1. Management of Multiple Dynamic Human Supervisory Control Tasks
MIT-Boeing Research Project Review Meeting – Feb. 2, 2005
Management of Multiple Dynamic Human Supervisory Control Tasks
Paul Mitchell – MIT Humans and Automation Lab

2. Outline
Motivation
Experiment Objective
Experiment Scenario
Experimental Design
Human-Automation Interaction Modeling
Wait Time Modeling
Key Experiment Display Elements
Expected Research Results / Benefits
Multi Aerial Unmanned Vehicle Experiment (MAUVE) Program Demo
HAL Lab Tour

3. Motivation
Consequences of NCW
Volume of information
Number of information sources
Operational tempo
Greater attentional demands on operators
Efficient attention allocation becomes critical to human & system performance!

4. Experiment Objectives
To investigate:
How operators cope with managing multiple HSC processes simultaneously
What kinds of decision support can aid operators in these situations
What effects human performance limitations have on the overall system

5. UAV domain has immediate military applications NCW concept of swarms
Experiment Scenario
UAV domain has immediate military applications
NCW concept of swarms
Subject is an operator supervising four separate, independent unmanned aerial vehicles (UAVs)
Objective:
To destroy a set of targets (which may change) within a certain time period, while taking minimum damage from enemy air defenses
Predator and Global Hawk, both UCAVs currently in use.

6. Experimental Design – Independent Variables
Level of Decision Support (Scheduling Assistance)
Between subjects
4 levels
Manual = LOA 1
Passive = LOA
Active = LOA 4
Super Active = LOA 6
Amount of Schedule Re-Planning
Within subjects
3 levels
None
Infrequent
Frequent
The Patriot missile battery, a prominent example of a high level of automation in use today.

7. Experimental Design – Dependent Variables
Primary task performance – Number/priority of deadlines missed
Performance score
Combines target and threat events
To provide insight into overall test session performance
Secondary task performance – Chat Box
Percentage correct answers Situation awareness metric
Average time to respond Workload metric
Wait Times
Result from deviations from “ideal” mission plan

8. Human-Automation Interaction Modeling
First proposed by Olsen and Wood (2004) with regard to traditional human-robot interactions
Interaction Time (IT)
The human operator is actively engaged in improving the performance of the vehicle, allowing overall mission accomplishment to occur
Neglect Time (NT)
The vehicle is operating autonomously, needing no input from an operator to continue its mission
Wait Time (WT)
The vehicle needs input from the human to execute its mission

9. Human-Automation Interaction Modeling

10. Workload Wait Times (W-WT) Result from operator overload
Wait Time Modeling
Wait times dramatically impact system performance and risk of failure in time-critical applications (eg. C2)
Two main categories
Workload Wait Times (W-WT)
Result from operator overload
Situation Awareness Wait Times (SA-WT)
Result from loss of situational awareness
Work-in-progress
Need to accurately model then measure individual WT components
WT can be further broken down

11. Key Experiment Display Elements – Mission Plan
Current mission plan for each UAV is shown
UAV that operator is currently interacting with highlighted in green
Active targets = red diamonds
Threat areas = yellow circles
Way Points = black triangles
Loiter Points = directed circles

12. Key Experiment Display Elements - Decision Support
Separate marching timeline for each UAV
Represents mission plan as laid out on map display
Current/future tasks color coded by action 1 3 2 4

13. Expected Research Results / Benefits
Validation of wait time models
Conclusions on how different types of wait times influence the overall cost function and fan out
Workload predictive model based on wait times
Further results on the validity of an imbedded chat box as a measure of secondary workload
An evaluation of the timeline decision support tool, with comparisons across its various levels and effectiveness under different re-planning conditions

14. MAUVE Demo & HAL Lab Tour
Questions?