Mission planning involves the assignment of discrete assets to prioritized targets, including the dynamic routing of those assets to their destinations under complex environmental conditions. This report presents a description of how artificial intelligence (AI) can be used to conduct mission planning and how AI methods compare with more-traditional operations research approaches.

Understanding the Limits of Artificial Intelligence for Warfighters

Volume 5, Mission Planning

Keller Scholl, Gary J. Briggs, Li Ang Zhang, John L. Salmon

ResearchPublished Jan 23, 2024

Cover: Understanding the Limits of Artificial Intelligence for Warfighters
Download PDF
Additional Downloads

Includes Research Summary

Mission planning involves the assignment of discrete assets to prioritized targets, including the dynamic routing of those assets to their destinations under complex environmental conditions. Because of the value of quick turnaround and the relative simplicity of the simulated operational environment, there has been considerable interest in improving the mission planning process with the addition of reinforcement learning techniques for artificial intelligence (AI), which could produce better, faster, or simply unique solutions for human consideration. This report provides a description of how AI can be used to conduct mission planning and how AI methods compare with more traditional operations research (OR) approaches.

One important aspect of mission planning is proper route planning, which can minimize risk to pilots and systems, reduce enemy information about U.S. assets, and increase the likelihood of successful mission execution. Although only a subset of all route planning, planning for an individual package to penetrate enemy airspace is a scenario that is frequently encountered by the Department of the Air Force (DAF). Using in-house modeling software, researchers explored the feasibility of applying AI to this task, comparing AI performance against an optimization approach, and assessing the limitations of this approach.

This report is the fifth in a five-volume series addressing how AI could be employed to assist warfighters in four distinct areas: cybersecurity, predictive maintenance, wargames, and mission planning. This report is aimed primarily at those with an interest in mission planning, operations research, and AI applications more generally.

Key Findings

  • Compared with OR approaches, AI typically performs worse. Given that OR involves solving well-posed optimization problems, this result is not surprising. However, AI can be more robust and responsive to a changing environment because an OR solution is only meant to solve a static problem.
  • AI is capable of helping out in some planning roles and using AI in this way will build capacity, experience, and user trust for future AI use. Mission route planning is one example of a narrow AI application that is particularly useful for dynamic threat environments, in which mission packages enter into a complicated air defense environment with pop-up threats.
  • AI for mission planning requires the development of infrastructure that efficiently connects a simulation environment with an AI framework, which is often written in a different coding language. Fortunately, this is a one-time investment for each simulation environment. The DAF should consider such investments and release the infrastructure to government and partner organizations.
  • Implementing AI in mission planning, and in warfare more broadly, is not just a matter of creating a standalone program. It is critical to support connections to other tools and continuously update those connections as new tools are invented. Without this ongoing support and effort, the actual use of AI will inevitably lag compared with near-peer adversaries.

Recommendations

  • The DAF should apply reinforcement learning (RL) mission planning to dynamic route planning for uncrewed systems that are reviewed and judged by operators. For now, the best use of RL in mission planning is as a fast-reacting management system that responds dynamically to threats. This applies both to an on-board drone and to a headquarters that can give an updated flight plan in a matter of seconds rather than minutes. Even when RL provides suboptimal plans, it can suggest immediate actions; operators can take the time gained from a rapid response to develop better plans using their current standard and preferred methods.
  • The DAF should train AI specialists who also have a deep understanding of military mission planning. RL is a difficult research area that relies on experience and heuristics. The research is further complicated by the need for application-specific knowledge. Those who lack familiarity in the area might not recognize undesirable states and behaviors, preventing them from crafting the suitable reward function.
  • The DAF must prioritize tools and software, not just by creating them but also by enabling those resources to be extensible and connectable to existing systems. Existing simulation tools should be extended to be compatible with AI frameworks.
  • The DAF should continually monitor the AI RL landscape. Although AI advancements occur rapidly in commercial and research communities, the DAF will need to remain vigilant about looking for opportunities to integrate new advancements.

Topics

Document Details

Citation

Chicago Manual of Style

Scholl, Keller, Gary J. Briggs, Li Ang Zhang, and John L. Salmon, Understanding the Limits of Artificial Intelligence for Warfighters: Volume 5, Mission Planning. Santa Monica, CA: RAND Corporation, 2024. https://www.rand.org/pubs/research_reports/RRA1722-5.html.
BibTeX RIS

Research conducted by

The research reported here was prepared for the Department of the Air Force and conducted within the Force Modernization and Employment Program of RAND Project AIR FORCE.

This publication is part of the RAND research report series. Research reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND research reports undergo rigorous peer review to ensure high standards for research quality and objectivity.

This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited; linking directly to this product page is encouraged. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial purposes. For information on reprint and reuse permissions, please visit www.rand.org/pubs/permissions.

RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.