Urban Reconnaissance through Supervised Autonomy (URSA)

Abstract

The Urban Reconnaissance through Supervised Autonomy (URSA) program developed and demonstrated new autonomous agents and techniques that support a Blue Force Commander in managing the complexity and ambiguity of urban spaces by rapidly identifying and discriminating among potential threats during missions ranging from minutes to hours. The program used perception-enabled autonomous vehicles to manage complexity and interactions with populations to drive down the ambiguity between peaceful civilians and threats. The program created a system of autonomous ground and air platforms operating in conjunction with U.S. ground forces that monitor an area overtly to detect hostile forces and establish Positive Identification (PID) before any U.S. troops come into contact. Military units follow strict rules of engagement (ROEs) that prescribe an escalation of force appropriate with the level of hostilities and confidence that an individual is engaged in nefarious behavior. This program established a Legal, Moral, Ethical (LME) working group comprising multiple experts (technologists, military, university professors, ethicists, legal experts) to engage in development of an ethical operations process (DevEthOps) to engineer Responsible Artificial Intelligence (RAI) principles into this supervised autonomous system. URSA explored scenarios and probed behaviors to enable identifying innocent civilians and individuals who pose a threat to U.S. Forces, allies, or non-combat civilians. This mission requires the integration and maturation of novel sensors, and unmanned ground and air vehicles which leverage current techniques in perspective and reactive autonomy to navigate cluttered urban environments. URSA developed new search and engagement behaviors to disambiguate human actions and serve as evidence that a potential target is a threat. It implemented new dimensions of evidence such as the human reactions to these engagements to improve confidence of operators in determining with high precision and low false positives who may pose a threat and who does not. While developed for Urban environments, other applications may include managing large populations of any kind to include supporting Military Police and detainee operations.

Document Details

Document Type
Accomplishment
Publication Date
Oct 01, 2025
Source ID
31c476cbfd7fa8437f968c60a772cd32

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Strategic Security Studies

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - DoD AI Strategy
  • Autonomy
  • Autonomy - Human-Robot Interaction
  • Autonomy - UAVs
  • Space

Related Documents