Wilamowski B.M., Irwin J.D. The Industrial Electronics Handbook. Second Edition: Industrial Communication Systems

Подождите немного. Документ загружается.

68-4 Industrial Communication Systems

that.are.hierarchical.as.well,.but.nevertheless.strongly.interconnected.with.each.other.(Figure.68.1)..

Obse

rving

.the.beha

vior

.with

.one.expe

riment

.cann

.lead.to.a.comp

lete

.desc

ription

.of.the.syst

em.

. is.

is.the.wron

.appr

oach,

.even.if.it.look

.stun

ningly

.simpl

.To.give.an.anal

ogy:

.biol

ogists

.do.not.expl

ain

owe

.by.visu

.feat

ures

.any.more

.but.they.look.for.func

tional

.unit

.by.consi

dering

.the.owe

.as.a.

“pro

cess.”

.Simi

larly

.[Dav

97],

.tryi

.to.desc

ribe

.mere

.the.beha

vior

.of.a.comm

unication

.proto

col

.unde

die

rent

.cond

itions

.is.a.lost.caus

.since.it.will.neve

.cove

.the.comp

lete

.abil

ities

.of.a.comm

unication

proto

col.

.But.it.is.possi

ble

.to.deve

lop

.a.func

tional

.mode

.of.a.proc

ess

.[HJ9

.(whe

.we.use.the.word.

“pro

cess”

.to.refe

.to.both.tech

nical

.and.biol

ogical

.occu

rrences)

.and.unde

rstand

.its.beha

vior

.base

.on.

this.mode

.If.we.want.to.buil

.robo

.or.othe

.syst

ems

.that.are.inte

lligent

.in.a.more.huma

n-like

.way,.

we.have.to.get.away.from.beha

vioral

.desc

ription

.and.empl

.func

tional

.deve

lopment

.inst

ead.

.Whil

this.is.well.unde

rstood

.in.comp

uter

.and.comm

unications

.engi

neering,

.some.psyc

hological

.scho

ols

.like.

beha

viorists

.hav

.a.sli

ghtly

.di

erent

.vie

.of.the.mat

ter.

68.4 automated Methods for Sensor and actuator Systems

Today’s.building.sensor.and.control.systems.are.primarily.based.upon.the.processing.of.sensor.informa-

tion

.usin

.prede

ned

.rule

. e.user.or.oper

ator

.den

es,

.e.g.,.the.rang

.of.valid.temp

eratures

.for.a.room.

by.a.rule

—when

.the.temp

erature

.value.in.that.room.is.out.of.rang

.(e.g..caused.by.a.defec

t),

.the.syst

reac

.(e.g.

.with.an.erro

.mess

age).

.More.comp

licated

.diag

nostics

.requ

ire

.an.expe

rienced

.oper

ator

.who.

can.obser

.and.inte

rpret

.real

-time

.sensor.value

.Howe

ver,

.as.syst

ems

.becom

.larg

er,

.are.depl

oyed

.in.a.

wide

.varie

.of.envi

ronments,

.and.are.targ

eted

.at.techn

ically

.less

-sophisticated

.users

.both.poss

ibilities

(rul

e-based

.syst

ems

.and.expe

.users

)

.becom

.prob

lematic.

. e.cont

rol

.syst

.woul

.requ

ire

.comp

rehen-

sive

.prio

.know

ledge

.of.poss

ible

.oper

ating

.cond

itions,

.rang

.of.value

.and.erro

.cond

itions.

. is.know

edge

.may.not.be.read

ily

.avail

able,

.and.will.be.dic

ult

.for.an.unsop

histicated

.user.to.inpu

.It.is.impr

actical

for.expe

rienced

.oper

ators

.to.direc

tly

.obser

.larg

.syst

ems,

.and.naiv

.users.can.not.inte

rpret

.sensor.value

.solu

tion

.to.this.prob

lem

.is.to.autom

atically

.reco

gnize

.erro

.cond

itions

.spec

ic

.to.a.give

.sens

[HLS

99],

.actu

ator,

.or.syst

.with

out

.the.need.of.prep

rogrammed

.erro

.cond

itions,

.user

-entered

.para

eters,

.or.expe

rienced

.oper

ators.

. e.syst

.shou

.obse

rve

.sens

.and.actu

ator

.data.over.time.[JGJ

S99],

cons

truct

.a.mode

.of.“nor

mality”

.[PL0

3],

.and.issu

.erro

.aler

.when.sens

.or.actu

ator

.valu

.vary.from.

norm

al.

. e.resu

.woul

.be.a.syst

.that.can.reco

gnize

.sens

.erro

.or.abno

rmal

.sens

.or.actu

ator

read

ings,

.with.mini

mal

.manu

.con

guration

.of.the.syst

.[SH0

3].

.Furt

her,

.if.sens

.read

ings

.vary.or.

dri.over.time

.the.syst

.coul

.autom

atically

.adapt.itse

.to.the.new.“nor

mal”

.cond

itions,

.adju

sting

.its.

erro

.cri

teria

.acc

ordingly.

68.5 the Diagnostic System

For.illustration,.a.diagnostic.system.utilizing.statistical.methods.(BASE.[SBR05]).is.compared.to.a.stan-

dard

.buil

ding

.auto

mation

.syst

.(BAS

. e.BAS.cons

ists

.of.a.numb

.of.sensor

.and.actu

ators

.conn

ected

by.the.LonW

orks

.eld

bus

.(LON

)

.[LDS

01].

.It.oer

.a.visu

.inte

rface

.usin

.a.mana

gement

.info

rmation

base.(MIB

)

.for.retr

ieving

.and.mani

pulating

.syst

.para

meters

.and.for.the.visu

alization

.of.syst

.mal-

unctions.

. e.diag

nostic

.syst

.BASE.is.based.on.stat

istical

.“gen

erative”

.mode

.(SGMs

. e.goal.of.the.

syst

.is.to.auto

matically

.detec

.sensor.erro

.in.a.runn

ing

.auto

mation

.syst

.[SBR

05,FS94].

.It.does.so.

by.obser

ving

.the.data.owi

.thro

ugh

.the.syst

.and.thus.lear

.abou

.the.beha

vior

.of.the.auto

mation

syst

em.

. e.diag

nostic

.syst

.buil

.a.mode

.of.the.sensor.data.in.the.unde

rlying

.auto

mation

.syst

em,

based.on.the.data.ow..From.the.opti

mized

.mode

.the.diag

nostic

.syst

.can.iden

tify

.abno

rmal

.sensor.

and.actu

ator

.value

. e.diag

nostic

.syst

.can.eith

.anal

yze

.hist

orical

.data

.or.direc

tly

.acces

.live.data

.use.a.set.of.stat

istical

.gene

rative

.mode

.to.repr

esent

.know

ledge

.abou

.the.autom

ation

.syst

em.

A.stat

istical

.gene

rative

.mode

.take

.inpu

.like.a.sens

.valu

.a.stat

.indi

cator,

.time.of.day,.etc.,.and.

retu

rns

.a.pro

bability

.bet

ween

.zer

.and.one

Processing Data in Complex Communication Systems 68-5

Using.SGMs.has.several.advantages..First,.because.the.model.encodes.the.probability.of.a.sensor.

valu

.it.prov

ides

.a.quan

titative

.meas

ure

.of.“nor

mality,”

.whic

.can.be.moni

tored

.to.dete

.abno

rmal

even

ts.

.Seco

nd,

.the.mode

.can.be.quer

ied

.as.to.what.the.“nor

mal”

.stat

.of.the.syst

.woul

.be,.give

an.arbi

trary

.subs

.of.sens

.read

ings.

.In.othe

.word

.the.mode

.can.“ll.in”.or.pred

ict

.sens

.valu

es,

whic

.can.help.to.iden

tify

.the.sour

.of.abno

rmal

.syst

.beha

vior.

. ird

.the.mode

.can.cont

inuously

be.upd

ated

.to.ada

.to.sen

sor

.dri

.

68.5.1 Error Detection

Given.an.SGM,.implementation.of.this.functionality.is.straightforward.. e.system.assigns.a.probabil-

ity

.to.each.newl

.obse

rved

.data.valu

.When.this.prob

ability

.is.high

.the.syst

.retu

rns

.that.the.new.

data.value.is.a.“nor

mal”

.value..When.the.prob

ability

.fall

.belo

.a.speci

c

.thre

shold,

.the.syst

.rate

.the.

valu

.as.“abn

ormal.”

. e.SGM.syst

.gene

rates

.alar

.even

.when.it.obse

rves

.abno

rmal

.sens

.valu

es.

 is.leav

.open.the.ques

tion

.of.how.to.assig

.the.thre

shold

.for.norm

ality.

.In.prac

tice,

.the.user.sets.the.

thre

shold

.using.a.grap

hical

.inte

rface.

.Init

ially,

.befo

.the.syst

.has.lear

ned

.norm

.syst

.beha

vior,

many.alar

.are.gene

rated,

.and.the.user.may.deci

.to.set.the.thre

shold

.to.a.valu

.near.zero

.As.the.

syst

.acqu

ires

.a.bett

.mode

.of.the.sens

.syst

em,

.the.thre

shold

.can.be.rais

ed.

.In.any.case

.the.thre

sh-

old

.para

meter

.tell

.us.how.impr

obable

.an.even

.shou

.be.to.rais

.an.alar

. e.syst

.can.also.use.a.

log-

probability

.sca

le,

.so.tha

.the.thr

eshold

.can.easi

.be.set.to.onl

.reg

ister

.ext

remely

.unl

ikely

.eve

nts.

68.5.2 Statistical Generative Models

 e.BASE.system.implements.a.number.of.SGMs.(see.Table.68.1).

.more.comp

lex

.mode

.add.addi

tional

.capa

bilities,

.or.rela

.assu

mptions

.in.comp

arison

.to.a.simpl

Gaus

sian

.mod

.[GL

S+00].

Histogram:

. is.is.a.very.gene

ral

.mode

.that.is.appr

opriate

.for.disc

rete

.sens

.valu

es,

.as.well.as.real

valued

.sens

ors

.with.an.arbi

trary

.numbe

.of.mode

.One.draw

back

.is.that.a.histo

gram

.requ

ires

.a.rath

larg

.qua

ntity

.of.dat

.bef

ore

.it.bec

omes

.usa

ble

.or.acc

urate.

Mixture of Gaussians:

. is.mode

.rela

xes

.the.Gaus

sian

.assu

mption

.that.the.dist

ribution

.has.only.one.

mode

.A.mixt

ure

.of.Gaus

sians

.is.comp

osed

.of.a.numbe

.of.Gaus

sian

.mode

ls,

.and.each.data.valu

.is.

attr

ibuted

.to.the.Gau

ssian

.mod

.wit

.a.wei

ghting

.giv

.by.a.“po

sterior

.pro

bability.”

.See

.e.g

.[Bi

s95].

Hidden Markov Model:

. is.mode

.is.the.equi

valent

.of.the.mixtu

.of.Gaus

sians

.mode

.or.the.histo

gram

mode

.but.with.the.addi

tion

.that.the.curr

ent

.sens

.valu

.can.be.depe

ndent

.on.prev

ious

.valu

.[RJ8

6,Sal00].

.diag

nostic

.syst

.uses.SGMs.of.autom

ation

.data.poin

ts.

.For.any.give

.data.valu

.x,.mode

.M.

assig

.a.pro

bability

.to.x: P

(x)→[0,1].

Note

.that.for.disc

rete

.dist

ributions

.such.as.a.hist

ogram,

.the.valu

.assi

gned

.to.x.by.the.mode

(x).

is.a.well

-dened

.proba

bility,

.sinc

.the.set.of.poss

ible

.assi

gnments

.to. x. is. nit

.For.a.proba

bility

dens

ity,

.such.as.a.Gaus

sian

.or.mixt

ure

.of.Gaus

sians,

.the.proba

bility

.valu

.assi

gned

.to.x.by.the.mode

is.the.proba

bility

.dens

ity

.at.that.valu

.In.orde

.to.conv

ert

.this.dens

ity

.to.a.proba

bility,

.the.prob-

bility

.of.gene

rating

valu

.with

.a.neig

hborhood

.±δ.arou

.x.is.comp

uted

.as.

P x d

( )+

−

∫

φ φ

,.and.

TABLE 68.1 Statistical.Generative.Models

Model Variable.Type Parameters

Gaussian Real

μ,

.σ

Histogram Discrete,.Real Bin.counts

Mixture

.of.Gaus

sians Real

µ σ π

i i i

, ,

Hidden.Markov.model Real

ij i i

µ σ,

Hidden.Markov.model Discrete T

,.Bin.counts

68-6 Industrial Communication Systems

approximated.as.2δP

(x)

for.small.δ..Alternatively,.the.probability.under.the.model.of.equaling.or.

exc

eeding

.the.obs

erved

.val

.can.be.com

puted:

P x x P x d

M M

′

≥

( )

= +

∞

∫

( φ) φ

(68

.1)

.dat

.x.can.be.a.sen

sor

.rea

ding

.suc

.as.an.air.pre

ssure

.sen

sor,

.con

tact

.sen

sor,

.tem

perature

.sen

sor,

and.so.on..Giv

.a.new.dat

.val

ue,

.the.sys

tem

.assi

gns

.a.pro

bability

.to.thi

.val

ue.

.Whe

.the.pro

bability

is.ab

ove

.a.gi

ven

.th

reshold,

.th

.sy

stem

.co

ncludes

.th

.da

.va

lue

.is.“n

ormal.”

Given

.a.seq

uence

.of.sen

sor

.rea

dings

.x.=.{x

,.….,x

}.from.times.1.to.T,.the.system.must.create.a.model.

of.“no

rmal”

.sen

sor

.rea

dings.

. e.sys

tem

.use

.an.onl

ine

.ver

sion

.of.the.exp

ectation

.max

imization

.alg

rithm

.for.max

imum-likelihood

.par

ameter

.est

imation.

.Giv

.a.mod

.M.wit

.par

ameters

.θ,.the.log

likelihood

.of.th

.mo

del

.pa

rameters

.gi

ven

.th

.da

.x.is.gi

ven

.by

L P x

( ) log ( | )θ θ=

(68.2)

where

.the. not

ation

.P(x|θ). den

otes

. the.con

ditional

.prob

ability

.of.x.giv

. the.cur

rent

. val

ues

.of. the.

par

ameters

.θ.

.max

imum-likelihood

.par

ameters

.are.de

ned

.as.the.par

ameter

.val

ues

.tha

.max

imize

.the.log

likelihood

.ov

.th

.ob

served

.da

ta:

θ θ

ML M

P x=

{ }

arg max log ( | )

(68

.3)

68.5.3 Online Parameter Updates

In.order.for.the.system.to.continually.adapt.the.model.parameters,.the.parameter.update.algorithm.

mus

. inc

rementally

. cha

nge

. the. par

ameters

. bas

. on. new

. obs

erved

. sen

sor

. val

ues.

. Suc

. “on

line”

upd

ates

.hav

.the.adv

antage

.tha

.the

.is.no.tim

.dur

ing

.whi

.the.sys

tem

.is.in.an.“opt

imization”

.pha

se,

and.un

available

.fo

.di

agnostics

.[K

al60].

For

.the.tes

.des

cribed

.her

.we.use

.mix

ture

.of.Gau

ssians

.mod

els.

.For.the

.mod

els,

.the.BAS

.sys

tem

use

.a.simp

.stoc

hastic

.est

imation

.met

hod,

.bas

.on.an.exp

ectation–maximization

.alg

orithm

.[NH

98].

As.eac

.new.dat

.val

.is.observed,.the.parameters.are.adjusted.in.a.two-step.process..First,.the.poste-

rior

.pro

bability

.of.eac

.ele

ment

.of.the.mix

ture

.giv

.the.dat

.val

.is.com

puted.

.Sec

ond,

.the.par

ameters

are.adj

usted

.so.as.to.inc

rease

.the.exp

ected

.join

.log

-probability

.of.the.dat

.and.the.Gau

ssian

.mix

ture

com

ponent.

.Se

ction

.2.

.of.[B

is95]

.fo

.de

tails.

68.5.4 results

 e.model.log-likelihood,.given.by.Equation.68.1,.is.a.measure.of.model.quality..As.the.parameter.val-

ues

.are.opt

imized

.onli

ne,

.the.log

-likelihood

.of.the.sen

sor

.mod

els

.incr

eases.

. e.log

-likelihood

.doe

not.cons

istently

.incr

ease,

.how

ever,

.due.to.the.onli

.tt

ing

.of.par

ameters,

.whi

.hap

pens

.sim

ultane-

ously

.wit

.rep

orting

.of.abno

rmal

.sen

sor

.val

ues.

.If.sen

sors

.rec

eive

.abno

rmal

.val

ues,

.the.log

-likelihood

dec

reases,

.unt

.the.val

ues

.ret

urn

.to.the

.nor

mal

.ran

.or.the.sen

sor

.mod

.ada

pts

.to.the.new.ran

of.val

ues.

.Figu

.68.

.sho

.an.exa

mple

.fro

.a.sin

gle

.sen

sor.

. e.upp

.gu

.sho

.the.sen

sor

.val

ue;

the.low

.gu

.sho

.the.cor

responding

.log

-likelihood

.as.a.func

tion

.of.tim

. e.lar

.dis

turbance

in.the.cen

ter

.(a.pow

.uct

uation)

.reg

isters

.an.ala

rm,

.jus

.lik

.the.two.sma

.“sp

ikes”

.nea

.the.end.of.

the.gra

ph.

Processing Data in Complex Communication Systems 68-7

68.6 Intelligent Surveillance Systems

Today,.surveillance.has.become.a.relevant.means.for.protecting.public.and.industrial.areas.against.

malicious.subjects.like.burglars.or.vandals..For.both.keeping.privacy.of.irreproachable.citizens.as.well.

as.enabling.automated.detection.of.potential.threats,.computer-based.systems.are.needed.that.support.

human.operators.in.recognizing.unusual.situations..A.suitable.approach.for.that.purpose.is.to.utilize.

a.hierarchical.architecture.of.semantic.processing.layers.distributed.in.a.network.of.nodes.. e.goal.of.

these.layers.is.to.learn.the.“normality”.in.the.environment.of.the.network,.in.order.to.detect.unusual.

situations.and.to.inform.the.human.operator.in.such.cases.. e.SENSE.project.[WSe06,BKV+08].imple-

ments

.such.an.architecture.. erefore,.we.will.briey.cover.it.here.as.an.example.of.how.hierarchical.

semantic.processing.can.be.used.for.surveillance.systems.

SENSE

.consists.of.a.network.of.communicating.sensor.nodes,.each.equipped.with.a.camera.and.a.

microphone.array.. ese.sensor.modalities.observe.their.environment.and.deliver.streams.of.mono-modal.

events.to.a.reasoning.unit,.which.derives.fused.high-level.observations.from.this.information.. ese.

observations.are.exchanged.with.neighbor.nodes.in.order.to.establish.a.global.view.about.the.commonly.

observed.environment..Detected.potential.threats.are.nally.reported.to.the.person(s).in.charge.

 ough

.most.of.the.methods.used.in.the.particular.layers.are.widely.used.in.many.applications,.the.

benet.lies.in.the.combination.of.them.in.order.to.let.the.messages.of.the.system.really.appear.meaning-

ful

.to.the.user.

68.6.1 architecture

In.this.case,.an.eight-layer.data.processing.architecture.is.adopted,.in.which.the.lower.layers.are.responsible.

for.a.stable.and.comprehensive.world.representation.to.be.evaluated.in.the.higher.layers.(Figure.68.3).

First,

.the.visual.low-level.feature.extraction.(layer.0).processes.frame.by.frame.from.the.camera.in.

2D.camera.coordinates.and.extracts.predened.visual.objects..At.the.same.time,.the.audio.low-level.

extraction.scans.the.acoustic.signals.from.a.linear.eight-microphone.array.for.trained.sound.patterns.of.

predened.categories..Due.to.limited.processing.capabilities,.this.layer.can.deliver.signicantly.unstable.

data.in.both.modalities..In.case.of.unfortunate.conditions.for.the.camera,.detected.symbols.can.change.

their.label.from.one.category.to.another.and.back.for.the.same.physical.object.within.consecutive.frames..

 e.size.of.detected.symbols.can.change.from.small.elements.to.large.ones.covering.tens.of.square.

–5

Log-likelihood Sensor values

FIGURE 68.2 Sensor.value.and.log-likelihood.for.a.single.sensor.from.the.system..Unusual.sensor.values.register.

as.drops.in.the.log-likelihood,.causing.alarms.

68-8 Industrial Communication Systems

meters.and.including.previously.detected.and.other.objects..Consequently,.the.challenge.for.higher.

levels.is.to.lter.out.signicant.information.from.very.noisy.data.

.modality.events.of.layer.0.(low-level.symbols.in.Figure.68.3).are.checked.at.layer.1.for.plausibility—

e.g.,

.the.audio.symbols.with.respect.to.position.and.intensity,.the.video.symbols.with.respect.to.position.and.

size..In.the.second.layer,.symbols.that.pass.the.rst.spatial.check.are.subdued.to.a.further.check,.regarding.

their.temporal.behavior.. e.output.of.this.layer.is.a.more.stable.and.comprehensive.world.representation.

including.unimodal.symbols..Layer.3.is.responsible.for.sensor.fusion.[BLS06].the.unimodal.symbols.are.

fused.here.to.form.multi-modal.symbols.

Layer

.4.is.the.parameter.inference.machine.in.which.probabilistic.models.for.symbol.parameters.and.

events.are.optimized.. e.results.of.this.layer.are.models.of.high-level.symbols.(HLSs).and.features.that.

User notication lter

Detected alarms and

unusual behavior

Trajectories

Sensor fusion (audio, video)

Audio Video

Inter-node communication

Parameter inference

Tracking of LLSS

Pre-processing of LLSS

Predened

alarm

recognition

Unusual

behavior

recognition

Layer 8

Layer 7

Layer 6

Neighbor node

High-level layers

semantic processing

Layer 5

Layer 4

Layer 3

Layer 2

Layer 1

Layer 0

Low-level

layer

What does the node belief at time t

Audio and video object

instances at time t

High-level symbol (HLS) instances

HLS and feature models

Local trajectories map

Local and neighbors knowledge

Low-level symbols (LLS)

FIGURE 68.3 High.level.semantic.processing.soware.architecture.

Processing Data in Complex Communication Systems 68-9

describe.their.behavior..In.layer.5,.the.system.learns.about.trajectories.of.symbols..Typical.paths.through.

the.view.of.a.sens

.node.are.store

.Laye

.6.mana

ges

.the.comm

unication

.to.othe

.node

.and.esta

blishes

a.glob

.worl

.view

. e.traje

ctories

.are.also.used.to.corr

elate

.obse

rvations

.betw

een

.neig

hboring

.node

In.laye

.7,.the.reco

gnition

.of.unus

ual

.beha

vior

.and.even

.take

.plac

.using.two.appr

oaches:

.the.rst.one.

comp

ares

.curr

ent

.obse

rvations

.with.lear

ned

.mode

.by.calc

ulating

.prob

abilities

.of.occu

rrence

.of.the.

obse

rvations

.with.resp

ect

.to.thei

.posit

ion,

.velo

city,

.and.dire

ction.

.It.also.calc

ulates

.the.prob

abilities

.of.

the.dura

tion

.that.symbo

.rema

.in.an.area

.prob

abilities

.of.the.move

ment

.alon

.traje

ctories,

.incl

ud-

ing

.traje

ctories

.acro

.node

.Obse

rvations

.with.prob

abilities

.belo

.den

.thre

sholds

.rais

.“unu

sual

beha

vior”

.alar

ms.

. e.seco

.part.of.this.laye

.is.conc

erned

.with.the.reco

gnition

.of.pred

ened

.scen

ar-

ios

.and.the.crea

tion

.of.alar

.in.case.pred

ened

.thre

.cond

itions

.are.met..Fina

lly,

.laye

.8.is.resp

onsible

for.the.comm

unication

.to.the.user

.It.gene

rates

.alar

.or.stat

.mess

ages

.and.lte

.them.if.part

icular

cond

itions

.wou

.be.ann

ounced

.too.oe

.or.the.sam

.eve

.is.rec

ognized

.by.bot

.met

hods

.in.lay

.7.

68.7 the Human Mind as an archetype

for Cognitive

autom

ation

A.long-awaited.breakthrough.in.technology.is.the.ability.of.machines.to.operate.in.an.everyday.human.

envi

ronment.

.Bein

.able.to.sens

.the.worl

.in.whic

.they.are.imme

rsed,

.robo

.may.act.in.a.way.that.is.

usef

.for.huma

.user

.Deca

des

.ago,.autom

ation

.syst

.desig

ners

.star

ted

.to.foll

.the.path.that.lite

ra-

ture

.had.dra

ed:

.int

elligence

.and.per

ception

.go.han

.in.han

 us,

.very.simp

.devi

ces

.can.only.carr

.out.task

.that.do.not.requ

ire

.sens

ing

.the.real.worl

.For.inst

ance,

movi

.a.certa

.piece.5.cm.ahea

.in.the.conv

eyor

.belt

.In.a.way,.this.is.like.driv

ing

.blin

.a.car:.if.the.piece.

falls.o.the.belt

.the.devi

.won’

.know.the.reas

.and,.ther

efore,

.won’

.be.able.to.nd.a.prop

.answe

Obviously,

.more.comp

lex

.acti

vities

.dema

.more.comp

lex

.devi

ces

.and,.in.such.case

.soph

isticated

perc

eption

.is.the.turn

ing

.poin

.that.allo

.autom

ation

.syst

ems

.to.coll

ect

.the.info

rmation

.need

.to.

cont

rol

.the

.act

ions

.and.the.con

sequences

.of.the

.the.foll

owing

.sect

ion,

.we.will.give.an.over

view

.on.how.mach

ine

.perc

eption

.has.been.addr

essed

.in.

the.past

.and.what.are.prom

ising

.appr

oaches

.for.perc

eption

.in.futu

.autom

ation

.syst

ems

.in.orde

.to.be.

able.to.ful

ll

.use

ful

.tas

.in.mor

.gen

eral

.env

ironments—as

.hum

ans

.can

68.7.1 Perception in automation: a Historic Overview

 e.term.perception.has.been.used.in.computer.and.automation.systems.from.the.1950s.onwards,.since.

the.foun

dation

.of.AI..It.mean

.acqu

iring,

.inte

rpreting,

.sele

cting,

.and.orga

nizing

.sens

ory

.info

rmation.

 e.topic.itse

.was.not.new.to.autom

ation,

.but.has.gain

.a.new.qual

ity

.from.the.mome

.info

rmation

proc

essing

.cou

.be.sep

arated

.fro

.ene

rgy

.ow.and.per

formed

.in.com

pletely

.new.way

.fore

seen

.time

frame

.for.revo

lutionary,

.huma

n-like

.appl

ications

.in.AI.(some

times

.refe

rred

.to.as.arti

cial

.gene

ral

.inte

lligences)

.has.been.esti

mated

.to.be.roug

hly

.10.year

s—for

.these.last.60.year

.Alre

ady

.in.

earl

.phase

.of.resea

rch,

.engi

neers,

.psyc

hologists,

.and.neur

opsychiatrists

.coop

erated

.in.orde

.to.expl

oit

syne

rgies

.and.achi

eve

.mutu

ally

.frui

tful

.resu

lts.

.Unfo

rtunately,

.it.soon.turn

.out.that.loose.coup

lings

betw

een

.scie

ntists

.from.die

rent

.eld

.woul

.not.be.enou

.to.nd.comp

rehensive

.solu

tions.

.In.1961

E.E..Davi

.wrot

.in.his.intr

oductory

.word

.for.an.issu

.of.the.Tran

sactions on Information eory

.[Dav

61]:

Not

.to. say. that.cros

s-fertilization

.of. engi

neering

.and. the.life

-sciences

. shou

.be.scor

ned.

. But.

ther

.must.be.more.to.thes

.atte

mpts

.than.mere

.conc

octing

.a.name

.gene

rating

.well

-intentioned

enthusiasm,.speculating.with.the.aid.of.brain-computer.analogies,.and.holding.symposia.packed.

with.“pre

liminary”

.resu

lts

.from.inco

nclusive

.expe

riments.

.A.bona.de.“int

erdiscipline”

.draw

.its.

vita

lity

.from.peop

.of.demo

nstrated

.achi

evement

.in.the.cont

ributing

.disc

iplines,

.not.from.thos

who.mer

ely

.app

.ter

minology

.of.one.el

.to.ano

ther.

68-10 Industrial Communication Systems

 e.authors.of.this.text.fully.agree.with.this.statement..It.is.of.great.importance.to.form.teams.with.

memb

ers

.fro

.bot

.sci

ences

.wor

king

.toge

ther

.on.a.dai

.basi

.rat

her

.tha

.mee

ting

.reg

ularly

.pre

sent-

ing

.own.the

ories

.abo

.the.oth

er’s

.pro

fession.

.An.ind

ication

.tha

.Dav

id’s

.sta

tement

.sti

.rem

ains

.tru

.in.

man

.ar

eas

.is.th

.th

ere

.ar

.st

ill

.no.ex

amples

.of.hu

man-like

.in

telligence

.in.au

tomation.

.dev

elopment

.in.mac

hine

.per

ception

.has.tak

.two.way

.one.rel

ated

.wit

.ind

ustrial

.pro

cess

con

trol,

.whe

.mac

hines

.are.des

igned

.and.bui

.in.ord

.to.inc

rease

.pro

ductivity,

.red

uce

.cos

ts,

.as.wel

as.enh

ance

.qua

lity

.and.ex

ibility

.in.the.pro

duction

.pro

cess.

. es

.mac

hines

.mos

tly

.nee

.to.per

ceive

a.wel

l-known

.env

ironment

.and.the

refore

.con

sist

.of.a.sel

ected

.num

ber

.of.ded

icated

.sen

sors.

. e.sum.

of.sen

sor

.vie

.com

poses

.the.mac

hines’

.vie

.of.the.wor

ld.

.Num

erous

.pub

lications

.hav

.bee

.wri

tten,

whi

.exp

lain

.mat

hematical

.way

.to.cop

.wit

.sen

sor

.dat

.to.ful

ll

.the.nee

.of.aut

omation

.in.the

res

pects

.[Is

e84].

.sec

ond

.dev

elopment

.pat

.was.and.is.con

cerned

.wit

.the.per

ception

.of.hum

ans

.[VL

S+08]

.and.

hum

.act

ivities,

.on.the.one.han

.and.wit

.imp

lementing

.per

ception

.sys

tems

.imi

tating

.hum

.per

ception

.for.bro

ader

.app

lication

.are

as,

.on.the.oth

er.

.Inv

olved

.res

earch

.el

.are

.amo

.oth

ers,

.cog

nitive

sci

ences,

.AI,.ima

.pro

cessing,

.aud

.dat

.pro

cessing,

.nat

ural

.lan

guage

.pro

cessing,

.use

.int

erfaces,

.and.

hum

.ma

chine

.in

terfaces.

Although

.the.goa

.of.per

ception

.sys

tems

.alw

ays

.was.to.cre

ate

.hum

an-like

.per

ception

.inc

luding

.all.

sen

sory

.mod

alities

.of.hum

ans

.[KB

S01],

.fro

.the.ear

.beg

inning

.of.com

puter

.sci

ence

.unt

.now

.the.

use

.int

erface—the

.fro

.end.of.the.com

puter,

.thr

ough

.whi

.com

puters

.can.“pe

rceive”

.hum

ans—nor-

mally

.does.not.oer.very.human-like.communication.channels..User.interfaces.up.to.now,.including.

toda

y’s

.so-

called

.tan

gible

.int

erfaces

.are.sti

.uni

ntuitive

.(ba

tch

.int

erfaces

.in.the.196

0s,

.com

mand-line

int

erfaces

.in.th

.19

70s,

.gr

aphical

.us

.in

terfaces

.fr

.th

.19

80s

.un

til

.tod

ay*).

.res

earch

.el

.con

cerned

.wit

.per

ceiving

.inf

ormation

.abo

.hum

.use

.is.cal

led

.con

text

.awa

sys

tems.

.Con

text-aware

.sys

tems

.are.use

.to.bui

.dev

ices

.in.the.el

.of.int

elligent

.env

ironments

.or.

ubi

quitous

.com

puting.

. e.com

mon

.vie

.in.the

.com

munities

.is.tha

.com

puters

.wil

.not.onl

.bec

ome

che

aper,

.sma

ller,

.and.mor

.pow

erful;

.the

.wil

.als

.mor

.or.les

.dis

appear

.and.hid

.by.bec

oming

.int

grated

.in.nor

mal,

.eve

ryday

.obje

cts

.[Ha

i06,Mat04].

.Tec

hnology

.wil

.dis

solve

.embe

dded

.into.our.sur

roundings.

.Sma

.obje

cts

.wil

.com

municate,

.coo

perate,

.and.vir

tually

.ama

lgamate

.wit

hout

.exp

licit

.use

int

eraction

.or.com

mands

.to.for

.con

sortia

.in.ord

.to.oe

.or.eve

.ful

ll

.tas

.on.beh

alf

.of.a.use

 ey.wil

.be.cap

able

.of.not.onl

.sen

sing

.val

ues,

.but.als

.of.der

iving

.con

text

.inf

ormation

.abo

.the.rea

sons,

.int

entions,

.desi

res,

.and.bel

iefs

.of.the.use

. is.inf

ormation

.may.be.sha

red

.ove

.net

works

.lik

.the.

Int

ernet

.and.use

.to.com

pare

.and.cla

ssify

.act

ivities,

.nd.con

nections

.to.oth

.peo

ple

.and

/or

.dev

ices,

loo

.up.se

mantic

.da

tabases,

.an

.mu

.mo

re.

One

.of.the.key.iss

ues

.in.con

temporary

.res

earch

.towa

.thi

.visi

.is.sce

nario

.rec

ognition

.[GJ

07].

Sce

nario

.rec

ognition

.tri

.to.nd.seq

uences

.of.par

ticular

.beh

aviors

.in.tim

.and.gro

ups

.it.in.a.way.

hum

ans

.wou

ld.

. is.can.ran

.fro

.ver

.simp

.exa

mples

.lik

.“a.per

son

.wal

king

.alo

.a.cor

ridor”

.up.

to.“t

here

.is.a.fo

otball

.ma

tch”

.in.a.st

adium.

references

[Bis95].Bishop,.C.M.,.Neural Networks for Pattern Recognition,.New.York,.Oxford.University.Press,.1995.

[BKV+08]

.Bruc

kner,

.D.,.J..Kasb

.R..Veli

.and.W..Herzner

.High-le

vel

.hierarc

hical

.sema

ntic

.pro

cessing

fram

ework

.fo

.sma

.sen

sor

.net

works,

.In:.Pr

oceedings of the HSI,

.Kra

kow,

.Po

land,

.2008.

[BLS06]

.Bey

erer,

.J.,.F..Puente.Leo

.and.K.-D..Somm

.(eds.),.Info

rmation

.fusio

.in.der.Mess-. und.

Sens

ortechnik,

.Uni

versitaetsverlag

.Ka

rlsruhe,

.Ka

rlsruhe,

.Ger

many,

.2006.

[Br

e02]

.Br

eazeal,

.C.,.De

signing Sociable Robots,

.Ca

mbridge,

.MA,.MIT.Pr

ess,

.2002.

[CR04]

.Cost

ello,

.M.C..and.E.D..Reich

le,

.LSDN

et:

.A.neural.netw

ork

.for.mul

tisensory

.percep

tion,

.In.Sixth

nternational Conference on Cognitive Modeling,

.Pi

ttsburgh,

.PA,.2004,.pp

.341–341.

http://en..wikipedia..org/wiki/User_interface

Processing Data in Complex Communication Systems 68-11

[Dam94].Damasio,.A.,.Descartes’.Error:.Emotion,.Reason,.and the Human Brain,.New.York,.Penguine.

Boo

ks,

.1994.

[Da

v61]

. Dav

id,

. E.,. Bionics. or. elec

trology?

. An. intr

oduction

. to. the. sens

ory

. infor

mation

. iss

ue,

. IEEE

Tra

nsactions on Information eory,

.8(2),.74–77,.1961.

[Da

v97]

. Dav

is,

. J.,. Biolog

ical

. sens

. fusio

. insp

ires

. nov

. system. design,. In:. Joi

nt Service Combat

Identication Systems Conference,

.San.Diego

.CA,.1997.

[ENF07]

.Dietric

.D.,.G..Fodo

.G..Zuck

er,

.and.D..Bruc

kner

.(eds.),.Sim

ulating the Mind—A Technical

Neuropsychoanalytical Approach,

.Sp

ringer,

.Vienn

.Au

stria,

.2009.

[FS94]

.Fas

olo,

.P..and.D.E..Seb

org,

.An.SQC.app

roach

.to.moni

toring

.and.faul

.detec

tion

.in.HVAC.cont

rol

system

.In:.Pr

oceedings of the American Control Conference,

.Ba

ltimore,

.MD,.1994.

[GJ07]

.Ge

orge,

.D..an

.B..Ja

ros,

. e.HTM.lea

rning

.alg

orithms,

.Num

enta,

.Re

dwood

.Cit

.CA,.2007.

[GLS+00]

.Gou

lding,

.P.,.B..Lenno

.D..Sando

.K..Smit

.and.O..Mar

janovic,

.Fau

.detec

tion

.in.cont

inuous

pro

cesses

.using.mul

tivariate

.sta

tistical

.meth

ods,

.Int

ernational Journal of Systems Science,

.31(11),.

1459–1471,.2000.

[Ha

i06]

. Hainic

. R.R.,. e E

nd of Hardware,

. A No

vel Approach to Augmented Reality,

. Boo

ksurge,

Cha

rleston,

.SC,.2006.

[HG06]

.Haw

kins,

.J..and.D..Geo

rge,

.Hierar

chical

.temp

oral

.memo

ry—Concepts,

.the

ory,

.and.termin

ology,

Numen

ta,

.Re

dwood

.Cit

.CA,.2006.

[HJ97]

.Hoo

.C.S..and.C..Ji,.Pro

active

.netw

ork

.faul

.detec

tion,

.In:.Pro

ceedings of the INFOCOM,

.Vol

.3,.

Kobe,.Japan,.1997,.pp..1147–1155.

[HLS99]

.Hou

se,

.J.,.W.Y..Lee

.and.D.R..Shin,.Cla

ssication

.tec

hniques

.for.faul

.detec

tion

.and.diag

nosis

.of.

an.air

-handling

.unit,.AS

HRAE Transactions,

.105(1),.1987–1997,.1999.

[Is

e84]

.Iser

mann,

.R.,.Pro

cess

.faul

.detec

tion

.bas

.on.mode

ling

.and.estim

ation

.meth

ods—A

.surv

ey,

Aut

omatica,

.20(4),.387–404,.1984.

aa97]

.Jaak

kola,

.T.S.,.Var

iational

.meth

ods

.for.inferen

.and.estim

ation

.in.grap

hical

.mode

ls,

.PhD.thesi

Depa

rtment

.of.Bra

.an

.Cog

nitive

.Scien

ces,

.MIT,.Ca

mbridge,

.MA,.1997.

GJS99]

.Jor

dan,

.M.I.,.Z..Gha

hramani,

.T.S..Jaak

kola,

.and.L.K..Saul

.An.intr

oduction

.to.vari

ational

.meth-

.fo

.gra

phical

.mo

dels,

.Ma

chine Learning,

.37,.183–233,.1999.

[Ka

l60]

.Kalm

an,

.R.E.,.A.new.app

roach

.to.linea

.lt

ering

.and.pre

diction

.pro

blems,

.Tra

nsactions of the

ASME,

.Ser

ies D:

.Jo

urnal of Basis Engineering,

.82,.35–45,.Ma

rch

.1960.

[KBS01]

.Kamm

ermeier,

.P.,.M..Bus

.and.G..Schmid

.A.system

.the

oretical

.mode

.for.huma

.percep

tion

in.mu

ltimodal

.pr

esence

.syst

ems,

.IEEE/AS

ME Transactions on Mechatronics,

.6(3),.234–244,.2001.

DS01]

.Loy

.D.,.D..Die

trich,

.and.H..Sch

weinzer

.(ed

s.),

.Ope

n Control Networks:

.Lon

Works/EIA 709 Technology,

Klu

wer

.Aca

demic

.Pub

lishers,

.Bost

on,

.MA/

Dordrecht,

.the.Net

herlands/London,

.U.K

.200

[Mat04]

. Mat

tern,

. F.,. Ubiqui

tous

. Com

puting:

. Sch

laue

. Alt

agsgegenstnde—Die

. Visio

. von. der.

Info

rmatisierung

.des.Al

ltags,

.Bu

lletin des SEV/VSE,

.19,.9–13,.2004.

[NH98]

.Nea

.R.M..and.G.E..Hint

on,

.A.view.of.the.EM.algo

rithm

.tha

.jus

ties

.increm

ental,

.spa

rse,

.and.

oth

.vari

ants,

.In:.Lea

rning in Graphical Models,

.M.I..Jor

dan

.(ed.),.Klu

wer

.Academic.Publi

shers,

Dor

drecht,

.th

.Net

herlands,

.1998,.pp

.355–368.

[Pa

l08]

.Palen

sky,

.B.,.Fro

.neuro-psy

choanalysis

.to.cognit

ive

.and.ae

ctive

.aut

omation

.system

.PhD.the-

.In

stitute

.of.Co

mputer

.Te

chnology,

.Vienn

.Uni

versity

.of.Te

chnology,

.Vienn

.Au

stria,

.2008.

L03]

.Pon

toppidan,

.N..and.J..Lar

sen,

.Uns

upervised

.con

dition

.cha

nge

.det

ection

.in.lar

.die

sel

.eng

ines,

In:. IEE

E Workshop on Neural Networks for Signal Processing,

. Tou

louse,

. Fra

nce,

. C.. Mol

ina,

T.

Adali,

.J..Lar

sen,

.M.V

.Hul

le,

.S..Dou

oglas,

.and.J..Rou

.(ed

s.),

.IEE

.Pre

ss,

.Pis

cataway,

.NJ,.200

pp..565

–574.

[RJ86]

.Rab

iner,

.L.R..and.B.-H..Jua

ng,

.An.intr

oduction

.to.hidden.Mar

kov

.mode

ls,

.IEEE ASSAP

Magazine,

3,.4–16,.Ja

nuary

.1986.

[Sal00]

.Sall

ans,

.B.,.Lea

rning

.facto

red

.repr

esentations

.for.par

tially

.obser

vable

.Mar

kov

.decisio

.pro

cesses,

In:.Adv

ances in Neural Information Processing Systems,

.S.A..Sol

la,

.T.K..Leen,.and.K.-R..Mül

ler

.(eds.),.

Vol

.12,. e.MIT.Pr

ess,

.Ca

mbridge,

.MA,.2000,.pp

.1050–1056.

68-12 Industrial Communication Systems

[SBR05].Sallans,.B.,.D..Bruckner,.and.G..Russ,.Statistical.model-based.sensor.diagnostic.for.automation.

system

.In:.Pr

oceedings of the IFAC FET,

.Pueb

la,

.Mexico

.2005.

[SH03]

.Sch

ein,

.J..and.J..Hou

se,

.App

lication

.of.cont

rol

.cha

rts

.for.detec

ting

.faul

.in.vari

able-air-volume

box

es,

.AS

HRAE Transactions,

.109(2),.671–682,.2003.

ol02]

.Sol

ms,

.M..and.Tur

nbull,

.O..e

Brain and the Inner World:

.An

Introduction to the Neuroscience of

Subjective Experience,

.Kar

nac/Other

.Pre

ss,

.Cat

.Mil

ler

.For

eign

.Rig

hts

.Age

ncy,

.Lon

don,

.U.K

.200

[Vel08]

.Veli

.R.,.A.bionic.mode

.for.huma

n-like

.machin

.percep

tion,

.PhD.thesi

.Ins

titute

.of.Com

puter

Tec

hnology,

.Vienn

.Uni

versity

.of.Te

chnology,

.Vienn

.Au

stria,

.2008.

[VLS+08]

.Vada

kkepat,

.P.,.P..Lim,.L.C..De.Sil

va,

.L..Jing

.and.L..L..Ling

.Mul

timodal

.app

roach

.to.huma

face

.dete

ction

.an

.trac

king,

.IEEE Tr

ansactions on Industrial Electronics,

.55(3),.1385–1393,.2008.

[WS

e06]

. Tsa

halis,

. D.T.,. SENSE-S

mart

. Embe

dded

. Netw

ork

. of. Sensin

. Enti

ties,

. Univ

ersity

. of. Pat

ras,

Avai

lable

.at.ww

w.sense-ist.org

.SENS

.pr

oject

.we

.sit