Slide 1

Slide 1 text

Room-by-Room Device Grouping for Put-and-Play IoT System Shigemi Ishida*1, Tomoki Murakami*2, Shinya Otsuki*2 *1Future University Hakodate, JAPAN *2Nippon Telegraph and Telephone Corporation, JAPAN

Slide 2

Slide 2 text

Outline n Motivation and background n Room-by-room device grouping system n Evaluation n Summary Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN 2

Slide 3

Slide 3 text

Motivation, Background Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN 3

Slide 4

Slide 4 text

Smart House 4 Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 5

Slide 5 text

Smart (?) House n Smart (?) speaker p Specify almost everything to use services n IoT device coordination p Provides smart (?) life by manually configuring IoT devices 5 Which light? Turn off the light Restroom There is no place named restroom Bathroom! Livingroom Bedroom Bath- room Wash- room Dining/Kitchen Nature Remo Switchbot Switchbot Hub Echo Flex Echo Flex Smart power strip There is no device named bathroom Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 6

Slide 6 text

Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 Put-and-Play (PnP) IoT n Configures automatically after installation and becomes in service 6 I m in living room This TV is in the same room The light in this room is this one 4 people here (After taking shower) Turn off the light Bathroom becomes empty Okay! (Turn off the light in the bathroom)

Slide 7

Slide 7 text

Goal 7 Network config Grouping Room-by- room grouping WPS Localize Ask user the location when used Coordinate Turn of the light Must be bathroom light Okay! (Turn off the bathroom light) Learn from actual usage Where is the microwave installed? Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 8

Slide 8 text

Goal 8 Network config Grouping Room-by- room grouping WPS Localize Ask user the location when used Coordinate Turn of the light Must be bathroom light Okay! (Turn off the bathroom light) Learn from actual usage Where is the microwave installed? Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 9

Slide 9 text

Room-by-Room Device Grouping System Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN 9

Slide 10

Slide 10 text

Key Idea n Group IoT devices based on CSI changes p Affected by human movement èAnalyze CSI changes when humans are moving around IoT devices 10 E and F are in the same room! v v A B C D F E G A D E G B C F Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 11

Slide 11 text

System Overview 12 Remove phase rotation by calculating sin, cos Room-by-room grouping using clustering Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 12

Slide 12 text

Feature Extraction Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 13 sin ∅!" , cos ∅!" instead of ∅!" mean median max min std p2p iqr

Slide 13

Slide 13 text

Feature Extraction Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 14 Group IoT devices by clustering based on feature vector Used k-means, but is not limited to

Slide 14

Slide 14 text

Evaluation Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN 15

Slide 15

Slide 15 text

Experiment Setup n 1BD smart house p AP, CSI collector in dining room p 2 of Galaxy S7 edge in each room l Height: 0 90cm p Sample CSI at 10Hz for 5 min p While human is walking p With doors opened/closed 16 Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 16

Slide 16 text

Datasets n Wildcard * represents combined datasets p NH/* = NH/OP + NH/CL p */OP = NH/OP + DN/OP + LV/ OP + BD/OP Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 17 Dataset (5 min each) Abbrv Human walking in Doors No human w/ opened doors NH/OP Opened No human w/ closed doors NH/CL Closed Dining room w/ opened doors DN/OP Dining room Opened Dining room w/ closed doors DN/CL Dining room Closed Living room w/ opened doors LV/OP Living room Opened Living room w/ closed doors LV/CL Living room Closed Bedroom w/ opened doors BD/OP Bedroom Opened Bedroom w/ closed doors BD/CL Bedroom Closed

Slide 17

Slide 17 text

Evaluation Metric n Adjusted Rand Index (ARI) p Clustering performance metric p −1 ≤ ARI ≤ 1 l 1 indicates perfect clustering l 0 indicates random n Not classification! p No confusion matrix can be derived Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 18 Living room Bedroom

Slide 18

Slide 18 text

Evaluation n Feature selection n Human location n Windowing p Win size p 𝑁!"# Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 19

Slide 19

Slide 19 text

Eval (1): Feature Selection n Win size = 10s, 𝑁!"# = 4 n Randomly select windows from each dataset and group IoT devices n Repeat 100 times, calculate mean ARI Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 20 Feature */OP */CL */* sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 mean 0.10 0.40 0.44 0.01 0.44 0.35 median 0.10 0.44 0.44 0.03 0.45 0.36 max 0.47 0.39 0.41 0.43 0.44 0.39 min 0.60 0.43 0.39 0.35 0.52 0.44 std 0.69 0.89 0.34 0.81 0.63 0.93 p2p 0.69 0.83 0.52 0.76 0.81 0.89 iqr 0.45 0.85 0.28 0.87 0.36 0.93

Slide 20

Slide 20 text

Eval (1): Feature Selection n Win size = 10s, 𝑁!"# = 4 n Randomly select windows from each dataset and group IoT devices n Repeat 100 times, calculate mean ARI Ishida Lab, Future Univ Hakodate, JAPAN Dec 4-8, 2022 21 Feature */OP */CL */* sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 mean 0.10 0.40 0.44 0.01 0.44 0.35 median 0.10 0.44 0.44 0.03 0.45 0.36 max 0.47 0.39 0.41 0.43 0.44 0.39 min 0.60 0.43 0.39 0.35 0.52 0.44 std 0.69 0.89 0.34 0.81 0.63 0.93 p2p 0.69 0.83 0.52 0.76 0.81 0.89 iqr 0.45 0.85 0.28 0.87 0.36 0.93 Features correspond to CSI amplitude changes are effective

Slide 21

Slide 21 text

Eval (2): Human Location n Win size = 10s, 𝑁!"# = 10 n Randomly select windows from each dataset and group IoT devices n Repeat 100 times, calculate mean ARI 22 Human location ARI 0.42 Dining room 0.27 Living room 1.00 Bedroom 0.29 Anywhere 0.95 Anywhere or no where 0.95 Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN Features correspond to many human locations are important

Slide 22

Slide 22 text

Eval (3): Windowing n # 𝑁!"# of wins p Big 𝑁()* increases ARI p Saturate when 𝑁()* = 15 èSufficient number of wins are required to extract influence of human n Win size p Big window increases ARI p Saturate when win size = 10 èSufficient length compared to human movement is required 23 Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 23

Slide 23 text

Summary Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN 24

Slide 24

Slide 24 text

Summary n Put-and-Play (PnP) IoT p Configures automatically after installation and becomes in service n Room-by-room device grouping p Group IoT devices based on CSI changes affected by human movement p Use feature vector including the influence of human in various location n Conducted experiment in smart house p Successfully grouped devices with ARI up to 1.0 25 Dec 4-8, 2022 Ishida Lab, Future Univ Hakodate, JAPAN

Slide 25

Slide 25 text

© 2022 Shigemi ISHIDA, distributed under CC BY-NC 4.0