Decentralized Path Planning For Air Traffic Management Wei Zhang

Transcription

1 Decentralized Path Planning For Air Traffic Management Wei Zhang Advisor: Prof. Claire Tomlin Dept. of EECS, UC Berkeley 1

2 Outline Background National Aviation System Needs for Next Generation Air Traffic Management Systems Air traffic control system from a control perspective Hierarchical Decentralized Flight Planning Problem Formulation Solution Procedure Advantages Over the Current Planning Procedure Simulation Results Conclusions 2

3 Motivations National Aviation System is a large-scale Cyber-Physical System 14,500 traffic controllers, 4,500 safety inspectors, 5,800 technicians, 19,000 airports, 600 traffic control facilities, 50, flights each day Physical components: large number of aircrafts, equipment and human agents Cyber components: traffic & weather measurements, computation, prediction and communications. Research Perspectives: FAA, traffic controllers, airline companies My focus: System-level modeling and optimization methods for en-route traffic management and terminal area operations 3

4 The Needs for Next Generation ATM Air traffic delays in 2007 has cost US economy $41 billion fuel: 740 million gallons, carbon dioxide : 7.1 million tons Staffing Emergency in major ATC facilities across the nation As of 2008: 11,077 certified controller s lowest level in 15 years 10,000 are expected to retire before 2015 Oakland Center: training ratio: 2-1 vs 12-1 in 2005 operational error: 30 vs 14 in FY07 planning to hire 12,000 before 2018 Jan, 2010 Certified TRACONs controllers plummeted more than 25% in the last six years New York reaches post-1981 low Situation gets much worse due to the expected two- to three-fold increase in air traffic Need to modernize, (semi)- automate the ATC system NOW 4

5 Challenges Legacy systems require continuous operations Critical Safety Standards Large number of competing users Human in the loop fear of new working conditions TRACON controllers are still using the same Radar system as they did in 1960s. Gradual change Respective the structure of the system 5

6 Background of ATM Plan negotiation Hold/ Clear FAA AOC/Pilot FMS Tower TRACON 6

7 Hierarchical Control Structure of ATM FAA-ATM Weather forecast Flight tracks clear hold AOC/Pilot plans FMS Aircrafts Switching control Spatially distributed local controllers Towers, TRACONs, Centers, Sectors 7

8 Lack of Collaborative Information Exchange ZLA27 ZOA33 ZLAED ZLA16 ZLC45 ZLALE ZLA38ZLA37 LAX ZLA40 ZLA34 ZLA39 SAN ZLC33 ZLA32 LAS ZLA36 ZAB92 ZLC34 ZDV24 ZAB67 ZAB50 PHX DEN CVG STL ZDV04 ZDV25 ZKC30 ZDV39 ZKC20 ZID91 ZID93 ZKC02 ZKC90 ZDV30 ZID92 ZKC22 ZKC28 ZKC29 ZID96 ZDV38 ZID94 ZKC06 ZAB71 ZKC23 ZKC27 ZME24 ZME61 ZTL15 ZAB70 ZME19 ZME23 ZAB95 ZAB93 ZAB97 ZFW49 ZFW50 MEM ZTL36ZTL40 ZME20 ZAB94 ZAB68 ZAB87 ZME32 ZFW47 ZFW48 ZME44 ZFW42 ZTL02 ZFW93 DFW ATL A major problem: lack of information exchange User does not know the traffic information only weather briefing is available before taking off FAA/ATM does not know users preferences Consistent situation awareness is needed 8

9 Benefit of Information Sharing ZLA27 ZOA33 ZLAED ZLA16 ZLC45 ZLALE ZLA38ZLA37 LAX ZLA40 ZLA34 ZLA39 SAN ZLC33 ZLA32 LAS ZLA36 ZAB92 ZLC34 ZDV24 ZAB67 ZAB50 PHX DEN CVG STL ZDV04 ZDV25 ZKC30 ZDV39 ZKC20 ZID91 ZID93 ZKC02 ZKC90 ZDV30 ZID92 ZKC22 ZKC28 ZKC29 ZID96 ZDV38 ZID94 ZKC06 ZAB71 ZKC23 ZKC27 ZME24 ZME61 ZTL15 ZAB70 ZME19 ZME23 ZAB95 ZAB93 ZAB97 ZFW49 ZFW50 MEM ZTL36ZTL40 ZME20 ZAB94 ZAB68 ZAB87 ZME32 ZFW47 ZFW48 ZME44 ZFW42 ZTL02 ZFW93 DFW ATL With the traffic information User can find the best path (according to its specific preference) to avoid traffic according Decide whether to delay the flight or take the best available detour 9

10 Towards a New Flight Planning Framework A framework with planning algorithm deal with large number of aircrafts in real time consider both weather and traffic restrictions, guaranteed safety with certain optimality for the nominal trajectories 4D trajectory (3D + time) practically feasible 10

11 Graph of Airways ( ) Spacial graph Gs = Vs, Es vertices (nodes): waypoints (Navigation aids, airports, virtual waypoint) Edges: airways of jetways Space-Time Graph G = ( V, E ) V = ( x, t) : x V, t = 1,, N { } G V E Nodes are disconnected within the t t +1 t + 2 Nodes are disconnected within the same layer Edges between layers determined by the dynamics of the aircraft t t +1 11

12 Planning Under Weather Uncertainty Link weight ( length ): t ( i, j ) l v v Fuel cost; expected turbulence based on weather forecast; infinite when crossing forbidden weather zone t +1 t + 2 Single aircraft path planning with weather data is a shortest path problem Departure nodes Departure time z 0 x 0 x f = t z f = 0 t f (, ; ) t 1 0 λ φ( ) (, t t t ) = + f J z u z l z u f Need to handle sector capacity constraints t Destination Latest arrival time 1 i S j i ( xt ) max sector counts, t j 12

13 Planning with Traffic Restrictions Current way for Traffic control: speed variation, ground delay program, holding pattern, vector for spacing, redirecting Traffic Regulation Function: λ ( j, t) [ tt ] 0 sector j open over, + 1 λ ( jt, ) = otherwise Each aircraft tries to minimize its own cost subject to the traffic rules specified by FAA ( ) ( ) i i i i i i 0, ; λ = φ + (, ) + λ, ( ) 1 J z u z i l z u z f i t t t t j j t Sj t infinite link cost if crossing forbidden weather zone infinite price if sector sold out over certain time period Safety and satisfy all sector constraints 13

14 Decentralized Path Planning Algo Planning /Rerouting Algorithm 1. Get weather data and traffic Update λ restrictions 2. Solve the shortest path problem 3. File the plan λ ATM approve and update traffic rules λ ATM Pilot/AOC Traffic & Weather ( ) λ j, t is a tool for the ATM to regulate traffic the above is First-Come-First-Serve rule can achieve certain fairness by using the historical data nominal plans are safe but capacity buffer is needed to cope with uncertainty 14

15 Distinctions and Advantages Traffic Flow Management Bertsimas 98, Waslander 08 Path Planning with Weather Uncertainty Nilim (ACC03), Pannequin (GNC07), Kamgarpour (CDC10) mostly centralized and only works for a small number of aircrafts require same taking off time does not consider traffic information Distinctions of our methods decentralized used for the entire NAS or different subregions of NAS planning considering i weather and traffic 4D trajectory (3D + time) guaranteed safety with certain optimality respect current planning procedure, practically feasible in the near future 15

16 Simulation Results I 30 sectors, 2 deterministic weather zones, 12 airports, 100 flights randomly select departure and arrival airports, random departure time plans are made and filed in the order of departure time 16

17 Simulation Results II ights Number of Fli Operational Evolution Plan (OEP) Airports -- about 74% passengers and 69% operations GMT (Hour) Flight schedules among OEP airports -- Aug. 24, 2005 We consider 34 OEP airports (except HNL) Consider flights depart between 7am EST and 5pm EST Proof of concept: the framework works for realistic traffic patterns and realistic number of flights no weather data and no comparison with real flight tracts assume all flights try to minimize travelling distance uniform grids corresponding to roughly 3 minutes flight time 17

18 Unconstrained Flight Plans 18

19 Traffic Regulation Results 50 1 Latitude Latitude ZSE12 ZSE07 ZSE03 ZSE31 ZSE11 ZLC20 ZLC17 ZMP23 ZMP24 ZMP25 ZSE01 ZSE02ZSE32 ZSE47 ZSE48 ZMP13 ZBW01 ZLC15 ZMP11 ZSE42 ZSE16 ZLC06 ZSE46 ZMP20 ZMP16 ZMP12 ZMP19 ZMP15 ZBW08 ZBW53 ZBW02 ZLC41 ZLC16 ZDV45 ZDV32 ZMP18 ZMP17 ZBW61 ZLC08 ZOB19 ZBW09 ZBW39 ZBW23 ZSE14 ZSE13 ZOB38 ZDV35 ZMP29 ZBW76 ZBW10 ZBW38 ZSE15 ZMP30 ZBW17 ZLC07 ZLC05 ZBW46 ZBW94 ZBW84 ZBW85 ZOB29ZOB26 ZDV34 ZDV33 ZOB79 ZNY34 ZOB47 ZBW20 ZDV09 ZBW18 ZBW33 ZOB49 ZOB59 ZMP43 ZNY49 ZNY75 ZLC42 ZLC40 ZLC03 ZDV14 ZMP42 ZBW91 ZBW45 ZBW47 ZNY73 ZNY42 ZNY56 ZBW31 ZDV16 ZDV08 ZOA36 ZOA31 ZOA43 ZLC04 ZKC92 ZID78 ZOB68 ZDV03 ZDV05 ZID97 ZNY10 ZID99 ZOB69 ZDV18 ZKC24 ZKC26 ZNY09 ZKC94 ZDC18 ZKC98 ZDC19 ZDC58 ZOA32 ZLC45 ZDV04 ZID76 ZDC04 ZDC59 ZLC33ZLC34 ZDV25 ZID95 ZDC42 ZDC97 ZDCVA ZDV39 ZKC30 ZKC20 ZDC12 ZOA33 ZOA34 ZDV30 ZKC02 ZKC90 ZID91 ZID93 ZDC37 ZDC10 ZKC22 ZID92 ZOA35 ZLALE ZKC28ZKC29 ZDV38 ZID96 ZDC98 ZDC72 ZDCVB ZOA13 ZOA15 ZLA32 ZDV24 ZKC06 ZID94 ZDC16 ZDC50 ZDCG1 ZLA16 ZAB71 ZKC23 ZKC27 ZME24 ZME61 ZOA14 ZLAED ZLA34 ZAB70 ZTL15 ZME19 ZLA36 ZLA26 ZLA27 ZAB67 ZAB93 ZAB95 ZME23 ZAB97 ZFW49 ZDC36 ZDCW7 ZFW50 ZTL36 ZTL40 ZDC38 ZTL28 ZLA38 ZLA37 ZME20 ZDC09 ZLA39 ZAB92 ZAB68ZAB94ZAB87 ZAB50 ZFW47ZFW48ZFW42 ZME44 ZME32 ZTL02 ZDC99 ZDCVE ZLA30 ZLA40ZLA60 ZFW93 ZLA25 ZABHM ZAB23 ZFW94 ZFW90 ZJX65 ZJX48 ZFW28 ZTL23 ZLA31 ZAB65 ZAB80 ZFW39 ZFW71 ZFW92 ZTL27 ZJX00 ZME43 ZAB89 ZJX52 ZFW82 ZFW46 ZFW86 ZTL08 ZFW65 ZFW89 ZJX67 ZHU81 ZJX49 ZJX88 ZAB78 ZHU74 ZHU46 ZHU26 ZHU37 ZJX34 ZJX33 ZHU82 ZJX11 ZJXW1 ZJX35 ZHU97 ZHU78 ZHU11 ZHU76 ZHU68 ZHU70 ZJX17 ZJX78 ZJX76 ZJXW3 ZHU95 ZHU24 ZJXW2 ZJX16 ZMA19 ZJX30 ZMA02 ZMA01 ZHU59 ZMA65 ZMA64 ZHU79 ZMA08 ZMA25 ZHU72 ZMA Longitude ZSE12 ZSE07 ZSE03 ZSE31 ZSE11 ZLC20 ZLC17 ZMP23 ZMP24 ZMP25 ZSE01 ZSE02ZSE32 ZSE47 ZSE48 ZMP13 ZBW01 ZMP11 ZLC15 ZSE42 ZSE16 ZLC06 ZSE46 ZMP20 ZMP16 ZMP12 ZMP19 ZMP15 ZBW53 ZBW08 ZBW02 ZLC41 ZLC16 ZDV45 ZDV32 ZMP18 ZMP17 ZBW61 ZBW09 ZLC08 ZOB19 ZBW39 ZBW23 ZSE13 ZOB38 ZSE14 ZDV35 ZMP29 ZMP30 ZBW76 ZBW10 ZBW38 ZSE15 ZBW17 ZLC07 ZLC05 ZBW46 ZBW94 ZBW84 ZBW85 ZOB29 ZOB26 ZDV34 ZDV33 ZOB79 ZNY34 ZBW20 ZOB47 ZDV09 ZBW18 ZLC42 ZLC40 ZLC03 ZBW91 ZBW33 ZOB49 ZOB59 ZDV14 ZMP42 ZMP43 ZNY49 ZNY75 ZNY73 ZNY42 ZBW45 ZBW47 ZNY56 ZBW31 ZDV16 ZDV08 ZOA31 ZOA43 ZLC04 ZKC92 ZID78 ZOB68 ZOA36 ZDV03 ZDV05 ZID97 ZNY10 ZOB69 ZID99 ZDV18 ZKC24 ZKC26 ZNY09 ZKC94 ZDC18 ZDC19 ZKC98 ZDC58 ZDC59 ZID76 ZDC04 ZDC97 ZLC45 ZDV04 ZDC42 ZDCVA ZOA32 ZLC33ZLC34 ZDV25 ZDV39 ZKC30 ZID95 ZKC20 ZDC12 ZOA33 ZID91 ZID93 ZOA34 ZDV30 ZKC02 ZKC90 ZDC37 ZDC10 ZID92 ZKC22 ZOA35 ZLALE ZKC28ZKC29 ZDV38 ZID96 ZDC98 ZDC72 ZDCVB ZOA13 ZOA15 ZLA32 ZDV24 ZID94 ZDC16 ZKC06 ZDC50 ZDCG1 ZLA16 ZAB71 ZKC23 ZKC27 ZME24 ZME61 ZLAED ZTL15 ZOA14 ZLA34 ZAB70 ZME19 ZLA36 ZAB95 ZME23 ZDC36 ZDCW7 ZLA26 ZLA27 ZAB67 ZAB93 ZAB97 ZFW49 ZDC38 ZFW50 ZTL36 ZTL40 ZME20 ZTL28 ZDC09 ZLA38 ZLA37 ZAB92 ZAB68ZAB94ZAB87 ZME32 ZLA39 ZAB50 ZFW47ZFW48ZFW42 ZME44 ZTL02 ZDC99 ZDCVE ZLA30 ZLA40 ZLA60 ZLA25 ZABHM ZLA31 ZAB23 ZFW93 ZFW94 ZFW90 ZJX65 ZJX48 ZFW28 ZTL23 ZAB65 ZAB80 ZFW39 ZFW71 ZJX00 ZFW92 ZME43 ZTL27 ZAB89 ZJX52 ZFW82 ZFW46 ZFW86 ZTL08 ZFW65 ZFW89 ZJX67 ZHU81 ZJX49 ZJX88 ZAB78 ZJX34 ZJX33 ZHU74 ZHU82 ZHU46 ZHU26 ZHU37 ZJXW1 ZJX35 ZJX11 ZHU97 ZHU78 ZJX78 ZHU11 ZHU76 ZHU68 ZHU70 ZJX17 ZJX76 ZJXW3 ZHU24 ZJXW2 ZHU95 ZJX16 ZMA19 ZJX30 ZMA02 ZMA01 ZHU59 ZMA65 ZMA64 ZMA25 ZHU79 ZMA08 ZHU72 ZMA Longitude Without constraints, 0.7 traffic concentrates on a 0.6 few sectors the majority of the rest under-utilized 40 sectors have counts above 8 at some time 0.8 With traffic control 0.7 meet capacity constraints at 0.6 all time 0.5 traffic in congested sectors 0.4 diffused into neighbors 0.3 increase 0.71% travel time

20 Result for Sector ZTL15 Aircraft Count Sector ZTL15 Without Capacity Constraint With Capacity Constraint EST (Hour) Satisfy capacity at all time The new sector count does not always stay below the old one 20

21 Conclusion Proposed a Hierarchical Decentralized Flight Planning Framework Respect user s preference and has potential to reduce delay and energy Future Work Further validating the framework using realistic weather data compare the fuel savings as compared with the real flight plans 21

22 Thank you very much! 22