IN GAME MAP GENERATION USING RANDOM PATTERN GENERATION

Abstract

In this work was conducted analysis in machine learning sphere for self-driving cars. Machine learning helps to extend spheres where of Artificial Intelligence where simple algorithm could not help.. That’s why developers are trying to create, find and form different data sets for their Artificial Intelligence training and improvement.

Every Artificial Intelligence needs their specific data set, for executing some actions. For example, for face identification we need dataset with faces, for text segmentation – text corpuses. That’s why Artificial Intelligence for self-driving cars needs their own data set.

Methods, which are used for its implementation, they need to build an iterative training on game maps, which will give results depends on training object on the map. In this work is described reinforcement learning method, which is a basis for building self-driving systems. Also was conducted examples of a training such Artificial Intelligence and displayed game map that are developed and available nowadays.

To provide such game maps, different algorithms are used to generate such game maps. Such algorithms are used in game industries for building levels that can have different sizes, event bigger than our planet. That’s why in this work were conducted descriptions to the companies and games that are using automatic game map generation.

In this work was conducted verification of an algorithm for road infrastructure generation and investigated algorithms for game map generation and as a result was implemented simple and fully automated algorithm for generating infrastructure. To build a full picture about such algorithm was  used game engine called Unreal Engine 4 and method optimizations for more descriptive illustration.

To make investigation more clearer was considered in details Unreal Engine 4 and made a comparison between next most popular game engine called Unity. So it is provided with advantages and disadvantages between those engines. Also pros and cons were described.

Keywords: Unreal Engine 4, game engine, reinforcement learning, map generation, OpenGym, algorithm, graph, Artificial Intelligence, game map.