1 of 13

A Layered Architecture for Adaptive Autonomy in Human-Robotic Team Interaction

Mark Allison∗ , Alice Gaspard , Max Parks

Department of Computer Science,

University of Michigan – Flint

Felicia Jefferson

Department of Biochemistry & Molecular Biology

University of Nevada - Reno

Funded by the National Science Foundation Award # 2220513.

2 of 13

Overview

  • Human – Robotic Team Interaction
    • Levels of Autonomy
    • Human Factors
  • Adaptive Autonomy Architecture
  • Preliminary results

3 of 13

Problem Domain – Shared Autonomy

  • Roles are based on strengths
  • Human becomes a single point of failure for the system – need for adaptive autonomy

4 of 13

Levels of Autonomy

1 Manually operated by human

2 Team offers decision alternatives

3 Selections are reduced

4 One alternative offered

5 Execution with approval

6 Team allows the human limited time to veto before automatic execution

7 Team executes then informs the human

8 Team informs human after execution if requested

9 Team informs human after action execution

10 Fully Autonomous Team

5 of 13

Human Factors

  • Situational Awareness
    • Perception
    • Comprehension
    • Projection
  • Mental Workload

6 of 13

Human Factors (Cont.)

Wickens’ Information Processing Model

7 of 13

Assessing Human Factors

8 of 13

Architecture

9 of 13

Behavioral Model

10 of 13

Experimentation - Simulation

  • Clear building problem
    • An infantry platoon is charged with clearing and securing a number of buildings.
    • The platoon is comprised of a team of small terrestrial robots paired with a human.
    • A cleared building needs to be manned to ensure it remains secure.
    • The active force becomes diminished after each clearing.
  • Repast Platform

11 of 13

Preliminary Results

12 of 13

13 of 13

References

C. D. Wickens and C. M. Carswell, “Information processing,” Handbook of human factors and ergonomics, pp. 114–158, 2021.