Robot Self-Assessment for Informed Resource Management

Abstract

This proposal centers around the key challenge of supporting informed human decision-making about robot resource consumption. For example, the range of a mule robot will be impacted by how much weight it carries, so soldiers may want to be strategic about when tocarry their own gear. Likewise, multiple soldiers using a delivery robot for logistics during camp operations will need to know what factors impact robot energy consumption. This challenge is more complex than a simple gas or battery level gauge since terrain, tasks, and other factors can influence energy use. In an ideal scenario, the robot will combine learned models of desired performance with self-awareness of expected energy consumption to generate predictions and behaviors that allow informed decisions by the soldiers.Specifically, we seek to address two use cases: (1) learning resource utilization preferences for both individuals and groups and(2) supporting energy resource decisions by groups of robot users who may not all share the same goals. These two scenarios are likely to occur across DoD applications but require basic science advances. For example, a robot mule may be assigned to a specific squad, but the preferences and performance goals will shift based on where and how the squad is deployed. Likewise, users within a teammay have subgoals that unknowingly interfere with overall team resource utilization goals. The robot can help mitigate this throughself-assessment, transparency, and communication. To achieve these research goals, the team seeks to extend prior work in two ways.First, we plan to extend our work from learning preferences in single-expert, singe-objective scenarios to groups of experts and for multi-objective situations. This is especially important when preferences are not a continuous curve, but a set of clusters (e.g.,large shifts in robot behaviors between scenarios). Second, we will focus on robots providing information to a team of users about its decision-making process on resource utilization. This is important since teams of humans can have different subgoals, while still pursuing an overall team goal. Project outputs will include new algorithms, research publications, and robot software.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 12, 2025
Source ID
N000142512174

Entities

People

  • Aaron Steinfeld

Organizations

  • Carnegie Mellon University
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Organizational Process Management (OPM).
  • Robotics and Automation.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - DoD AI Strategy
  • Autonomy
  • Autonomy - Autonomous System Control
  • Autonomy - Human-Robot Interaction