Lessons learnt from ERATO

1
Lessons learnt from ERATO

• M. Leroux
• CENA

2
Two metaphors in tools design
Information
Cognitive Engineering
EC PC Tools
Prosthesis
Decision
3
A model-driven approach

• Default reasoning sentry parameters monitoring
• Ambiguity elimination process relevant
  parameters monitoring
• Tightly packed data flows processing
• Interruptions processing
• Memorization tasks
• Cognitive resources management

4
MILESTONES

• Launch end 1985
• 1991 First experiments on the cognitive model
• ...
• Sept 1997-May 1998 Evaluation of the JHMS in
  simulated environment
• Method of working,
• Usability of the interface,
• Algorithms,
• Training.
• End 2001 Evaluation of the JHMS on real traffic
• 2004 Operational deployment

5
ERATO ENVIRONMENT
6
THE TRIALS

• 8 teams of 2 controllers
• 3 weeks each
• 2 weeks training
• 1 week trials
• 2 days cognitive activity and cooperation
• 90 minutes traffic simulation
• deferred interviews using simulation replay (4-5
  hours)
• 1 day usability
• 1 day fatigue

7
DATA INFORMATION FLOWS
TID
TID
EC
PC
NEIGHBOUR SECTORS
PILOTS
8
DATA INFORMATION FLOWS
ambiant sound microphone
19 
telephone
TID
TID
TID
PIL2
EC
PC
PIL1
radio
9
DATA ENCODING

• Data automatically displayed on ECs HMI
• Input and queries made by the EC on HMI
• Transcription of real time speech (EC, PC,
  Pilots, Neighbours)
• Encoding these transcription by means of
  activity predicates
• Data automatically displayed on PCs HMI
• Input and queries made by the PC on HMI
• Transcription of deferred verbal protocols

10
Verify that the tools supports cognitive
mechanisms (1)

• With reference to the cognitive model
• Identify samples of cognitive activity,
• For each sample, describe the role of each
  function
• As support,
• As a potential cue that has triggered cognitive
  activity,
• As a potential consequence of cognitive activity.

11
Verify that the tools support cognitive
mechanisms (2)

• Describe the interaction between cognitive and
  perceptive activity
• Evaluate the pertinence of decision aids
• Improve the HMI
• Improve the working method
• Improve training
• Elicit the laws governing the JHMS

12
The main tools

• Problem driven information filtering
• Less than Conflict Probe
• makes easier conflict detection, but doesnt show
  conflicting A/C
• More than Conflict Probe
• displays the resolution context
• Only on ATCOs request

13
The main tools

• The reminder
• A friendly MTCD
• Acting as a sketch that must be adapted by the
  ATCO
• Combined with a friendly Task Scheduler
• 94-95 trials showed how the working method is
  critical

14
RESULTS

• Frequencies - Duration
• Differences between controllers
• Differences between filtering and reminder
• Differences between EC and PC

15
RESULTS

• Default reasoning
• Verify that tools support DR
• Efficient use of both functions is based on DR
• A/C expected behaviour vs trajectory prediction
• Minimise data presentation
• Efficient use vs handling
• All data are interpreted (may be mis-)
• Evoking is more powerful than describing
• description is limited and clumsy,
• evocation meets default reasoning and implicit
  cooperation, values skills and know-how, creates
  complicity

16
RESULTS

• The reminder
• 23 of problems deleted at the first consulting
• Well accepted (mental picture)
• But preserve stability of data presentation
• Avoid over reactivity of the algorithms
• The dynamics of the JHMS must be operator driven,
  not machine driven
• 37 problems created (4 )
• Based on the (expected) intents of the
  controller(s)

17
RESULTS

• The time lines
• The genesis of all the clearances
• stories
• Different strategies
• For cooperation
• To use the tools
• In relation with the demand
• The cognitive YIELD of tools handling and of
  cooperation

18
Conclusion

• Evoking is more powerful than describing
• To make the operator feeling comfortable,
  preserve
• Non determinism,
• Interruption management,
• Focus of attention,
• Cognitive resources management,etc
• We cannot verify or validate a tool independently
  of the working method