A MULTI-AGENT SYSTEM FOR INTELLIGENT

BUILDING CONTROL

Norm Approach

Jarunee Duangsuwan and Kecheng Liu

Informatics Research Centre, The University of Reading, Building 4, Whiteknights, PO Box 241, Reading, RG6 6WB, U.K.

Keywords: EDA model, Norm aware agents, Intelligent agents, Intelligent buildings, Multi-agent systems, Norms.

Abstract: Most previous research in the intelligent buildings have proposed the controlling systems that can change

building environmental conditions automatically in order to save energy consumption and also to increase

an occupant’s satisfaction. Decreasing energy consumption and increasing occupant comfort are important

factors to indicate an intelligent building’s performance because it is a particular way to improve

productivity resulting in the business benefits. By applying agent technology, an intelligent building control

system provides a practical application that can minimize energy consumption levels, while keeping a

satisfying response to an occupant’s comfort. This paper proposes an abstract extended-EDA (Epistemic-

Deontic-Axiologic) model which is enhanced capability in order to make decision under norms: obligations,

permissions and prohibitions. The model is represented in terms of an individual agent that is prepared for

the multi-agent system of intelligent building control. The multi-agent system is proposed to combine the

comfort condition control with an energy saving strategy.

1 INTRODUCTION

According to a definition of intelligent building

(Himanen 2003), an intelligent building is a building

equipped with an integration of advanced

technology, especially the computer-based artefacts

and systems, in order to support automatic

adaptation to the changing environment conditions,

and to provide comfortable living conditions for the

current occupants as well. Most previous research in

the intelligent building control systems have been

is software that continually processes the input it

gets from its environment to determine the output it

should send back to the environment. Furthermore,

an intelligent agent has the following characteristics:

reactivity, pro-activity and social ability. Therefore,

such characteristics make an intelligent agent has

capable of autonomous actions in the environment in

order to meet its goals. A multi-agent system

(D'Inverno and Luck 2004) comprises at least two

agents. The agents in the multi-agent system can

interact with others so that this interaction must

result from one agent satisfying the goals of another.

The projects contributed by Magnus Boman et al.

(Boman, Davidsson et al. 1998; Boman, Davidsson

et al. 1999; Davidsson and Boman 2000; Davidsson

and Boman 2005) have been implemented in the

multi-agent system approach to implement building

control system. Such projects have the main goal to

increase energy saving and to meet customer

preference by automatic control of lighting and

temperature according to occupant’s requirement.

The multi-agent system composes of four categories

of agent. Personal comfort agents record the

personal preferences and try to increase occupant’s

satisfaction. Room agents represent and control a

particular room to maximize energy saving and to

make occupant feel satisfy at the same time.

Environmental parameter agents monitor and control

the environmental parameters in a particular room.

Badge system agents keep a track of location in the

building where the occupants are situated.These

agents are the BDI agents that are based on the

theory of practical reasoning stating that the agent’s

goals drive the agent’s behaviours. However,the

multi-agent system lacks the capability of learning

and predicting the occupants’ behaviors.

For intelligent building approach, an important

issue concerns with the energy saving policies.

Although, the multi-agent system for intelligent

building control proposed in our research is designed

Duangsuwan J. and Liu K. (2010).

A MULTI-AGENT SYSTEM FOR INTELLIGENT BUILDING CONTROL - Norm Approach.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Agents, pages 22-29

DOI: 10.5220/0002720000220029

 SciTePress

to adapt the environment factors such as

temperature, lighting, humidity according to the

preferences of occupant. However, such factors are

set under optimal energy. For a simple case, when a

private area such as office room is occupied by a

person, the environmental conditions should be

adjusted according to the current occupant’s

satisfaction. Furthermore, in a multi-occupant

scenario, making decision about the conditions is

more complex than a single-occupant scenario

because the conditions should be set by the certain

values in order to make most occupants in the room

feel comfortable as much as possible. Basically, by

using the average values to set the environment

conditions is the simple and reasonable method.

However, for our research it has been conducting on

a particular agent model that can make decision via

normative reasoning, and can be membership of

multi-agent system supporting the building control

system in order to provide both condition

recommendations for occupants’ comfort conditions

and saving energy conditions. We decide to enhance

the EDA model proposed by (Filipe 2000) to

construct a multi-agent system for controlling an

intelligent building because the model allows agent

to make decision under normative consideration.

The EDA model has been contributed by

combination between norms and corresponding

attitudes for supporting the organizational semiotics

approach. By the original EDA model, it was

proposed for normative reasoning in business

domain, and most agents were referred to human-

agent. However, the agents in our research domain

are both human-representing agents such as

occupant agents and artefact-representing agents

such as zone agents then the traditional EDA model

has been adjusted to support our research domain.

The following section begins by reviewing

background of relevant literature on BDI agent

model, norms, and normative agent. The framework

overview of multi-agent system for intelligent

building control in a single-occupant scenario, and

an extended-EDA model are represented in section

3. The final section provides a conclusion and the

future of our work.

2 RELATED STUDIES

2.1 BDI Model

The BDI model (Wooldridge and Jennings 1995)

proposed by Rao and Georgeff in 1991 is an agent

model that emphasizes an intentional notion of

agency. The original of this model is developed by

Michael E. Bratman via a theory of human practical

reasoning stating that an agent’s behaviour is driven

by its goals. According to (Jarvi 2004), the practical

reasoning is reasoning directed towards actions, so it

involves to the process of determining what to do.

By contrast, the theoretical reasoning most directly

affects beliefs. Therefore, the BDI model is behaved

in the same way like human practical reasoning by

adopting mental attitudes of Belief (B), Desire (D)

and Intention (I), respectively representing the

information, motivation, and deliberative states of

the agent. Beliefs represent the information of the

agent about the world, in other words its beliefs

about the world. Desires of agent refer to the

motivational state of the agent or may be thought of

as the task allocated to the agent. Furthermore, the

desires represent objectives or situations that the

agent would like to accomplish or bring about.

However, all of agent’s desires may not be achieved.

The agent’s intentions represent desires which the

agent has chosen and committed to.

2.2 Norms

Norms have been used in several words such as

‘pattern’, ‘standard’, ‘type’, but the meaning of

norms is been defined unclear because it is used in

many different senses (Wright 1963). Wright

categorized norms into three main types: rules,

prescriptions, and directives. Rules, e.g. rules of a

game, rules of grammar, are the explicit

standardized patterns then the rules can determine

which are right or wrong. Prescriptions or

regulations, e.g. military commands, traffic rules, are

commands or permissions that are issued by a norm-

authority to a norm subject. For directives or

technical norms, there are concerned with the means

that is used for attaining a certain end. ‘Directions of

use’ is an example of directives. According to

(Stamper, Liu et al. 2000), a norm is a field of force

that has been used to govern the behaviors of the

members in a society. Stamper et al. divide norms

according to social psychology classification into

perceptual, evaluative, cognitive and behavioural

norms. These four types of norms are respectively

associated with four distinct attitudes: ontological,

axiological, epistemic, and deontic. Besides, these

norms are elaborately outlined in (Liu 2000).

Perceptual norms concern with the ways of seeing

the world, the ways to receive signals from the

environment via human being’s senses through

media such as light, sound and taste. Cognitive

norms can help to incorporate the beliefs and

A MULTI-AGENT SYSTEM FOR INTELLIGENT BUILDING CONTROL - Norm Approach

knowledge of a culture, to interpret what is

perceived, and to obtain an understanding based on

existing knowledge. Evaluative norms are required

to evaluate the state of affairs, and also to explain

why people have certain beliefs, values and

objectives. The evaluation may be based not only on

the physical boundary but also on a social boundary.

Behavioural norms determine human being’s

behaviours within regular patterns. These norms

determine how agent should behave under given

certain conditions, and define what an agent should

perform to achieve a state of affairs under given

certain conditions. Denotative norms direct the

choices of signs for signifying, these depend on the

culture that may be a community or a social group.

According to norm classification by Wright, the

building policies are categorized as the prescriptions

or regulations issued by norm authority such as a

building owner, a local council, a government and so

forth. The regulations must be complied with norm

subject such as the occupants, the facility

administrators, the building managers, and so on in

order to control energy consumption. Usually, the

policies are expressed as pre-defined rules so these

rules are static and can represent as following simple

form:

IF nobody is present in the room

THEN the lighting and heating should be turned off

IF a room is the public area AND at least one person

appears in the area

THEN lighting is turned on AND the temperature is

set to 20 degree Celsius

Furthermore, the character, which is a

component of prescription, expresses the effects of

norm so the policies can be prohibited, permitted or

obligated. Therefore, the examples of rule

represented in table1 are declared to identify these

effects.

Table1: The examples of rule.

Category example

Obligation IF An occupant appears in a room

AND he is a room owner

THEN The BMS is obligated to set the

current room conditions according

to the current person’s preferences

Prohibition IF Nobody in present in the room

THEN The lighting and heating are

forbidden to turned on

Permission IF Time of weekdays is 6pm AND

room is occupied

THEN The room temperature are permitted

to set 20 Celsius

In addition, we distinguish the policies into four

priority levels:

Safety: to ensure that the environmental

conditions are set at the safe level, for example the

temperature of the building should be set at a

particular level in the winter to protect the water

pipes from freezing and cracking.

Security: to ensure that the environmental

conditions keep at a security level, for example in

case of an emergency circumstance such as fire

alarm, the emergency door should be opened.

Energy saving: to ensure that a low energy

consumption is achieved by dynamical controlling

so that if a room is unoccupied, the heating and the

lighting are switched off.

Comfort: to ensure that the room conditions are

set according to the occupants’ preferences

For the first two levels, we assume that the

controlling system is set to keep the both levels all

time then the later two levels are under our

considerations. Although the building policies are

defined to control energy consumption, the most

important aspect of an intelligent building is to take

consideration about personal preferences of the

people living or working in the building. The

personal preferences represented as the dynamic

rules provide comfort conditions that obtain from

behaviour learning of occupant in a particular room.

However, the goals of the building policies and the

personal preferences are conflict. In addition,

another conflicting goal situation is the adjustment

of environmental conditions in a shared area such as

a meeting room, a common room where the

occupants have the different preferences. For the

first conflict, we provide an agent that acts as a

negotiation agent to solve the conflict. The agent

determines whether the energy saving policy is not

affect the comfort of the current occupants, partially

affect the current occupant’s comfort conditions but

with in the limit of occupant’s preference thresholds,

or fully affect the current occupant’s comfort

conditions which means the current conditions

beyond the limit so it can be summarized as follows:

1) If the occupant’s preferences are not affected

then the energy saving policies are applied to set the

currently environmental conditions

(energy saving level).

2) If the occupant’s preferences are partially

affected then the standard conditions are used to set

the currently environmental conditions

(default

conditions set by BMS supporting both energy saving and

occupant’s well being).

3) If the occupant’s preferences are fully affected

then the currently environmental conditions are

adjusted according to the occupant’s preferences

(comfort level).

In case of a shared area, the conflict occurs

whenever at least two persons present in the shared

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

area. Therefore, we adopt norm approach, which is a

partial factor of decision making about the currently

environmental conditions, to reconcile contending

preferences from the different occupants. We

employ norms in two main tasks: 1) to assign the

priority to a particular occupant 2) to assign the

weight to normative goals and individual goals.

Whenever a room is occupied by more than one

person, we claim that a human being society, either

formal or informal, is set up, and leads to assigning

of role for each person in the society. For instance,

in a private area, such as an office room scenario, an

informal society is formed when anybody comes

into the room then the role of this scenario is

categorized into owner and visitor(s). On the other

hand, in a common area such as a meeting room, the

members of conference compose with a chairman, a

secretary, a president, a board of director etc. These

roles are considered as a formal role because it is

defined by the organizational structure in terms of a

hierarchical relationship. In a common room where

anyone takes a break for drinking, eating, relaxing

etc., an informal society occurs. Therefore, the

society is a lateral or peer-to-peer relationship. By

using a role concept, it will be adopted to classify

the different priority between particular occupants.

The high the occupant presents in the level of a

hierarchical relationship, the high the occupant gets

priority in the public area. Therefore, a relative

ordering on the values of occupants’ preferences is

created.

Norms represent what the members in society

ought to do, and their fulfilments can be seen as a

benefit of the overall system. However, in some

circumstances, individual goals conflict with the

norms. In a multi-occupant situation, each occupant

has the personal preferences that can be seen as the

individual desires. By contrast, the normative goals

are the certain environmental conditions that make

all occupants as comfort as possible. Therefore, two

steps of conflicting resolutions are needed. The first

step is to resolve conflict among the occupants for

finding the optimal preferences the make most

occupants feel comfortable. The optimal preferences

are evaluated by the system whether they fall in

which level: saving energy level or comfort level. If

the result falls in saving energy level then the second

step is not happen because the individual goals are

not conflict with the normative goals. In contrast, the

second step is initiated to resolve the conflict. In our

research, we adopt negotiation to reconcile the

conflict then a goal of negotiation is to maximize

occupant comfort and minimize energy

consumption.

2.3 The BDI Model and Norms

Norms, claimed by (Torre 2001), are used for

linking the gap between an agent level and a multi-

agent system level. This means that a role of norms

and obligations can support an agent society so

many previous researchers tried to enhance an

agent’s ability by proposing a novel agent model

that can make decision under norms: obligations,

permissions and prohibitions. Although BDI is the

most widely known model that is used to implement

an agent for individual and intentional decision

processes, this model was not been represented

nothing about the social aspects of agent being in the

multi-agent systems. However, the BDI model has

some limitation such as lacking of policy and norm

supporting ability then many researchers have

proposed the extended-BDI models for example;

extension of BDI model with norm (Dignum 1999),

Belief-Obligation-Intention-Desire (BOID) model

(Broersen, Dastani et al. 2001) ,normative agent

architecture (Lopez and Marquez 2004), EDA model

(Filipe 2000), and so on. These extending

architectures are the normative model addressing the

usage of norms and policies for reasoning and social

interacting. An agent that can reason about norms

and obligations is called a normative agent

(Verhagen 2000) or a deliberate normative agent

(Castelfranchi, Dignum et al. 1999). The deliberate

normative agent has explicit knowledge of the

enacted norms for reasoning, and can make a

decision whether to comply with norms or not in

some cases.

Broerson et al. (Broersen, Dastani et al. 2001)

have proposed BOID architecture that composes

with four basic components. The architecture

focuses on conflict resolution among informational

and motivational attitudes. The possible conflict is

classified into two types: internal and external

conflicts. The internal conflicts occur within beliefs,

obligations, intensions, and desires, and can be

categorized into four unary subtypes: B, O, I, D. The

external conflicts occurring between these

components can be distinguished into multi

subtypes: BO, BI, BD, OI, OD, ID, BOI, BOD, BID,

OID, BOID. The conflict resolution mechanism is

based on Thomason’s idea of prioritization which

can determine the type of an agent simultaneously.

The BOID architecture is discussed more detail in

(Broersen, Dastani et al. 2001). Lopez et al. (Lopez,

Luck et al. 2001; Lopez and Marquez 2004) have

proposed a framework representing the adoption of

norms towards the BDI agent. The authors present

an abstract normative agent architecture designed by

A MULTI-AGENT SYSTEM FOR INTELLIGENT BUILDING CONTROL - Norm Approach

merging the BDI agent architecture to three

components that can make the agent reason about

norms. Furthermore, the processes are included:

norm adoption, norm deliberation, and norm

compliance. Norm adoption is a process that decides

whether norms must be adopted or not. If the norms

should be adopted, then they can be represented in

terms of norm instances. Norm deliberation decides

which intended norms are fulfilled or unfulfilled by

agents. Norm compliance process applies both

intended norms and rejected norms to update the

agent’s goals which are affected by norms. After the

goals are updated, the intentions of agent might be

changed. The outcomes from those three processes

have proposed as the three mental attitudes that are

norm instances, intended norms, and rejected norms.

Besides, motivation is a component which agents

employ for assigning the preference values over

their goals. However, these components of norms

must be partly considered by the agents when a

normative decision must be taken at that time.

3 MULTI-AGENT SYSTEM FOR

INTELLIGENT BUILDING

CONTROL

3.1 Agents in the Multi-agent System

Due to the nature of intelligent buildings, when a

person is present in a room, the building control

system is initialized by setting the room conditions

via a standard set of preferences. In case of the

existing occupant who is recognized by the system,

the environmental conditions are adjusted according

to the occupant’s preferences retrieved from a

repository of the system. However, an occupant can

change the currently environmental conditions to

desired conditions when she/he feels dissatisfaction.

The agent who observes the occupant behaviours

and lifestyle must be provided for learning occupant

preferences and anticipating what the occupants

want. Although the system allows the occupants to

change the environmental conditions, the conditions

are controlled under the system’s goals that support

both energy conservation and occupant comfort. To

decrease energy consumption without affecting the

comfort of the building occupants, it is necessary to

determine an occupant breakdown of the energy

consumption in the building. In our research, a

multi-agent system used to implement the

controlling system consists of a collection of agents

that monitors and controls the building. The system

is situated in some environment, and that is capable

of autonomous action in the environment in order to

achieve its objectives. The following agents have

been proposed and will be implemented in our

research: 1) Occupant agent corresponds to a

particular occupant in a multi-agent system. The

occupant agent presents some personalities or

characters, monitors and adapts to the user's

activities, learns the user's styles and preferences.

The agent aims to maximize user’s preferences by

learning these preferences from observing user’s

behaviours. The occupant agent can reside on the

various tools for example; a personal computer, a

badge, a mobile phone, a radio frequency

identification (RFID) technology and so on. 2) Zone

agent corresponds to and controls a particular zone.

Basically, there are many sensors embedded in each

zone such as temperature sensors, lighting sensors,

blind sensors etc. In addition, the zone agent acts as

a negotiating agent to reconcile the conflict between

its goals and the occupant agent’s goals. We define

the smallest logical unit in the building as a single

room so a zone is a single room, or composes with

more than one room. 3) Manager agent directly

interfaces to building management system (BMS) by

sending the final decision for governing the

location’s environment to BMS. 4) Environmental

control agent monitors and controls different

environmental parameters in each zone. For

example, a temperature agent can read the

temperature sensor, and can control the actuator in a

zone.

3.2 Single-occupant Scenario

Normally, the preferences of a particular occupant

are set when the occupant enters the room at the first

time. As illustrated in figure1, when an occupant

signs into the system, an occupant agent

corresponding to the occupant provides a zone agent

with the personal information and preferences. The

zone agent decides the new conditions under

building policies and occupant’s comfort. Hence,

these conditions are passed to a manager agent for

changing to BMS-format commands. The manager

agent sends these commands to the BMS to enforce

these commands requested by the multi-agent

system. However, if an occupant does not change

the environments, for example by changing the

temperature or by adjusting the heating, a system

assumes that the current environments are

comfortable for the occupant. By contrast, as

represented by figure2, if an occupant changes the

current conditions by using the occupant agent, the

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

new changed conditions are observed and learned by

the occupant agent. The occupant agent calculates

the new preferences that will forward to the zone

agent for evaluation whether the new preferences

conflict with the building policies or not. If the

conflict occurs; for example, the occupant agent A

desires to set temperature at 20°C. The A’s desire

may conflict with a normative goal of the zone agent

Z which set the temperature at 25°C for energy

conservation (building policy); then the zone agent

reconciles the conflict. Subsequently, a result of

reconciling is sent to the occupant agent for updating

the occupant’s preferences, and is sent to the

environmental control agents to adjust the

environmental conditions.

Figure 1: Sequence diagram of multi-agent system for IB

in a single occupant scenario.

Figure 2: Sequence diagram of multi-agent system for IB

in a single occupant and learning scenario.

For a multi-occupant scenario described in section

2.2, a conflict of preference may or may not occur.

According to the capability of an intelligent building

control system, the environmental conditions of a

building are set at the optimal comfort conditions.

We assume that if the occupants do not take any

actions, they prefer the current conditions then a

conflict does not happen. However, if someone

requests to change the current conditions,

negotiation among occupants is needed. We adopt

norms to resolve the conflict occurring between the

normative goals and the individual goals. A

particular occupant in a multi-occupant scenario has

the individual preferences then a preference conflict

among the occupants may happen. Therefore, the

negotiation that will be implemented by the

blackboard concept is used to reconcile the

occupants’ conditions. We implement the

negotiation between the occupants by using

blackboard system that can be compared with

technologies such as message queues and databases

where a simple API for the manipulation of entry

objects: insert, read, delete can be used when some

type of object is available at (Creswell

http://www.dancres.org/cottage/javaspaces.html).

The preliminary negotiating design has been

proposed in (Duangsuwan and Liu 2008) but in this

paper we pay attention to how to use norm approach

reaches a joint preference of the occupants. To

design agent architecture for negotiating and making

decision under norm consideration, we were inspired

by the EDA model so an extended-EDA model has

been proposed to enhance the EDA model.

3.3 The Extended-EDA Model

The concept of EDA model has been contributed by

combination between norms and corresponding

attitudes. The main components of model are

epistemic component (E-component), deontic

component (D-component), and axiologic

component (A-component). Furthermore, two

external components are included: a perception

interface obtaining and interpreting external events

from the environment, and an action interface

sending the output actions to the environment. The

EDA model has its own beliefs represented in E-

component that contains current beliefs or facts

about the world. The obligations, rights and

behaviours of agent are set in D-component where a

set of plans is declared in terms of the interesting

behaviours of agent. A-component is an evaluating

component for assigning a preference relationship

among the available plans in D-component. The

component provides a dynamically value-setting

method for agent in order to assign the importance

of norms. Therefore, the constituted obligations are

assessed through axiology then the committed

intentions are established.

The EDA agent is allowed to reason about norms

then the components and processes involving norms

are included into the extended-EDA model. From

our perspective, norms are the external forces that

A MULTI-AGENT SYSTEM FOR INTELLIGENT BUILDING CONTROL - Norm Approach

might influence the agent capabilities or behaviours.

Initially, an agent may hold a set of capabilities and

certain desires to deploy these capabilities but by

adopting norms, the agent’s capabilities are partly

determined by obligations which the agent must

comply with, prohibitions that restrict some kind of

capabilities which the agent can pursue, or

permissions that expand the capabilities for the agent

to choose. A multi-agent system forms an agent

society. Whenever an agent joins a society or an

organization, it will be bound to certain rules and

regulations. The adoption of norms will specify

agent with a specific social position or role. This

role is annotated with certain duties, privileges,

authority, responsibility etc. Therefore, the agent’s

behaviours are behaved according to the adopted

norms ascribed to an agent’s role. The building

regulations and policies are presented as the external

forces pushed by the building owners or

administrators to save commercial cost, or pushed

by the government to encourage efficiency of energy

consumption.

Figure 3: An abstract extended-EDA model.

We propose an abstract extended-EDA model

including the new components for norm reasoning.

Figure 3 illustrates the proposed model prepared for

a multi-agent system to control the intelligent

buildings. The figure represents the internal

components and processes of a particular agent

which is the member of a multi-agent system. The

new two components, relevant norms and selected

norms, are added and merged to the EDA model. In

addition, the processes are depicted in order to

present data and control flow among the

components. Similar to other agent models,

changing of the environment is observed by the

sensors and is sent towards the system for updating

the beliefs in E-component via belief revision

process. E-component stores beliefs or facts that

correspond to the information that an agent have

about the world. The first process involving norms is

norm identification that responses to verify that

which norms in a norm base defined in advance

relate to the current beliefs held by an agent and may

be adopted for an agent’s reasoning later. A set of

possibly relevant norms is an outcome that will be

sent towards the next process. Norm adoption

process decides whether which norms must be

adopted to normative decision. The result of this

process is represented in terms of instantiate norms.

D-component represents goals, objectives, or any

states of affairs which an agent want to bring about.

According to the EDA model concept, D-component

is where the interesting behaviours of agent are

defined. The behaviours may be represented as the

partial plans at different levels. A goal is a very high

abstract plan, whereas a sequence of elementary

actions defines a plan at the instance level.

However, goal generation has one more step further

because it creates the goals based on norms. By

merging goals from goal generation process to the

selected norms, it can decide whether what goals are

brought about by an agent under the currently

selected norms. All candidate goals, which agent has

to bring about, will be kept in D-component. The

partial plans for achieving these goals are pre-

defined in a plan library which is a repository of all

plans that an agent knows. Once one of these plans

is adopted for execution, it is considered an intention

and an agent is committed to do it.

To make decision when goal conflicting occurs,

the agent’s goals are associated with preferences

because they are used to make decision in our model

not only to choose the goals for pursuing, but also to

decide the goals for preferring and achieving first.

The EDA is modelled to support this scenario by

proposing the A-component. The generalized goals

in D-component need to be assigned the value of

preference by preference evaluation process. An

outcome from this process is a preference set which

is different for each agent, thus an agent shows their

individual preferences towards the particular goals.

We represent the relevance between a preference set

and the particular goals by using the relationship that

will be created and stored in A-component. Because

decision making of the EDA agent depends both on

the available goals in D-component and a preference

relationship in A-component, both components are

combined to choose a goal. When a goal is chosen, it

becomes an intention. A-component concerns with

deciding whether which goals are placed in agenda.

The goals in the agenda become intentions of agent.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

The intentions are executed, and then the actions are

sent out to the environment.

4 CONCLUSIONS

We have designed a framework of multi-agent

system that controls the environmental conditions of

an intelligent building. A particular agent in the

multi-agent system is modelled under an extended-

EDA architecture enhanced the capabilities to

support normative decision making. Different from

the other systems, we have included norm concept

that promotes an increased flexibility towards the

policies and the preferences of occupants in the

building. Therefore, the multi-agent system in our

research gives a good support for extensions and

adaptations in the building’s policies that used to

control the energy consumption, and also makes the

building’s occupants feel comfort as much as

possible by using the ordering preferences supported

by role aspect to set the environment conditions of a

particular area. In the future, we will deploy our

extending model to test our framework that will be

implemented by eclipse software to build an agent-

based prototype for verifying the proposed model.

REFERENCES

Boman, M., P. Davidsson, et al., Eds. (1998). Energy

Saving and Added Customer Value in Intelligent

Buildings. Proceedings of the 3rd International

Conference on the Practical Applications of Agents

and Multi-Agent Systems (PAAM-98).

Boman, M., P. Davidsson, et al. (1999). Artificial

Decision Making Under Uncertainty in Intelligent

Buildings. the Fifteenth Conference on Uncertainty in

Artificial Intelligence, Stockholm, Sweden.

Broersen, J., M. Dastani, et al. (2001). The BOID

Architecture - Conflicts Between Beliefs, Obligations,

Intentions and Desires. In Proceedings of the Fifth

International Conference on Autonomous Agents,

ACM Press: 9-16.

Broersen, J., M. Dastani, et al. (2001). An alternative

classification of agent types based on BOID conflict

resolution. Proceedings of the 13th Belgium-

Netherlands Artificial Intelligence Conference: 79-87.

Castelfranchi, C., F. Dignum, et al. (1999). Deliberate

normative agents: principles and architecture. In

Proceedings of the Sixth International Workshop on

Agent Theories, Architectures, and Languages

(ATAL-99), Springer Verlag: 364-378.

Creswell,D.(http://www.dancres.org/cottage/javaspaces.ht

ml). "Getting started with JINI TM 2.x." Retrieved

20 May 2009.

D'Inverno, M. and M. Luck (2004). Understanding agent

systems, Springer.

Davidsson, P. and M. Boman (2000). A Multi-Agent

System for Controlling Intelligent Buildings.

Proceeding of the Fourth International Conference on

MultiAgent Systems. Boston, MA: 377-378.

Davidsson, P. and M. Boman (2005). "Distributed

monitoring and control of office buildings by

embedded agents." Information Sciences 171(4): 293-

307.

Dignum, F. (1999). "Autonomous Agents with Norms "

Artificial Intelligence and Law 7: 69-79.

Duangsuwan, J. and K. Liu (2008). Multi-agent Control of

Shared Zones in Intelligent Buildings. Computer

Science and Software Engineering, 2008 International

Conference on.

Filipe, J. B. L. (2000). Normative Organisational

Modelling Using Intelligent Multi-Agent Systems.

Staffordshire, Staffordshire University.

Himanen, M. (2003). The Intelligence of Intelligent

Buildings : The Feasibility of the Intelligent Building

Concept in Office Buildings. Department of

Surveying. Helsinki, Finland, Helsinki University of

Technology: 498.

Jarvi, J. (2004). BDI & Reasoning. http://www.cs.uta.fi/

sat/lectures/lecture-28-02/sat-lecture-28-02.pdf

Liu, K. (2000). Semiotics in Information Systems

Engineering, Cambridge University Press.

Lopez, F. L., M. Luck, et al., Eds. (2001). A framework

for norm-based inter-agent dependence. Proceedings

of the 3rd Mexican International Conference on

Computer Science.

Lopez, F. L. and A. A. Marquez (2004). An architecture

for autonomous normative agents. Computer Science,

2004. ENC 2004. Proceedings of the Fifth Mexican

International Conference in.

Stamper, R., K. Liu, et al. (2000). "Understading the roles

of signs and norms in ogranizations : A semiotic

approch to infomation systems design." Journal of

Behaviour & Information technolgoy 19: 15-27.

Torre, L. v. d. (2001). "Contextual Deontic Logic:

Normative Agents, Violations and Independence."

Annals of Mathematics and Artificial Intelligence 37:

33-63.

Verhagen, H. J. E. (2000). Norm Autonomous Agents.

Department of Computer and Systems Sciences.

Stockholm, Sweden, The Royal Institute of

Technology and Stockholm University.

Wooldridge, M. J. and N. R. Jennings (1995). Intelligent

agents: Theories, Architectures, and Languages.

Germany, Springer-Verlag.

Wright, G. H. v. (1963). Norm and action - A logical

enquiry. London, Routledge & Kengan Paul.

A MULTI-AGENT SYSTEM FOR INTELLIGENT BUILDING CONTROL - Norm Approach