Free-Flight Flight & Air Traffic Control. Prof Peter Lindsay Boeing Professor of Systems Engineering Director, ACCS

Transcription

1 Free-Flight Flight & Air Traffic Control Prof Peter Lindsay Boeing Professor of Systems Engineering Director, ACCS

2 Outline of talk Some terminology & concepts The future of Air Traffic Control ATC as a complex system Multi Agent System models Boids & flocking behaviour Evaluation of ATC design options The ATC Workload project Modelling operators Summary & conclusions 2

3 Current Air Traffic Control 3

4 ATC trends & challenges Changing nature of Air Traffic Control: Australian system now entirely computer-based Datalink will (partially) replace radio communications ADSB + GPS will enable radar-like surveillance of whole continent Automated Dependent Surveillance - Broadcast Massive savings possible if airlines can choose own trajectories ( free flight ) Free flight is a fundamental change to operational concept How do we ensure that safety is not compromised, & objectives are achieved? 4

5 A simple ATC sector

6 Terminology & concepts En-route flight phase: > 200km from airports Different separation standards apply: Lateral: 5NM horizontal distance Vertical: 1000 Longitudinal (aka in trail ): 30NM when on same path Also soft standards : eg 10NM intervention standard separation violation: the separation standard is not met Flight plan: 4D trajectory, including time at waypoints 6

7 ATC: The Australian context Australian ATM Strategic Plan released 2002 Plan through to 2017 & beyond Stakeholders include: AirServices Aust (ASA): ATM providers CASA: regulators Airlines international, domestic, regional, Airports International standards bodies: IATA, ICAO Dept of Transport & Regional Services (DOTAR) Lobby groups: Aviation Industry Forum (ASTRA),

8 Key drivers for change Reduced infrastructure costs eg use GPS & satellite comms instead of radar Reduced flight times, fuel use, noise, More flexible airline operations eg negotiate slots on day Desire for increased system predictability eg more reliable arrival times; know how much fuel to carry ATM service market becoming global increased flight ranges sector charge differentials

9 Australian ATM Strategic Plan 7 key strategies including User Preferred Trajectories (UPTs) User = airline Flexible Use of Airspace (FUA) civilian use of military airspace reservation system Conflict Management Move from distance & time separation standards to more flexible risk management approach Decision Information Network Increased information sharing, more negotiation 15 year lead times: 5 ConOps, 5 functional architecture, 5 test & prove

10 User Preferred Trajectories (UPTs) Goal is to optimise for flight distance, time, fuel usage, weather, Applies to en-route control rather than airport vicinity Mainly for international & long domestic (E/W) flights Plan is for staged introduction Flex tracks: to take advantage of jet streams (seasonal) Full freedom (eg great circle routes) will require total rethink of separation paradigm Note findings of Tasman simulator trial

11 Outline of talk Some terminology & concepts The future of Air Traffic Control ATC as a complex system Multi Agent System models Boids & flocking behaviour Evaluation of ATC design options The ATC Workload project Modelling operators Summary & conclusions 11

12 ATC as a complex system A non-linear system Changes at one level can have unanticipated effects at other levels Eg conflict alert Propagation of delays Effect of weather

13 Future National Air-Space (NAS) model NAS-wide traffic patterns Regional traffic patterns Traffic Management Flight trajectories Flight parameters ATC AOC Flight Deck

14 Tackling the free flight problem noise Flight plans ATC 14

15 Tackling the free flight problem (2) Approach to Free Flight ATM: Agents: aircraft(/airlines), weather, airport controllers, traffic controllers, flow controllers Behaviour: flight plans, procedures, Connections: proximity, communications, Emergent properties: safety, congestion, throughput, System safety managed primarily through offline negotiation of flight plans (+ design of procedures etc) Optimise system robustness in pre-flight planning

16 Tackling the free flight problem (2) Provide tools for situation awareness & decision support eg visualise downstream effect of changes Understand how local decisions cascade to global consequences

17 Flocks of aircraft? One way to reduce controller workload would be to have aircraft fly in a cluster Moving sectors Aircraft do their own separation assurance within the cluster Possible application of flocking behaviour? Flocking behaviour follows from 3 simple rules See Craig Reynolds on Boids

18 3 simple steering behaviours Separation: steer to avoid crowding local flockmates Alignment: steer towards the average heading of local flockmates Cohesion: steer to move toward the average position of local flockmates From Reynolds 18

19 Predator & school of fish Can add a 4 th rule: steer away from predator

20 Outline of talk Some terminology & concepts The future of Air Traffic Control ATC as a complex system Multi Agent System models Boids & flocking behaviour Evaluation of ATC design options The ATC Workload project Modelling operators Summary & conclusions 20

21 Evaluation of ATC design options Need a way of evaluating & predicting risk associated with new operational concepts Human operator will play a role in ATC for a long time to come Role of ground-based controller will change from active control to passive monitoring But what exactly should the role be? How to divide responsibility between ground & air? What tools will they need?

22 How to evaluate HCI design choices? Challenge with existing systems is to evaluate safety of different Human Computer Interaction design choices: eg Which software tools to make available & when? What settings? What procedures & protocols to use? How to train operators? Note: existing Human Reliability Assessment techniques are inadequate for this purpose Mostly designed for well-designed sequences of essentially independent activities ATC task is highly interleaved, concurrent & memory-based

23 ATC software tools include Short route probe: indicates predicted position of aircraft at selected time intervals into the future Bearing & range lines: shows aircraft's distance & bearing from selected points can be another aircraft Estimated time of passing: shows time & point where aircraft will come closest Conflict alert warning: indicates that separation violation will occur if aircraft maintain current speed & bearing

24 Overview of SafeHCI approach Our approach to comparing HCI design choices: understand & model operator s cognitive processes as stochastic processes hypothesize what factors affect likelihood and duration of activities within the task conduct experiments to calibrate the models hypothesize effects of design interventions on individual activities predict system-level effects of the design interventions

25 ATC-operator cognitive processes

26 Operator Choice Model

27 Calculating system risk Design individual experiments to calibrate the different parts of the model as functions of the environment & history (= memory) probability of transitions duration of transitions Have developed tool that will calculate overall likelihood of user-specified indicators for user-supplied scenarios e.g. whether a separation violation occurs Use to estimate system-level effect of proposed design interventions vary task & predict/estimate effect on individual transitions calculate effect on indicators

28 Modelling the conflict detection task A conflict is a pair of aircraft that will come within 5NM while at same Flight Level unless controller intervenes What follows is a proof of concept study still underway Objective is to develop a model that emulates operator performance & effect of different tools

29 The different tools The 4 operational concepts being modelled: Baseline: unaided Conflict alert: automated tool that indicates conflict when 50NM apart DOMS Predictor tool: user-invoked tool that calculates Distance Of Minimum Separation (DOMS) Both tools in use 29

30 Operators as stochastic processes Model probability & duration of transitions, & probability of outcomes Use experiments to develop formulae for the above Main factors: DOMS, angle, time to min separation (ttms) Eg Timing for the baseline model (in minutes): scan attend i 1 if IC 0 if NC classify i 0.1

31 Postulated effect of tools on timing Timing for the Conflict Alert case: scan 0.25 attend i 0.1 if alerting 0.25 otherwise 1 if IC 0 if NC classify i 0.1 Timing for the DP tool case: scan attend i classify i 0.5

32 Postulated effect of ttms on accuracy ttms = time to minimum separation Roughly corresponds to urgency ttms: 2, 5, DOMS=0 DOMS=5 DOMS=10

33 Predicted effect on operator performance Likelihood that operator will classify the pair as being in conflict model baseline conflict_alert doms both Free Flight & Air Traffic DOMS Control

34 Outline of talk Some terminology & concepts The future of Air Traffic Control ATC as a complex system Multi Agent System models Boids & flocking behaviour Evaluation of ATC design options Quick introduction to the ATC Workload project Modelling operators Summary & conclusions 34

35 The ATC Workload project Collaboration with UQ s Key Centre for Human Factors & Applied Cognitive Psychology & Airservices Australia Aim: To develop a model that can: Measure the flow of traffic through an air sector Predict the level of workload that an average controller will experience The challenge: model the effect of controller interventions on traffic Also, controllers adapt their behaviour to moderate future workload

36 High fidelity simulations

37 Video-cued recall Medium workload about to decrease

38 Video-cued recall (2) Medium workload about to increase

39 Agent-based modelling We are developing stochastic agent-based models of the full task

40 Agent-based modelling (2) Conflict detection in 3D

41 Agent-based modelling (3) Conflict resolution

42 ATC simulator architecture Sector, route and Waypoint database Aircraft characteristics Flight plans Prerecorded aircraft positions ATC Simulator - updates the position of aircraft over time according to their specified flight plans Track data and amended flight plans Aircraft Agents Pre-recorded Controller outputs ATC Engine Controller Agents Workload and Traffic Flow metrics Visualisations

43 Summary & conclusions Air Traffic Control is undergoing fundamental changes Free flight has the potential for substantial savings & efficiency gains, provided it can be made safe The technology is available, but a lot of research is still needed before a workable operational concept can be implemented Nature can give us inspirations Modelling & simulation has advantages over than experimentation But how can we be sure our models are valid? 43

44 Summary & conclusions (2) ARC Centre for Complex Systems Theme: computation in and by networks Simple agent behaviour + connection topology = complex system behaviour (emergent properties) Emergent properties in this case are safety, efficiency, orderliness, predictabilty, Methods & tools for understanding, managing & controlling complex systems Evaluation of system design options Later: decision support systems Application of complex systems science to ATC eg flocking behaviour

45 Acknowledgements Air Traffic Control program: ATC Workload project is a collaboration between ACCS, UQ s Key Centre for Human Factors & Applied Cognitive Psychology & Airservices Australia (Andrew Neal, Project Leader) Simon Connelly & Junhua Wang stochastic modelling tool Scott Boland traffic replay tool Penny Sanderson & Martijn Mooij hi-fi simulator experiments Katie Duzcmal & Peter Robinson agent model in Prolog Colin Ramsay trajectory modelling Tim Rudge 3D visualiser Jacki Wicks & Rachel Chitoni calibration of conflict detection models Ariel Liebman risk-based conflict management Plus many more