In military Modelling and Simulation (M&S) there is an increasing need for Computer Generated Forces (CGFs) with machine learning capabilities for use in training or decision-support applications. Machine learning based CGFs have benefits for the implementation of adaptive enemies for human-in-the-loop training; the development and evaluation of TTPs (tactics, techniques and procedures) for military platforms; or supporting course of action (CoA) planning and analysis in mission simulation.
Machine learning introduces a radical new paradigm for behavior modelling of CGFs. When modelling a behavioural task for a CGF, rather than handcrafting individual decisions and actions, learning algorithms allow for the mere specification of the underlying goal of the task, where it is up to the algorithm to learn how to achieve this goal. In this paper we the explore the implications for this paradigm shift. Our contribution is threefold.
First we describe the benefits and potential drawbacks for introducing CGF machine learning capabilities for involved stakeholders, such as end users (trainees, instructors, scenario developers and operators) and modellers (subject-matter experts, designers and developers).
Second we present a framework infrastructure to support machine learning capabilities for CGFs in military simulations. It addresses required capabilities for (1) the simulation environment to act as a data-provider and (2) CGF models to provide suitable sensors and actuators to facilitate the required situational awareness and decision-making interfaces for a learning algorithm.
Third we propose a method for training CGFs to perform domain-specific tasks in a simulation environment, hereby making CGFs ‘fit-for-purpose’ for an application domain. Inspiration is taking from the field of Adaptive Instructional Systems (AIS), originally intended for training tasks for humans.
The ideas in this paper have been implemented in a military simulation system geared towards the air-to-air combat domain. Based on this implementation, challenges, lessons learned and future directions are discussed.