Collection Mémoires et thèses électroniques
Accueil À propos Nous joindre

CHAPTER 11: General Conclusion and Future Work

Table des matières

In this chapter, we give a general conclusion about the research done in this dissertation. We then present the major contributions of our work. Finally, we propose various directions for future research works.

In this dissertation, we proposed a generic method that can be followed to develop 2D and 3D multiagent geosimulation applications of phenomena in georeferenced virtual environments. One of the advantages of this method lies in the fact that it captures recent progress in two promising domains, multiagent based simulation and geographic information systems. This method contains ten steps, which are summarized as follows:

The first three steps of the method aim to (1) define the geosimulation users’ needs, (2) identify the characteristics of the phenomenon to be simulated, as well as its environment, and (3) create the geosimulation models using the multiagent paradigm. The fourth step aims at selecting the simulation tool/environment/language that is used to develop the geosimulation. In step five, we collect the data which feed the geosimulation models. In this step, we analyzed the collected information in order to define some patterns of the behaviors of the phenomenon to be simulated. In the sixth step, we develop the geosimulation prototype, on the selected simulation platform, using the collected data. In step seven, we collect information about the course of the simulation, once again using the multiagent paradigm. In this step, we deal with the non-spatial and spatial data, generated by the simulation using several analysis techniques: Classical or traditional analysis techniques, our own analysis technique/tool, and the OLAP (On Line Analytical Processing) and SOLAP (Spatial On Line Analytical Processing) technique. In order to ensure the correctness of the simulation models, as well as to enhance the confidence of the simulation users, we need to verify and validate the simulation models. The verification and validation are the purpose of the eighth step of our method. In the ninth step, we test and document the simulation, while in the last step users can use the multiagent geosimulator in order to make efficient spatial decisions about the phenomenon to be simulated or about the configuration of the simulation environment.

Our method is illustrated throughout this dissertation, using the customers shopping behavior in a mall as a case study. The information used in the simulator has been collected from a survey on customers’ behavior of Square One mall, in the Toronto area.

As shown all along this dissertation, our method is suited to develop 2D-3D geosimulation of phenomenon in georeferenced environments. This method is intended to be generic; even it has been illustrated using the shopping behavior case study. This illustration makes it a promising method that can be followed to develop other geosimulation applications of similar phenomena in different geographic environments. The main advantage of the method with respect to the other methods and approaches existing in the literature, is that it takes into account and focusses on the spatial characteristics of the phenomena to be simulated and the geographic features of the simulated environments. The characteristics of the method’s steps, as discussed all along this dissertation, make is easily applicable to simulate other phenomena in various geographic environments.

Our method has currently a limitation which: it is intended to simulate phenomena in geographic environments represented at only one level (e.g., simulation of shopping behavior in one floor of the mall). It would be interesting to improve it in order to take into account the multi-level aspects of simulated geographic environments.

In this thesis, we presented a generic method that can be used to develop multiagent geosimulations of spatial phenomena in geographic environments. Based on our literature review, we can mention a noteworthy absence of methods or approaches to develop simulation applications in the geosimulation field. Our proposal of such a method can be considered as a significant contribution since geosimulations are becoming increasingly useful in several different domains, and a method to systematically develop and use geosimulations is a step forward to help users from different fields to take advantage of theis powerful approach.

In order to illustrate this method, we used as a case study the human shopping behavior in a mall. Studying and simulating such behavior required that we carried out in-depth studies in several disciplines: this which gives a multi-disciplinary flavor to our research work. The results of this work, which have been presented in this dissertation make our research findings as original in the computer simulation field as in the other fields we have investigated.

The main scientific contributions of this dissertation are the following:

■ The main contribution of this thesis is the proposition of a new generic method that can be followed to develop 2D-3D multiagent geosimulations of phenomena in georeferenced virtual environments. What distinguishes our method from those proposed by (Drogoul et al., 2002) and (Ramanath and Gilbert, 2003), is that it focusses on the spatial characteristics of the phenomenon to be simulated and the geographic features of the simulated environment. What’s more, our method contains some steps which are useful, but are often neglected by other methods and approaches. As examples, we can cite: the step in which we collect and analyze input data aiming to feed the geosimulation and the step aiming to analyze the outputs generated by the geosimulation. Since geosimulation is considered as a new field, our method can be considered as original. This method was published in (Ali and Moulin, 2005a) and (Ali and Moulin, 2005b).

■ In this thesis, we illustrated our method using the shopping behavior case study. In this illustration, we created some useful models. As an example, we created a model of the individual’s shopping behavior based upon the models of consumer behavior proposed by (Engel et al., 1968), (Howard and Sheth, 1969), and (Nicosa, 1966). In this model, we integrated some factors (variables) that influence the shopping behavior, as well as some processes that compose the shopping behavior. These factors and processes were not presented in the models presented by the aforementioned authors. We also proposed an agent-based model of the individual’s shopping behavior in a mall. This model stands out from those proposed by (Dijkstra et al., 2001) and (Timmermans et al., 2003) and (Ben Said et al., 2001) because it is not limited to presenting the shopping behavior in a mall as a pedestrian activity, but it presents it as a whole behavior containing several activities which can be spatial, social, cultural, etc. In addition, in our agent-based model, the shopping behavior contains some spatial and knowledge-based processes such as perception, memorization, decision-making, etc. Finally, based upon our literature review, we can affirm that there are no research papers involving the simulation of shopping behavior of group in a mall. The majority of works that deal with groups and collective decision-making are restricted to social psychology. In our work, we propose a simple model to represent a group of shoppers. We also proposed a simple model representing a crowd of shoppers in the mall.

■ Without data, computer simulation does not work (Anu, 1997). Since we deal with geosimulation, the spatial aspect of the simulation data is a very relevant one. In our method, we introduced two steps that take into account this data. The first step deals with the simulation input data, while the second concerns the data generated as outputs. These steps were presented in (Ali and Moulin, 2005a) and (Ali and Moulin, 2005b).

For the input data : We presented a step in which we collected and analyzed data before using it to feed the geosimulation models. This step is, generally, neglected by research works dealing with simulation methods and approaches such as (Anu, 1997), (Fishwick, 1995), (Allen et al., 2001), (Groumpos and Merkuryev, 2002), (Drogoul et al., 2002), (Ramanath and Gilbert, 2003), etc. The introduction of such a step in our method can be considered as an original work in the field of simulation. What’s more, (1) developing a software to digitalize non-spatial and spatial simulation data at once and (2) analyzing these two kinds of data can also be considered as original work. We also presented a survey method to collect spatial and non-spatial real data about a phenomenon for geosimulation. This method which is illustrated using the shopping behavior case study can be considered as a contribution, because we did not find any similar method or technique that can be used to collect spatial empirical data about a specific phenomenon which operates in a geographic context.

For the output data : we presented a promising technique that can be used to collect spatial and non-spatial output data from a geosimulation prototype. This technique is based upon the concept of observer agents whose mission is to observe some aspects of the geosimulation execution and generate data in specific files or databases. The advantages of using such a technique represent a contribution of this work to the simulation field. We also presented our new analysis technique and tool that can be used to analyze the outputs generated by the geosimulation. This technique/tool differs from the existing techniques/tools such as those presented by (Sanchez, 2001), (Kelton, 1997), (Alexopoulos et al., 1998), (Alexopoulos, 2002), and (Seila, 1992) because it takes into account the spatial aspects of the data to be analyzed. We emphasized the fact that the spatial data is fundamental to a geosimulation study. What’s more, the analysis results are presented and exploited spatially on the simulated environment which is closer to the users’ mental models. The characteristics of such a technique/tool let us consider it as a contribution in the simulation field. We also presented how we exploited an existing technique called OLAP/SOLAP in order to analyze and explore spatial and non-spatial data generated from geosimulation. The idea of coupling a multiagent geosimulation application and a SOLAP tool is also original in the simulation field.

■ In this thesis we illustrated our method by developing a geosimulation prototype that simulates the shopping behavior in a mall. This prototype stands out from the simulation prototypes existing in the literature by its ‘realism’ and ‘usefulness’:

□ A more ‘ realistic ’ prototype: In the simulation, the shopper agents are equipped with advanced spatial knowledge-based capabilities which can increase the realism of the behavior simulation. What’s more, the simulation is presented in 2D and 3D modes. It is known, that simulating in 3D mode increases the realism of the visualization of the simulation. In the literature, we did not find any geosimulation prototypes exhibiting, simultaneously, these characteristics. Our prototype was presented in (Ali and Moulin, 2005c).

□ A more ‘useful’ prototype: If we look at simulation applications proposed in the literature to simulate human behaviors in spatial environments (Raubal, 2001), (Dijkstra, 2001), (Koch, 2000), and (Moulin et al., 2003), we note that they are used only to visualize the simulation course. Thus, they are not used to compare simulation scenarios, nor to generate results about the comparisons of these scenarios. This represents a noteworthy gap in the usage of a simulation. In order to fill this gap, we developed a multiagent geosimulator of shopping behavior that can be used to compare several simulation scenarios. This prototype generates analysis results related to these comparisons, and can be applied by users (mall managers) in order to make efficient decisions about the phenomenon to be simulated, or about the simulation environment. To sum up, our shopping behavior geosimulation tool stands out from other human behavior simulation tools because it can be used by end-users to make spatial decisions about the configurations of the simulated environment. Hence, our shopping behavior geosimulation prototype can be considered as a contribution to the field of computer simulation, and especially, to the field of multiagent based-simulation.

In this thesis, we reached the following objectives:

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

Even if it is illustrated using the customers’ shopping behavior case study, our method is intended to be generic and can be followed to develop geosimulations for other phenomena in different geographic environments.

- ‘ To design the simulation models: the models for the environment (shopping mall) and those for the shopper ’:

In this thesis, we developed several models that are used by the shopping behavior geosimulation. For example, we developed: an initial version of the shopping behavior model based upon a large literature review, an initial version of the multiagent model which is independent of the tool used to execute the simulation, and an agent-based model created according to the selected platform used to develop the geosimulation. All these models are related to the individual shoppers and to the simulated environment representing the mall.

- ‘ To develop, using empirical data, a multiagent geosimulation prototype that simulates human shopping behavior in a mall ’:

This led us to the development of a 2D-3D multiagent geosimulation prototype of the shopping behavior in a mall. Unfortunately, this prototype does not simulate the shopping behavior of groups of shoppers due to a lack of data ccollected in the surveyx and related to groups.

- ‘ To propose an analysis technique to efficiently exploit (in terms of easiness and rapidity) the data generated by the multiagent geosimulation prototype ’:

In this thesis we proposed and developed an analysis technique and tool that can be coupled with the geosimulation in order to analyze its outputs. These techniques and tools focus on the spatial aspect of the output data generated by the shopping behavior geosimulation. We also tested the use of an existing technique called OLAP/SOLAP which is coupled with the geosimulation. This technique was then, compared with our proposed technique/tool.

As future works, we intend:

1. To apply our method in order to simulate the shopping behavior in other geographic environments. We plan to use our method to simulate the shopping behavior in a large scale environment, such as a city (in the shopping streets), and to simulate the shopping behavior inside a store.

2. To develop a simulation prototype that simulates the shopping behavior in several floors of a mall.

3. To use the method in order to simulate other phenomena and behaviors in different geographic environments.

4. To develop a simulation prototype for the shopping behavior in a Quebec City’s mall, called Place De La Cite. We have already conducted a survey in this mall in 2004 and gathered data, just as we did at Square One. We would then be able to compare the shopping behavior of Toronto residents with that of Quebec City customers. This comparison can be interesting in order to understand the differences or common points between the shopping habits and behaviors of Toronto residents and those of Quebec City residents. Understanding these differences or common points can be relevant to (1) retailers (stores or others) in both Toronto and Quebec City (e.g., The Bay, Wal-Mart, Old Navy, etc.) and to (2) companies that own shopping malls in the two areas (e.g., OMERS that owns Square One mall in Toronto and Place Laurier mall in Quebec City).

5. To conduct a new survey concerning the groups of shoppers using an enhanced version of the questionnaire already developed for groups. This would enable us to simulate group behaviors in our geosimulation environment and to study the effects of this kind of behavior on the customers’ shopping patterns.

© Walid Ali, 2006