Source of Acquisition NASA Ames Research Center With the imkuducfiun of the Federn1 Anation Admilnhtrafion’s Adwinced Tcchnology and Oceaaic Proceduris system at the OaWsind Oceaxliic Center, a level of automation now exists in the oceanic environment fb potentially begin accommodating increased user preferred mutifig mqacslu, This paper gr4~enlu Ihe results uf BQ initial frasibility assessment which examines the potentiat benefits of transitianing from the fixed Central East Pacific routes to user preferred routes. A$ B surrogate for the actual user-provided routing requests, a minimum-travel-time, wind-optimal dynamic prngrannming aigwithm , was developed and utilized is this paper. After first describing the characteristics (e,g. origin airport, destination airport, vertical distribution and t%mparald istribution) of the westbound flights utilizing the Central East Pacific routes on Dee, 14-16 and 19-20, the results of both a flight-plan-based simulation and a wind-crptimsl-based sirnslatioq are presented. Whereas the lateral and Iong~tudimld istribatiort of the aircraft trajectories is these YO simulations varied dramatically, the number of simulated first-lobof-separation events remained relatiyely constant. One area of concern that was .qwwergg ip,_jhjpliaitiax analysis was a potential workload issue associated with the redistribution a€ traffic in the oceanic sectors due to the prevailing wind patterns. I. Xlaitroduction Th e Oakland Oceanic Flight Information Rsgion (FIR) (or Center), which is shown in Fig. 1, controls approximately 21.3 miilion square miles nf airspace and borders the Anchorage FTR tn the North, the Tokyo FIR to the East, the Aukland FIR to the South, and the coasthe of the contiguous United States on the West[l]. In contrast, the twenty Air Route Raffic Control. Centers {AKTCCs) in the contiguous US.e ncanpass roughly 3 million square miIes. Despite ths vast amount of aimpace Fontrotled by this Center, flights, for the most part, are required to fly along fixed route structures and adhere to lateral separation standards that extend up to 100 mi, Imghdmal. separation standards sxwiding up to 15 minutes, and vertical separation standards of 1000 fi 121. These stringent scpadon standards ate requited because of the limited ~U~V~~~IXcIapCaGb ilities in the ocean and the FAA’s legacy Oceanic Display and Planning System (ODAPS). It is worth noting that both the lateral and longitudinai separation standards can be, significantly reduoed for Automatic Dependent Surveillance - Contract (m3-C) equipped flights that are opersting in regions of the Pacific that are designated for Requked Havigatfon Perforname {W}[Z ]. ‘R esearch Scientist, Automation Concepts Research Branch, Mail Slop 2 10-10,E -mail: [email protected]. Chief, Automation Coiiccpki Rcscarch Branch, Mail Stop 210-10, Pellow ALAA Project Manager, Advanced Air Transportation Twhnologies Office, Mail Stop 210-15 Student, Dept ofM echanic American Institute ofA exonau~csa nd Astronautics Figure 1. QaMilnd Oceanic Center Display Including Sector Boundaries and Coastal Mag With the introduction of the FAA’s new Advanced Technologies & Oceanic Procedures (ATOP) at Oakland CmteI in 2005, increased rout0 flexibility and reduced separation standards can potentially &w be accommodated. Sonic of thc mior ATOP innovations flexibility arc (1) fuIly intsgratcd flight- and radar- data-goc&%ng capabilities, (2) con s, and (3) dellite &ti-iirik communication and surveillance capabilities [3]. This paper presents the results o f ai nitial study, which examines the potential benefits a dc onsequences of allowing user-prefemd routing in place ofthe fixed Central. East Facific (CEP) routes. The CEP routes shown in Fig. 2 connect the wast coast of the Vnited States to Hawaii. The alphanumeric designaturn for these S ~ Y rEou~te s are R463, R464, R465, WX5, R576, and 5577- Sectors UC-3 and OC-4 in the Oakland Oceanic FIR handles lraffic along these routes. Routes R464 and R576 are used exclusively for westbound traffk, whilc routcs F.465, R585, and R577 ate used cxzlusively for eastbound ttaffjc. Routes R463 and R578 can accommodate bi-directional traffic. The RNP for aircraft flying on the CEP routes is 10, or W-IO, which implies that the total horizontal position enor of the aircraft cannot exceed 10 nmi for more thm 95% of the flight tame [2], 141. Those flights can bo identified by the “Et” ~~quipmernsut ffix appearins in their Intsrnationd Civil Aviation Organization (ICAO) flight plans. [5], Based on the required equipage level for flights operating on the CEP routes, the lateral separation standards are 50 mi, the longitudinal. separation standards vary between 5 and 10 minutes, and the vertical separation standardv are fU00 A. To our knowledge, this is the fist study designed to explore the banefits of Qsnsitioning from the fixed CEP- baed routa $tmcture to a mor&f lexible user-preferred routing strumre. Though the flight routing application in this study is believed to be unique, the general area of flight plan routing and flight path design bas a long history [6]. Recent advatices in this area include the conflict-free, wind optimal. routing work that WES introduced in Rcf 7, thc dynamic network flow remuting approach introduced in Ref. 8. arid the: fleet. assignment and routing approach introdwed in Ref, 9. A comprehensive summary of many of the earliest flight routing techniques can be found in Kef. 10. 2 “ American bnstitlrtc: of Aeronautics and Astronautics Figure 3, Central East Pacific (CEP)R outes The rest oft his paper is organized as follows. Section TI describes the wind-optimal routing methodology that was adopted for this paper. 'RIGun modified characteristics of the C routes are describkd in Section XI1 and the modtiirig rcsults arc 11. Mudding Methodology As r>rz~ious!y mentioned, the FAA's f i ~ ;A TOP system offew 23 anprecebanted level of zitlaxiition in comparison to the legacy DVL~JsSy stem that it replaced. With this new system, it is now conceivable to transition &am The fixed CEP route struoturs in the Pacific to c set of routes that accommodate the airline user's preferences. Tu begin, understanding the potential. implications of this transition, a minimum-travel-time, wind-optimal dynamic programming algorithm was developed to simulate possibie user route pr&erences. Though the business models, schedules, and aerodynamic perform-mcc charactcriatics (c.g,, cflginc typc, wcight, etc.) of the aircraft in the airline's fleet will ultimately govern the design of the optimal trajectories each airline wishes to fly, a simple minhum-the wind-optimal model, as adopted in this study, should initially sufi3ce to understand this change in rohting philosophy. For the purpose of this study, the pasition of an aircraft dong a minimumthe whd-optima1 route at stage k+ 3 can be related to the position at stage k via the followir~ge quation: XkCl = Xk f y (1) Here uk is the decision vaxiabIe at stage B For the purpose of this study, wind optimal routes are calculated on a grid composed of latitude and longitude values that encompass the region of airspace in which each analyzed aircraft travels, Though the grid is individually tailored for each flight, the grid is roughly bounded between fhe 20"N latitude to Ma south, 40" N latitude to tha north, 156W longitude to the west, and 12O"w longitude to the east. This grid is sufficient for most flights that either depart or arrive at the west coast of #G United States. For ff ights departing ox amivine; at hn-Xand aUpvrts, such as Chicago O'Haw International Airport, the grid is increased accorduigly. The stages, ts within our problem formulation refer to the available longitudinal valuas; the states xk refer to the latitude values; and the decision variables, uk,a xe the changes in the latitude values that are permitted at each stage k. The bounds on the decision vzriables and the states are 4Wk Q) ;txlla,k 5 Xk 5 { uk € U, where U,= 4-5 5 uk 5 5) (31 ,hdditional bounds tux applied to xkatt he, first and last stages ds at the mtry and exit points of the CEP routes into the Oakland Oce&c FIR. lied to enforce any procedural constraints that might apply as an aircraft transitions to/&om a U.S. Air Route Trafic Control Center (ARTCC) and the oceanic envijronment. Using tho principle of optirnality, the rninkmm cost function at stage k, f(x,k),c an be caicuk&ed from the + minimum cost fuxlction at stage k+l, I(x, u,, k i- 1) using the following expression [I 11: Here C(xk,& ik,ki)s the cost associated with Wmsiti~nkgf rom state xt to z ~us+ing~ th0 decisioa variable i~~ at stage k. Fox our study, the cost is equd to the amount of the required for an aimaft to fly from the current State, xk, to the next state, xkA1T he details of the cost function calculations me provided in Appendix A. "he ~ - sequence of controls &e. uo,ztLt... ,uN-J resulting from the solution of Eq. 5 for 0 5 k s rv' 1. is used to construct the minimum-time, wind-optimal routes beONem the origin and destination airport for each flight. For all routes gen5ratad using Eq. 4, the minimum cost fundon at k=N(i-e. I(x,N)1 h as been set to zero. To ilhstratc thc usc of this algorithm, wind optimal trajectoiics for tlic wcstbcruiid aircrefl that were nominally flying on the CEP routes are presented in .the fourth section of this paper. 1x1. Unmodified Flow Chara'ctsristics The prin~.iyp urpoue ofrhis saction is to describe rhe dam SOUTCLS used in this study and to familiarize the reader with the nominal characteristics (erg.,n umber of aircraft, vertical distribution, etc.) of the westbound traffic flying dong the CEP ro~tes,A m& flight plans, flight plan ammdaats, departure time$, arrha! times, md positior! reports were obtained froIxl an Aircraft Situation DBpIay to Industry (ASDI) dntn feed [12], [I31 for DCC.1 4-16 and 19-20 of 2005, For reference, Dec. 14* is a Wednesday and Dec. 20' is a Tuesday, To filter the westbound CEP route traffic fiom the national data mosaic provided by the ASDX data feed, only flights that exp1icitly contained R463, R464, R576, and R578 in eirher a Right pian or flight plan amendment message were retained For further processing. The resulting aircraft counts for this five-day period are shown in Fig. 3. A total of 325 westbound flights were analyzed during this time period. The average daily traffic count was 65 and the standard dkation was 4.6. Figure 3. Westbound Traffic Counts on the CER Routes To assess the vertical distribution of the westbound flights, fie actual flight level (FL) of each aircrafi as ir crossed 1-40"w est longitude, which is roughly halfivy between Hawaii and the U.S. mainland, was rworded and the results are diuplayed m the leftmost im3g-e in Fig 4. As can be seen from this figwe, the most pr'edominmtly used flight l e dw as FL340 {or 34,000 ft), arid approxhately 73% of the aircrafi 'IVm betwean FL320 and FL360. The rightmost image in Fig. 4 depim the 'termpnral distribution of the westbound flights at I@" west Iongituds. As cm be seen from this 5pre, the largest concentration offlights cross this iongimde at rou&ly 19r00 UTC, which corresponds to a west-mast departure rime of approximately 1630 UTC (or 890 PST). The trflic colmnt distribution on each of tbe analyzed days shows little variation, but slightly more aircraft w0ra observed to cross this longitude at €200U TC and 21; OOU TC on 12/19/05, which is a Monday. Figure 4.VerticaI (left} and Temporal. (dght) Ristribution of the West-i3oudd Frights st 140 degrees West Longitude The leftmost image in Pig. 5 depicts the departure airport usage statistics for the wegtbound CEP Bights, while the rightmost hag^ shows the corresponding destination airport statistics. For the dsparturo statistics, on& the top 15 airports xe illlustrated, a1tkoug.h flights were observed to depart from PO 6i-p during the five-day perrod 5 Arnsrican Instinrtc?o f Aeronautics and Amom&S analyzed in this study. Koughly 42% of all westbound flights dqmed from Los Angeles International Airport, m y an d San Francisco hternatbna1 Airporf K$3pO. The names sissociated with tho remaining departure airport acrofiyms in Fig. 5 follow: Chicago O'Hat-e (X(\oRD),M cCman Inti. (KLAS), Phoenix Sky Harbor Intl. (IEPHX), John Wayne Airport {KSNA), Dallas/Fort Worth Intl. (Iu3PW), Metropohn Oakland htl. &OAK), San Diego htl. {KSAN), Memphis Id. (ICMENI), bntwio Inti. (KONT), Seattle-Tacoma Intl. (KSEA), Norman X. Mhinata Sa Jose htl, KSJC), PortJand IntL (KPDX), and Sacramento htl. CKSMP). The arrival airport statistics that are shown in the rig$mnost image in Fig. 5, illwirate that approximately 60% of all flights destined for Hawaii landed at Honolulu Xntl. Airport (WJPI,). In contrast, only one flight was observed to arrivs at Lanai {PHNY) and Hilo (PHTO) during the same period. Of the remaining fl ighb, 24% landed at Kahului (PHOG), 1. I% landed at KondKealnole Kailua (PIKO), and 4% landed at LihueKauai Island (PHLI), Figure 5. Top Origin and Destination Airports for the Westbound ETights The dimibution ufthr: flights amongst the ava&ble wcstbound CEP routes is shown in Fig- 6. Aa shown by this figure, flights traveled predominately along R4M and R576, which are both undirectionat routes. $igni&antly, less traffic vas observed along 8463 and R578, which are bi-directional routes. The control of flights along these routes was distributed roughly evenIy between Oakland Oceanic Sectors 3 and 4. As iilustrated by Fig. X, raffic on R463 and 8.464 is controlled by Sector 3, while tr&c on R576 and R578 is controlled by SEC~4O. ~ Rgwe 6- Westbound CEF Route Usage Statistics IV. ,ModSified Flow Characteristics The results from an initial set of sirnutations that were designed to access the potentiai benetlt~a nd consequences of allowing flights to deviate from the fixed CEP route structure in favor of user-prefem presehted io this section. U~ingth e ,45131 data files for Dec. 14-16 and 19-20 of2005, two sets of simul run in the Future ATM Concepts Evaluation Tool (FACE'I") [14]. In the first set of sim were allowed to fly dong their fikd flight plans, as specified in the ASDI files, SCCO f simulations a wind optimal trajectory was cakulareb for each flight using the method din li. For both s w o f simulations, only aircraft that filed flight plans on one of the four westbound &e. R463, R464, R576, and RS78) CEP routes were rerained from the original ASDI files. The mag13lplin3de and direction ofthe winds in the Pacific O G Eth~at were used to simulate the trajectory of the aircraft on both the original flight plan route and the wind optimal route wer0 obtaiacd .from the Global Forecast System {tiM] Atmospheric Model. tl5l that is produced by the National. Oceanic and Atmospheric Administration (NOBA), For reference, GFS is c global ahnospheric model with a horizontaf resolution of approximately-0.5' x 0.5' Iatituddlongitude and an unequally spaced vertical resolution starting at 1000 rnb (surface) and extending up to 100 mb. Updates to the GFS model. are available every six hours and forecasts are available up to 16 days into the futurc. As an sxample of the wind data used in the shulation, ffir wiiid magnitude contours at 250 mb, which rou$bsftly includes the altitudes betwaan 31.000 ft and 36,000 I?, are shown in Fig. 7, For reference, the Hawaiian islands are shown ai; he bomm left ofthis figure and the wes1: coasr of the United States is shown on the right side of this image. A region of very strong, easterly winds {l I20 knots) is designated by the magenta polygon in the cefitei ofthis Egiiie aiid k id irectly abooiie ?WO ofthe ~&3i5otind CE? roiita (U76 X I ~R5 78). Figure 7. W-ind contours at 250 rnb io the Central, East pacific The aircraft position (latitude and longitude) histories for the flight plan and wind optimal based simulations that were run using the ASDI data for Dec. 14, 2005 are displayed in Pig. 8. The leftmosf image in this figure depicts the westbound poskion histories fur aircraft on tha CEP-based flight plan routes, and the rightmost imap depicts the position histories for the flight3 on wind-optimal routes. Only minor deviations in the magnitude and direction of the wind-field were observed m the Central. East Pacific on Bec. 24, 2005, so only a single GFS file (0;OO UTC, zcro-hour forccast fib) was used to grsnerare these remits. The most striking feature differentiating these hvo images is the tremendous latitudiilal *adl angitudinal dispersion of the aircraft position histories resulting from the wind-optimal routing simulation. Not only are the histories dispersed but, with only one exception, the routes have all. been shifisd north because ofthe strong easterly winds present Ixl Sector 3, see Fig. 8. When generating the wind optimal routes, esch aircrafr has been allowed to fly a wind optimal trajwtory Eom the origin akport to the destination airport. Alfmmtive wind apfimnl sPcite@e$ lo be explored in thefinar draft o f this paper incIude the following: 7 Amcncan institutept Acmnauties and Astronautics Fipre 5. F!igHt P!m md WiDd 8pfjmdS wim for Westboz~dF lights The sector counts as a fLmction of time for Oakland Oceanic Sec3ors 3 md 4 are skoowr! it! Fig. 9 fi~Sro & kbke CEP-based and wind-optimal routing simulations. The thick red and blue CU~GSin this figure correspond to the CEP-based routing simulations, while the thin rad and blue curves correspond to the wind optimal routing simulation. As cm be seen from this Egure, the impact of transitioning to wind optimal routes results in a decrease in the peak aircraft count in Sector 3 from eleven to one, and an increase in the peak aircraft count in Sector 4 from I8 to 27. From a controller-workload pmptrctive, the 50% increase in the peak traflk count in Sector 4 is a bit. dkconcetthng. To aceommodate& C ehsng*s &e .aaffic psttcrn~th at could potsntkd~yb e associated with adopting u5er preferred tmjectoriev, ~ESOU~CsGc heduling algorithms [I61 or dynamic re-sectorization [I71 may prove beneficia!, Both these areas of~mwkl!! be expk~redi n. more &%ai! in fi~Pmsm dies. h i Figure 9. Aircraft Counts in Oakland Sectors 3 and 4 with wind optirnaJ and CEP muting Changes in the traffic patterns in the Central East Pacific can not only increase the workload of a wntroll.~b~u,t also can give rise to an increase in the number of possible losses of sepparation. .Using FACET, the simulated trajectories for all westbound flights were examined for possible losses of separation on both #e nominal. CEF- based flight plan routes ide separation criteria: (1) 30 oft 8 American Xnstimte of&nmauucs ana ~svmaut~cs mi latml spacing, 50 mi longitudinal spacing and 1000 fi vertical spacing. For the remainder of this section, the first of these No separation standstds will be referred to the “301’30’’ separation standard atld lhc: hkr Of these two will be referred to as the “50/50” separation Ytandard, Both of these separation standards are technically Icss restrictive than the sfiandards that currently &st far ff ights 011 the CER routes, though the 50/50 separation standards are achievable with the cutrent ATOP system and thc 30/3Q separation standards are proposed reductions in the separation standcitds [2]-[3]. The geographical location of each unique, fitst-loss-of-separat~ratio*e1v ent is depicted h Fig. 10. The leftmost image this figure: depicts these events with the 30/30 separation standards, while the rig:htrk~usli mage depicts these events with the 50/50 separation standards. Red circles are used to denote t hC~E P- based routing first-loss-of-separation events and blue squares are used far the wind-optimal events. Under both separation sc%n&x, the wind-optimal routing changes the location and frequency of the potential first-loss-of- separ~tione vents. The most strking feature to notz regarding the location of the events is the large number of events that occur immariincely upon sntcring .the Oakland Oceaaic Ccntcr. This would seem to indicate that additional depawe constraints are required far each flight to ensure that the separation standvda are met prior to passing control of flights from h ed omestic Cenrers to Oakland Oceanic. In addition, more first-loss-of-separation wents Re observed in Oakland Oco-mic Sectors 4 a d 7 under thc wind-optimal routing scenario than the CEP- based routing scenario. This behavior is to be expected given the Lack of structure md the numerous oonvarging flight paths that are observed in Fi5. 8 for the flights on wind-optimal routes. In regards to the frequency OF ~~t-I~ss~o~-s~peavreanttsi,o thne 30/330 scenarios yielded five events mder the CEP-based r~utbgsi mdation and four events for &e whd-optimal routing simulation. A5 would be cxpeckd, the number of first-loss-of-separation events increased under the 50/50 scsnario, yielding six events for the CEP-based routing simuiation and 12 events for the wind-optimal routing simulation. V. Conclusion Wjth the introduction ofthe FA4’s new ATOP cpbm at Oakland Center, increased routs flexibility a dm duccd separation standards can potentially now be accbmmobated. The results of an initial study that were designed to (I) understand the characteristics of traffic on the westbound CEP routes and (2) to assess the feasibility of transitioning from tht: CEp routes to user-preferred ones were presentcd. To accomplish thr; first of these two rash, #e ASDX data for Dec. 14-16 and 19-20 was used to examine the following: origin and destination of flights on the CEP routes, usage of each route, vertical distribution oftraffic on all routes, and the temporal diskibution offfights on the roiites. As a surrogate for *e user-prefmed routes, a dynamic programming based rnethodoIogy was presmted and used to cafculate wind optimal trajectories for the westbound flights that originally flew on CEP routes. The secror counts for O&!and Oceanic Sectors 3 and 4 and the location ofthc simulated first loss of sepratiun events were compared f ~ frlig hts on both the wind optimal routes and the nomina1 CEP-based flight plan routes. Though the lateral and longitudinal distributioe of flights was obszmed to change dramatically with &E wind optimal routkg the ruunber uf r~sl-luss-of-separatione vents was not observed to change significantly. Appendix A For the cwent study, the cost associated with transitioning from an inirial sate?, xk, to a final state, xkCIi,s quai to the travel. time between these two states. If the idcia1 ~ ~ t ~ ~ d e /po~sitoion~ i~s dien~otdede by (Ai,TZa)n d the final pGsition is denated by (hf,Sft)he n the costttnvsl time 3s given by [IS] C(Xk,Uk,k) = t = d/Vg where ? I.;' + Vg = y2 Her0 Rcerui2a the radius ofthe Earth, which is takcn to he 3,444.046647 nm, and ths horizontal components of ~G velocity are denoted by k and 9. These vdo&y components are calmlaTed from the horizontal velocity of rhe xcow7 aircraft, v,, the horizontal component of the wind veiocity, W,, the aircraft's commanded heading, and &e xh, horizontal wind direction, using &e Mlowing expressions: and The aircraft's command heading, )&m, is related the course angle for great circle navigation via the following expression: = XGC - sin''. (( wh /',) ' sin(36h - %G, >) (XI