Modelling Complexity of Economic System with Multi-Agent Systems

Pavel Čech, Petr Tučník, Vladimír Bureš and Martina Husáková

Faculty of Informatics and Management, University of Hradec Králové, Rokitanského 62, Hradec Králové, Czech Republic

Keywords: Virtual Economy, Complexity, Agent, Multi-Agent System, Self-Organization, Simulation, Modelling.

Abstract: Agent-based computational economics (ACE) is a multidisciplinary area using the agent-based approach for

deeper understanding of economic phenomena occurring in the micro or macro-level. This paper

investigates the application of multi-agent systems for modelling and simulation of virtual economy for

research of self-organizing principles and adaptability of economic subjects. The proposed agent-based

model uses four basic types of autonomous agents. Each one is responsible for crucial activity (consuming,

production, mining, transporting) ensuring existence of the modelled virtual economy. Presented model is

simplified in several aspects, for example banking operations or activities of government are not included in

the model, but the model provides useful basis for the research of economic processes and progress of the

city of Hradec Králové.

1 INTRODUCTION

Complex systems dispose a lot of highly

interconnected components able to influence each

other on the basis of its state or state of the

environment. Interactions between these components

can lead to the occurrence of emergent properties

which are not a part of these individual units.

Complexity is investigated with the aid of different

modelling approaches (Kotzian et al., 2011). Multi-

agent systems are stochastic and decentralized

systems which are used as a bottom-up modelling

paradigm for better understanding of these complex

systems. This approach can be applied in the study

of economic subjects’ behaviour (consumers,

households, companies, congregations or countries)

that can exhibit complex behaviour (Tučník, 2010),

(Čech and Bureš, 2009). Agent-based Computational

Economics (ACE) uses computational tools in the

study of microeconomic or macroeconomic

processes for deeper understanding and effective

influencing of the economic systems. ACE is

defined according to (Tesfatsion, 2006) as: “The

computational study of economic processes modeled

as dynamic systems of interacting agents.“ Agent-

based paradigm is a natural approach for study of

economic processes, because it is able to model and

simulate a large number of interacting entities and

study adaptive or self-organized systems with the

emergence. The economy can be perceived as

complex adaptive and evolving system (Tesfatsion,

1999); (Bruun, 2006).

This paper investigates the application of multi-

agent systems for research self-organizing principles

and adaptability in virtual economy. The proposed

model of virtual economy is intended to be modular

and its complex behaviour should emerge through

mutual interactions of a large number of agents.

These agents use standard economic behaviour to

pursue their goals, such as maximizing profit or

utility, and are able to self-organize into structures

allowing processing of resources in the given

environment and creating production and supply

chains with maximum efficiency as much as

possible. Although the proposed system is in several

aspects simplified – banking sector and government

are not included in the model – the presented model

provides useful basis for the research of adaptation

and self-organisation mechanisms of virtual

economy.

The paper is structured as follows: the section 2

mentions related applications of agent-based

approach in the study of economic systems, the

section 3 describes the proposed conceptual model

of virtual economy, the section 4 introduces the

initial computational model of virtual economy, the

section 5 mentions future research activities and the

section 6 concludes the paper.

464

Cech P., Tu

cník P., Bureš V. and Husáková M..

Modelling Complexity of Economic System with Multi-Agent Systems.

DOI: 10.5220/0004624304640469

In Proceedings of the International Conference on Knowledge Discovery and Information Retrieval and the International Conference on Knowledge

Management and Information Sharing (KMIS-2013), pages 464-469

ISBN: 978-989-8565-75-4

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

2 APPLICATIONS OF AGENTS IN

ECONOMICS

The agent-based modelling and simulations are

applied for solving different problems related to the

economy or management. For instance, Kasbah

platform is developed at the MIT Media Lab for

automation of business processes, e. g. selling,

buying and negotiation are realized only by

autonomous agents. The user specifies the initial

parameters for agents (description of trading

product, prices of products or acceptable price of

products) (Chavez and Maes, 1996). Project

MAGMA proposes the prototype of virtual market

in which buying, selling, usage of bank services and

advertising are taken into account (Tsvetovatyy et

al., 1997). Agents are used also in the VEMMA

model – the virtual electronic marketplace using

mobile agents which are able to negotiate about

products prices. Decision-making of autonomous

agents is based on their preferences and local

parameters (Aklouf and Drias, 2006). An agent-

based architecture for e-business is introduced in

(Jain et al., 2011) where agents sell and buy selected

kinds of fruits. The multi-agent approach can be

used also in modelling adaptive behaviour of firms

which compete with each other on the shared market

(Guessoum et al., 2004). Economic multi-agent

system is used in (Vidal and Durfee, 1998) to

determine when agent should behave strategically

(i.e. learn and use models of other agents), and when

it should act as a simple price-taker. Their results

show how savvy buyers can avoid being cheated by

sellers, how price volatility can be used to

quantitatively predict the benefits of deeper models,

and how specific types of agent populations

influence system behaviour. Damaceanu and

Capraru (2012) focus their attention on banking

market. In their study, they conduct 11 computer

experiments and study the evolution of various

banking market indicators such as total amount of

money, savings, wallets, or bank reserves. Due to its

complexity, Sinha et al., (2011) studied and created

model of petroleum supply chain. Dosi et al., (2008)

develop an evolutionary model of output and

investment dynamics yielding endogenous business

cycles. The model describes an economy composed

of firms and consumers. Whereas firms belong to

two industries, consumers sell their labour and

consume their income. Simulation results show that

the model is able to deliver self-sustaining patterns

of growth characterized by the presence of

endogenous business cycles.

3 CONCEPTUAL MODEL

Conceptual model of virtual economy is proposed in

this paper. The model represents the production and

consumption processes in real economies. The

general aim is to study adaptive and self-organized

principles which are behind the real economy.

Economic principles of effective price and quantity

settings under specific demand and capacity

constraints (Pennings, 2001) are modelled and

simulated with the aid of multi-agent systems. The

focus is on trading products, services and offering

work on a labour market. Virtual economy

simulation is similar to the work of Deguchi, et al.,

(2001), however, in that representation the

considered entities are more specific producing more

complicated net of relations than necessary. Trust

issues as discussed for example in (Gazda et al.,

2012) and similar concerns are not of primary

attention in our virtual economy. The presented

model of virtual economy consists of four types of

autonomous agents:

 Consumer (C-agent);

 Factory (F-agent);

 Mining (M-agent);

 Transport (T-agent).

Banking and government sector are not included in

the model because of the simplicity and clarity of

relations. The basic architecture of the virtual

economy model is depicted in the Figure 1.

Consumer agents are economic entities

consuming products and services (i. e. goods) and

offer work. They are able to buy goods based on

their wealth. The wealth of this agent is a product of

work and qualification (the higher qualification the

faster accumulation of wealth). A consumer agent

makes a trade-off between investment into higher

qualification (e

) and consumption. The

consumption function embodies the combination of

consumed products and the speed of consumption.

The combination of products forms a pattern of

consumption that can be used to divide consumers

into three categories: low income, middle income

and high income consumers. The pattern determines

the ratio of goods that the consumer agent is buying.

Three types of goods are considered in our model:

necessity, normal and luxurious goods.

Factory agent (a company) is responsible for

transforming input to output (i. e. material and other

products to final product that is bought by consumer

agent or sub-product that is used by another factory

agent). The consumption function determines

materials and their proportions. The production

ModellingComplexityofEconomicSystemwithMulti-AgentSystems

465

Figure 1: Conceptual model of proposed virtual economy.

function determines the portfolio of goods produced.

Production requires workforce i.e. employing

consumer agents. The production depends on the

technological level e

and qualification of the

workforce i.e. employed consumer agents e

. The

production equation is as follows:

















































(1)

Let k

con

be the speed of consumption of a material xi

and WF is the workforce; e

is qualification level of

a consumer agent and e

technological level factory

agent; k

pro

to be the speed of production of a product

Mining agent is responsible for transforming

resources into raw material that is used by factory

agents for production of products or services. The

cost of mining is determined by the consumption

function in which the energy and technology

necessary for mining is reflected. Each mining agent

supplies only one type of raw material. Raw

material, as transformed from resources, is stocked

in order to be later sold to transport agents.

Transport agents are intermediaries between

mining agents and factory agents. The cost of

transportation is given by the distance. The task of

transportation agent is to find the most economical

route in case of barriers on the road. The

performance of a transportation agent is determined

by the speed of mobility, capacity and technology.

Transport agent does not have any wealth and buys

material on behalf of a factory agent. The

KMIS2013-InternationalConferenceonKnowledgeManagementandInformationSharing

466

technological advancement of a transportation agent

is also the same as for the factory agent.

Transportation agent is always buying all available

material up to the capacity of transportation.

Transported material that is not used directly in

production is stored in factory agent warehouse.

The proposed model of virtual economy contains

also the representation of a society of agents which

is called “colony” in this context. The colony is a

compact formation of consumer, factory and

transportation agents. Mine agents are not a part of

the colony. They are distributed in the environment,

depending on the resources they process. The colony

has the two basic characteristics: position in

environment and the size of population. Colonies

compete for resources that are supposed to be scarce.

The success of a colony can be measured by

wealth. The wealth of a colony is given by the sum

of wealth of all agents. Due to different colony

populations the comparison among colonies requires

computing wealth per agent. The formula is as

follows:







∑



,



,











(2)

where



,







(3)

4 COMPUTATIONAL MODEL

Initial model of the virtual economy is developed in

the java-based NetLogo modelling platform that is

suitable for modelling complex systems (Brabenec

and Tučník, 2012), see Figure 2. The following

processes are simulated in the NetLogo:

 Raw material extraction by M-agents;

 Receiving raw material by T-agents from M-

agents;

 Transportation of raw material by T-agents into

the colony;

 Detection of obstacles during the transportation

(swamp, forest, water reservoir, sand);

 Production of products by F-agents;

 Consuming of products by C-agents;

 Buying/selling products by C-agents (the lowest

price is accepted).

The user can set up mainly the number of colonies

(1 – 4), size of each of the colony (1 - 100 C-agents),

position of colony, number of M-agents (0 – 5),

terrain (type of obstacle), number (type) of sources,

production chains, number of outputs produced by

F-agents. FIPA ACL extension developed by Ilias

Sakellariou (Sakellariou, 2008) is used for modelling

communication between agents, because the pure

NetLogo tool supports only reactive agents without

specific communication language.

Elementary coordination principles, competitive

behaviour during the sources management,

distribution of sources, living standard or level of

background can be investigated with this multi-

agent-based model. Couple of disadvantages of the

NetLogo were discovered during the model

development. More complex behaviour patterns are

hardly implemented and long-term planning cannot

be adequately included. GUI can be chaotic in case

of a lot of control or monitoring elements (monitors,

buttons, input/output fields, graphs, sliders,

switches). Hence, the different multi-agent

environment is going to be selected for modelling

more complex behaviour of virtual economy. Multi-

method simulation software AnyLogic is considered

for this purpose.

Figure 2: Initial model of virtual economy in NetLogo.

5 FUTURE RESEARCH

Proposed model of virtual economy is simplified.

Capital market, labour market or governmental

sector is not a part of our initial model. Future

research is focused on the development of more

complex model with different strategies of agents

and research of effectiveness of agent´s communities

in different settings. Our aim is to investigate self-

organizing capabilities of the system using models

ModellingComplexityofEconomicSystemwithMulti-AgentSystems

467

of behaviour from economic theory with the

application of the multi-agent approach. The final

model should correspond with the behaviour of the

economic subjects existing in the city of Hradec

Králové and contribute to development of the “smart

city” concept (Mikulecký, 2011). On the basis of the

realized state of the art, no other study aimed at the

modelling the economy of the city of Hradec

Králové was not found. The intended system must

have the following attributes:

 Stability: the system should be able to maintain

itself for as long as possible, ideally for

unlimited time;

 Efficiency: the system should be able to optimize

resource allocation considering production

capabilities and provide capacity for suitable

distribution of products;

 Adaptation and dynamism: the system should be

able to be flexible due to changing conditions of

the environment (i. e. resources reallocation or

reorganization its structure).

Real data are crucial part in multi-agent modelling

and simulation. It is necessary to have model that is

reflective of reality as much as possible. Actual

NetLogo model is not based on the real data. Real

data are going to be used in the next model.

Database of the Czech Statistical Office (CSO,

2013), e. g. data related to the city of Hradec

Králové, is the main source of real data. A lot of

datasheets are freely available.

The economical context in which proposed

system operates is used to maximize the operational

potential of the whole multi-agent system. Economic

notions like “customer satisfaction” or

“maximization of profit” are excellent to define

target parameters for performance of individual

entities in the system. The autonomous entities

(agents) may then adopt different strategies in order

to achieve appropriate level of efficiency in their

actions. It is our intent to achieve desired behaviour

of the whole system by behavioural patterns that will

emerge by interaction of large amount of small

individual agents with each other. The work is

divided into two phases. Case of single multi-agent

community (colony) is investigated in the first

phase. Activities in the first phase are focused on the

efficiency and productiveness of the colony and its

self-organizing capabilities. The second phase is

focused on multiple colonies interacting with each

other within the given environment. There will be

several (or all) resources in the environment present

in a limited amount only. Individual colonies have to

negotiate distribution (or ownership) in this case. It

is assumed that colonies would be encouraged to

specialize in their production, according to given

allocation of resources in the environment. This will

lead to increase of mutual dependency of colonies

on each other and emphasise the need for their

efficient cooperation. Application of scenarios

focused on maximizing performance of colonies

against each other is planned in the final phase of the

project.

6 CONCLUSIONS

Actual state of the art shows that there is a strong

tendency for formal defining of organizational

structures and policies in self-organizing systems.

Multi-agent systems are natural alternative for

modelling complex systems occurring also in the

economic theory.

The initial model of virtual economy is

developed in the NetLogo, but the intended model

have to more sophisticated, modular, adaptive and

based on the real data. Consisting of individual

agents of transparent architecture, the complex

behaviour emerges over time as a result of their

mutual interactions. By defining consumption

patterns for consumer agents, system output could

be modified to produce selected type of goods or

services. This will allow us to study and investigate

wide range of scenarios under changing conditions.

ACKNOWLEDGEMENTS

This paper is created with the financial support of

the specific research project “Cooperation

mechanisms of network organisations” funded by

the University of Hradec Králové, Czech Republic.

REFERENCES

Aklouf, Y., Drias, H., 2006. Designing a Generic

Marketplace Architecture Using Multi-Agent Based

Technology. The International Arab Journal of

Information Technology, vol. 3, no. 3, July 2006, 249-

255.

Brabenec, M., Tučník, P., 2012. Autonomous Systems in

Virtual Economy Environment. University of Hradec

Králové, Faculty of Informatics and Management,

2012. 97 p. Diploma thesis. Supervisor Petr Tučník.

Bruun, Ch., 2006. Agent-Based Computational Economics

– An Introduction. Handbook of Research on Nature-

Inspired Computing for Economics and Management.

KMIS2013-InternationalConferenceonKnowledgeManagementandInformationSharing

468

Ed. Jean-Philippe Rennard, vol. 1 Idea Group

Reference, 2006, 183-197.

Čech, P., Bureš, V., 2009. Advanced Technologies in e-

Tourism. In Proceedings of the 9th WSEAS

International Conference on Applied Computer

Science, Genoa, Italy, pp. 85-92.

CSO (Czech Statistical Office), 2013. Homepage of the

Czech Statistical Office, (the last update: 23. 4. 2013).

http://www.czso.cz/eng/redakce.nsf/i/home (online).

Damaceanu, R.C., Capraru, B.S., 2012. Implementation of

a Multi-Agent Computational Model of Retail

Banking Market Using Netlogo. Metalurgia

International. 17(5), 230-236.

Deguchi, H., Terano, T., Kurumatani, K., Yuzawa, T.,

Hashimoto, S., Matsui, H., Sashima, A., Kaneda, T.,

2001. Virtual Economy Simulation and Gaming - An

Agent Based Approac. New Frontiers in Artificial

Intelligence. 2253, 218-226.

Dosi, G., Fagiolo, G., Roventini, A., 2008. The

microfoundations of business cycles: an evolutionary,

multi-agent model. Journal of Evolutionary

Economics. 18(3-4), 413-432.

Gazda, V., Gróf, M. Horváth, J., Kubák, M., Rosival, T.,

2012. Agent based model of a simple economy.

Journal of Economic Interaction and Coordination.

7(2), 209-221.

Guessoum, Z., Rejeb, L., Durand, R., 2004. Using

adaptive multi-agent systems to simulate economic

models. In Proceedings of the Third International

Joint Conference on Autonomous Agents and

Multiagent Systems (AAMAS). New York, USA, pp.

68–75.

Chavez, A., Maes, P., 1996. Kasbah: An Agent

Marketplace for Buying and Selling Goods. In

Proceedings of the First International Conference on

the Practical Application of Intelligent Agents and

Multi-Agent Technology. London, UK, pp. 75-90.

Jain, M. B., Rao, M. V., Shukla, P., 2011. An Agent Based

Architecture for E-Business Application with Multi

Agent Systems. International Journal of Advanced

Engineering and Application, Jan 2011, 205-209.

Kotzian, J., Konečný, J., Krejcar, O., 2011. User

Perspective Adaptation Enhancement Using

Autonomous Mobile Devices. Lecture Notes in

Artificial Intelligence, 6592, 462-471.

Mikulecký, P., 2011. Learning in Smart Environments -

From Here to There. In Proceedings of the 10th

European Conference on e-Learning. Brighton, United

Kingdom, pp. 479-485.

Pennings, E., 2001. Price or quantity setting under

uncertain demand and capacity constraints: An

examination of the profits. Journal of Economics,

74(2), 157-171.

Sakellariou, I., 2008. Extending Netlogo with BDI and

FIPA ACL Support (software project).

http://users.uom.gr/~iliass/projects/NetLogo/index.htm

l (online).

Sinha, A. K., Aditya, H. K., Tiwari, M. K., Chan, F. T. S.,

2011. Agent oriented petroleum supply chain

coordination: Co-evolutionary Particle Swarm

Optimization based approach. Expert Systems with

Applications, 38(5), 6132-6145.

Tesfatsion, L., 1999. Agent-based computational

economics: A constructive approach to economic

theory. In Tesfatsion and Judd, 2006, chapter 16,

pages 831 – 880.

Tsvetovatyy, M., Gini, M., Mobasher, B., Wieckowski, Z.,

1997. MAGMA: An Agent-Based Virtual Market for

Electronic Commerce. Journal of Applied Artificial

Intelligence. 11(6), 501-523.

Tučník, P., 2010. Multicriterial Decision Making in

Multiagent Systems - Limitations and Advantages of

State Representation of Behavior. In Proceedings of

the 9th WSEAS International Conference on Data

Networks, Communications, Computers, Faro,

Portugal, pp. 105-110.

Vidal, J. M., Durfee, E. H., 1998. Learning nested agent

models in an information economy. Journal of

Experimental & Theoretical Artificial Intelligence.

10(3), 291-308.

ModellingComplexityofEconomicSystemwithMulti-AgentSystems

469