two Months
• Setup the software and hardware to
actuate the smart motes on hand
• Characterize the motes signals
• Collect data of target office space using
one or several motes
• Characterize motes failure patterns for
individual motes
• Build algorithms for feature
identification and extraction
• Search for the accurate and efficient way
to sense occupancy
30. Plans for the Next
Six Months
• Build up mote sensor networks in
the target office space
• Benchmark test the networks
• Characterize motes failure patterns
for mote networks
• Evaluate appropriate validation and
fusion algorithms
• Determine best locations for motes
31. Plans for the Future
• Implement the mote validation and
fusion algorithm to real time
validating and fusing
• Refine the mote validation and
fusion algorithm
• Evaluate the possibility of using
motes to actuate dimming ballast
directly
32. Benchmarking Research
Goals
• Verify the need for a smart lighting
system based on human interactions
with their environment
• Develop design guidelines for a smart
lighting system
• Propose a smart lighting system for the
BEST Lab, (6102 Etch.)
33. Benchmarking Research
Deliverables
• Benchmark the current switching and
occupancy patterns in the BEST Lab
• Discuss potential energy savings based on
the results of this benchmarking
• Perform a usability study to determine
user preferences with respect to smart
lighting
• Propose a system that will personalize
lighting based on occupancy and save on
electricity costs
34. Occupancy in Work Area
Average Total Occupancy vs. Time of Day
4
3.5
3
Average occupancy (people)
Wednesday
2.5
Thursday
Friday
2
Saturday
1.5 Sunday
Monday
1
Tuesday
0.5
0
-1 4 9 14 19 24
-0.5
Time of day (military time)
35. Occupancy in
Conference Area
Average Conference Area Occupancy
3.5
3
2.5
Average Occupancy
Wednesday
2
Thursday
Friday
Saturday
1.5
Sunday
Monday
1
Tuesday
0.5
0
0 5 10 15 20 25
-0.5
Time of Day (military time)
36. Switching Patterns
in BEST Lab
Switching Patterns
120.0
100.0
Probability That Light Will Be On
80.0
Monday
Tuesday
60.0
Wednesday
Thursday
Friday
40.0
Saturday
Sunday
20.0
0.0
0 5 10 15 20 25
-20.0
Time of Day (military time)
37. Potential Energy Savings
• Calculate current energy usage in lab
• Calculate energy usage for lights only
being used when and where they are
needed
• Compare current and potential costs
38. Usability Issues
• What level of manual control and
override will users need to feel
comfortable with the system?
• How will users enter personal lighting
preferences into the system and when
(initially or once a problem is detected)?
39. Occupant Preferences and
Perceptions
• Goal: Determine the illuminance
ranges over which occupants perceive
the lighting at their desk to be
– too bright,
– too dark,
– or just right