6 . 1 Conclusions

Modelling and simulations in the geo-spatial domain is being carried out using the techniques of Cellular Automata (CA) to model processes in space and time. Modelling of land-use / land-cover dynamics in both space and time is bounded by various causal factors driving the changes that can have varied relations in space and time apart from the inherent physical state dynamics. However it is seen that CA models do not address the external driving factors that are also responsible for the dynamics involved directly. Typically, the CA models need to account for the external drivers that also drive a land-use / land cover change which are not accounted in the transition rules of the automata. Certain externalities can be system wide or specific to certain locations, for which the CA models have to evolve to address such requirements. Further there are also different significant processes that take place in the region in question, apart from those represented in the CA model. Integrating the agent-based models (multi-agent systems) with the CA models would address these inadequacies as evident from the case study of modelling land-use / land cover dynamics by incorporating different drivers / agents.

The focus of the current research was on overcoming the limitations of the CA models by integrating CA with the agent-based models considering the issues of scales for synchronization of these models in space and time. The main objective of the research was achieved through the design of the framework for the integration of agent-based and a cellular automaton model. With the possibility of defining all these processes as agent-based models, the agent-automata described for this scenario over a specific period of time, can be coupled in over all simulation, which would enable in the visualization of such ‘what if' scenarios more effectively.

An attempt was made in this work to integrate agent-based models and CA model for synchronizing in their respective spatial and temporal scales. Accordingly, this research has contributed the Agent-Based Cellular Automata (ABCA) framework for integrating the agents and CA models at respective scales, by incorporating a geo-spatial analyser (GSA) to handle the spatial synchronization and the HLA framework to handle the temporal synchronization. The key assumptions that govern the framework are as follows:

• Agent-based models attributed with the spatial relationships defined over a cellular space representing discrete entities of the phenomenon.
• The models in time are assumed as discrete-time stepped models.
• In order to have congruity with the geo-spatial data obtained by satellite remote sensing which are also in regular tessellated square or rectangular cells, the scope of the present research confines to the geometry of the cells used in CA as regularly tessellated square or rectangular cells. Subsequently, the spatial relationships to the agents are defined with their geometry associated as cells.
• The simulation time for the CA and agents are assumed to have a single time advancement mechanism. However, each of the agent-based models could be defined separately with different time advancement mechanisms depending upon the events for which they are defined, as discrete-event models. Then a significant issue would be to address the synchronization of these models with different time advancement mechanisms, this however is beyond the current scope of research.

With the development of the simulation framework integrating the agent-based models and the CA models, the research aimed at the applying the simulation framework for simulating for a use case of urban sprawl, which is very dynamic in nature. The phenomenon of urban sprawl is very dynamic in nature. A complex of activities involving the economics, infrastructure addition, population growth and so on mainly attributes the urban sprawl dynamics. The different drivers or the agents for inducing sprawl can manifest the sprawl by the complex of interactions and responses amongst them. The interactions among the agents of sprawl like the population or infrastructure can be complimentary to each other. Subsequently the reactions of these agent manifestations can lead to fuel further sprawl. Further, these agent-behaviours are not same with space and time; in effect these agents have an impact over system dynamically both in space and time. Visualizing such multi-scale space-time dynamic phenomenon like the urban sprawl is still not well handled by the traditional GIS (Batty, 2003).

Simulating a dynamic phenomenon like the urban sprawl using the agents in conjunction with the CA models was attempted in this research. The framework for ensuring the simulations developed in the Chapter 3 and the dynamics of urban sprawl studied and modelled using CA and agents in the Chapter 4 are implemented. Accordingly, there are two agent-based modelling approaches for simulating urban sprawl. In the first approach, the simulation of the model demonstrating the radial urban sprawl was carried out using the tool – NetLogo (Wilensky, 1999). This revealed the pattern of growth that takes place under different scenarios. From the model interface, it was evident that the agent actions can be continuously obtained in terms of their numbers over time. This model can also be of use for teaching and demonstration purposes. For the application of agent-based and cellular automata models for a real situation, the case of Mangalore city, Karnataka, India was considered and applied. The Mangalore city is currently experiencing high rates of urbanisation as evinced from the study. For a scenario of an infrastructure initiative like the creation of a ring road, the implications of this are depicted using the combination of agent-based and CA models. The simulations revealed the nature of likely growth in the region due to infrastructure initiative. Based on the simulations combining the CA and agent automata for the cell transitions, the visualization of the future scenario of urban growth by the creation of an infrastructure was successful. However, for an effective utilization of the agent hood in the simulation, adequate agents could be devised to report the nature of changes during the iterations. In this work, only one agent type was demonstrated while for an effective realization of the agent properties in the geo-spatial simulations, more agents can be incorporated in the future research.

In order to achieve the main research objective of developing a framework for the integration of agent-based and CA models, the research was divided to address specific research questions. The first research question concerned with the conditions and methods for the integrations of agents and CA models. Accordingly, the ABCA framework is developed for the integration with the key assumptions noted above as the conditions for enabling the integration. Considering the methods for integrations of the agent-based models with the CA, the research noted that the available tools for building agent-based models can be used with appropriate interfaces to be developed for enabling these tools to communicate with the geo-spatial data. Another approach is by enabling the agent hood properties to the geographic objects and attaching the behaviour of the discrete entity of the phenomenon for which the agent is being attributed. The scheme for building agents by defining the agent extents and duration is suggested. The subsequent research question addresses the requirement for the framework to be interoperable and reusable for different processes to counter any proprietary use of the framework. Accordingly the developed framework is adopts already established architecture of HLA for time management with the geo-spatial inputs subject to the standards specified by the OGC. Even with the availability of a standard for the agent-based modelling specified by the FIPA, it currently lacks any spatial component and is inept for handling multiple simulations with heterogeneous time advancement mechanisms. The ABCA does not recommend any specific software or programming environment for undertaking such geo-spatial simulations. An integrated architecture combining the OGC, HLA and the FIPA would help in the realization of the overall goal of achieving the interoperability across the domains. The last research question aimed at identifying the calibration and validation techniques for evaluating and ensuring the usability of the simulations. It was noted that the CA model upon which the agents were to be integrated could be calibrated, while the agents as such were used to predict the scenarios and not already observed phenomenon. And so, the overall simulations had to be validated by calibrating only the CA model for which the suitability images prepared played a crucial role.

6 . 2 Scope for Further Research

The research developed the framework for the integration of agent-based and CA models for geo-spatial simulations. The present research treats the agent-based models to represent the discrete entities of the phenomenon for which they are modelled. However in reality agents can also represent the continuous phenomenon like the movement of pedestrian or vehicles apart from the discrete entities depicted in the research. Thus for an effective and more realistic representations of the different entities of the phenomenon all the entities, discrete and continuous should be coupled. Accordingly the research suggests improvising the generic framework to handle both discrete and continuous entities of the geo-spatial phenomenon.

The models in time considered in the research were as discrete-time stepped models with a single time advancement mechanism. However in reality, the models that can be represented by the agents can be discrete-event and continuous time models. Subsequently, with the inclusion of heterogeneous time models and heterogeneous time advancement mechanisms, research has to substantially contribute to the current framework to mature for enabling such simulations.

With the definition of the agents as geographic objects and subsequent association of agents to cells corresponding to that of CA, inhibits the features to be represented appropriately. Torrens (2000) notes that not all features can have regular tessellations, and so the integration methods should also be ideally addressing the geometry of the agents as irregular lattice structures.

Modelling and simulating the dynamics of urban sprawl as complex systems can now be addressed by using the agent-based models to generate the complex non-linear behaviour depicted by these systems. And so, the future research has to contribute to study the patterns of sprawl attributing the direction of likely growth and the rate of such sprawls using the combination of agent-based and CA models. However, research has to contribute significantly to determine the entities to be modelled as agents in case of urban sprawl dynamics in developing countries where the complexities of abrupt policy decisions and lack of prior planning result in fuelling sprawl.

An important characteristic of the agent-based models in the geo-spatial domain can be for visualising space-time variant phenomenon. The agents can be devised to report the nature of changes for all the entities in question. Apart from the utility of the agents in geo-spatial simulations as explored in this research, utility of agents embedded within the GIS to depict space-time variant phenomenon should be addressed by the future research. Such advances in the geo-spatial domain would fill the much need gap of addressing the temporal dimension of the geo-spatial databases. A research in this direction would answer the questions on analysing patterns of change through time (Peuquet, 1999), for which the current GIS is still inept. Thus a complete convergence of agent technology and GIS should be a major research agenda in the geo-spatial domain for incorporating the temporal dimension of the entities to the geo-spatial attributes.