Imitation of Human Behavior in 3D-Shooter Game

Ch. 9. P. 64–77.

Makarov I., Mikhail Tokmakov, Lada Tokmakova

There are many algorithms for 3D-shooter game artificial intelligence that are based on automatic target recognition with the help of ray casting, scriptbased event systems and many others. In such models game agents have an advantage over human player because computers do not need time for enemy recognition and target aiming processes. The main goal of our research is to imitate a realistic enemy player under PC control, which acts similar to a human reaction. We consider such aspects of shooter in-game processes as visual recognition of players, the delay for the response on each game event and distributed strategies of decision making. The paper presents method of visual recognition of player's 3D-model with the use of geometry knowledge of the three-dimensional maze and target sighting with respect to human motor reflexes. The implementation is made in Unreal Engine 4, which provides large variety of methods for programming game agents.

Keywords: нейронные сети, Game Artificial Intelligence, Visual Recognition, игровой искусственный интеллект, Визуальное распознавание, Neural Network

In book

Supplementary Proceedings of the 4th International Conference on Analysis of Images, Social Networks and Texts (AIST'2015)

Issue 1452. , Aachen: CEUR Workshop Proceedings, 2015.