Simulated Laser Weapon System Decision Support to Combat Drone Swarms with Machine Learning
Abstract
This thesis demonstrates an application of machine learning for enabling automated decision support to warfighters operating laser weapon systems in complex tactical situations. The thesis used the NPS Modeling Virtual Environments and Simulation (MOVES) Institute's Swarm Commander modeling and simulation software environment to develop simulated datasets of wargaming scenarios involving a shipboard laser weapon system defending against drone swarm threats. The simulated datasets were used to train a machine learning algorithm to predict the optimum engagement strategy in a complex battlespace with heterogeneous drone swarms. Multiple machine learning techniques were evaluated, and the classification tree technique was selected as the preferred approach. The final algorithm had an overall accuracy of 96 percent in correctly predicting engagement outcomes based on drone threat types, quantities, and the laser weapon system attack strategy. The research results demonstrate (1) the utility of modeling and simulation for supporting the development of tactical machine learning applications, (2) the potential for machine learning to support future tactical operations, and (3) the potential for machine learning and automation, in general, to reduce the cognitive load on future warfighters faced with making critical decisions in complex threat environments.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2021
- Accession Number
- AD1164253
Entities
People
- Daniel M. Edwards
Organizations
- Naval Postgraduate School