← Back to portfolio

Man On The Loop (MOTL) ISR Supervision

Publication

Investigated a “Man-On-The-Loop” approach for supervising autonomous agent teams in an ISR mission.

Role: Autonomy Research Lead, Simulation Architect, Full Stack Developer & Primary Author

Focus: Autonomous Agents · Control System Supervision · ISR Mission · Multi-Agent Systems · Reinforcement Learning

Outcome: Published at IEEE SMC-B demonstrating effective high-level supervision of agent teams via cultural trait adjustments.

At a Glance

Investigated a “Man-On-The-Loop” approach for supervising autonomous agent teams in an ISR mission.
Used high-level cultural parameters (e.g., risk aversion, individualism) as global control inputs.
Demonstrated that adjusting these group traits significantly influenced collective behavior and mission success.

Visual highlights

Multi-agent ISR supervision concept

Agent role exchange and partner selection workflow — Role exchange workflow illustrates adaptive task redistribution.

The Problem

Individual oversight of large agent teams is impractical in time-critical ISR scenarios.
Direct micromanagement cannot scale, creating the need for new high-level control methods.
No existing method allowed indirect influence on group behavior via global parameter tuning.

The Solution

Built a custom agent-based simulation to test the MOTL concept in a surveillance mission.
Modeled two cultural dimensions (uncertainty avoidance and individualism) affecting all agents’ decisions.
Provided a user interface for adjusting trait sliders and observing group performance in real time.

Architecture Overview

Simulated a team of autonomous agents on a mission with adjustable cultural parameters.
Implemented a supervisory console where the operator could adjust global trait values.
Tracked group and individual performance metrics to evaluate each trait configuration.
Agents followed utility-based norms that enabled cooperative, group-oriented behavior.
A front-end dashboard visualized updated state metrics as parameters changed.

Results and Impacts

Small trait adjustments caused significant shifts in team behavior and mission outcomes.
Validated the MOTL approach: global parameter changes steered agents effectively without low-level commands.
Operators learned trait–outcome relationships and tuned parameters to improve coordination.

Skills and Tools Used

Technique/Skill	Tools/Implementation
Multi-agent simulation	Custom testbed development
Cultural modeling	Hofstede cultural dimensions
Optimization analysis	Utility theory, Pareto rules
Data analysis	Statistical analysis of simulation outputs
Research and collaboration	Academic writing (IEEE publication)

Cross-Project Capabilities

Introduced the MOTL supervision concept, which shaped later multi-agent control research (MAS-2).
Developed agent-based modeling techniques applied to other domains (human factors, robotics).
Combined AI with social science modeling, an interdisciplinary approach reused in other projects.

Published Papers/Tools

Published "Influencing Agent Group Behavior by Adjusting Cultural Trait Values" at IEEE Trans. SMC-B.Paper
Prototype multi-agent simulation platform for MOTL experiments (USAF-funded research tool).