Simulating Human Activities to Investigate Household Energy

Consumption

Edouard Amouroux

, Thomas Huraux

1,2

, Franc¸ois Semp

, Nicolas Sabouret

and Yvon Haradji

LIP6 - Universit

e Pierre & Marie Curie, Paris, France

EDF Research & Development, Clamart, France

Franc¸ois Semp

e AE, Paris, France

Keywords:

Agent-based Modelling, Social Simulation, Energy Control, Inhabitants’ Dynamic Behaviour.

Abstract:

This paper presents a multi-agent model and simulator for the investigation of inhabitants’ behaviour in re-

lation with energy consumption in a household, at a ﬁne-grain level. We present the meta-model that allows

non-computer specialists to deﬁne the household activities, and the SMACH simulator that helps specialists

analyse behaviour and manage and control energy consumption. We illustrate on concrete example scenarios

how the system’s adaptation mechanisms allow us to outline behavioural patterns.

1 INTRODUCTION

Energy efﬁciency and consumption reduction is a ma-

jor challenge for our society due to awareness raising

to the greenhouse effect and growing tensions on the

energy market. European Union (EU) set the ambi-

tious objective to divide by four greenhouse gas emis-

sions by 2050. Yet, a large part of the ﬁnal energy

consumed in Europe, 26.6%, is used within residen-

tial sector, according to the EEA (European Environ-

ment Agency, 2008). For this reason, several pro-

grammes in the residential sector have been proposed

such as building renovation, deﬁnition of low energy

building (LEB) and smart energy controller (Freire

et al., 2008), consumer awareness to cost and carbon

emissions thanks to real-time feedback (Rogers et al.,

2011). At a larger scale, smart grid have been pro-

posed (Massoud and Wollenberg, 2005) in order to

optimise the production and the distribution of elec-

tricity depending on consumption.

Nevertheless, individuals’ behaviour also strongly

inﬂuences energy consumption. In France, accord-

ing to (Poquet and Dujin, 2008), despite a moderate

increase of household electricity consumption (8%

from 1973 to 2003), a strong increase of the spe-

ciﬁc uses of electricity (i.e. electricity used except for

heating) occurred (85% for the same period). In this

context, the study of propositions to diminish speciﬁc

uses of electricity becomes a crucial issue.

One aspect of the problem is to study the con-

sumption of electrical appliances in real-life situa-

tions. That aspect is at the core of the REMODECE

european project, that provides knowledge on elec-

tricity waste due to inefﬁcient appliances and standby

consumption. A complementary approach is to study

household activities. Several works have been pro-

posed in this direction such as (Duy Ha et al., 2006;

Mahdavi and Pr

oglh

of, 2009; Kashif et al., 2012),

but they usually focus on electricity demand predic-

tion, so as to optimise energy control systems. Never-

theless, one might want to consider inﬂuencing peo-

ple’s consumption behaviour through incentives such

as variable electricity pricing, awareness campaigns,

etc. For instance, (Newsham and Bowker, 2010) sug-

gests that a critical peak price program might reduce

by 30% the peak load.

Several models allows to take into account hu-

man activities and their resulting electricity demands.

Many of them rely on statistical approaches such

as (Wid

en and W

ackelgard, 2010) and (Yamaguchi

et al., 2011) that use hidden Markov chains to gen-

erate human activities related to electricity demands.

Despite their statistical validity, they focus on repre-

senting an “average” familly and scenario and, thus,

do not allow to investigate on the activity dynam-

ics related, for instance, to speciﬁc events. Actu-

ally, agent-based models are best ﬁtted for such anal-

ysis, as recently exhibited by the model proposed

in (Kashif et al., 2012). This model relies on the

Brahms language (Sierhuis et al., 2000) to allow ex-

http://remodece.isr.uc.pt

Amouroux É., Huraux T., Sempé F., Sabouret N. and Haradji Y..

Simulating Human Activities to Investigate Household Energy Consumption.

DOI: 10.5220/0004225100710080

In Proceedings of the 5th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2013), pages 71-80

ISBN: 978-989-8565-39-6

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

plicit representation and organisation of human activ-

ities. Nevertheless, the parametrization of this model

appears to be very complex. They are hardly acces-

sible to ﬁeld experts and do not allow simulation of

complex and long situations.

In this paper, we present a meta-model and a sim-

ulator, called SMACH, that allows experts to model,

simulate and study the household activities and their

relation with electrical consumption depending on

speciﬁc pricing policies or appliance use. This system

can be used to evaluate possible incentives to dimin-

ish peak hours electricity demand.

In numerous modelling efforts, expertise collec-

tion is combined with deﬁnition of the technical ar-

chitecture which often limits the scalability of the

model. In contrary, our approach, incremental and

situated modelling (Semp

e and Gil-Quijano, 2010) ,

clearly separated the two processes: (1) we propose

a meta-model generic enough to (2) let ﬁeld experts

integrate their knowledge and evaluate their hypothe-

ses by deﬁning scenarios (for instance, in the present

application : a household of two adults, two kids, one

infant, their activities their habits, house environment

and speciﬁc events). Once a new question emerges or

an incoherence is pointed out by domain experts, the

cycle starts again by making amendment to the previ-

ous meta-model. In practice, our meta-model and the

dedicated simulator (SMACH) allow experts to de-

ﬁne ﬁne-grain activities (ranging from few minutes to

hours) and to study their evolution over long period

of time (one year). Using an advanced and intuitive

user interface, the simulator allows experts to under-

stand the respective inﬂuence of activities and energy

consumption, and the co-inﬂuence of activities.

In this paper, we focus on the presentation of the

meta-model of human activities within the household

(presented in Section 2) and the dynamics of this

model which includes communication, beliefs, action

selection and behaviour adaptation (detailed in Sec-

tion 3). Section 4, introduces the simulator GUI and

provides two distinctive examples of household emer-

gent adaptation forced by action competition and vari-

able electricity pricing. Section 5 concludes the paper

and presents the ongoing work.

2 META-MODEL OF HUMAN

ACTIVITY

Our meta-model of human activity follows two pur-

poses that are tightly connected. First, it aims

at describing human behaviours in the context of

household electricity consumption. This descrip-

tion must remain easy to manipulate, even for non-

computer specialists. To this purpose, we followed

an individual-centered agent-based approach in which

human activity is decomposed into tasks. Second, it

serves as agent description model for our multi-agent

simulator (section 3). This simulator allows to study

the evolution of human behaviour depending on the

characteristici.e.s of energy pricing and household ap-

pliances, that are deﬁned in the environment.

2.1 Environment

Let [d

] be the considered time period. En-

ergy prices can be speciﬁed to different time periods

{[d

i+1

], price

} with d

= d

and price

∈ R

, the

kilowatt hour price (kWh) for the i

sub-period.

The house is composed of several rooms where

electrical appliances are spread. Let R be the set of

rooms and E the set of appliances. For each room

r (r ∈ R), we note E

⊂ E the set of electrical ap-

pliances in this room. In our model, each appliance

can only be in one room, ﬁxed for the duration of the

simulation.

2.1.1 Appliances

Each appliance e ∈ E is characterised by its electrical

consumption θ

. We note power : E → R

the func-

tion that returns the current electrical consumption of

an appliance. In our model, we consider two different

kinds of appliances:

State-based appliances are deﬁned as a tuple

{θ

,θ

,st} where θ

and θ

are the electrical con-

sumptions (in R

) when e is running or in standby

mode, and st ∈ {off,standby,on} is the state of the

appliance, modiﬁed by individuals in the house dur-

ing their activities. Heaters, TVs, fridges... can be

represented by state-based appliances.

Program-based appliances are deﬁned as a tuple

< P

,st > where P

is a set of operating programs

characterised by load curves. Each program p is a

couple (τ

,φ

) where τ

is the program duration and

: [0,τ

] → R

gives the appliance consumption

at each time of the program. The status st of the

appliance is then deﬁned by a couple st = (p

)

where p

is the currently selected program and t

the time since the beginning of this program. Thus,

power(e) = φ

). Ovens can be represented by this

kind of appliances. Note that state-based appliances

can be seen as speciﬁc cases of program based appli-

ances (with three programs and constant load curves).

All energy consumption proﬁles comes in our

model from real data from the REMODECE project.

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

2.2 Household Members and their

Behaviours

Individuals represent the household members. Each

individual i ∈ I is characterised by its responsibility

level res

and the set of actions it can perform A

(see

below).

The responsibility level allows us to distinguish

between children, youngsters, adults and elderlies in

the simulation (see Section 3). Moreover, at each time

of the simulation, we note:

• room(i) ∈ R the room in which the individual is

located;

• action(i) ∈ A

the individual’s current activity;

• B

the individual’s beliefs set (see Section 3.1.1);

• C

the individual’s communications set (see sec-

tion 3.1.2);

2.2.1 Tasks

A task t ∈ T represents a generic activity in the house,

from which individual actions are derived. A task is a

tuple < τ

min

,τ

max

,bene,visi,coll,E

pre

> where :

• τ

min

,τ

max

the minimum and the maximum task du-

ration;

• bene ∈ {alone,collective} describes whether this

task, once achieved, can be used by other individ-

uals or not (see preconditions below);

• visi ∈ {none,individual,room,house} describes

the visibility (by other individuals) during and af-

ter the task realisation (see visibility of actions in

the following subsection);

• coll ∈ {none,allowed,mandatory} describes

whether this task cannot, can or must be per-

formed alone or with other individuals.;

• E

= {< e, p >} | e ∈ E & p ∈ {e.Pe} the set

of appliances used during this task realisation and

the associated programs (for program-based ap-

pliances);

• T

pre

⊂ T is the set of pre-conditional tasks of t.

An individual can perform a given task if and only

it has the information that all pre-conditional tasks

have been achieved by himself (if bene = alone)

or by others (if bene = collective). The precondi-

tion relation maps into a graph of tasks;

2.2.2 Actions, Rythms and Events

An action a ∈ A represents the instanciation of a task

by an individual. We denote actor(a) the unique i ∈ I

such that a ∈ A

An action is deﬁned by a tuple a = ht,w,R

,sti

where t ∈ T is the associated task, w ∈ W is the

action rhythm (see below), R

∈ R is the set of

rooms in which this action can be performed and

st ∈ undone,done is the action status, whose dynam-

ics is described in section 3.2.2. When an action has

been achieved (st = done), we note rooms(a) the set

of rooms in which it has been done.

Rhythms. In order to express habits of humans, the

concept of rhythm has been introduced in the meta-

model. A rhythm w (w ∈ W ) allows to deﬁne, for

each action, a frequency and preferred periods. The

preferred period allows to specify the month, day of

the week and the time slots in which an agent may

perform a task. To each of these speciﬁcations is as-

sociated a force indicating if the system should more

or less take into account these periods. More formally,

a rhythm is a tuple hper

, f req

,var

,Pp

i where :

• per ∈ {day, week,month, year} is the period con-

sidered;

• f req ∈ N

is the frequency for this period;

For instance, per = day and f req = 2 means

“twice a day”.

• var ∈ {weak, medium,strong} is the frequency

variability;

• PP is a set of preferred sub-periods in the

considered period. Each preferred period

is a tuple with h[d

],strengthi where

and d

deﬁne the sub-period, strength ∈

{weak, medium,imperative} the strength of the

preferred period;

For instance, let us consider the task “eating”. We

might want to deﬁne the default rhythm for this task

as being performed three times a day in most situa-

tions (per = day, f req = 3, var = weak) with pre-

ferred periods pp1

< [7 : 00, 9 : 00],imperative >,

=< [12 : 00, 14 : 00], weak > and pp

=< [19 :

00,22 : 00], strong >, which means that our individual

will eventually eat in the morning, possibly at lunch

and very probably at dinner. Note that the number of

preferred periods does not need to be equal to f req

(in which case some executions of the task will re-

main completely free over the period).

Events. Events represents exceptional situations

(such as holidays, sickness, etc.) during which usual

actions cannot be performed. An event q ∈ Q can

be deﬁned as a couple ([d

],W

with W

a set of

When different individuals can perform the same task,

each one is associated with a different instance.

SimulatingHumanActivitiestoInvestigateHouseholdEnergyConsumption

couples (a,w). Concretely, an event associates new

rhythms w to some actions a for a given time period

]. For instance, during holidays, the rhythm

associated with actions “work” is null.

2.2.3 Collective Aspects

Our meta model was designed so as to integrate

the collective aspects in family organisation, such as

helping the children, sharing domestic tasks or spend-

ing time together during certain activities. The coll

and bene parameters in tasks play a key role in lead-

ing to a cooperation mechanism among individuals,

following the simulation model that is presented in

the next section. As an illustration, there is a syn-

chronisation mechanism between agents when coll =

mandatory so that individuals wait for each other to

begin the task. Similarly, the fact that some tasks with

collective beneﬁts can be used by other agents allows

to really describe complex cooperation in the house.

3 SIMULATION OF HUMAN

ACTIVITIES

Based on the meta-model presented in the previous

section, human activity is simulated in a multi-agent

simulator called SMACH. In this simulator, agents

correspond to the human individuals in the model and

their procedural loop is presented in the next two sub-

sections. The adaptation of this loop over time to the

scenario constraints is presented in subsection 3.3.

3.1 Multi-agent Model

To each individual i ∈ I corresponds an agent in

SMACH. All agents run synchronously in the plat-

form and each step consists into selecting the best cur-

rent action based on the agent’s beliefs. In this sub-

section, we present the agent beliefs and the commu-

nication model. Next subsection presents the agent’s

dynamics.

3.1.1 Beliefs

A belief b of an individual is a tuple b = hd, a, si

such that d is the creation date, a ∈ A an action and

s ∈ {undone, running, done} is the believed status of

this action. < d,a,s >∈ B

means that, at time d,

individual i had an information about action a being

in status s (with s = running when ∃i ∈ I such that

action(i) = a). This belief is kept in i’s beliefs base

as long as no observation or deduction contradicts it.

The initial beliefs set B

of an agent i is that all

actions have the status : undone. We also consider

individuals know from the start which one can per-

form which task. The dynamics of the beliefs set is

presented in section 3.2.1.

3.1.2 Communications

In addition to individual actions, people in the house-

hold must communicate to exchange information or

request the participation of others in tasks. This is

represented using a multi-agent communication ap-

proach such as FIPA-ACL (FIPA consortium, 2003),

based on speech-act theory by Searle (Searle, 1969).

A conversation c ∈ C

is an ordered set

of n messages, msg

,...,msg

with msg

d,to, f rom, per f ∈ Per f ,c > where d is the date of

the message, to and f rom are the sender and the re-

ceiver of the message, per f is the message performa-

tive and c its content.

In our model, the main considered performatives

are : I do ..., Are you ready to ... ?, What are you do-

ing?, Did you ... ?, Can you ... ? and their following

replies No, Yes, Later. The content of the message,

when required, is always a single task t ∈ T .

The selection of messages, based on the agents be-

liefs, is presented in the next subsection.

3.2 Agents’ Dynamics

This section ﬁrst presents how beliefs are maintained

in the agents base. Then, we present how actions

are selected, based on individual’s preferences, be-

liefs and communications.

3.2.1 Beliefs Update

Beliefs in the agent’s base B

are added or removed

at each turn, based on the agent’s perception (either

direct perception in the visibility zone of an action, or

via a message) and on automatic inferences on actions

status and individual’s possible rooms.

More formally, let d

be the current time stamp

and let us note +

b and −

b the addition and removal

of belief b in B

We ﬁrst deﬁne the visibility zone of an action

a ∈ A, depending on its status and task’s visibility,

as follows (all cases are exclusive):

• if a.t.visi = none, then it is not visible: visi(a) =

• if a is active (i.e. ∃i ∈ I such that action(i) =

a) and has visibility a.t.visi ∈ {individual,room},

then visi(a) = {room(i)}

∀i|action(i)=a

;

• if a is active and a.t.visi = home, then visi(a) = R

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

• if a is ﬁnished (i.e. a.st = done) and a is visi-

ble on on individuals (a.t.visi = individual), then

visi(a) = room(actor(a));

• if a is ﬁnished and a.st.visi ∈ {room,home}, then

visi(a) = rooms(a);

The agent belief base B

is updated when observ-

ing actions from their visibility zone:

∀a|room(i) ∈ visi(a),@hd

,a,status(a)i ∈ B

⇒

,a,status(a)i

Similarly, when receiving a message with per-

formative I do or an answer to a message Did

you in a conversation, a new belief is added:

,a,status(a)i, with a the action corresponding

to the sender and the task (content) of the message.

Moreover, a coherence mechanism removes all

previous incompatible belief:

• An action can only have one status:

,a,status(a)i ⇒ ∀d

< d

,−

,a,s

status(a)i

• An agent can only do one action at a

time: +

,a,status(a)i ⇒ ∀a

|actor(a) =

actor(a

),−

iand +

,donei;

• A task “must” be over after its maximum duration:

if ∃b ∈ B

such that b.d + b.a.t.τ

max

> d

, then −

and +

,b.a,donei;

• Two actions associated with one sin-

gle “exclusive” task (a

.t = a

.t = t and

t.coll = none) cannot be active simultaneously:

−

,runningi and +

,donei

Note that when a previous belief is updated, its sta-

tus is moved to done (actions that can no longer be

performed are considered as achieved by default ).

3.2.2 Action Dynamic

The status of an action st (from undone to done) is

actually more complex than a simple done/undone

boolean. It is controlled by four internal variables:

• real

number

∈ N

is the number of realisation al-

ready conducted within the current day. The value

of a.st ∈ {done,undone} used in beliefs is directly

derived from a.real

number

≥ 1;

• real

max

∈ N

is the planned maximum number of

realisation for the current day (which directly de-

pends on w);

• τ ≤ a.t.τ

max

is the elapsed duration when a is run-

ning;

• pr ∈ [−1; 1] is the action priority.

The priority a.pr evolves in time (see below, section

3.2.5) and the agent selects the action with the high-

est priority amongst all possible actions. An action

is possible if and only if all the following criteria are

met (see also Table 1):

• Every precondition tasks have been conducted;

• All needed appliances are available;

Table 1: Action availability conditions.

(a) ∀t

∈ a.t.T

pre

,∃a

∈ A|a

.t = t

∧ a

.st = done ∧

(actor(a

) = actor(a) ∨ t

.bene = i.e.collective)

(b) ∀e ∈ a.t.E

,@a

.running ∧ e ∈ a

Note that a.st cannot be computed directly by the

agent. On the contrary, the agent will use its belief

base to decide whether actions are done or undone

and whether they can be performed. In the algorithm

in table 1, a.st = x means ∃ < d,a,done >∈ B

. How-

ever, when starting an action, the simulator will check

all the actual preconditions and an agent will not be

able to perform an action that it believed possible if

table 1 is not checked. In this case, all inconsistent

beliefs are removed from the agent base.

3.2.3 Interruption and Realisation

The running action of an individual (a = action(i)),

can be interrupted for two reasons: 1) another (pos-

sible) action, a

, has a higher priority or 2) a became

impossible. When this occurs, another action will be

selected and the status of a is modiﬁed depending on

its duration time :

• If a has been executed for a duration τ ≥ τ

min

, its

number of realisation a.real

number

is incremented

and a.τ is reset to 0. The action status a.st

switches do done.

• If a.τ < τ

min

The action is put in a “standby

mode“: a.τ is blocked at its current value, its num-

ber of realisation is not increased. For instance, a

child is interrupted during homework, (s)he does

not loose the beneﬁt of having started working.

3.2.4 Daily Reset

The action’s state evolves over time depending on the

individual behaviour and his/her environment. It is

reset at midnight except for the sleep action which

is reset at noon. In practice, a.st is reset to undone,

a.real

number

, a.τ and a.pr are reset to 0, and the maxi-

mum number of realisation is re-computed depending

on the action rhythm if the new day contains a pref-

erential period, ppi, and if some realisation are to be

done (w.real < w.real

max

SimulatingHumanActivitiestoInvestigateHouseholdEnergyConsumption

3.2.5 Action Priority

An action priority takes into account the individual’s

internal state: preference

, commitment to the current

action (see table 2 for details) and the inﬂuence of the

associated rhythm (a.w, see also 3.3). In addition to

the individual’s internal state, the priority of an action

may be inﬂuenced by external factors: energy price

level (variability of the energy price is summarised as

3 levels), other individuals invitations (through mes-

sages) and events inﬂuence. Table 2 describes the

computation of this raw priority, called pr

raw

= pre f + in f real + in f pp + in f eng +

in f ener + in f inci + in f eve

All the coefﬁcient values have been determined

empirically, only their relative order is signiﬁcant. For

instance, the individual preference is much less im-

portant than its commitment/lassitude to his/her cur-

rent action.

In order to allow the representation of sequential

behaviour (e.g. leave for work requires to be suit up

which requires to have taken a shower), the action pri-

ority must also consider the priority of the most urgent

depending action, i.e. the maximum of priorities of all

actions for which a is in precondition:

dep

= max

a.t∈a

pre

.pr}

Thus, the action priority a.pr is the maximum of these

two values:

a.pr = max{a.pr

raw

, pr

dep

}

3.3 Behaviour Adaptation

Adaptation is essential in our model as we want to

study the possible reactions of a household to vari-

able energy pricing, habits emergence and other sin-

gle events (e.g. holidays, sickness, etc.). Adaptation

is conducted trough re-organisation of the household,

i.e. when tasks are made and who does which collec-

tive task.

We implemented two adaptation mechanisms dur-

ing a simulation: (1) everyday life obligations or

strong habits are represented through rhythms with

imperative frequencies and/or preferred period (PP)

and (2) habits may emerge thanks to the controlled

mobility of actions.

3.3.1 Everyday Life Strong Constraints

Some everyday life actions may require to be con-

ducted at speciﬁc time. Such constraint is controlled

minimal factor enabling complete differentiation be-

tween actions with similar priorities

Table 2: Raw priority external factors.

(a) pre f ∈ [−0,01;+0,01] is the individual preference for

the action;

(b) in f

real

∈ {−0,2;0} is the penalty if the action already

reached its maximum number of realisation (a.real

max

);

∈ {penalty,0, bonus} is the action’s rhythm in-

ﬂuence, with penalty and bonus computed as described

below in section 3.3. in f

= 0 if a.pp =

0 and is neg-

ative (penalty) when the current date is outside a pre-

ferred period and positive (bonus) otherwise.

(d) in f

eng

∈ [−0.1; +0,1] is the commitment or lassitude

inﬂuence. Its value is set to 0 for all non running ac-

tions. Otherwise, it is deﬁned by a linear interpola-

tion from 0.1 when the action just began, to 0 when

it reaches τ

min

and −0.1 when at τ

max

(e) in f

ener

is the electricity price inﬂuence, deﬁned by:

– in f

ener

= 0 if a.t.cons = 0 or price = medium;

– in f

ener

= −0.1 if price = high;

– in f

ener

= +0.02 if price = low.

Tasks that consume energy are strongly penalised when

the energy price is high, but only marginally favored

when it is low.

(f) in f

inci

∈ [0; +0,1] computes the inﬂuence of other in-

dividuals, through messages with performative Are you

ready to and Can you, also called invitations. When

a.t.coll = none, in f

inci

= 0. Otherwise, it is increased

by +0,02 per invitation (with a maximum of 0.1).

(g) in f

eve

∈ [−0, 1; +0,1] is the events’ inﬂuence. For each

q ∈ Q such that d

current

∈ [d

] and ∃(a,w) ∈ W

in f

eve

is modiﬁed by ±0.05, depending on the new ry-

thm w’s properties compared to the action’s initial ry-

thm a.w.

through the PP inﬂuence (see Table 2): positive dur-

ing a PP and negative outside of any PP, in order to

favour within PP and penalise outside PP realisation

respectively. Unfortunately, as in the everyday life,i.e.

several actions may compete for a realisation during

an overlapping timeslot (e.g. have breakfast and take

shower both should be done between 7 to 8am). To

lessen such competition between imperative rhythms,

each action at the end of the day reevaluates its bonus

and penalty values according to the following policy:

bonus must remain between 0 and 0.15;

penalty must remain between −0.15 and 0;

When the action could not be executed during its PP

this day, bonus is increased 0.02;

When the action was achieved during its PP for this

day, the bonus is decreased by 0.01. This ensures

a slow re-organisation capacity.

When the action was executed outside a PP, bonus is

increased by 0.01;

When the system could not reach the target reali-

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

Figure 1: SMACH GUI (analysis mode).

sation number for this action (i.e., real

number

real

max

), penalty is decreased by 0.02, since we

want to give priority to action achievement over

PP constraints.

In other words, actions that could be done during

their PP during day n will be less favoured on their

PP in day n + 1 and vice-versa. Moreover, actions

that missed some executions will be encouraged. Al-

though this policy cannot guarantee the satisfaction of

all realisation constraints (which may not be possible

at all), it provides modeller more freedom regarding

rhythms deﬁnition.

3.3.2 Everyday Life Habits

Outside the competitive timeslot, habits may emerge

as in everyday life. This is achieved in SMACH

via the automatic creation of non-imperative rhythms

during simulation. In practice, if an action has

been triggered at a given period, this period will be

favoured the following days (using a limited mem-

ory window)

. As a consequence, all actions with no

speciﬁed PP are allowed to ﬁnd the most suitable pe-

riods for their realisations and to keep them as long

they are suitable.

The maximum values for bonus and penalty are

The implementation detail of this mechanism is not de-

tailed in this paper but it is very similar to the one presented

in the previous subsection

lesser than those of imperative PP, so this kind of ac-

tions can change its PP more easily. The next section

will give an example of such adaptation: in conjunc-

tion with the priority penalty during periods with ex-

pensive energy, the household organisation is altered.

4 IMPLEMENTATION AND

EVALUATION

The SMACH simulator has been implemented in

Java and several example scenarios have been pro-

posed to validate our approach. We evaluated dif-

ferent generic modelling and simulation platforms

(i.e. Repast (North et al., 2007), Netlogo (Wilensky

and Evanston, 1999) and GAMA (Taillandier et al.,

2010)) but none would meet our two most impor-

tant requirements. First, a tailored modelling helps

to ease deﬁnition of simulation scenario by ﬁeld ex-

perts. Programming-like modelling language are out

of the question for non-computer-scientists. Second,

a domain-related simulation analysis tool is also re-

quired by ﬁeld experts and the implementation of such

a tool in a constrained environment can be very costly.

For these two main reasons, we developped our own

agent platform for the SMACH simulator, which of-

fer a detailed simulation analysis GUI. This GUI is

detailed in the following subsection. We will also

discuss two examples that illustrate the ability of the

SimulatingHumanActivitiestoInvestigateHouseholdEnergyConsumption

Figure 2: Activity diagram: (a) ﬁrst week and (b) working day pattern.

model to express rhythm constraints and, at the same

time, and the actions reorganisation process. The ﬁrst

example focuses on action competition whereas the

second one demonstrates how is handled household

adaptation in the situation of a change in electricity

pricing policy.

4.1 Simulation Analysis GUI

The SMACH simulation analysis GUI shown in ﬁg-

ure 1 can be detailed as follows. The overall electric-

ity consumption is shown in panel (1). Each color rep-

resents a different appliance’s consumption (the cen-

tral gap correspond to holidays). Panel (1’) presents

some global variable (total electricity consumption,

associated price and social indicators). Below, panel

(2), details individuals activities over the week se-

lected in panel 1 (S45) whereas panels (3 and 3’)

presents a zoom for even more details (one rectangle

representing one minute). Finally, panel (4) allows

to investigates actions actual rhythms (only some ac-

tions are selected). Concretely, the horizontal axe rep-

resent the passing of days whereas the vertical axe

represen the time of the day. For instance, we can see

that the work actions (blocks centered horizontally)

occurs only during week days.

4.2 Dealing with Competing Actions

Our ﬁrst example considers a family of 3 with a con-

ﬂict situation: they have to make diner, have diner

and have shower within the 7 to 9 pm time slot. This

time slot imposes a constraint due to the actions dura-

tions and the exclusiveness of the have shower action

(one at a time). The adaptation process will organise

progressively the actions to ease their execution.

In the simulation ﬁrst week, individuals usually

fails to have shower as shown in Fig.2 (a). This dia-

gram represents the individuals’ acitvities as colored

In addition, another simulation analysis GUI mode is

also available in order to compare several simulations: over-

all evolution and speciﬁc period of time can be compared

with side by side diagrams.

bar over time. In particular, the sleep action (straw-

berry pink) and the work action (brown) can serve as

reference points. These actions have regular sched-

ule since the beginning (no adaptation needed) as they

have no strong competitors and large preferred peri-

ods (around 8 hours). In the conﬂict timeslot, diner-

related actions are occuring whereas have shower is

not (only the mother do it on monday). The advantage

of diner-related actions is due to two factors: (1) make

dinner is a collective beneﬁt task (one individual can

cook for 3) and (2) have dinner is a coordinated one

which imply strongly incentive invitations.

The conﬂict is resolved through adaptation of the

preferred periods (PP) and results in the following

working-day pattern showed in Fig. 2 (b). After

work, the mother takes her shower while the father

and Liz make diner. Once diner is ready, the father

or Liz invite the others to have diner. After diner, the

father takes a shower followed by Liz.

This example demonstrates how the adaptation

process can organize the actions in relation with PP

constraints. Although, this organization is not fully

stable, due to the continuous adaptations of the PP, it

can last several weeks.

4.3 Adaptation of Habits in Relation

with a New Pricing

The second example is based on similar parameters

but shows how PP modiﬁcations work with agents

sensitivity to the electricity price. In this example,

the simulation spans over 10 weeks from January to

March. In order to evaluate the capabilities of adap-

tation to variable electricity price we make the fol-

lowing hypotheses. In January, the electricity price

is ﬁxed to a medium level. The following month, we

introduce a variable price policy: high from 6 to 8

pm and low for the rest of day (everyday). Finally,

in March, the price is switched back to the original

policy (ﬁxed price at medium level).

Several actions, i.e. watch TV, use computer, gar-

den, take shower, play chess, without speciﬁed PP are

also added to the ones presented in the previous ex-

ample. Some of them consume electricity and are,

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

Figure 3: Long term activity diagram: (a) watch Tv and (b) homework.

thus, sensible to electricity price. The variation of

electricity pricing induce actions mobility as show on

the long term activity diagrams (Fig. 3). In Fig. 3 (a),

we focus on the watch TV action (in black).Although,

two actions have been added as time landmark : have

breakfast (in the bottom) that takes place between

6:30 and 8:15 and have diner (in the top) between

19:30 and 20:30. In this simulation, individuals do

watch TV during 4 periods: (1) After breakfast, (2)

during the afternoon on week-ends and wednesday (3)

before and (4) after diner.

In January, individuals mostly watch TV before

the diner though a slow change from after diner to

after breakfast habit is appearing all along this month.

When February comes, with its special pricing, the

before diner period is abandoned to the beneﬁt of the

after breakfast and after-diner periods due to the en-

ergy price penalty. One may notice that a slim black

line remains before diner in February. Indeed only

2 agents have completely quit watching TV before

diner. The third one continues to do it but for a shorter

duration. It means that the other inﬂuences, mainly

the strength of the before-diner PP, have balanced the

energy-price penalty. It is interesting to notice that

the effect of the penalty is progressive (it may com-

pletely stop or just shorten an action) and vary from

one agent to another. In March, when the special pric-

ing is stopped, we can observe a motion from after-

breakfast and after-diner periods to the before-diner

period as there is no electricity price penalty. Nev-

ertheless, the last two weeks of simulation still bear

this penalty inﬂuence (as adaptations are a progres-

sive process): the before-diner period happens later

and lasts shorter. This evolution demonstrates the ca-

pacity both to create and to modify habits thanks to

the adaptation of the PP.

Note that, the movement of the watch TV action

implies that other action have moved too. On the sec-

ond long-term diagram (3 (b)), we track the evolution

of the do homework action (red). As the time slot oc-

cupied by the watch TV action in January is free in

February, the action do homework happens more of-

ten during the before-diner period. We observe that

the return of an uniform pricing does not push back

the do homework action: Energy special pricing ef-

fects persists even after its end.

5 CONCLUSIONS AND

CURRENT WORK

In this paper, we presented the model and implemen-

tation of SMACH, a multi-agent simulator of human

activity. This simulator supports the investigation of

activity adaptation and energy consumption evolution

in response to different appliances or pricing policies.

We could not present all the details of this simula-

tor that results from a 5-years collaboration between

the French EDF energy company and AI researchers,

but we outlined its main features: accessibility of the

model to non-computer specialists and intelligent in-

terface for activity analysis and explanation. SMACH

also comprises a machine learning algorithm for auto-

mated management of temperature in the house (Gil-

Quijano and Sabouret, 2010) and an interactive mod-

elling system that helps reﬁning the agent’s behaviour

(Semp

e and Gil-Quijano, 2010).

This model has been evaluated in several exam-

ple scenarios. The two examples presented in this pa-

per illustrated the ability of our model to represent

both the constraints and some of the “degrees of free-

dom” of everyday life. Furthermore, with the help

of the two adaptation processes (actions competition

SimulatingHumanActivitiestoInvestigateHouseholdEnergyConsumption

and habits emergence), agents are able to explore new

organisation and to discover pattern of actions in rela-

tion with time constraints and energy price for saving

purpose. In addition to this evaluation, our simulator

is provided with a participatory-simulation user inter-

face (introduced in (Haradji et al., 2012) ) that allow

to give control of one or several agents to users. Stu-

dents in our lab “played their own role” in our test

scenarios, which allowed us to validate the believabil-

ity of the model. In most situations, students could

not distinguish between artiﬁcial agents and human-

controlled ones.

We are currently extending the SMACH model to

study the activity of groups of families in different en-

vironments, over long period of time (one year) and

taking into account external temperature and build-

ing’s thermodynamical properties. This lead us to re-

consider the action rhythm model and to use multi-

level agent systems for individuals, families and ac-

tivities. Our long-term goal is to allow energy com-

panies to be able to investigate incentive to reduce or

to have a better prediction of consumption peaks us-

ing simulation of human activity.

REFERENCES

Duy Ha, L., Ploix, S., Zama

ı, E., and Jacomino, M. (2006).

A home automation system to improve household en-

ergy control. In Proceedings of The 12th IFAC Sym-

posium INCOM2006.

European Environment Agency (2008). Energy and envi-

ronment report. Technical Report 6.

FIPA consortium (2003). FIPA Communicative Act Library

Speciﬁcation and FIPA ACL Message Structure Spec-

iﬁcation. Technical report.

Freire, R. Z., Oliveira, G. H., and Mendes, N. (2008). Pre-

dictive controllers for thermal comfort optimization

and energy savings. Energy and Buildings, 40(7):1353

– 1365.

Gil-Quijano, J. and Sabouret, N. (2010). Prediction of hu-

mans’ activity for learning the behaviors of electrical

appliances in an intelligent ambiant environment. In

Proc. of the IEEE/WIC/ACM IAT’10 int. conf., pages

283–286.

Haradji, Y., Poizat, G., and Semp

e, F. (2012). Human Activ-

ity and Social Simulation, pages 416–425. CRC Press.

Kashif, A., Ploix, S., Dugdale, J., and Le, X. H. B. (2012).

Simulating the dynamics of occupant behaviour for

power management in residential buildings. Energy

and Buildings, To appear.

Mahdavi, A. and Pr

oglh

of, C. (2009). Toward empirically-

based models of people’s presence and actions in

buildings. In Proceedings of Building Simulation 09,

pages 537–544.

Massoud, A. S. and Wollenberg, B. (2005). Toward a smart

grid: power delivery for the 21st century. Power and

Energy Magazine, IEEE, 3(5):34–41.

Newsham, G. R. and Bowker, B. G. (2010). The effect of

utility time-varying pricing and load control strategies

on residential summer peak electricity use: A review.

Energy Policy, 38(7):3289 – 3296.

North, M., Howe, T., Collier, N., and Vos, J. (2007). A

declarative model assembly infrastructure for veriﬁca-

tion and validation. In Advancing Social Simulation:

The First World Congress, pages 129–140. Springer

Japan.

Poquet, G. and Dujin, A. (2008). Pour les m

enages, la

recherche du confort prime encore sur les

economies

d’

energie. Consommation & Modes de Vie (CRE-

DOC), 210.

Rogers, A., Maleki, S., Ghosh, S., and Nicholas R, J.

(2011). Adaptive home heating control through gaus-

sian process prediction and mathematical program-

ming. In ATES’10, pages 71–78.

Searle, J. (1969). Speech Acts. Cambridge University Press.

Semp

e, F. and Gil-Quijano, J. (2010). Incremental and sit-

uated modeling for multi-agent based simulations. In

Proc. of the IEEE RIVF’10 int. conf., pages 1–6.

Sierhuis, M., Clancey, W. J., van Hoof, R., and de Hoog, R.

(2000). Modeling and simulating human activity. In-

telligent Agents for Computer Supported Co-operative

Work Workshop.

Taillandier, P., Vo, D.-A., Amouroux, E., and Drogoul, A.

(2010). Gama: A simulation platform that integrates

geographical information data, agent-based modeling

and multi-scale control. In PRIMA’10, pages 242–

258.

Wid

en, J. and W

ackelgard, E. (2010). A high-resolution

stochastic model of domestic activity patterns and

electricity demand. Applied Energy, 87(6):1880–

1892.

Wilensky, U. and Evanston, I. (1999). Netlogo. center

for connected learning and computer based modeling.

Technical report, Northwestern University.

Yamaguchi, Y., Fujimoto, T., and Shimoda, Y. (2011). Oc-

cupant Behavior Model For Households To Estimate

High-temporal Resolution Residential Electricity De-

mand Proﬁle. In Proceedings of BS2011, pages 1548–

1555, Sydney, Australia.

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence