The intelligent habitat and everyday life activity support

kneewastefulΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 4 χρόνια και 11 μέρες)

54 εμφανίσεις

The intelligent habitat and everyday life
activity support

Hélène Pigot
,
André Mayers
, Sylvain Giroux

D
é
par
e
tment
de mathématiques et d’informatique
,

Universit
é
de
Sherbrooke, Canada.

{pigot, andre.mayers,
sgiroux
}
@dmi.usherb.ca

Abstract

Dementia causes cognitive

deficits producing functional impairments.
Continuous care and monitoring are thus compulsory to keep at home elder
s

suffering
from

dementia. Intelligent habitat can play a central role toward a
global and integrated solution

and alleviate

relatives from
th
e care

burden
.
The general idea is twofold. On the one hand, the physical environment
could supplement elder cognitive impairments by providing personalized
environmental cues that assist
him

in achieving

his

tasks. On the other hand,
the intelligent hou
se could maintain a link with relatives and medical care
system to inform them of the evolution of the disease and to alert them in
case of emergency. This paper show
s

how intelligent houses can deliver
such
cognitive assistance to elders, prolonging the t
ime they can remain at
home. First
we derive
the requirements for cognitive assistance by an
intelligent habitat from

the impact of the Alzheimer disease in the daily
living of elders.
S
ubsequently

we describe the layered computer
infrastructure needed to
implement a distributed intelligent house
information system. The implementation of such a pervasive system raises
many issues that are not trivial from a computer science perspective. In this
paper, we focus on mode
l
ling issues. Finally
a simple scenario
is used to

exemplify
the

interactions
between the intelligent house and the elders.


1

Introduction

In
1991
, 10.6% of the population
of Canada
was age
d

of
65
years
and
over
.
T
his age group is
estimated
to increase to 14.5% in 2011 and to 21.8% in
2031

[1]
.
For obvious reasons, most of the elders
would
prefer to stay at
home as long as possible. Public administration policies
also
consider
maintainin
g people at home
as
an efficient
way

to
control

medical costs.
However d
ementia
str
i
kes

up to 8% of
the 65 and over age group
. Dementia
of the Alzheimer’s type
ac
counts for about 50% of all forms of dementia

[10]
.
It

causes cogn
itive deficits
provok
ing functional impairments.
But

technolog
y
may
provide assistance
.
In
deed
intelligent

habitat can play a
central role
toward a

global and integrated solution.

The
approach

is
twofold. On

the one hand, the

physical environment
c
ould
sup
plement

cognitive
impairments by providing
personalized
environmental cues
t
hat

assist the elder
in

achieving
his
tasks.
On the other hand, the intelligent
house could maintain
a link

with relatives

and medical care system to inform
them of the evolution o
f the disease and to alert them in case of emergency.

This paper
show
s

how intelligent houses can
deliver

cognitive
assistance

to

elders
,
prolong
ing

the
period

they can remain

at home.
First we
analyse the impact of the Alzheimer disease in the daily livi
ng
of elders
and
we list
the resulting
requirement
s

for
cognitive assistance

by

an intelligent
habitat


2
)
.

Next

we
present features of the
remediation
system
s


3
-
4
)
.
Dec
ision

making and actions
for cognitive assistance

are
also

described


5
).
Then

we
sketch

the
layered
computer infrastructure
needed to
deploy

a
n

intelligent
distributed
information system inside the
house

6
)
.
The
implementation of such a pervasive system

raises many issues that are not
trivial from a computer science perspective
.
In this paper, we focus on
modeling issues


7
)
.
Finally a simple scenario is used to exempli
fy the
interactions between the intelligent house and the elders
(
§
8
).

2

Dementia, cognitive deficits and risks

The Alzheimer disease is
characteriz
ed

by deterioration of the intellectual
capacities.
It

evolves during 7 to 10 years in average.
At
the beginning

of
the
disease,
the elder often

hides the cognitive losses

by

avoid
ing

embarrassing
situations.

But
as disease evolves, cognitive deficits beco
me more obvious.


I
f assistance is provided
,

the elder
may

remain
at home
in most cases
more
or less

four years
. Beyond
this point,
supervision

is required
.

Functional performance is one criteria used to diagnose Alzheimer
disease.
Initially deficits
impe
de

memory, attention and planning
. Th
is

is

noticeable
in

difficulties to deal with new and complex tasks. As the disease
evolves the elder is less and less able to cope with familiar tasks. Spatial
orientation is also impaired. It is
noticeable

first when
the elder is lost

in new
locations and then in familiar settings
[11]
.
Cognitive losses are grade
d

in 7
levels
in

the
G
lobal
D
eterioration
S
cale (GDS)
[13]
.
The intelligent habitat
we are developing i
s designed for levels 3 to 5.
At level
3,

the elder
suffering
from

Alzheimer disease exhibits difficulties in solving problems. Balancing
budget or converting recipe ingredients
are

beyond his
reach
.
At level 4,
difficulties
appear
in preparing a
n

unusual

plate

for guests.
At level 5,
he

is
no more able to plan the cooking time
for

two meals neither to pay attention
to two meals
cooking simultaneously
.
At levels 6 and 7, he needs

supervision to
eat
,

and physical assistance is
therefore
required
.

At home, s
everal

risks could
result from

Alzheimer related

cognitive
deficits,
e.g.

memory

lack
or judgement

loss
.
The most obvious
risk
s are
f
ire
s
, wound
s
, burns, cuts, food and medication poisoning or falls.
Bad

hygiene and malnutrition
c
ould
as well
occur
. The de
cision
to

maintain

or
not

the elder at home
mainly
depends on the control

of

these
risks.

3

Remediation by the p
hysical environment

Continuous care and monitoring are thus compulsory to keep at home elder
s

suffering from
Alzheimer

dementia. This burden is
too often simply
delegated to

their

relatives.

As a result

relatives
frequently

suffer from
distress, fatigue

Since t
he
environment
significance increases when
people

becomes physically or cognitively impaired,

physical environment strategies

can

lessen
t
his
burden

on relatives
. Environmental strategies
can

control
inappropriate behaviours such as wandering or agitation.
They can also
enhance the performance of activities of daily living
(ADL)
[2]
.
Recent
p
rogress in science and

technology
can help deliver
such aids
and
foster

their use
. Actually
support

technology is
primarily

designed to increase
independence
with respect to

mobility and communication
[8]
.
Nonetheless

in a near future

ubiquitous comp
uting and
networks

will
drastically
transform

houses

[18]
.
H
abitat
s

will

assist
people

in their

ADLs

in an
“intelligent” way.
A pervasive information system

will

g
ather

information
from the environment
,

analyse it and
then

inter
vene according to
people

needs and preferences.

In the n
ext
two sub
sections
, we

present

approaches
towar
,
d this direction

either
montoring
at distance (
§
3.1
) or
assistance
in situ


3.2
).
For each app
roach,
current research project
s

are reviewed.

Next
section

4
)
gives a high
-
level perspective on our approach that integrates
distance monitoring and on
-
site
automated
cognitive assistance.


3.1

Telemonitoring

Individual and ho
us
e

security
c
ould be enhance
d

by settling a telemonitoring
system.
Such a system

alerts

people

located outside the house in case of
emergency. The house is
usually
equipped with sensors
detecting

presence,
falls and
collecting
physiological parameters. When
data extracted from the
sensors
seem

abnormal the telemonitoring system transmits a warning to the
caregivers.
Usually research

projects
focus on a specific

patholog
y

:

cardiac,
asthma
[4]

or disabled conditions
[16]

[9]

and Alzheimer disease

[3]
.

3.2

ADLs

support

A complementary approach is to provide cognitive
assistance

to help
people

achieving ADLs. The house is sti
ll equipped with sensors. Data extracted
from the environment
is

analysed to infer activities in process.
Data analyses
involve complex methods.
I
nformation given by the sensors are noisy,
uncertain and incomplete.
Once the system has inferred the current
activities
under process, it can supply

cognitive assistance
by
interven
ing directly on
the devices (
shutting

down the stove) or by providing environmental cues
(sounds, images…)
. Th
is kind of

activity assistance
is

stud
ied

in few
research centres. Kautz
Error! Reference source not found.

presents an
assisted cognition project designed for people suffering
from
cognitive
disorders.
In
this project,
an activity compass reduce
s

spatial disorientation
and an adaptive prompter help
s

people
to
carry out
ADLs
. Enhan
cing daily
living performance is also the goal of the I.L.S.A project
[5]

and the
A
ware
H
ome project
[7]
.

4

Everyday life support system

A close analys
is

of the
hardware
and software
infrastructure and
of
the
information processing needed to achieve telemonitoring and

cognitive

assistance reveal
s

that
,

in most case
,

information
may

be

acquired from the
same
categories of
devices,
then
analysed by similar processes
to feed
akin

models needed for reasoning.
On the contrary
, t
elemonitoring and cognitive
assistance
differ
though

in the
course of
actions
they
take.

Telemonitoring is
rather a passive process while cognitive assistance is proactive.
Accordingly

we are designing an
d implementing a pervasive distributed information
system that factors common
models and
processes and
then nourish

both
ways of conceiving

interventions
.
On the one hand, telemonitoring is used to
enhance security

and confidence
, by

report
ing

disease
evol
ution
or
by
alerting
the elder
external
supporting network
of frail elders

family
,

caregivers, or medical staff.

On the other hand, cognitive assistance is used
to
generate

appropriate environmental cues to support or enhance the
individual performance.

T
hese two approaches
are

active

simultaneously
and
cooperate to provide the best
of both world
s

[12]
.

5

Decisions and a
c
tions

in cognitive assistance

Actions done by the intelligent
habitat

can range from “doing nothing” to
drastic

operations
such
as
shutting down
electrical
system
. Advi
sing

stands
at

the middle
of

these extremes.
Since the
support
system will be deployed
in very diverse context
s
,
ad hoc
specification
of decisions and actions

is
neither

practical
nor

feasible

since
all

context
s

of operations cannot be
known beforehand
.
Hence t
he system must provide
models
to describe:



what to do
.
G
eneric models of a
ctions must be
predefined
.
I
nterventions
are
characterized

by content,
degree of assertiveness
, and frequency.



how to do

it
.

G
eneric personalization mechanisms must choose and
adapt actions to the current context (people, device, disease state
…).
The personalization process is guided by the following parameters:

o

cognitive abilities such as the ability to remember events in
the
environment or the language abilities to understand
pieces of
advi
c
e
;

o

h
abits in order to preserve the elder’s way of life
;


o

i
nteraction preferences such as visual or acoustic cues
;

o

previous

ADL performance.



w
hen to do it

:
The


minimal intervention


pr
inciple
steers

interventions

for cognitive assistance.

Intervention

must
occur

sole
ly

when necessary
.


Too often


may lead to
the

acceleration o
f the deterioration process of
cognitive functions

since

the elder may wait for the support system
pieces of adv
ice instead of
acting

by himself.


T
oo scarce


may lead to
dangerous situations.
Similar
principles

govern external interventions

in
telemonitoring

: too scarce interventions are potentially dangerous but
too frequent
may

upset

supporting people.

The Kitch
en Task Assessment is the starting point for the first prototype
[1]
.

The

testbed
ADL
used

is the
preparation
of a meal
.
This activity

requires
complex cognitive
challenges

such as
planning
, memory and attention.
H
ome risks such

as burns and cuts

are
also
likely to
happen
.
Moreover

“preparation of a meal”

is
one of
the canonical
ADL
in occupational therapy
.

6

A layered c
omputer infrastructure

The infrastructure is
divided in

three layers:
hardware, middleware and
application layer
s

[12]
. The hardware layer
organization
is inspired

from

an
experimental Smart Home
designed and built
by the AFIRM team of the
TIMC
-
IMAG laboratory (University J. Fourier, Grenoble, France)

[14]

[15]
.

The
middleware layer links
all the agents of the system from the lower
to the higher level.
Sensed data are percolating from lower reactive level to
higher cognitive level.

This
organization

is
fault
-
tolerant and enables
grac
eful failures
.

The

framework for spontaneous
network
ing belongs to this
layer.
The overall system is completely distributed.


The
application

layer
contains

two

integrated
services
, one for

telemonitoring,
and the other for cognitive
assistance to the
ADLs
.
Figure

1
shows the interaction between all the components of the
application

layer.




Figure
1

The
ADLs

support system.

Metamodels contain generic knowledge.

Their instantiations represent a given person who performs precise ac
tivities
in a specific environment.

7

Personalization
:

model
s

and actions

To

make

personalized

interventions the
cognitive assistance

system needs
models
describing

an individual performing an activity in his environment
.

Three models are d
rawn from the comp
etence model
[17]
.


The person model contains
knowledge on
cogniti
on
,
on
physiolog
y

and

on
life
habits and preferences
. The cognitive knowledge concerns attention,
planning
,
and
memory. The physi
ologic
al

knowledge
focus on

some
critical
physiological data such as threshold for blood pressure or heart
beat
.
Knowledge on life

habits and preferences
is
used for
personalization
.

The task model contains
successful

and
failed

scripts of
ADL

to be
compared to the one performed by the p
erson.
Failed

scripts describe pre
-
selected erroneous situations due to cognitive impairments.

The environment model contains generic description of
devices
:
sensors, electrical appliances, visual and acoustic effectors and ways to
connect to external res
ources.

As an activity
proceeds
, isolated events are detected by sensors.
Data
are percolating up to the integrator.
Along the percolation path, a
ctivities are
inferred
.
The competence model establishes if the elder is in a competence
or in a handicap si
tuation.
As long as the elder is in competence situation,
the system remains mute
.

B
ut
w
hen handicap situation occurs, messages are
sent to the environment to deliver
personalized

interventions
.
Cues

are
generated to help the
elder

to achieve his activity.

In case

of risk, the
telemonitoring system sends

message
s

to the
medical staff and the relatives
.


For a detailed discussion on models, see
[12]
.

8

Scenario

Let take a scenario to highlight how the
cognitive assistance

system
wor
ks and i
ntervenes.
As autumn is on, afternoons are colder and darkness is
falling soon. Mrs. Smith
enjoys

to take coffee at 4.00 p.m. On Friday
afternoon as usual Mrs. Smith decides to cook a veal stew. Water is boiling
for coffee and veal stew is simmerin
g when the phone rings.

A
fter her phone
call Mrs Smith feels tired. She decides to rest for a while in her bedroom.

Suddenly the
fire alarm of the
cognitive assistance
system
’s

is
trigger
ed..
.

8.1

The i
nstantiated

person model

of Mrs Smith

Mrs Smith is a 73 ye
ars old female with Alzheimer disease and osteoporosis.
She lives alone. Her cognitive abilities
range

from mild to severe deficits.

8.2

The i
nstan
t
iated environment model

of Mrs Smith’s house

Mrs Smith’s house is equipped with sensors to detect presence in
m
ost

rooms

: kitchen, lounge, bathroom and bedroom. Her stove
is a

high
-
tech
device

enhanced with
a
small
screen and
audio
. The stove could be turn
ed

off automatically.
Each

heating
element

ha
s also
its

own
security
switch.

8.3

The i
nstantiated

ADL script
s of M
rs Smith on
-
going activities

Derived from sensor information, two meal preparation activities are in
process at 4.00 p.m.

: veal stew activity and coffee activity. A third
activity

starts when the phone rings.

Ten minutes later, the phone activity is
compl
eted

as Mrs Smith hangs up. The two meal preparation activities are
still in process. A fourth activity is inferred as the bedroom sensor detects
Mrs’s Smith presence.

8.4

Supervisor

At 4.00 p.m. the supervisor diagnoses handicap situation based on three
facts
. Three activities are in process. Mrs Smith’s attention is overloaded
beyond two activities. Meal preparation activities are
at

risk of fire. The
supervisor decides not to intervene but
keeps an eye on

the situation. The
supervisor is even more
in

alert w
hen Mrs Smith goes to her bedroom
without checking the meals. But the activities are still in control and the
supervisor decides to wait again.

8.5

R
eflex decisions

As soon as
internal
fire alarm
of the cognitive assistance systems is
trigger
ed
, the stove is t
urned off.
The decision is taken locally at the stove
level.
Such a local
ity

of
decision
is implemented to
guarantee

security.
No
matter how long it takes for the supervisor to intervene
Mrs Smith and her
the house
are

at least
protected against fire.
More
over if network is down,

Mrs Smith

security

is not compromised
. Simultaneously, the
hardware

layer
send
upward two messages

: 1) the fire alarm
is
triggered, 2) the stove is
turned off. The supervisor
receives this information and
faces a handicap
situatio
n.
I
ntervention
may be necessary and
must be
personalized
.

If a prompt is
enough

to help Mrs Smith remembering her cooking
activities, the stove is turned on
.

T
he prompt is
done

using music since

Mrs
Smith prefers
to w
a
ke up in
music. T
he supervisor
conti
nues to watch the
stove to be sure
Mrs Smith
does the necessary action to remain

in security.
The

intervention
has
follow
ed

the minimal intervention
rule
.


If a prompt is
not enough

to
remember to

Mrs Smith
she was preparing

coffee
before the phone call,

t
hen
the supervisor turns on the

stove
element

used for

the veal stew preparation
while
the
element used for
the

coffee

remains off
. This intervention respects the security and the
personalization

principle
taking into account

Mrs Smith life

habits.


If Mrs

Smith presents too severe deficits to remember the two cooking
activities the supervisor decides to let the stove off.


In all cases information is transmitted to the person model in order to
built history for further interventions and to warn caregivers
if
cognitive
deficits related to

the Alzheimer

disease have

increase
d
.


An analogy could be drawn between the withdrawal reflex and the local
and supervisor decisions. When a hand feels heat source, immediately it
withdraws. Further when information reache
s the cerebral cortex by
ascending pathways conscious movements could be directed to let the hand
off the heat source or not, like the supervisor decisions.


9

Conclusion

Two kinds of environment interventions are provided to assist the elders
suffering from

Alzheimer disease and to lessen the burden caregivers. The
first one is situated inside the home. It is aimed to help the elder
complete

failed ADL due to cognitive deficits. The second type of interventions is
send to external caregivers, the medical tea
m or the family. It is aimed to
warn in case of risks and to inform about disease evolution.

To provide appropriate interventions it is necessary to model the
relationship between the person, the activities and the environment.
Distributed architecture is
also needed as data come from several sensors
spread

around

the house
. Data is

percolating up to the cognitive assistance
application. Like
a
withdrawal reflex
,

decisions could be drawn locally and
then controlled by the central supervisor. Actually one AD
L is understudy,
the meal preparation activity. But as the system evolves more ADL
s

will be
supported in order to increase the elder autonomy.

References

[1]

Baum, C., Edwards, D.F. (1993). Cognitive performance in senile
dementia of the Alzheimer’s type

: t
he Kitchen Task Assessment.
The
American Journal of Occupational Therapy
, 47 (5), 431
-
436.

[2]

Bonder, B.R., Wagner, M. (2000)
Functional performance in older
adults
. 2nd Edition. F.A. Davis.

[3]

Chan M,
Bocquet H, Campo E, Val T, Pous J.

(1999) Alarm
communicatio
n network to help carers of the elderly for safety
purposes: a survey of a project.
Int J Rehabil Res
.

1999
.

22. 131
-
13
6.

[4]

Finkelstein J, O’Connor G, Friedman RH. (2001)
Development and
implementation of the Home Asthma Telemonitoring (HAT) system to
facili
tate asthma self
-
care.
Proc MEDINFO 2001
. London, 810
-
814.

[5]

Haigh, K., Geib, C., Miller, C., Phelps, J. Wagner, T. (2002). Agents
for recognising and responding to the behaviour of an elder.
Proc. of
the AAAI workshop
on
automation as caregive
rs
. 31
-
38.

[6]

Kau
tz, H.,
et al
.

(2002). An overview of the assistede cognition project.

AAAI
-
2002 Workshop on Automation as Caregiver: The Role of
Intelligent Technology in Elder Care
.

[7]

Kidd, C.D.,
et al
., Newsletter, W. (1999) The aware home: a living
laboratory for ubiqui
tous computing research.
Proc. of the second
International Workshop on cooperative buildings


CoBuild’99
.

[8]

Lange M. L., Smith, R. (2002) Technology and occupation

:
contemporary viewpoints. The future of electronic aids to daily living.
American Journal of

Occupational Therapy
, 56

(1). 107
-
109.

[9]

Lind L, Sundvall E, Ǻhlfeldt H. (2001) Experiences from development
of home health care applications based on emerging Java technology.
Proc MEDINFO 2001
. London. 830
-
834.

[10]

McDowell, I.
(1994) Canadian Study of Health

and Aging Working
Group. Canadian Study of Health and Aging : Study methods and
prevalence of dementia
.
Canadian Medecine Association Journal
. 150
(6). 607
-
609.

[11]

Passini, R., Pigot, H., Rainville, C., Tétreault, M.H. (2000).
Wayfinding
in a Nursing Home fo
r Advanced Dementia of the Alzheimer’s Type.
Environment and Behaviour
.

32
(
5
)
, 684
-
710.

[12]

Pigot H., Lefebvre B., Meunier J.G., Kerhervé B., Mayers A., Giroux
S. (2003) The role of intelligent habitats in upholding elders in
residence.
5th international conf
erence on Simulations in Biomedicine

,
April 2003, Slovenia.
WIT Press.

[13]

Reisberg, B., Ferris, S. & Crook, T. (1982). Signs, symptoms and
course of age
-
associated cognitive decline. In S. Corkin, K.L. Davis,
J.H. Growdon, E. Usdin, R.J. Wurtman (Eds.),
Agin
g
:
Alzheimer’s
disease : a report of progress
.

New York. Raven Press.
177
-
182.

[14]

Rialle, V.,
et al
.
(2002).
Health smart home: Information technology for
patients at home.
Telemedicine Journal and E
-
Health
. 8 (4), in press.

[15]

Rialle, V. et al.,
Habitat intel
ligent pour la santé (HIS). Laboratoire
TIMC
-
IMAG.
Faculty of Medecine
and

CHU de Grenoble.
Online at
http://www
-
timc.imag.fr/AFIRM/SiteHIS/habitat
-
capteurs.html

[16]

Roth A, Carthy Z, B
enedk M.

(1997)

Telemedicine in emergency home
care


the Shahal experience.
J Telemed Telecare
. 3(1). 58
-
60
.

[17]

Rousseau, J., Potvin, L., Dutil, É., Falta, P. (in press). Model of
Competence: A Conceptual Framework for understanding the Person
-
Environment In
teraction for Persons with motor disabilities.
Journal of
Occupational Therapy in health Care.

[18]

Skackel, B. (2000) People and computers


some recent highlights.
Applied Ergonomics
.

31
.

595
-
608.

[19]

Statistiques Canada (19
94),
Répartition du revenu au Canada se
lon la
taille du revenu
. Cat 13
-
207