Human Supervisory Control Issues in Unmanned Vehicle

Operations

Mary (Missy) CummingsHumans and Automation Laboratory

http://halab.mit.eduAeronautics & Astronautics

(617) 252-1512MissyC@mit.edu

Humans & Automation Lab

•Former U.S. Navy officer and pilot

•Systems engineer with a cognitive focus

•Research Interests: Human supervisory control, decision support design, human interaction with

autonomous systems, design of experiments technology development, social impact of

technology

HAL Director

Task

Displays

Human Operator(Supervisor)

Computer Actuators

Sensors

Controls

• Humans on the loop vs. in the loop

• Supporting knowledge-based versus skill-based tasks

• Network-centric operations & cognitive saturation

Human Supervisory Control

Ten Areas of Concern

• Information overload • Attention allocation• Appropriate levels of automation• Adaptive automation • Decision biases• Distributed decision-making through team

coordination• Complexity • Supervisory monitoring of operators  • Trust and reliability• Accountability

Information Overload

0

4.5

1.5

Workload

Pe

rfo

rma

nc

e

Low HighModerate

Good

Poor

1414 1313 1414N =

Target Interarrival Time

2 Minutes4 Minutes

Ove

rall

Pe

rfo

rma

nce

Me

an

(+

/- 1

Sta

n.

De

v)

260

240

220

200

180

160

140

Missiles

8

12

16

Attention Allocation

• Multiple HSC tasks = Divided attention problem

• Information uncertainties & time latencies• Preview times & stopping rules• Primary task disruption by secondary task

– Chat

Level Automation Description

1 The computer offers no assistance: human must take all decision and

actions.

2The computer offers a complete set of

decision/action alternatives, or

3 narrows the selection down to a few, or

4 suggests one alternative, and

5executes that suggestion if the human

approves, or

6 allows the human a restricted time to veto before automatic execution, or

7 executes automatically, then necessarily informs humans, and

8 informs the human only if asked, or

9 informs the human only if it, the computer, decides to.

10 The computer decides everything and acts autonomously, ignoring the

human.

Appropriate Levels of Automation

Adaptive Automation

• Dynamic role allocation– Mixed initiatives

• A problem of intent

• Cueing mechanisms– Psychophysiologi

cal– Decision

theoretic– Performance-

based

Decision Biases

• Naturalistic Decision Making– Dynamic ill-

structured problems with shifting goals (i.e., NCW)

– Heuristics good & bad

• Biases– Confirmation– Recency– Automation

Distributed Decision-making & Team Coordination

• The move from hierarchical, centralized to decentralized control

• Team mental models & shared situation awareness (SA)

• Decision support– Automated agents as

team members

• Not just an issue for human teams– Swarming UAVs

Complexity

Supervisory Monitoring• Nested supervisory control• Two basic issues: Recognizing &

intervening• Interventions

– Redistribute workload– Adding team members (both human &

computer)– Modify mission objectives

Trust & Reliability

Accountability

The Future of UVs and NCO

• We can’t do it without automation & intelligent autonomy– Bounded Collaboration

• Human-centered design vs. mission-centered design– Unmanned systems do not really exist– The systems engineering process must

consider humans early– Robust systems are needed for both human

and automation brittleness considerations