1 of 13

Lights Out

Final Presentation

Presented by B2: Diva-Oriane Marty, Malavika Krishnamurthy, Ryan Gess

2 of 13

Application Area

In the motion detection lights market, many systems are limited to few sensors
Often turn on more lights than required in a large room

Wasted energy caused by having all the lights on in room

Commercial buildings consume 2.5 kWh/square foot *

Often fail to stay on while a person is still in the room

i.e. light motion detectors shut off when people are seated

*Commercial Building Energy Consumption Survey (CBECS)

3 of 13

Solution Approach

Goal: Turn on lights in specific occupied section(s).

Save energy
Improve user experience (lights remain on with no motion detected)

+

Solution:

Motion detection (PIR) + audio detection (microphone)
Simplified hardware solution: reduced number of sensors
Expanded software solution: combining sensor and timing values +

Added web application privacy settings and manual control of lights

4 of 13

Hardware & Communication

Sensing:

Microphone: (Daoki TS-US-115-CA)

PIR: (Adafruit Motion Sensor )

2m radius, 120° FOV

Communication (WiFi):

Node MCU Mini

Mosquitto Protocol (MQTT Protocol)

Computation/Localization:

Raspberry Pi

Led light strips: LED light strips

Localizing: Weighting + Coupling sensors data

App: Django

PIR sensor

Led Strips

Microphones

RPi

App

Node MCU

5 of 13

Web Application Solution

Django (Python based web framework) running locally, will be hosted on AWS in the coming week
Google OAuth service to implement authentication and registration for users
Python, JavaScript, HTML, CSS
MQTT protocol for bidirectional communication

Subscribe to topic from RPi (communication broker) to listen to incoming messages
Send light commands to separate topic for main threads to handle

Manual control of lights, protection mode, and manual weighting of PIR/Microphones
GUI displays which work station lights are currently active

6 of 13

Device Logic

Modify Score

Send ON or OFF?

7 of 13

Design Trade-Offs

Sensitivity of microphone threshold
PIR distance/range threshold
Value of score decrement score with time (energy consumption and user experience)

Larger the score, the faster the light turns off when person leaves vicinity
Larger the score, more energy conscious but may not provide best user experience

Score increment from PIR or microphone signal

Larger the score, the faster light turns on when person enters vicinity
Larger the score, takes longer for light to turn off when person leaves the vicinity, less energy conscious

False Positives

PIR/Mic Threshold

False Negatives

8 of 13

Testing, Verification, and Metrics

Requirement	Testing	Metric
Localization latency + communication latency	Person enters zone of detection; lights turn on within 2 seconds	< 2 seconds
Detect location of individuals having a conversation (or significant noise)	Individuals are still and talking at conversation level	< 1m radius
Detect individuals moving	Individuals are not speaking while walking across the room	< 1m radius
Detect individuals with low sound and low movement	Individual enters room then does not move or make a sound	> 50%

9 of 13

Latency (Test 1)

Requirement: Lights turn on within 2 seconds of person entering area

Ensures: Computation and communication latency from signal detection to output is less than 2 seconds

Test: Film a person entering area. Record the time from when entered area to lights turning on

Success Metric: Lights turn on within 2 seconds

Results: Average latency = 1.028s (5 trials all below 2 seconds)

10 of 13

Moving and Still (Test 2 and 3)

Requirement	Turn on lights when individuals moving	Turn on lights when individuals are sitting or standing still but making noise
Test	Enter and exit 1m radius area	Enter 1m radius and have conversation while sitting for 1 min
Success Metric	Turn on lights when person enters area / Turn off when person exits area / Keep on while person is still in area	Lights remain on while individuals are in area but not moving
Results	10 Trials Successes: 10 / 10 / 8	10 Trials Successes: 8

11 of 13

Low Sound and Low Movement (Test 4)

Requirement: Turn on lights in low movement low sound environment

Test: Enter a 1m radius area of two work area and individuals are silent and still (completing work on laptop) in one work area

Success Metric: Lights remain on for 1 min while individuals are there

Results: 10 trials

Successes: 8*

* 4 false positives meaning other work areas also stayed on, this comes at cost of erroring on the side of keeping the lights on when uncertain

12 of 13

Schedule

Week 4	Week 5	Week 6	Week 7	Week 8	Week 9	Week10	Week11	Week12	Week13	Week14

Project Proposal

Order Parts

Design Presentation

User Behavior Testing

App Front End

Wifi connections

Simple Localizing

App Oauth

Weighing relative PIR

Weighing sound+PIR

Connect App to RPi

Integrating Multi Sensors

Combine all Weights

Final Pres

Testing

Ryan

Diva

Malavika

Everyone

Ryan+Diva

Modified Scheduled in last weeks:

Web App communication to and from Main Program
Scaling up from 2 to 4

User Survey Results
Testing
Reconsidering weights of PIR and microphones base on testing

13 of 13