A Model Based Approach to Cognitive Work Analysis and Work Process Design in Air Traffic Control Chr

1
A Model Based Approach to Cognitive Work
Analysis and Work Process Designin Air Traffic
ControlChristoph Möhlenbrink, Bernd
WertherGerman Aerospace Center (DLR), Institute
of Flight Guidance
2
Outline

• 1. Background and Motivation
• DLR project Remote Airport Tower Operation
  research RApTOr
• Why executable models for work process design?
• Aim Better understanding of the controllers
  decision strategies
• 2. FAirControl Formal Airport Control Model
• Human-Machine-Model
• Microworld Experimental Environment
• 3. Formal Analysis Experimental Data
• 4. Conclusion

3
Background and Motivation Remote Airport Tower
Operation research

• Target Application Remote control of a small
  airports
• Experimental System 4-screen videopanorama plus
  pan-tilt-zoom camera
• New Work System Work Process Design

Controller Assistance Augmented
VisionSupport Image Processing/Camera Control
Remote tower controller work station
4
Background and MotivationBasis for Work Process
Design and Tailored Controller Support

• Understanding the temporal- and causal-
  interdependencies at the controller's work
  position
• Understanding controllers strategies in dealing
  with multiple concurrent task demands and
  objectives
• Work definition controller strategy
• A rule for mapping observed traffic situations to
  controller decisions and actions (controller
  clearances)

5
Cognitive Work Analysis (Vicente 1999)for
Leipzig Tower Control
omitted
Cognitive Work Analysis (CWA) results in
FAirControl Model
(5) Tower Controller Competencies Analysis
Role definition Single tower controllerCoordin
ation Radio, Far view Predefined
(4) Organizational Coordination Analysis
(1)-(5) Phases of CWA according to Vicente
(1999)
Goal handling with limited cognitive
resources Goal set
(3) Strategies Analysis
Pre- and post-
conditions, TaskTrig,
TaskCoord,
TaskDel
(2) Control Task Analysis
Cognition

(1) Work Domain Analysis
Aircraftmovements Traffic limitations
Availableinformation
Radio Com.,Far View
Constraints through...
Work Domain
6
FAirControl ModelHuman-Machine-Model
Architecture (Cacciabue (1998))
7
FAirControl Model Transfer results CWA ?
executable Model
8
FAirControl Model Microworld
Experimental Environment
Interaction Model (IA)
FAirControlGUI
HumanModel (PG)
Aiport Process Model (AP)
Controller Subjects
Coloured Petri Net Model
Model Validation
9
Microworld FAirControlExperimental Environment
10
Microworld FAirControlExperimental Environment
Waiting for CL
Waiting for CL
InboundTraffic
OutboundTraffic
DeliverCL
11
Microworld FAirControlExperimental Environment

• cognitive resource constraints

• direct resource conflict (airport resources)

Runwaybottleneck
Apronbottleneck
12
Microworld FAirControl Experimental Environment
13
FAirControl Video
14
2. Microworld FAirControlDecision step ladder
model (Rasmussen 1986)
15
Question of interest Strategies and Goal
Priorization?1. Which effects does a strategy
have?2. What are controllers decision
strategies?
Two strategy hypotheses

• Simple CL Priority List
• 1. Landing CL
• 2. Taxi In CL
• 3. Taxi In Apron CL
• 4. Push Back CL
• 5. Taxi Out CL
• 6. Take Off CL

16
3. Formal State Space Analysis Example for the
generation of a reduced reachability graph
N3
N7
E9
A1

• a)
• single simulation run

17
3. Analysis Example for the generation of a
reduced reachability graph

• a)
• single simulation run
• complete reachability graph

18
3. Analysis Example for the generation of a
reduced reachability graph

• a)
• single simulation run
• complete reachability graph

• b)
• clustering nodes with same reachable end nodes
• identification of branches

• c)
• reduced reachability graph

19
3. Analysis Case Study Identification of
critical decision states

• Scenario (consisting of 6 aircraft)
• Strategy Effects
• 1.) Strategy 1, Strategy 2
• Concurreny Effects
• 2.) GoalSelect, GoaI_Ident
• 3.) ObsRWY, Outbound
• Reduction
• State Space Branching Graph
• Nodes 3713 Nodes 52
• Arcs 4416 Arcs 28
• Status full Status reduced

20
FAirControl Test Trials Comparison of modelled
strategies and subjects responses
21
FAirControl Test Trials Comparison of modelled
strategies and subjects responses
Subject acts likemodelled strategies
Subject acts unlikemodelled strategies
22
Subjects tendencies in priorization
23
Subjects tendencies in priorization
ApronArea
RunwayArea
24
Conclusion

• Model based approach to CWA
• Executable models based on CWA allow to
  investigate temporal- and causal
  interdependencies in highly dynamic environments
• State space analysis/reduction technique permit
  to focus on critical decision states crucial for
  the outcome of interactive scenarios
• Microworld FAirControl
• The experimental environment helps to validate
  the implemented model and CWA results
• The experimental environment allows to compare
  strategies of human subjects to strategies
  implemented in the cognitive model
• Next steps
• Experimental trials with subjects