analytics as-a-service-io t-asia-arpanpal

19
1 Copyright © 2014 Tata Consultancy Services Limited Analytics as a Service for IoT Dr. Arpan Pal Principal Scientist, Innovation Lab Tata Consultancy Services Ltd. India Sunday, March 20, 20 22

Upload: arpan-pal

Post on 08-Aug-2015

36 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Analytics as-a-service-io t-asia-arpanpal

1 Copyright © 2014 Tata Consultancy Services Limited

Analytics as a Service for IoTDr. Arpan PalPrincipal Scientist, Innovation LabTata Consultancy Services Ltd.India

April 15, 2023

Page 2: Analytics as-a-service-io t-asia-arpanpal

2

The Internet of Everything

Humans

Physical Objects and Infrastructu

re

Computing Infrastructu

re

Peo

ple

Con

text

Dis

cove

ry

PhysicalContext Discovery

INTERNET OF EVERYTHING

Physical Context

DiscoveryWhat is happening,

where and when

People Context Discovery

Who is doing what, where and when, who is

thinking what

Internet of

Digital

Internet of

Things

Internet of

Humans

ABI Research. May 7, 2014

Page 3: Analytics as-a-service-io t-asia-arpanpal

3

Understanding the Physical Context

New Business / Pricing Models, Always On–Anytime–Anywhere, Secure, Context-aware - need to guarantee ROI for sustainability

Enables real-time monitoring to reduce downtime, reduce cost of maintenance and improve personnel safety, predicts wind-speed to improve productivity

Enables crop scouting and mapping of farmland to improve productivity of the farmers

Page 4: Analytics as-a-service-io t-asia-arpanpal

4

Understanding the People Context

Non-intrusive, un-obtrusive sensing

Identity, Location, Activity, Physiology

Understand Behavior – Individuals / Groups

Quantified Self

Customer becomes the focus, not the product or service – key is understanding the Customer, Extend B2B to B2B2C

Page 5: Analytics as-a-service-io t-asia-arpanpal

5

Platform Requirements for IoT

TCS Connected Universe Platform (TCUP)A horizontal platform for addressing the IoT Software and Services market

Applications need support for

VisibilityCapture & store data from sensors

InsightsPatterns, relationships and models

Control Optimize and actuate

TCUP Platform

Analytics is the Key

Page 6: Analytics as-a-service-io t-asia-arpanpal

6

IoT Analytics – what does it really mean?

http://www.ciandt.com/card/four-types-of-analytics-and-cognition

Page 7: Analytics as-a-service-io t-asia-arpanpal

7

Challenges for IoT Analytics

Scalability – Distributed Computing

Affordability – Reusability

Fusion – Sensor Data and Error Modeling

Ease-of-Development – Address Complexity

S

A

E

F

A sensor techie An embedded programmer A cloud programmer An algorithm expert A domain specialist An infrastructure expert

The App Developer needs to be

Page 8: Analytics as-a-service-io t-asia-arpanpal

8

Analytics-as-a-Service

Distributed Computing Infrastructure for IoT (Fog Computing)

Analytics Algorithm Repository for IoT (Algopedia)

Planning Prognostics

Causal Analytics

Behavior Sensing Measurement Anomaly

Detection

Algorithm Recommendati

on: Ease-of-Development and Fusion

Analytics Libraries: Affordability via Reuse

Base TCUP Platform (Sensor Data Transport, Storage and Analysis)

Compute Scalability: Utilize Edge

Devices

Prescriptive DescriptiveDescriptiveDescriptivePredictive Diagnostic

Page 9: Analytics as-a-service-io t-asia-arpanpal

9

Model-driven-development for IoT – Separation of Concerns through Knowledge Modeling

• Knowledge models include rules, ontologies, Information flow graphs, physical models

• Ratified / Augmented by experts (domain, sensor, algorithm and infrastructure)

Page 10: Analytics as-a-service-io t-asia-arpanpal

10

Proposed Architecture

Algorithm repository

TCS Connected Universe Platform

infrastructure sensors

Scheduler/Execution Engine

Analytic Service Layer

Workflow Engine Algorithm Recommender Partition Recommender

Knowledge Base (Algorithm,

Infrastructure)

PlanningPrognostic

sBehavior Sensing

Measurement

Anomaly Detection

Applications

Domain and Sensor Knowledge Base

Causal Analytics

Rule and Reasoning Engine

Page 11: Analytics as-a-service-io t-asia-arpanpal

11

Model-driven Framework for IoT Analytics

Page 12: Analytics as-a-service-io t-asia-arpanpal

12

Sensor-agnostic Anomaly Detection – Remote Health Monitoring

Sensed data – PPG, ECG, HR,

BP, Heart Sound, Smart-

Meter …..

Outlier Detection

Information Measure

Generate Alerts

based on critical

information

Preventive Healthcare

Promote WellnessSensor agnostic outlier analysis

library

Refer to Doctors

Being Tested on ECG, PPG and EEG Data• Anomaly within same source, same

time• Anomaly within same source,

different time• Anomaly between different sources• Can also be used for Adaptive

Compression

Page 13: Analytics as-a-service-io t-asia-arpanpal

13

Behavior Sensing – Crowd sourcing of people context using mobile phones

Indoor Localization – Bldg, Mall• Entry-Exit and Zoning• Fine-grained positioning

Activity Detection - Wellness• Walking / Brisk Walking / Jogging /

Running• Calorie BurntTraffic Sensing – City Authority• Congestion Modeling• Honk Detection• Road Condition Monitoring

Driving Behavior - Insurance• Hard Cornering / Breaking

People web-behavior - Telecom• Location-based clustering

Magnetometer – Entry/Exit

WiFi -Zoning Bluetooth -Proximity

RFID Fusion

98% 97% 96% 99.7%

(Accuracy ~2m)

(Accuracy ~ 98%)

Mobile phone sensors – Magnetometer, Wi-Fi, Bluetooth, Accelerometer, Microphone, GPSKnowledge – Sensor Noise Models

Page 14: Analytics as-a-service-io t-asia-arpanpal

14

Measurement – using Camera Vision for Physical World Metrics

eGarment Fitting – Online Retail

• Web cam based affordable system at home• Real-time 3D reconstruction is a challenge

Accident Damage Assessment - Insurance

• Mobile phone camera based Insurance Application• Template based damage assessment

Postal Packaging Automation - Transportation• Mobile Camera based System

• Camera vision based approach• 3D reconstruction from 2D images• Affordable, quick to deploy

systems

Sensors - Mobile Phone Camera, WebcamsKnowledge – Physical Object 3D Models (Human, Car, Box)

Page 15: Analytics as-a-service-io t-asia-arpanpal

15

Other Analytics Services – Causal Analysis, Prognosis, Planning

Causal Analysis - Vehicular Telemetry• Fault Detection - Automatic switch-over

to another sensor when one sensor fails• Information flow graph based

knowledge modeling• Telemetry Sensor data from OBD port

Prognosis – Remote Health Monitoring

• Knowledge Ontology from Web and experts on Disease to Symptom to Sensor Observation mapping

• Learning cum abductive reasoning based inference to prognose disease from sensor data

• Sensor data from Pathological and Physiological Devices

Planning – Emergency Evacuation• Knowledge in form of building floor plan• Graph analytics based optimization• Sensor data from BMS and Mobile

phone localization

Page 16: Analytics as-a-service-io t-asia-arpanpal

16

Vision: Democratizing IoT App Development

I only know the business logic, I do not know how to code, nor

do I understand analytics algorithms…

I know how to code, but I do not know algorithms, nor do I know about the

business logic…

Oh, I know algorithms, but I can’t code for

your mobile devices…

I have all these cloud and edge nodes which you can use to deploy

the app…

Need of the Day - Knowledge-driven Framework for IoT App Development

Page 17: Analytics as-a-service-io t-asia-arpanpal

17

Publication List

Anomaly Detection and Compression1. A Ukil, et. al., “Adaptive sensor data compression in IoT systems: sensor data analytics based Approach”,

ICASSP 20152. One more

Crowd-sensing via Mobile Phones3. Nasimuddim Ahmed et. al., ""SmartEvacTrak: A People Counting and Coarse-Level Localization Solution for

Efficient Evacuation of Large Buildings“, CASPER'15 workshop of IEEE Percom 20134. Sourjya Sarkar et. al. “Improving the Error Drift of Inertial Navigation based Indoor Location Tracking” , IPSN

20155. Vivek Chandel et.al., "AcTrak - Unobtrusive Activity Detection and Step Counting using Smartphones“,

Mobiquitous 20136. Ghose, Avik et. al., "Road condition monitoring and alert application: Using in-vehicle smartphone as internet-

connected sensor.“, Percom Workshops 2012.7. Tapas Chakravarthy et. al., “MobiDriveScore — A system for mobile sensor based driving analysis: A risk

assessment model for improving one's driving”, ICST 20138. Maiti, Santa, et al. "Historical data based real time prediction of vehicle arrival time." ITSC 2014

3D Vision based Measurements9. Saha, Arindam et. al.,"A System for Near Real-Time 3D Reconstruction from Multi-view Using 4G Enabled

Mobile." IEEE Mobile Services (MS), 201410.Brojeshwar Bhowmick et. al., “Mobiscan3D: A low cost framework for real time  dense 3D reconstruction on

mobile devices”, IEEE UIC 2014

Model-driven Development11.A. Pal et al., “Model-Driven Development for Internet of Things: Towards Easing the Concerns of Application

Developers,” IoT as a Service (IoTaaS), 201412.S. Dey et al., “Challenges of Using Edge Devices in IoT Computation Grids,” ICPADS 2013

IoT Platform13.P. Balamuralidhara et al., “Software Platforms for Internet of Things and M2M,” Journal of. Indian Inst. of

Science 14.www.tcs.com/about/research/Pages/TCS-Connected-Universe-Platform.aspx

Page 18: Analytics as-a-service-io t-asia-arpanpal

18

TCS at a Glance

Bangalore, India1

Chennai, India2

Cincinnati, USA3

Delhi, India4

Hyderabad, India5

Kolkata, India6

Mumbai, India7

Peterborough, UK8

Pune, India9

2000+

Associates in Research, Development and Asset Creation

1 2

3

4

597

6

8

10

Singapore10

iCity Lab - Collaboration with Singapore Management University – Elderly Care, Mobile Sensing

46+

13.44

Billion US$ in Q1-FY15 revenues *

305,431

119

55+ Countries where TCS has presence

Employees*

Nationalities

Source: Figures from TCS Analyst Report FY Q1-15 Employee count includes that of TCS subsidiaries

Years in Business

3.694

Billion US$ in FY14 revenues

Innovation @ TCS

Page 19: Analytics as-a-service-io t-asia-arpanpal

19

Thank [email protected]