Simulating Cyclists in a Simulator with the use of Behaviour Trees

Abstract

The thesis starts by explaining the theory behind the AI agent. This includes a study of cyclist laws, rules and behaviours in Norway, and a state-of-the-art explanation of behaviour trees in games. After this, the thesis explains how this theory is used to implement an AI agent in Unreal Engine 4. This includes an explanation of how the AI agent controls the cyclist model and how behaviour trees are used to simulate different behaviours.

The analysis of the results show that the project has successfully made an autonomous cyclist agent. The AI agent can control the cyclist model and simulate behaviours for a child, transport or leisure cyclist. This is categories of cyclists defined in the project. Each cyclist type was simulated by the AI agent. Then the cyclist types participated in several tests, in different traffic scenarios. The tests are defined and implemented as part of this project, and is explained in detail in the thesis. The test results have been recorded and analysed. The videos, together with the implementation, was then used to conclude the project.