B6: HumanAutomation Interactions in Rail Network Control

1
B6 Human-Automation Interactions in Rail Network
Control

• University of Nottingham and University of
  Southampton

Sarah Sharples, John Preston, Nora Balfe, John
Armstrong, Laura Millen, David Golightly Alex
Stedmon
2
RRUK B6 aims and activities

• Explore and define technical and organisational
  systems that best allow hybrid control
  (automation plus people)
• Identify decision and communication support that
  make best use of staff skills to interpret,
  prioritise, intervene optimise
• Activities
• Analysis of workload with different levels of
  rail signalling automation
• Identify whether ironies of automation apply
• Apply cognitive work analysis and ethnography to
  map joint cognitive system in rail control
• Identify appropriate current and future
  automation concepts

3
Example of Automation rail signalling

• Machine assumes a task previously performed by a
  human

4
Automation in rail control ARS

• Automatic Routing System
• Runs off the timetable
• Checks timetable database for the timetable
  associated with a train approaching the control
  area
• Sets route for train according to the timetable
• Conflict Resolution Algorithms
• Decision mechanism to decide which train to route
  first when trains are running late or out of
  sequence

5
B6 activity a recap

• A 5 level framework for describing different
  levels of rail automation
• A coding strategy for real-time observation
• Video observation, interviews and structured
  real-time observation
• Apparent differences in level of activity under
  ARS
• Provisional indications of a non-linear
  relationship with workload
• Variation in attitudes
• Trust and a tool of the job
• A nuisance
• Switch it off
• Variation in strategy
• Some support for the ironies of automation

6
Recent activity Extended Observation Studies

• Aims
• To establish levels of intervention in ARS
• To investigate individual differences in the use
  of ARS
• To investigate differences between signal boxes
• 5 basic activities
• Monitoring
• Interaction
• Planning
• Communications
• Quiet Time
• Sampling every 5 seconds
• Observations carried out at same time of day for
  90 minute period
• Signaller observation studies at 6 workstations
  in 3 IECCs
• Comparison with non-ARS VDU (4 workstations in 2
  boxes), and NX (2 panels in 1 box 1 box to do!)

7
Results VDU with ARS
8
Results VDU without ARS
9
Results NX data
10
Provisional interpretation

• With ARS
• No apparent difference in time interacting with
  workstation panel
• People are swapping monitoring for quiet time
• Could be taking them out of the loop
• Non-ARS and NX
• Very similar patterns of behaviour
• Possibly less communication with NX
• VDU has no shared display
• Variability
• Through signaller strategy, and workstation
  difficulty
• But maybe also influenced by level of automation

11
Recent Activity Cognitive Work Analysis

• A means to describe work domains, functions and
  tasks
• Is both holistic and formative
• Application in process control, aviation,
  military and beyond
• For design and procurement
• Uses five representations
• Work domain analysis
• Control Task Analysis
• Strategies analysis
• Social, organisational and co-operation analysis
• Worker competencies analysis
• For an introduction see http//www.cognitivesystem
  sdesign.net/
• For our work see Millen et al (2009) The
  Application of Cognitive Work Analysis in Rail,
  Proceedings of Rail Human Factors Conference,
  Lille, 2009

12
Understanding the work domain
13
Analysing strategies
14
B6 within a broader context
15
Remaining work

• Collect more NX data
• Statistical analysis
• Comparison with workload data
• Does change in perceived workload vary due to
  automation?
• Review latest developments in automated and
  hybrid train control systems
• Supporting Network Rail in applying CWA to
  understand future rail control

16
Any Questions?