Foundations and Light Compass Case Study

1
Foundations and Light Compass Case Study
Design Techniques for Sensor Appliances

• Jennifer L. Wong, Seapahn Megerian,
• Miodrag Potkonjak
• University of California, Los Angeles
• DAC June 3, 2003

2
Organization

• Sensor Network Applications
• Sensor Appliances
• Quantitative Sensor-centric Design
• Light Compass
• Models and Abstractions
• Problem Formulation
• Sensor Fusion
• Design
• Experimental Results
• Future Work
• Conclusions

3
Embedded Networked Sensing

• Micro-sensors, on-board processing, and wireless
  interfaces all feasible at very small scale
• can monitor phenomena up close
• Embedded Networked Sensing will reveal previously
  unobservable phenomena
• Borrowed from - Wireless Sensor Networks From
  Rhetoric to(ward) Rigor by D. Estrin

Seismic Structure Response
Contaminant Transport
Ecosystems, Biocomplexity
Marine Microorganisms
4
What is a Light Compass?

• Sensor Appliance small system with a set of
  sensors organized to provide a particular sensing
  functionality.
• Light Compass a sensor appliance capable of
• measuring
• predicting
• amount of light in an arbitrary
• direction
• position
• in an enclosed environment.

5
Light Compass Applications

• Light Monitoring
• Environment Control
• Personal lighting preferences
• (i.e. minimal light into monitor), museums,
  greenhouses
• Energy Conservation
• Dimming of light in unnecessary areas
• Visual Security Privacy
• Home security

6
Sensor Node State-of-the-Art

• Components of a sensor node
• Processing
• Storage
• Communication (Radio)
• Power
• Sensors
• Actuators

7
Quantitative Sensor-Centric Design

• For a given task
• What types of sensors, and how many?
• Design
• How to interpret the data?
• Sensor Fusion
• Where should they be placed?
• Deployment

8
System Development Phases

• Phase 1 Deployment
• Typical situations for application,
  boundary/extreme cases
• Statistical precision for application

• Phase 2 Design
• Typical environments (dimensions, lights,
  obstacles)
• Shape/structure/device size best suited

• Phase 3 Sensor Fusion
• Method to determine phenomena at devices
  position
• Estimate phenomena at arbitrary position in
  environment
• Determine accuracy reliability of results

9
Organization

• Sensor Networks
• Sensor Appliances
• Quantitative Sensor-centric Design
• Light Compass
• Models and Abstractions
• Problem Formulation
• Sensor Fusion
• Design
• Experimental Results
• Future Work
• Conclusions

10
Models and Abstraction

• Physical World (environment)
• Space time, plus their interaction
• Resolution
• Phenomena
• Sources rules of propagation
• Sensors
• Sensors (position, size, shape)
• Accuracy, resolution, error, dynamic range and
  latency

11
Models and Abstraction

• Appliance
• Physical dimensions, shape
• Rigidity
• Operational Mode of Appliance
• Single/Multiple instances
• Type of output

12
Equation-Based Models

• Fundamental Laws such as
• Newtons Mechanical Laws
• Theory of Relativity
• Electromagnetics
• Quantum Mechanics
• Have proven to work exceptionally well in
  practice
• All approximations

13
Point Source Physical Laws
14
A Light Sensing Problem

• Given n light sensors
• Sensor positions and orientations
• Sensor measurements

• Problem
• Find the number of point-light sources, their
  positions, and intensities in the environment.

• Alternate Problem
• Predict light intensity at a given point from a
  given direction.

15
Point Light Model Formulation
For each sensor-light pair
16
But Errors Will Be There

• Sensor Measurement Errors
• Calibration
• Noise
• Errors in Parameters
• Locations of nodes
• Orientation of nodes
• Modeling Errors
• Solution Errors
• Algorithmic
• Numerical

17
System of Equations
Resulting in a system of non-linear equations
18
Ways to Solve

• Non-linear function minimization with non-linear
  constraints (non-linear programming)
• General application
• Address many properties of solution, etc
• Significantly longer runtime
• Lower likelihood of optimal solution
• Non-linear function minimization
• Constraints are part of objective function
• No guarantee constraints are satisfied
• Shorter runtime

19
How to Minimize Non-linear Function?

• Non-linear numerical optimization methods
• Newtons Method
• Steepest Descent variations
• Conjugate Gradients with random restarts
• Singular Value Decomposition
• Combinatorial optimization
• Number of Lights

20
Light Compass Design

• Sensor placement - distance
• Blockage between sensor on device

21
Required Number of Sensors
2 Lights
Lights Sensors 1 5-7 2 15-20 3 40-60 4 60-
120
22
Model of Light-Sensing Device

• Minimal number of sensors needed
• Single Light need at least 4 sensors
• System of equations has 4 unknowns (x,y,z,I) of
  light source
• Only consider structure with 4 sensors
• Where to place sensors?
• Guides No Blockage, high probability of light
  placement and intensity
• Sensor placement in unlimited forms

Parametrizable Structures
23
Example Simulation Experiment

• 1 light source
• 1000 x 1000 x 500 room
• Device is in the middle on the floor
• Light can be anywhere s.t.
• 100 z 500
• 1 Device
• 5 sensors (max)
• Pyramid and cut-pyramid structures
• Assume sensors are at same position

24
5-Sensor Light Appliance ExampleAve. source
position error all solutions vs. pyr. angle
25
5-Sensor Light Appliance ExampleFraction of
valid solutions vs. pyramid angle
26
Light Compass Prototypes

• Sensors
• Miniature silicon solar cell
• Photovoltaic element
• does not require external bias
• 0.78cm x 0.58cm x 0.18cm
• 400 mV in moderate light

27
Light Compass Prototype

• ADC Hardware
• Maxim MAX186 ADC
• 8 analog channels
• 12-bit resolution
• 10?s conversion time
• 8-bit Latch
• PC parallel port
• Effective sampling 10Hz

28
Light Appliance Advantages
Small Example (10m x 10m x 5m room, 1 light)
sensor errors 1 measurement, 2 cm position, no
orientation errors (flat)
Reduced Number of devices by 6 (50)
29
Light Compass Prototype
30
Future Work

• Multi-modal appliances
• Distributed appliance
• Sensor placement strategies
• Statistical error modeling

31
Conclusion

• Sensor networks and sensor applications are
  emerging
• Sensor-centric design
• Quantitative techniques for Light Compass
• Simulation and experimentation demonstrate
  effectiveness