Análise de qualidade de Steering Behaviors e desempenho da Unity 5

Abstract

Artificial intelligence is a fascinating field and filled with practicable areas, one of these areas is the area of digital games. In this area it is possible to find several techniques of artificial intelligence, being one of them the steering behaviors, directional behaviors that aims to help autonomous characters to move in a more realistic way, using simple techniques involving vectors. In this paper we discuss three different steering behavior techniques that have been implemented and modified in the Unity 5 game engine, such as Arrival, Collision Avoidance and Wall Avoidance. This work aims to analyze the quality of these steering behaviors as well as analyze the performance of this game engine. To achieve these objectives a number of quality and performance metrics were collected and analyzed after simulations were performed in nine scenarios adapted from a benchmark for Unity 5 that can represent real world scenarios. At the end of the work it was possible to verify notable differences in quality metrics due to the different scenarios, number of agents and behaviors, as well as to verify that the frame rate per second is more sensitive to the number of agents and their behaviors while it was only partially possible to prove that the total memory usage of the system is sensitive regarding the number of game objects (agents, obstacles, etc.), number of agents and their behaviors.