77
Ivan Stojmenovic 1 IoT/CPS M2M communication, actuation and coordination challenges IoT = Internet of Things CPS = Cyber Physical Systems M2M=Machine to Machine Communications Ivan Stojmenovic www.site.uottawa.ca/~ivan

Ivan Stojmenovic ivan

  • Upload
    lycong

  • View
    254

  • Download
    2

Embed Size (px)

Citation preview

Page 1: Ivan Stojmenovic ivan

Ivan Stojmenovic 1

IoT/CPS M2M communication, actuation

and coordination challenges

IoT = Internet of Things

CPS = Cyber Physical Systems

M2M=Machine to Machine Communications

Ivan Stojmenovic

www.site.uottawa.ca/~ivan

Page 2: Ivan Stojmenovic ivan

Ivan Stojmenovic 2

Ordinary objects

are instrumented;

Autonomic terminals

are interconnected;

Pervasive services are intelligent.

Concept of IoT (slide: Huadong Ma)

Based on the traditional information carriers

including Internet, telecommunication network

and so on, Internet of Things (IOT) is a network

which can interconnect ordinary physical

objects with identified addresses.

Page 3: Ivan Stojmenovic ivan

Ivan Stojmenovic 3

1. Efficient Interconnection and Data exchange

among large-scale heterogeneous network elements

Requirement: Global network convergence and local regional autonomy;

weak-state interconnection of weak ability elements e.g. sensors, RFID

2. Intensive information processing

Utilization of uncertain sensory data; Multi source and type data fusion;

authorization and privacy protection; interaction and adaptation;

3. Comprehensive intelligent service

Service delivery; adapting software design; service adaptation; modeling

Goals of IoT

Page 4: Ivan Stojmenovic ivan

Ivan Stojmenovic 4

CPS

Cyber physical systems (NSF USA)

system featuring a tight combination of, and

coordination between, the system’s computational

and physical elements

integration of computer- and information-centric

physical and engineered systems

Interconnection and addressing not required

IoT subset of CPS

example CPS which is not IoT ?

Page 5: Ivan Stojmenovic ivan

Teleautomation

1898: radio remote controlled submarine

1926 Nikola Tesla

‘When wireless is perfectly applied, the whole

earth will be converted into a huge brain, which

in fact it is… and the instruments through which

we shall be able to do this will be amazingly

simple compared with our present telephone. A

men will be able to carry one in his vest pocket.’

Ivan Stojmenovic 5

Page 6: Ivan Stojmenovic ivan

Machine-to-Machine communications

M2M uses a device (sensor or meter)

to capture an event (temperature, inventory

level, etc.),

which is relayed through a network (wireless,

wired or hybrid)

to an application (software program) that

translates the captured event into meaningful

information, which can trigger an actuation

=Networked control systems

Ivan Stojmenovic 6

Page 7: Ivan Stojmenovic ivan

Networked control systems

Ivan Stojmenovic 7

Central point for data collection and actuation

Network nodes for communication only

M2M well describe most existing CPSs

Page 8: Ivan Stojmenovic ivan

Ivan Stojmenovic 8

CPS beyond control theory ?

Embedded systems and Control theory:

single robots, space shuttle etc are CPS

Provide methods e.g. equation solvers

What are new challenges and new methods?

Control theory over single robot

Wireless networks: networked control systems

Coordination

In-network computation

In-network actuation

toward large scale CPS

Page 9: Ivan Stojmenovic ivan

Distributed traffic control systems

Ivan Stojmenovic 9

Page 10: Ivan Stojmenovic ivan

M2M communications toward large

scale CPS

Modeling

Security and privacy

In-network regional

data aggregation

coordination and

actuation

Ivan Stojmenovic 10

Page 11: Ivan Stojmenovic ivan

Ivan Stojmenovic 11

Modeling CPS/ IoT/ M2M

model of a CPS comprises models of physical

processes as well as models of the software,

computation platforms, and networks

A single model of CPS/IoT probably does not

exist

Taxonomy of models, and component modeling

e.g., model cyber-physical dependencies

Model participating networks, gateways,

communication channels, mobility …

Page 12: Ivan Stojmenovic ivan

Network model example

Ivan Stojmenovic 12

Sensors and gateways

(data aggregators)

Page 13: Ivan Stojmenovic ivan

Two-tier architecture

DA= Data Aggregators

Ivan Stojmenovic 13Gu, Lin, Chen, WCNC 2013

Page 14: Ivan Stojmenovic ivan

Cloud based M2M communications

Ivan Stojmenovic 14

Tseng, Lin, Chen, 2013

Page 15: Ivan Stojmenovic ivan

Ivan Stojmenovic 15

Actuation in CPS/IoT/WSN/M2M

Actor=actuator can act on the environment,

itself (e.g. controlled movement, turn, video..),

and networks elements e.g. sensors

Static actuators e.g. traffic lights

Mobile actuators e.g. robots

Combined cyber-physical elements

Page 16: Ivan Stojmenovic ivan

Ivan Stojmenovic 16

Large scale CPS with actuationWark et all, ACM IPSN 2007

Prevent bulls from fighting in a farm

Page 17: Ivan Stojmenovic ivan

Ivan Stojmenovic 17

•Bulls are nodes in network, carrying collars with

sensing and actuation capabilities

•Actuation:

• stimuli when two bulls come near each other.

Page 18: Ivan Stojmenovic ivan

Data dissemination in M2M

mobility,

intermittent connectivity,

collisions,

QoS for different messages and levels of

urgency, and

event distance dependent requirements.

Ivan Stojmenovic 18

Page 19: Ivan Stojmenovic ivan

reporting M2M mechanisms for

real-time monitoring

Ivan Stojmenovic 19

M2M node senses subset of data;

Gateways schedule data transmissions

Fu, Chen, Lin, Fang 2012

Page 20: Ivan Stojmenovic ivan

Cooperative access stabilization

Ivan Stojmenovic 20

Lien, Liau, Kao,

Chen JSAC 2012

access probabilities ?

to receive one report by each BS

Page 21: Ivan Stojmenovic ivan

Ivan Stojmenovic 21

Sensor-actuator model

Sensors provide input

Certain input triggers certain action

Single sensor controlled

Networked sensor controlled

Smart building (temperature, humidity…)

Page 22: Ivan Stojmenovic ivan

Ivan Stojmenovic 22

CPS/IoT taxonomy by physical

Vehicular networks

Smart cameras

Smart power grids

Smart city (bus arrival time, crowd-sensing)

Human based CPS

Smart living (context inference, assistance)

Page 23: Ivan Stojmenovic ivan

23

Smart camera and networks

Weijia Jia, City U HK &

Shenzhen

Page 24: Ivan Stojmenovic ivan

24

Internet

2G/

3G/

4G

Wi-Fi

WLAN

Actuation: camera rotation

coordination of multiple cameras

Page 25: Ivan Stojmenovic ivan

Ivan Stojmenovic 25

Extending sensor networks to IoT/CPS

Temperature, humidity, light data are extended to smoke,

carbon, fuel consumption, electricity, location, audio, etc..

Actuation? Coordination ? GreenOrbs

Tianmu mountains

Lin'an City

5000+ sensors in Lin’An and WuXi

Collaborative monitoring based on multiple networks;

Smart city applications

Page 26: Ivan Stojmenovic ivan

Ivan Stojmenovic 26

Forest Fires

• Smart cameras can focus, rotate, based on event discovery, start video

• Improved prediction and management of forest fires

Sensors Measure Temperature,

Relative Humidity, Wind

Speed and Direction

Page 27: Ivan Stojmenovic ivan

Social networks and computing

Ivan Stojmenovic 27

Human as physical systems

Online social networks with mobile access

Crowd sourcing/sensing, crowd computing

Mobile social networks

Social media and networks

(all of us are generating data)

Page 28: Ivan Stojmenovic ivan

2/24/2010 28

MobiSN, MobiClique: Creating self-configurable mobile ad

hoc social networks

Spontaneous, Opportunistic,

Smartphone networks

Page 29: Ivan Stojmenovic ivan

Human in the loop CPS

Ivan Stojmenovic 29

Page 30: Ivan Stojmenovic ivan

Restoring fundamental autonomy

Ivan Stojmenovic 30

Page 31: Ivan Stojmenovic ivan

31

Smart vehicles are upon us

2013-11-

1831

Page 32: Ivan Stojmenovic ivan

3232

• Cooperative traffic

signal preemption - Preemption for

ambulance/fire truck/police

car

- Fleet for special tasks

Hazard WarningRear-end collision

•Collaborative traffic and

construction event

alerting-Construction alert (RSU)

-Traffic event alert (OBU)

Curve

Hit from behind

Cen

terEmergency Vehicle Approaching WiMAX

WiMAX

DSRC

Alert Systems

Yi-Bing Lin, Taiwan

Warning delivery in vehicular networks

Page 33: Ivan Stojmenovic ivan

Leading car driven by driver; other cars use radars,

communications and swarm intelligence to follow

Coordinated automatic car driving

Page 34: Ivan Stojmenovic ivan

Vehicular social networking

architecture

Page 35: Ivan Stojmenovic ivan

Data gathering in VANET: 3G and V2V

Ivan Stojmenovic 35

Zhao, Zhu et all, IEEE Sensors 2013

3G access has

limited budget

V2V: delay due

to intermittent

connectivity

Gather data from

cars, delivery vs

delay

Page 36: Ivan Stojmenovic ivan

Ivan Stojmenovic 36

Smart Grid

Sensors, Computing, Communication

Page 37: Ivan Stojmenovic ivan

Modeling cascading failures in smart grids

using interdependent complex networks and

percolation theory

Ivan Stojmenovic 37

Huang, Chen, Ruj, Nayak, Stojmenovic

Page 38: Ivan Stojmenovic ivan

Ivan Stojmenovic 38

Page 39: Ivan Stojmenovic ivan

Security architecture: Ruj, Nayak, ISRTU= Remote Terminal Unit; KDCs for type of users, appliances, power sources ..

Privacy preserving data aggregation: Paillier additive homomorphic encryption

Access control: Lewko-Waters

Page 40: Ivan Stojmenovic ivan

Ivan Stojmenovic 40

Networked robots/actuators

Page 41: Ivan Stojmenovic ivan

Ivan Stojmenovic 41

Robots move to connect other nodes

Page 42: Ivan Stojmenovic ivan

Ivan Stojmenovic 42

Robocopters (TU Berlin)

Page 43: Ivan Stojmenovic ivan

Ivan Stojmenovic 43

Flying robots

aerial imaging and mapping

Internet-embedded 3D model of

human habitat

Remotely controlled quadcopters

Aerial computer vision

TU Graz 2010 IEEE Computer

Collaborative coordinated ??

Page 44: Ivan Stojmenovic ivan

04/19/2012 Wang, Tan, Xing, IPSN'12, Beijing, China

Aquatic Sensing via Robotic FishCollaborative diffusion profiling

Unocal oil spillSanta Barbara, CA, 1969

http://en.wikipedia.org

BP oil spill,Gulf of Mexico, 2010

http://en.wikipedia.org

Chemicals/Waste Water PollutionUK, 2009, Reuters

Page 45: Ivan Stojmenovic ivan

Ivan Stojmenovic 45

Robot dispersionMcLurkin, Smith 2004

Robots move along vector sum of forces

Attracted by centroid

Repelled by borders

Attract/repel by other robots

centroid

borders

Page 46: Ivan Stojmenovic ivan

Ivan Stojmenovic 46

Coordinated movements by robots:

Vector sum of repelling forces

Page 47: Ivan Stojmenovic ivan

Ivan Stojmenovic 47

Heterogeneous

robot and sensor networks

Page 48: Ivan Stojmenovic ivan

Ivan Stojmenovic 48

Mobile sinks as robots/actors

Moving to collect sensor readings

Design routes for actors to optimize energy/mobility

and collect reports periodically

Sensors may assist by sending measurements to

randevouz points

Page 49: Ivan Stojmenovic ivan

Robot collects sensors’ event data

Tour visiting one sensor in each region

Space-time, delay-tolerant

Xu, Luo, Zhang 2010

Page 50: Ivan Stojmenovic ivan

Robot localizes sensors

Li, Mitton, Simplot-Ryl TPDS 2012

Page 51: Ivan Stojmenovic ivan

Robot self-localization

Eckert et all, IEEE TIM 2011

Ivan Stojmenovic 51

Page 52: Ivan Stojmenovic ivan

Networked robotics

Ivan Stojmenovic 52

Page 53: Ivan Stojmenovic ivan

Ivan Stojmenovic 53

Robot to Robot

Communication aspects of

coordination in robot wireless networks

Page 54: Ivan Stojmenovic ivan

Nearest robot(s) selection

Page 55: Ivan Stojmenovic ivan

Ivan Stojmenovic 55

distributed ‘auction protocol’:Event reported to one robot. Which robot should act?

Melodia et all, Mobihoc’05: Centralized integer non-linear program

Each actor reports back to originating actor the offer to

provide service and the cost of it

R2 and R1 send separate reports to R3

Delay ?

Page 56: Ivan Stojmenovic ivan

Ivan Stojmenovic 56

Auction aggregationMezei, Malbasa, Stojmenovic IEEE Robotic and Automation Magazine Dec. 2010

Communication aspects of robot-robot coordination

Restricted search for best robot to respondOne robot needed to attend a single generated task

Collecting robot may have high cost or busy

Tree expansion and tree contraction phases

Robots retransmit if they suspect better robot in neighborhood

Otherwise stop expansion and respond

Responses aggregated and only best offer proceeds

Robots use k-hop knowledge

Flooding could be limited to p hops

Page 57: Ivan Stojmenovic ivan

Auction over response tree: requests

5

8

18

12 10

19 18

15

20

11

‘20’ too far, do not ask

19 and 19 too far

React to first

request only

Page 58: Ivan Stojmenovic ivan

Auction over response tree: best responses

5

8

18

12 10

19 18

15

20

11

‘20’ too far, do not ask

19 and 19 too far

React to first

request only

Page 59: Ivan Stojmenovic ivan

Auction over response tree: the winner moves to the event

‘20’ too far, do not ask

5

8

18

12 10

19 18

15

20

11

19 and 19 too far

React to first

request only

Page 60: Ivan Stojmenovic ivan

Dispatching robots to eventsLukic, Stojmenovic, IEEE MASS 2013

Ivan Stojmenovic 60

Stable matching

Preserve strong

ties

But not optimal

and not balanced

Page 61: Ivan Stojmenovic ivan

Pairwise exchanges: matching

Ivan Stojmenovic 61

Simplifies

Hungarian algorithm

Can balance load

Page 62: Ivan Stojmenovic ivan

Sequence dispatch

Ivan Stojmenovic 62

Distributed

implementation

Auction based on

local knowledge

+

negotiation

Page 63: Ivan Stojmenovic ivan

Ivan Stojmenovic 63

Sensor deployment

Sensor placement:

deploy sensors at best locations

Self-deployment or carrier based

Sensor relocation: mobile actors/sensors

move to replace failed monitoring sensors or

improve functionality (monitoring efficiency)

Mobile sensor self-relocation, or

Robots relocate sensors

Page 64: Ivan Stojmenovic ivan

Focused Coverage

Page 65: Ivan Stojmenovic ivan

Maximize sensor area coverage

robots carry sensors

Ivan Stojmenovic 65

Page 66: Ivan Stojmenovic ivan

Ivan Stojmenovic 66

Snake-like coverage

Chang et all IEEE WCNC 2007; IEEE TVT 2009; IEEE T SMC-A 2009

Robot moves within area in snake=like order and drops sensors at vertices of

hexagonal tilling

Page 67: Ivan Stojmenovic ivan

Ivan Stojmenovic 67

Snake-like deployment with

obstacles

Chang et all

IEEE WCNC07

Page 68: Ivan Stojmenovic ivan

Ivan Stojmenovic 68

Problem: uncovered holes

Fletcher, Li,

Nayak,

Stojmenovic

IEEE SECON

2010.

Page 69: Ivan Stojmenovic ivan

Ivan Stojmenovic 69

Single robot backtracking

Page 70: Ivan Stojmenovic ivan

Ivan Stojmenovic 70

Shortcut back to uncovered area

Page 71: Ivan Stojmenovic ivan

Ivan Stojmenovic 71

Single robot trajectory

Page 72: Ivan Stojmenovic ivan

Ivan Stojmenovic 72

Carrier-based coverage repair

Falcon, Li, Nayak, Stojmenovic 2010

Robot carries one or more sensors (capacity)

Mobile robot replaces damaged sensors with

spare ones; minimize tour length

One-commodity traveling salesman problem

with selective pickup and delivery

Genetic Algorithm Ant Colony Optimization

Centralized algorithm

Localized algorithms in progress

Page 73: Ivan Stojmenovic ivan

active spare faulty base station

Capacity=1

Page 74: Ivan Stojmenovic ivan

active spare faulty

Capacity=3

Page 75: Ivan Stojmenovic ivan

Nature-Inspired Optimization Methods

Powerful meta-heuristic optimization techniques that draw inspiration from the self-organizing principles of living systems.

ant colonies

bird flocks

bee hives

firefly swarms

fish schools

bat clouds

Page 76: Ivan Stojmenovic ivan

Ivan Stojmenovic 76

Movement for energy optimal

routingMobile sensors

relocate to improve

service, e.g.

video monitoring:

sensor reports

continuously to sink

Establish initial route

with mobile sensors

as interim nodes and

move to optimize

energy for routing

Page 77: Ivan Stojmenovic ivan

Robot social networks ?

Ivan Stojmenovic 77questions?