Main Subcommittee Category: Simulation

How would you label your submission?: Academia

Abstract:

This paper introduces a novel Beyond Visual Range (BVR) air combat simulation environment using the open-source Godot game engine, focused on evaluating Multi-Agent Reinforcement Learning (MARL) alternatives for military training and simulation. Traditional military simulations often have restricted access, hindering research and innovation in autonomous agent development. Open environments based on open-source tools like Godot can be highly beneficial for sharing with the community and ensuring reproducibility and scalability. Our approach leverages Godot's high performance and simplified script-based development capabilities, offering a cost-effective and highly customizable solution for creating air combat simulation scenarios tailored for training and analyzing the development of autonomous agents. Godot's seamless integration with existing reinforcement learning frameworks enables researchers to quickly evaluate the possibilities for desired agent behaviors. Our solution introduces new integrations for state-of-the-art MARL algorithms, simplifying their adoption and application. This empowers researchers and military organizations to explore cutting-edge techniques in autonomous agent development within complex BVR scenarios. The proposed BVR environment aims to replicate essential aspects of air combat, including radar detection, weapons engagement, and tactical maneuvers, facilitated by Godot's robust physics engine and flexible scripting capabilities. While not overly realistic, the environment serves as a valuable testbed for rapidly prototyping and evaluating AI development approaches. This paper aims to highlight the potential of open-source tools and advanced machine learning techniques in accelerating the development of autonomous agents for military applications, presenting our BVR environment as a reference for the possibilities of such solutions and fostering discussion within the community.

Keywords: AI;AIR AND MISSILE DEFENSE;MACHINE LEARNING

Additional Keywords not in the list above: Air Combat, MARL, Godot

Andre Kuroswiski

Affiliation

Biography

Refine Results

Open-Source MARL for Autonomous Agent Research: A New Godot-based Environment for BVR Air Combat Simulation