G-PY: A Game Playing AI to Simulate Real Time Close-Quarter Firefights Using 3D First-Person-Shooter Games

This paper is regarding the proposal of the use of virtual environments to create, test and improve complex algorithms that power battle-ready drones which are deployed in actual war environments. An overview of the drone's operation would include the use of extensive image recognition to identify and differentiate enemies and friendlies, the environment and the objectives, static and interactable objects, all possible through training, using thousands of samples. The navigation of the drone is made possible using a specialized area recognition, memorizing and tracing algorithm. The weapons system used is similar to that of an average human militant, for the purpose of training, testing and calibration, and it employs a ‘Flick Shot’ algorithm to maximize the hit probability even in processors with lower computation power. The damage and loss of property sustained during deployment of these drones during field work, and the difficulty in training them with dummies which never are a 100% imitation of the enemies' technology, and the cost incurred, all can be avoided by using the proposed model.