Collection Mémoires et thèses électroniques
Accueil À propos Nous joindre

CHAPTER 1: General Introduction

Table des matières

This chapter introduces the context of our research which is related to the multiagent based simulation field, and more precisely multiagent geosimulation. It identifies the motivations, the problems, and the research questions that we address in this dissertation. It also presents our hypothesis, objectives, and the research methodology of our work. Finally, it presents the organization of the rest of this thesis.

This thesis is about multiagent geosimulation ( MAGS ), which is a sub-field of multiagent-based simulation ( MABS ). It is widely recognized that MABS is one of the main topics of the multiagent systems (MAS) domain. MABS differs from other kinds of traditional computer-based simulations such as discrete event simulation ( DES ), continuous event simulation ( CES ), and object oriented simulation ( OOS ), in that (some of) the simulated entities are modeled and implemented in terms of agents (Davidson, 2000). The agents’ capabilities (e.g., autonomy, social ability, reactivity, pro-activeness, etc.) make MABS more attractive than traditional simulation approaches. In the literature, several applications have been created using the MABS paradigm in order to simulate various kinds of systems/behaviors in different areas/domains. In this dissertation, we concentrate our research on the use of MABS to build simulations of human behaviors in virtual geographic environments[1]. The simulation of human behaviors in space is a very interesting and powerful research method to advance our understanding of human spatial cognition and the interaction of human beings with the environment (Frank et al., 2001). Several researchers used the MABS paradigm to simulate human behaviors in geographic environments. For example, (Raubal, 2001) and (Frank et al., 2001) presented an application which simulates human wayfinding behavior in an airport. (Dijkstra et al., 2001) and (Timmermans et al., 2001) presented an application which simulates pedestrian movements in a mall. (Koch, 2001) simulated people’s movements in a large-scale environment representing a town. (Moulin et al., 2003) simulated, in a geographic environment representing a part of Quebec City, crowd movements using thousands of virtual agents. Although these applications emphasize the spatial features of the simulation environment (SE), they are distinct in the ways they represent them. Some applications use cellular automata (CA) to represent the environment, while other approaches use geographic information systems (GIS). The large number of simulation applications that emphasize the spatial features of the environment gave birth to other simulation sub-fields such as spatial simulation and urban simulation (Benenson and Torrens, 2004). Recently, a new form of simulation, called geosimulation , became popular in geography and the social sciences. Geosimulation is a useful tool for integrating the spatial dimension in models of interactions of different types: economic, political, social, etc. (Mandl, 2000). This form is supported by advances, both in the geographical sciences and in fields outside geography (Benenson and Torrens, 2003). (Mandl, 2001) and (Moulin et al., 2003) present multiagent geosimulation (MAGS) as a coupling of two technologies: the MABS technology and that of GIS. Based on the MABS technology, the simulated entities are represented by software agents that autonomously carry out their activities. What’s more, they can interact and communicate with other agents, and they may be active, reactive, mobile, social or cognitive (Koch, 2001). Using the GIS technology, spatial features of geographic data can be introduced into the simulation. The GIS plays an important role in the development of geosimulation models. New methodologies to manipulate and interpret spatial data developed by geographic information science and implemented in GIS. They have created added-value for these data (Benenson and Torrens, 2003). Progress in the multiagent and GIS fields makes multiagent geosimulation a promising paradigm which can be used to simulate complex systems and behaviors in geographic environments.

In summary, our research context consists in the simulation of human behaviors in geographic environments using the multiagent geosimulation paradigm .

The motivation for this thesis is threefold:

  • Multiagent Geosimulation aims to simulate, using agent technology, spatial phenomena or behaviors in geographic environments. Since this field is a recent, we did not find any paper dealing with approaches[2] to develop multiagent geosimulation applications. Thus, our first motivation is to propose a generic method which can be followed to develop multiagent geosimulation applications that simulate human behaviors in geographic environments .

  • Our second motivation is to illustrate our method by developing multiagent geosimulation models and applications which simulate human shopping behavior in a geographic environment representing a mall .

  • It is known within the computer simulation community that simulation output analysis is an important step in a simulation study. This step is necessary to test different ideas and learn about the simulation model and the corresponding simulation system (Anu, 1997), (Alexopoulos, 2002), (Seila, 1992), (Sanchez, 2001), and (Kelton, 1997). Multiagent geosimulation deals with non-geographic and geographic data related to the system/behavior to be simulated and its environment. The spatial characteristic of geographic data used and generated by multiagent geosimulations, make classical analysis techniques, (such as statistical and mathematical techniques) inefficient in providing a better (i.e., easy and rapid) exploitation of simulation outputs. Thus, our third motivation is to develop a new analysis technique, which is more efficient than the existing ones, in order to easily and rapidly analyze multiagent geosimulation outputs. Of course, this technique must take into account both non geographic and geographic simulation outputs .

The first problem addressed by this thesis is the lack of a generic method, which can be followed to develop multiagent geosimulation applications . Our literature review revealed that there are few papers dealing with methodological issues in the multiagent based simulation field. Furthermore, we did not find any paper that deals with approaches or methodologies which can be followed to develop multiagent geosimulation applications. Thus, the first research question that we address is: Using the actual progress in the multiagent and GIS fields, is it possible to develop a generic method that can be followed to systematically develop 2D and 3D multiagent geosimulations of systems/behaviors in geographic virtual environments? If yes, what are the important features and steps of such a method?

The second problem that we explore in this thesis involves the simulation analysis techniques which are used to analyze outputs generated by the simulation applications . Based on our literature review, we identified two sub-problems: The first concerns the simulation output generation, and the second is related to the exploitation and manipulation of the simulation output.

  • The simulation output generation : The literature reveals that, even with the existence of simulation tools that allow output generation such as the SWARM system, the majority of multiagent simulation applications do not profit from this function to generate output data (Minar et al., 1996). Most frequently, the only feedback that exists consists of a mere visualization of the simulation course on a screen (Alexopoulos, 2002). This type of feedback does not aid simulation users to make decisions about the system/behavior to be simulated or the simulation environment. To make efficient decisions, users need output data which can be efficiently stored, analyzed, and well-presented.

  • The exploitation and manipulation of the simulation output : Some simulation applications generate output data and analysis results for their users. Although this data may or may not be spatial, these applications use classical analysis techniques, such as statistical and mathematical techniques, to analyze and exploit the generated data. Classical analysis techniques are too limited for spatial analysis (no spatial analysis, no spatial visualization, no map-based exploration for spatial data, etc.) (Yougworth, 1995). For a domain such as multiagent geosimulation, geographic data becomes an important issue for decision-makers. Thus, we need more sophisticated analysis techniques that can be used to analyze complex simulation models involving geographic data. These analysis techniques must generate analysis results which can be easily exploited by users. They must also take into account, both the spatial and non-spatial aspects of the output data to be analyzed.

This second problem brings out our third research question which is: Among the existing analysis techniques, is there an appropriate one which could (i) be used to exploit multiagent geosimulation outputs (non-spatial and spatial), and (ii) present analysis results to users in an efficient (easily and rapidly) manner? If yes, how can we couple this technique with a multiagent geosimulation paradigm?

The main objectives of this thesis are:

  • To propose a generic method that can be followed to develop 2D-3D multiagent geosimulation of systems/behaviors in geographic environments. This method will be illustrated by developing a prototype that simulates the human shopping behavior in a mall.

  • To design the simulation models: the models for the environment (shopping mall) and those for the shoppers (individual shoppers, group of shoppers, and crowd of shoppers).

  • To develop, using empirical data, a multiagent geosimulation prototype that simulates human shopping behavior in a mall.

  • To propose an analysis technique to efficiently exploit (in terms of easiness and rapidity) the data generated by the multiagent geosimulation prototype.

Since we deal with the computer simulation field, at the beginning of this thesis, we studied several research works done in this field, and especially in the field of multiagent simulation. Our literature review revealed that only a small number of works exist, which simulate human behavior in geographic environments. Specifically speaking, there is a noteworthy lack of works that deal with methodological issues for the simulation of systems/behaviors in geographic environments. For this reason, we had the idea to propose a new and generic method which can be followed to develop 2D-3D multiagent geosimulations in geographic environments. In order to illustrate our method, we used the case study of shopping behavior in a mall.

In addition, we noticed that the human behavior simulation applications did not focus on the simulation output generation. The latter is important to end-users who want to use the simulation as a decision-making tool. For this reason, and in order to avoid falling into the same trap, we made some researches in the field of data analysis techniques in order to find efficient techniques which can be used to generate and analyze geosimulation output data. After an in-depth comparison of several analysis techniques, we found that the one most appropriate for multiagent geosimulation is OLAP (On Line Analytical Process) for non-spatial data, and SOLAP (Spatial On Line Analytical Process) for spatial data (Bédard et al., 2001). Thus, we coupled the multiagent geosimulation paradigm and OLAP/SOLAP analysis technique for efficient simulation output analysis and, therefore, for a well-supported decision-making process.

In the next chapter, we present a literature review with regards to computer simulation. We start by examining some computer simulation sub-fields. Next, we present some simulation applications simulating several systems and behaviors in various domains/areas. Afterwards, we present a number of simulation tools/platforms/languages that can be used to develop simulation applications. Finally, we discuss various methods and approaches that can be followed to develop such tools and applications. It is also relevant to note that our work implicates additional research fields and domains. The literature reviews related to these fields or domains are presented gradually throughout the following chapters of the thesis.

Chapter 3 briefly presents the principal steps of our method that can be followed to develop 2D-3D multiagent geosimulation of systems/behaviors in geographic environments. It also presents the simulation case study that will be used to illustrate the proposed method. This simulation case study consists of developing a multiagent geosimulation application which, in turn, simulates human shopping behavior in a mall.

In chapters 4 to 10, we present the details of each step of the proposed method using the case of human shopping behavior in a mall. In some of these chapters we present certain literature reviews that complement the one presented in Chapter 2.

Chapter 4 presents the illustration of the first two steps of the proposed method. In this chapter, we specify the needs of the primary users of the shopping behavior simulator for decision-making purposes. We also present the specification of the characteristics of the human shopping behavior in a mall. This specification is based on a large review of the literature produced by several domains and fields related to the shopping behavior.

Chapter 5: In this chapter, we create the agent-based models of the shopping behavior simulation. These models are created, based upon the literature study results obtained in the previous chapter.

Chapter 6: This chapter illustrates the step which aims to collect and analyze the empirical input data for the shopping behavior simulation.

Chapter 7: In this chapter, we select the platform which is used to execute the simulation shopping models. We also present our use of the selected platform and the simulation execution results. In this chapter, we give a complementary literature review which aims to briefly present certain simulation tools and criteria that can be used to select the most suited simulation tool for a given application.

Chapter 8: This chapter aims to present how we use software agents to collect and analyze the outputs of the shopping behavior simulation. In this chapter, we present a complementary literature review which focuzes on the existing data analysis techniques that can be used to analyze simulation outputs.

Chapter 9: This chapter illustrates two steps of our method. It first presents the illustration of the method’s step in which we verify and validate the shopping behavior simulation models. Secondly, it illustrates the step in which we test and document the simulator of the shopping behavior simulator. Again, in this chapter we present a literature review concerning the verification and validation techniques that can be used to verify and validate the simulation systems or behaviors.

Chapter 10: This chapter illustrates the final step of the proposed method. In this chapter, we present how the end-users of the shopping behavior simulator (who are mainly shopping mall managers) can use this shopping behavior simulator to make efficient decisions about the spatial configuration of their mall with respect to customers’ use.

Chapter 11 begins by summarizing the work done in this thesis. Next, it presents the results and main findings/contributions of our research. Finally, it concludes with possible directions for future works.



[1] In this thesis, the term spatial can be used instead of the term geographic .

[2] The term method in this thesis can be used instead of the terms approach or methodology .

© Walid Ali, 2006