Autonomous Driving in Games
Autonomous Driving in Games

Summary/Abstract: With the aid of advanced computer technologies and software capabilities, the world is gradually moving towards automation and artificial intelligence in all engineering and life domains. One of those fields is the auto industry. The cars of our days have options like automatic parking, keeping a safe distance to the nearest front car and use of navigation paths. The target is to breed autonomous cars that are able to consistently drive by themselves and to treat routinely all the sensor information regarding the environment, dropping off the possibility of accidents. Compared to the real case, the employment of autonomous cars in games has a consistent history. We present in this paper the attributes of using autonomous driving in games, as well as our knowledge in the implementation of these features. The Deep Learning with Convolutional Neural Networks we are using in driving games comes along with major benefits. The simulation of real autonomous vehicles in a virtual environment allows a better training and exposure to a multitude of conditions that would require lots of resources and a long testing time in real-life conditions. A second direct benefit comes from their integration in the AI of the cars from the driving games. The Deep Learning replaces creatively the simple predefined steering and path selecting methods that become repetitive and predictable. An indirect benefit is the introduction of these new technologies to the people, increasing the degree of acceptance towards the autonomous cars, which will become a steady partner of our existence.

• Details
• Contents