Artificially intelligent agents in VR military training

Abstract

Virtual Reality (VR) is the experience by which users feel immersion in a virtual simulated world through the use of hardware and software. Developers of these worlds can then create environments where the user can freely move and interact with objects or people without needing to be present physically. VR technology has been progressively used more in military and police training to provide more immersive and realistic simulations. VR training is used to provide users with a safe and controlled environment in which their expertise and decision-making skills can be practiced without risking heavy injury or wasting financial resources. These simulations are currently used for practicing tactical training, quick decision-making skills and even marksmanship training. It is also used to put users in scenarios that will similarly match real-life missions they might have to execute thus allowing trainees to gain experience and confidence before facing real-world scenarios. It is of utmost importance to undergo training for these situations, primarily since the military and police challenges that this project focuses on entail significant risks. In the future, VR technology will thus revolutionize military and police training by providing a safer and more effective way to prepare trainees for real-world scenarios. As VR technology continues to advance, it is likely that it will become an even more important tool for training in these fields. The main issues with current simulated training environments used nowadays is that the actions taken by the synthetic AI agents created are done using rule-based and purely reactive models which possess almost negligible intellect (Karr et al., 1997). These models lack the ability to adjust to external actions and surroundings, making them less suitable as training companions since their sole purpose is to assist users in improving their skills. Another large issue is that the combinations of user actions along with the poorly made agent models still require highly intensive computational power. In this project, machine learning is applied to the creation of artificially intelligent agents to be used for future military and police training simulations. This was done in addition to showcasing how most commonly used rule-based agents lack the necessary knowledge to be used for more challenging training scenarios.