Analysis of Design Patterns on an AI Game Environment

This Barrett thesis seeks to analyze software design patterns’ effects on a software system. To achieve this, the author specified a game environment that lets users write their own artificial intelligence (AI) algorithms for simulation in the environment. Afterwards, the author designed an architecture implementing the game system and designed components implementing the architecture. In software design, engineers use design patterns to develop components since software patterns generally apply to object-to-object interactions; architecture patterns apply to component-to-component interactions, and while they greatly influence software design, they are out of this project’s scope. To design the objects comprising this thesis system's event-driven model-view-controller (MVC) architecture, the author used the Adapter pattern to interface with other libraries, the Publisher-Subscriber pattern to pass information between objects, the Singleton pattern to enforce the existence of single state objects, the Dependency Injection pattern to build generic and composable functions, the Observer pattern to directly alert objects of observed objects’ changes, the Factory pattern to abstract object initialization, the Monad pattern to express complex computations without explicit branch control logic, and the Facade pattern to unite the game objects’ disparate interfaces into a single interface for AI developers. The implementation, integration, and synthesis of these pre-existing design patterns is the primary contribution of this project. After designing the software system, the author implemented the design using the TypeScript programming language, the Babel transpiler, the Webpack code bundler, and the Babylon.js graphics library. The author then performed a static evaluation on the implemented game system files by describing the overall dependency hierarchy and measuring each file’s lines of code, maintainability index, cyclomatic complexity, and Halstead difficulty score. Furthermore, the author compared these measurements with those collected from the Babylon, Phaser, and Lodash JavaScript libraries. The goals for reporting these measurements were to help show the game’s design enabling the system’s maintainability, usability, and expandability quality attributes and underscore software development as a creative and artistic discipline grounded in computational science. This thesis highlights the need for further research including developing methods with tools for evaluating behavioral aspects of design patterns relative to their quality attributes.