The Centre for Transport Studies Imperial College London .... Arnab Majumdar.pdf · 1 Imperial College London The Centre for Transport Studies Imperial College London: Developments

1Imperial CollegeLondon

The Centre for Transport StudiesImperial College London:

Developments in airspace capacity and safety research

Dr. Arnab [email protected]


Content• The en-route airspace capacity problem

• How is capacity estimated?

• Capacity analysis projects– a framework for estimating en-route airspace capacity

RAMS simulationcapacity curve

– cross-sectional time-series analysis of controller workloadComplexity factors

• Safety implications


The En-Route Capacity Estimation Problem

ATC Guild Nov 1-2 2004 Delhi


The Indian Context• Rapid growth in air traffic in India:

– Consequences.– predicted traffic growth;

• Infrastructure growth?

• Need for airspace capacity growth: – en-route regions– controller workload

There is a need in an Indian context to:• Prepare for future traffic growth;• Understand controllers crucial;• Workload• Develop methodology to estimate airspace capacity.


The Airspace Congestion Problem• Rapid growth in air traffic in Europe & USA:

– Consequences, e.g.US$ 5bn.– predicted traffic growth;

• Airspace capacity needs to be increased: – en-route controller workload

• New CNS/ATM concepts - new technologies and procedures, e.g. direct routes- future capacity estimation => workload

There is a need to:• Understand the drivers of airspace capacity;• Develop a consistent method to estimate airspace capacity.


The controller workload problem!

What makes en-route capacity different?


– What is controller workload?Confusing term.Many definitions.

– How is it measured?Many methods.

– What is an acceptable level?

Lyons and Shorthose (1993) : shortcomings of capacity measures

Three questions on controller workload


Classification scheme for capacityPerceived Workload Based Measured Workload Based

Declared Capacity MBB Method

MACE Capacity Estimate Task Time Methods

FAA Order 7210.46 “Schmidt” Workload Model

Air/Ground Communications Link*The CARE-INTEGRA Method

* Workload measured indirectly by Air/ground communications link method.


• En-route controller workload:– determines en-route sector capacity

• Current capacity estimation:– controller workload simulation;– workload threshold value.

Sector Capacity:No. of aircraft entering the sector, respecting the peak

hour pattern, when controller workload is 70% in that hour.

Current situation


• Sector entry is the only variable:considerable dispersion.

The problem - I

Workload vs. number of flights entering sector in each hour for 46 sectors in the CEATS region.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0 10 20 30 40 50 60 70

Number of aircraft

Wor

kloa

d, s

econ

ds


• What about other variables?possible relationships;additional effects;univariate vs. multivariate.

The problem - II

Workload vs. total flight time in sector in each hour for 46 sectors in the CEATS region.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0 5000 10000 15000 20000 25000 30000 35000 40000

Total flight time, seconds

Wor

kloa

d, s

econ

ds

Workload vs. Number of Neighbouring Sectors Entry in each hour for the 46 sectors in the CEATS Region.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0 2 4 6 8 10 12

Number of neighbouring sectorsW

orkl

oad,

sec

onds


• Airspace capacity defined by sector entryUseful BUTConsiderable variance.

• Need to consider other variables?Interactions; Quadratic effects.

How is capacity estimated for airspace planning?

The problem - III


International airspace capacity estimation methods



Plan

• MAEVA.

• FTS Approach.

• Analysis approaches.

• Other Issues.

• Conclusions


MAEVA Approach


• Discrete-event simulation model:

- Simulate traffic flow- Events associated => tasks- Task time workload- Controller involvement

• Note:

- events => tasks=>?- validation issues

FTS Approach I


FTS Approach II


FTS Model – RAMS I• RAMS not overtly cognitive, but :

- captures observable tasks- also mental tasks e.g. resolution- workload thresholds - controller based

• RAMS:- > 25 years use in European airspace planning

• Controller:- task input- realistic conflict detection and resolution- simulation & output verification


FTS Model – RAMS II

SECTOR

Sector corner points Sector boundaries

Number of flight levels Number of navigation aids

Number of airports

AIR TRAFFIC

Aircraft type Aircraft performance Flight plan of aircraft

Rules for "cloning" aircraft

CONTROLLERTASKS

Controller tasks Task categories

Task tim ings Conflict resolution strategies

INPUTS

OUTPUTS

RAMSSimulation model

FLIGHT HISTORY

CONFLICT HISTORY WORKLOAD

Actual flight profiles flown ATC interventions to flights

Aircraft involved in conflict Type of conflict

Resolution applied

W orkload recorded for controlling each flight, per controller

W orkload discrim inated by category


Analysis Approaches - TAAM

DFS CENA

90 Workload Points Threshold

Workload BUT No ThresholdTraffic PatternsConflict Figures

Vertical movements

NO workload

TAAM WorkloadThreshold Based

TAAM WorkloadsBUT no threshold

Workload threshold use diminished

Skyguide NAV Canada MLIT Japan

Geometric separations

DFS Workload Formula

FAA

No TAAM workloadmeasurement

N.B. DFS completely modified TAAM workload model

Formula No Formula


Analysis Approaches - RAMS

Formula No Formula

SITCA NAV PORTUGAL LFV AENA

70% Workload Threshold 70% Workload ThresholdTraffic PatternsConflict Figures

Workload BUT No ThresholdTraffic PatternsConflict Figures

Vertical movements

PUMAWorkload

Model

Events

RAMS WorkloadThreshold Based

RAMS EventsOnly

Workload threshold use diminished


Formula ApproachesAirspace Capacity depends upon controller workloadi) C = tWC = Airspace capacity t = thresholdW = controller workload

ii) W = f(X)X = factors affecting workload

“Complexity issue”• DFS Workload Model X= many variables. • EUROCONTROL X = one variable;


DFS Workload Formula I

Number of vertical movementsLevel change (WL4)

Quality and number of the coordination unitsCoordination (WL3)

Number and quality of potential conflictsDistance of closest approach (Severity)

Conflict (WL2)

Number of flights per hourAverage flight time per sector

Movement (WL1)

SymbolWorkload


DFS Workload Formula I I• Used by Germany + Switzerland:

- Well established threshold- Controller acceptance- Airspace assumptions

=> technology=> procedures

• NOT used by Canada + France:- Inaccurate results- Canada => coordination workload- Baseline

• No controller workload issues in FAA + MLIT!


FTS Post-Processing

• TAAM DFS workload model:- max post-processing- Detailed formula

• RAMS:- post-processing - one variable

• Not much post-processing

• Move to complexity- Dynamic density


FTS vs. RTS

• MAEVA process:- FTS => RTS

• RTS expensive:- resources

• Need for formula based:- less need for RTS- DFS, Skyguide

• No threshold:- more need for RTS- CENA, FAA


Controller Resources

• All FTS methods:- Controller input- Validation

• Formula– based methods:- controller support essential- many simulations

• Capacity Panel:- max controller involvement- no workload threshold- CENA, FAA


Other Issues

• Belgium Capacity:- cont.wkld acceptable- technology problems- System capacity?

• AENA Capacity Estimation:- tasks + cognitive- sophistication- resource expensive- PITOT


Conclusions

• MAEVA approach.

• FTS approach.

• Analysis- Formula/ No Formula- Lack of post-processing

• Controller resources:- RTS

• AENA Approach

Moves towards a complexity formula


Airspace capacity: complexity based analyses using simulated

controller workload data



Estimation of en-route capacity of Europe:• simulation modelling

- RAMS;- methodology;

• workload - factors;- analysis;

• capacity curves

Plan


Airspace Capacity Again

Airspace Capacity depends upon controller workloadi) C = tWC = Airspace capacity t = thresholdW = controller workload

ii) W = f(X)X = factors affecting workload

• Analyse factors affecting workload;• Then determine impact on capacity.


What affects controller workload?

CONTROLLERWORKLOAD

RESULTMEDIATING FACTORS

QUALITY OFEQUIPMENT

INDIVIDUALDIFFERENCES

CONTROLLERCOGNITIVESTRATEGIES

SOURCE FACTORS

ATC COMPLEXITY:AIR TRAFFIC

PATTERN ANDSECTOR

CHARACTERISTICS

FACTORS AFFECTING CONTROLLER WORKLOADSource : Mogford et al. (1995), page 5


Literature on variables affecting workload

Air Traffic Factors Sector Factors Total number of aircraft Sector size Peak hourly count Sector shape Traffic mix Boundary location Climbing/ descending aircraft Number of intersection points Aircraft speeds Number of flight levels Horizontal separation standards Number of facilities Vertical separation standards Number of entry and exit points Average flight duration in sector Airway configuration Total flight time in sector Proportion of unidirectional routes Average flight direction Number of surrounding sectors

Previous research indicates:

See Hilburn (2004); Cummings et al. (forthcoming)


En-Route Capacity Estimation IMain features of simulation:• Traffic levels varied systematically

Current (1996) base traffic;Future traffic.

• 122 ATC sectorsContinental European airspace

• Bordeaux Task Base

• Sectors at capacity rules:“Nominal” capacity

Imperial CollegeLondon

Ordinary Least Squares (OLS)

Estimated Generalised Least Squares (EGLS)

Maximum Likelihood (ML)

Prior Studies

Covariogram

Interaction terms

Formulate a model

Test for variance in data

Test for spatial autocorrelationConsequences

Estimation MethodsSpatial autocorrelation present

Assumptions

RAMS OUTPUTAIRCRAFT ATTITUDEWORKLOAD

Assumptions

Capacity Estimation - Analytical Procedures


En-Route Capacity Estimation IIIRAMS output:• Workload;• Flight history.

Functional model formulation:• OLS;• Test assumptions;• Maximum Likelihood.

Spatial correlation:• Estimation;• Variogram.


En-Route Airspace Capacity IV

Factors that affect controller workload:

• Cruise;• Ascend; • Descend x Cruise;• Ascend x Cruise;• Descend x Ascend.


Variable Parameter SE t Intercept 148.54 54.73 2.71

Cruise 56.95 6.25 9.11

Ascend 46.54 8.527 5.46

Descend x Cruise 4.27 0.746 5.73

Ascend x Cruise 1.67 0.634 2.62

Descend x Ascend 4.98 0.947 5.26

Adjusted R2 0.9241

Current Demand Pattern - WLS

N.B. Surface uses Bordeaux Task Base

Results


The Capacity CurveFor current ATC/ ATM environment


The Capacity Curve - Uses

What does the capacity curve predict?• Number of descending traffic in declared sectors

Sector DECLARED TOTAL Declared Diff. WLS (+) Diff. MLE (+)Maastricht 51 12.8 3.1 4.1Luxembourg 41 6.2 12.8 14.3Munich 36 12.6 20.2 21.5Milan 41 11.1 13.4 15.5Reims 28 1.1 27.1 29.5

N.B. Cruise and Ascend traffic same as declared


Methodology issues

A framework to estimate airspace capacity:• Simulations using RAMS:

systematically vary traffic;

• Analytical framework:Assumptions;Spatial analysis.

• Methodology provides:Capacity curve;

• Framework applicable to other scenarios


Simulation analysis extended

Current analysis:

Q. What are the affect of variables on workload in sectors throughout the day?

1. Interview controllers into factors.

2. Use RAMS simulation – based methodology

3. Functional analysis of output using cross-sectional time-series.


CEATS Airspace Regions


MFF Airspace Regions


CEATS vs. MFF

FewManyNeighbours

Very LargeSmallSector Size

Long (15 mins)Short (5mins)Transit times

ManySmall-MedTraffic

MFFCEATSFeature


Individual models:Time Series: Wt = α + X'βt+ εit => for each sectorCross Sectional: Wi = α + X'βi+ εi => for each timePool the data?

Method


Fixed Effects Time-Series Cross-Sectional Model (sector level)wit= αi + x'itβ + εit

wit = workload in sector i at time tαi = effects of var. peculiar sector i, constant over timeX'it = variables in sector i at time tβ = coefficientsεit assumed:

i.i.d. over individuals i (the sectors) and time; mean zero and variance σε2

Estimators from T-S C-S model are more accurate:Greater efficiency cf. c-s or t-s; Note estimated αi

Model specification - test residuals:Temporal correlation;

Method - II


Why Panel Data?Baltagi (1995):

Control for individual heterogeneity.

More informative data, more variability, less collinearityamong the variables, more degrees of freedom and more efficiency

Study the dynamics of adjustment

Identify and measure effects not detectable in pure cross-sections or pure time-series data.

Construct and test more complicated behavioural models

Gathered on micro units, such as individuals, or in the case of capacity analysis, ATC sectors.


CEATS variablesMajor findings:

• Flight profiles significant:Cruise-descend => +37 secsCruise-climb => +12.5 secsClimb-climb => +49 secs

• 1 sec of total flight time => +0.012 secs workload;• Increase of 1 FL => -1 second workload;• 1 nm/h speed diff => +0.32 secs workload;• Neighbouring sectors entry significant: • Entry attitudes significant BUT not exit attitudes: • Temporal correlation statistics:

Modified D-W and Baltagi-Wu; Indicates temporal autocorrelation.


MFF ResultsModel indicates:

• Flight profiles significant:Cruise-CruiseCruise-climbCruise-descendClimb-climb

• Speed difference variable• No FL effects;• No exit point effects• No Neighbouring sectors entry/ exit effect:

Could be MFF region effectNot many neighbours

• No Temporal Effect?


Panel data issues• Example MFF variables => not same as CEATS

• MFF analysis indicates:Flights profiles significantspeed differential quotient significantSectors – neighbouring not significantSectors - FL differences not significant

• MFF airspace area characteristics?Flights profiles significantBUT not sector


• RAMS Simulation methodology:CEATS and MFF Region;Variable outputs;Better geographical output.

• Hour-by-hour analysis more complicated than peak hour:Greater dispersion in data.

• Panel data analysis:More variables than for peak hour;Aircraft and sector variables;New variables.

Panel Data Methodology valid• Variables differ between CEATS and MFF Regions• External validity

Conclusions


Safety impacts of controller workload



• Increase in traffic:

– Controller workload increase;

– Are skies unsafe?

• Evidence from incidents:

– UK;

– New Zealand.

A Safety Problem


• MATS Part I/ MORS occurrence categories:

– conflict;

– infringement;

– loss of separation;

– ATC Overload => v.rare loss of separation;

– level bust

• Controller caused AIRPROXES

UK Occurrence Categories


Loss of separation occurrences 1998-2000• Main causes : Percentages

1/3 cases of misjudgement:- speeds and/ or descent or climb rates of the aircraft- update of performance

12

20

51

134

1999

76High controller workload

1925Poor coordination / handover

5248Failure to separate/ poor judgement

128141Number

20001998


Controller CAT Airprox Causes 1998 -2000Cause Grouping Total Number Percentage

FAILURE TO SEPARATE/POOR JUDGEMENT 92 55.09ERROR BY TRAINEE CONTROLLER 11 6.59LACK OF CO-ORDINATION BETWEEN CONTROLLERS 11 6.59FAILURE TO ADHERE TO PRESC'D PROCED'S/OPERAT INSTR'S 8 4.79INADEQUATE SUPERVISION 7 4.19UNDETECTED READBACK ERROR 6 3.59CONFUSION OR POOR COORDINATION AT HANDOVER 5 2.99INAPPROPRIATE ATC INSTRUCTIONS, USE OF INVALID FL 5 2.99COLLAPSED-SECTOR WORKING(BANDBOXING)/HIGH WORKLOAD 4 2.4FAILURE TO PASS OR LATE PASSING OF TRAFFIC INFO 4 2.4INADEQUATE AVOIDING ACTION/LACK OF POSITIVE CONTROL 4 2.40DISTRACTION / FAILURE TO MONITOR 3 1.80OTHER ATC CAUSE 2 1.2CALLSIGN CONFUSION 1 0.60ERROR BY SUPPORT STAFF 1 0.60INADEQUATE/INCORRECT DATA DISPLAY 1 0.60MISUNDERSTOOD RT/APPARENT LANGUAGE DIFFICULTY 1 0.60UNFAMILIARITY WITH EQUIPMENT/LACK OF TRAINING 1 0.6


Based on MORS reports:Overloads I

However no sector characteristics data


Main features:

* 1998 one overload for weather = 9hrsTSF = 80%Theoretical Capacity

Overloads II

Year Number TMAs/ airports

Average (minutes)

Maximum (minutes)

Overloads where TSF not exceeded (%)

1998 59 8 52.95 (44.09)* 540 58.33 1999 54 8 40.31 120 51.61 2000 43 8 31.28 90 61.54


Main causes: PercentagesOccurrence example: Overloads III

9.7611.1116.92Weather and

winds

13.9515.7918.18Procedures,

coordination

30.1936.8416.92Traffic Complexity

0.991.080.95Aircraft/ min

in bunch

40.3241.4640.00Bunching Cause

200019991998


New Zealand Loss of Separation• Loss of separation incidents 1994-2002• N.Z. < U.K. air traffic

Loss of separation

18.71

14.39

4.871.3313.19

17.51

2.21

11.5Task overload+ Informationoverload + Time shortageRisk misperception

Lack of knowledge.

Poor instructions procedures.

Inadequate checking

Poor concentration or lack ofattentionNegative task transfer, habits

Psychological other


New Zealand vs. UK• %age due to high workload > in NZ

•“Sperandio” effect?- high workload situations

- efficient strategies

• Evidence from UK supports US evidence:- not many errors at high workload.

• Current analysis:- what causes UK overloads- one sector analysis.


Not high workload?• Evidence from US, Canada, NZ errors due to:

- low workload situations- lack of attention, boredom- beware new technologies

• Realise air traffic controller crucial:- analyse any errors made- human factors approach- use errors to learn, not punish- reporting scheme- “just culture” of safety

• Current analysis:- air traffic incident precursors in ATC- seven nation analysis.


A just culture for safety


Conclusions

– Rapid growth in aviation

– Recognise crucial role of air traffic controller

– Controller workload for capacity

– Understand why errors are made

– Available analytical methodologies

– Human Factors

– A Just Safety culture


Acknowledgements

Big Thank You to

ATC Guild India

Airports Authority of India


Websites

Dr Arnab Majumdar [email protected]

Dr Washington Y. Ochieng [email protected]

Centre for Transportation Studies web-sites:

– www.cts.cv.ic.ac.uk

– www.cts.cv.ic.ac.uk/geomatics

Documents

The Centre for Transport Studies Imperial College London .... Arnab Majumdar.pdf · 1 Imperial College London The Centre for Transport Studies Imperial College London: Developments