Simulating ASW: Agent models and game theory

Successful anti-submarine warfare (ASW) operations require intelligent use of both platforms and sensors. In the real world this is accomplished through training and usage of highly skilled personnel, with the help of tactical decisions aids. Simulations of these operations will therefore also require intelligent decisions to be made in the virtual world by the simulated agents, though necessarily requiring them to be decided automatically by the simulation program. The difficulty of modelling these behaviours increases with the fidelity of the virtual environment, making striking the correct balance one of the main goals of any modelling endeavour.

In anti-submarine warfare operations, sonar remains the most important sensor class for detecting submarines under water. Modelling this in a realistic manner will require a high degree of fidelity in the virtual sonar environment, thus also requiring more complex decision-making processes for the simulated agents. By using an appropriate acoustic model, the agents can calculate the probability of detection based on relevant configurations and the current environment.

We demonstrate the possibilities of this approach with the anti-submarine warfare simulation framework Rattus, where the agents use game theory to intelligently pick their sonar depths to achieve superior performance. This gives the simulation results more credibility and weight for their use in evaluating the performance of both current and future systems. At the same time, by increasing the simulated environment fidelity, any intelligent decision-making processes developed for the simulations can be more easily adapted to real world operations, potentially providing decisive tactical advantages.