TRAFFIC ANALYSIS AND SYNTHETIC SCENARIO GENERATION FOR ATM OPERATIONAL CONCEPTS EVALUATION

TRAFFIC ANALYSIS AND SYNTHETIC SCENARIO GENERATION FOR ATM OPERATIONAL CONCEPTS EVALUATION Juan A. Besada, Javier Prtill, Gnzal de Miguel, Rafael de Andrea, U. Plitécnica de Madrid, Madrid, Spain Jse M. Canin, U. de Las Palmas de Gran Canaria, Las Palmas, Spain Abstract This paper describes a pair f systems which can be used t btain realistic traffic samples in a Sectr/TMA frm a given real traffic database. Thse are a Traffic Analyzer and a Traffic Pattern Generatr. These tw systems allw the ATM engineer t bth gain insight n the traffic structure f the area under analysis and t btain statistically significant samples fr the evaluatin f peratinal cncepts and prcedure changes, perfrm analysis f ATM perfrmance under traffic changes, Intrductin With the advent f new ATM paradigms such as Trajectry Based Operatins and the evlutin in Cllabrative Decisin Making, changing rles f pilt and cntrller it is imprtant t have realistic traffic simulatins t design and validate new ATM peratinal cncepts. It is especially imprtant nw, with the rapid changes appearing in SESAR and NextGen prjects. In this paper we will describe a pair f systems with can be used t btain realistic traffic samples frm a given real traffic database. Thse are a Traffic Analyzer and a Synthetic Traffic Generatr. Traffic Analyzer, taking the real flight infrmatin file and sme additinal infrmatin, btains the patterns (statistical, determinist, r mixed), t be further used as a pssible input t the Traffic Pattern Generatr blck. Traffic Pattern Generatr generates the infrmatin abut the initial cnditins f the aircraft t be cnsidered in the simulatin exercise, including preferred trajectry and dynamic state. It prduces Synthetic traffic file with the generated aircraft features and flight plans assciated. This file is t be used in a realistic simulatin t define traffic scenari under analysis. The paper will describe thse systems in detail, their related HMIs enabling fr a cmprehensive analysis f the traffic under analysis, and fr graphical validatin f the generated traffic. There are systems similar t these, such as Eurcntrl Traffic Sample Generatr (TSG) [1], r TRAFGEN, within PITOT by AENA[2] and similar effrts, based in mdificatins f recrded data, has been published recently [3]. Many f the systems fr scenari preparatin are based n real traffic clning r deletin by a human user. This is a very cstly and errr-prne methd. Please nte in the current definitin f thse systems they are riented t the analysis/simulatin f TMA traffic, althugh many f the cncepts invlved may be extraplated t the simulatin f enrute sectrs r t ther gegraphical areas. This wrk is a MATLAB prttype that has been develped fr INDRA under Spanish ATLANTIDA[4] prject, and it will surely becme a tl integrated in near future INDRA ATM prduct line. High Level System Descriptin The main bjective f the traffic analyzer and generatr is t prduce a file with the infrmatin f the flight initial cnditins t be generated in a specific spatial distributin (in a TMA). It is imprtant t nte that the final utput f the synthetic traffic generatr is nt the time evlutin f the real flights, but the initial infrmatin necessary (i.e. initial cnditins and flight plans) t feed an ATM simulatr (either a real time r a fast time simulatr). Figure 1 shws the main wrking blcks f the prpsed systems. All data is described in XML files t ensure readability and prcessing capabilities. 978-1-4244-4078-8/09/$25.00 2009 IEEE. 2.A.1-1

Figure 1. Blck diagram f the Analyzer and Synthetic Traffic Generatr Next, a brief explanatin f the blcks in the figure is presented. Traffic Analyzer Traffic analyzer, taking the real flight infrmatin file and additinal infrmatin, btains the patterns (statistical, determinist, r mixed), t be further used as a pssible input t the Traffic Pattern Generatr blck. The infrmatin input / utput t this blck is: Real traffic data. It is a file with the infrmatin f real flights t analyze. The pattern analyzer t extract the traffic patterns will examine this infrmatin. Operatinal cntext: These files cntains the TMA physical structure: Entry pints (STARs starting pints), STARs (waypint list), TMA exit pints (SID ends and starting f verfly rutes), SID (waypint list), dmestic rutes, airprts, In additin, sme characteristics f aircraft are included in a set f aircraft types files, t be used fr classificatin f flights. User preferences fr the analysis. This is anther file cntaining the infrmatin necessary t define the analysis preferences. It will cntain a pair f flags indicating if the traffic analysis must try t btain deterministic traffic patterns r nt, and if the traffic analysis must include the analysis f verfly flights; a set f threshlds used t define if the pattern under analysis has a deterministic frm r nt; a list f special days where the traffic des nt fllw the same pattern as thers due t special circumstances (public hlidays, lng weekends ); and a plar grid used t grup airprts utside f the TMA with similar gegraphic lcatin. It prduces the fllwing infrmatin (which culd, ptentially, be input f traffic generatr): Mdels t generate traffic. A file, with infrmatin and frmat detailed in a further sectin with the infrmatin extracted frm the real traffic analyzed. This file may als be ttally r partly generated by the user, using an adequate 2.A.1-2

GUI, and it culd als be edited t adapt system behavir t desired traffic patterns. Undefined rutes: they are a descriptin f rutes fllwed by traffic but nt described in the peratinal cntext definitin. Synthetic Traffic Generatr This blck generates the infrmatin abut the initial cnditins f the aircraft t be cnsidered in the simulatin exercise, as a starting pint t the real traffic generatin (trajectries) realized by the fllwing units. The infrmatin input / utput t this blck is: Operatinal cntext: The same input as traffic analyzer. Undefined rutes: As prvided by traffic analyzer. Mdels t generate traffic: As we have mentined befre, instead f using the patterns extracted by the pattern analyzer, the user culd replace them (r a part f them) with user-defined parameters. It culd be useful, fr instance, t perfrm a cntrlled change f the traffic generatin parameters (i.e., an aircraft arrival rate t a TMA entry pint) in rder t see the effects in the simulatin exercises. Additinal Mdels t generate traffic: Sme traffic features cannt be extracted frm the real traffic file analysis, and must be directly prvided by the user. User preferences fr the generatin: Thse are sme additinal parameters t be used at generatr, such as: the time interval t be used fr generatin; a list f special days where the traffic des nt fllw the same pattern as thers due t special circumstances (public hlidays, lng weekends ); a set f traffic validatin threshlds; the flight duratin between injectin time and arrival t TMA; and the minimum time separatin at entry pints. It prduces the fllwing infrmatin: Synthetic traffic generated initial data. A file with the generated aircraft features and flight plans assciated. Detailed Functinality f Traffic Pattern Analyzer This sectin is a detailed descriptin f the functins cmprising the traffic pattern analyzer. First f all, we shuld cmment a general feature f the traffic patterns. It is reasnable t suppse that an imprtant part f the traffic in the TMA (bth arrivals and departures) will have an imprtant deterministic cmpnent. Many f the arrival and departure flights will repeat n a daily basis with the same flight plan. Superpsed n that deterministic pattern there will exist a randm pattern which will nt fllw such a predictable mdel. Taking int accunt the preceding paragraph, we prpse t cnsider the fllwing traffic pattern analysis mdel (see Figure 2): Mixed mdel. In this case, the traffic pattern analyzer will extract the parameters f a mdel that has a deterministic and a randm part. We will generate the deterministic part cnsidering the regularity f flights (i.e., we will analyze a number f similar weeks and extract the deterministic pattern fr each weekday). The randm part f the mdel will be analyzed after extracting the deterministic part. Prbabilistic mdel. In this case, the analyzer perfrms a bypass f the deterministic mdel estimatin functins and treats the entire database file in rder t btain prbabilistic functins fr all the necessary functins. Figure 2. Traffic Pattern Analyzer 2.A.1-3

In rder t describe the functins, we will assume we have a database with a number f representative flight weeks enugh t perfrm an accurate estimatin f the traffic functins and their parameters. There are fur different kinds f flights t be included in the traffic mdel: Arrival Flights. Departure Flights. Dmestic Flights (Internal t TMA). Overfly flights (thse with an rigin and destinatin airprts ut f the TMA). Slightly different mdels must be defined fr each f thse types, and all thse mdels will be prvided in a unique analyzer utput. We will detail the functinality fr arrival traffic. Arrival Traffic Analysis This functin has fur stages, related t identificatin f deterministic/randm flight patterns arrival t an airprt, assignatin f entry pint and STAR, and mdeling f initial cnditins f arrival aircraft (see Figure 3). Figure 3. Traffic Analyzer fr Arrival Aircraft These functins will be detailed next. Deterministic Arrival Flights Determinatin This functin will btain the deterministic pattern f incming flights. Frm the analysis f the input flight database, a deterministic pattern f flights with an ETA and destinatin airprt may be inferred. In rder t estimate the deterministic cmpnent, the functin will perfrm a flight identificatin and ETA crrelatin ver the database infrmatin in rder t detect repetitin patterns f flights in the fllwing cases (with a threshld related t a given percentage f ccurrences): Every day f the week Several days f the week One day f the week Special days marked by the user We will grup rigin airprts using a plar grid with respect t TMA center, s that all airprts in a cell f that grid will be equivalent. A lgarithmic law will define ranges defining the grid. Each cell in the grid will be treated as a cmmn rigin. In additin we assume deterministic flights have a distributin f aircraft types assciated, with a discrete prbabilistic law assciated, which will be estimated. We will have an equivalence f aircraft types frm the aircraft table, s that we can aggregate equivalent aircraft. The infrmatin extracted by the deterministic cmpnent will be a list f deterministic flights, with: Flight Identificatin Origin (grid cell) Destinatin airprt ETA t the airprt Repetitin pattern, ne f the fur cases previusly cnsidered. Aircraft type list: Fr each aircraft type used fr this flight: Identifier f the type Percentage f entrances Once the deterministic cmpnents are btained, with the remaining infrmatin, the randm cmpnent f the traffic mdel will be estimated. Please nte this functin may be bypassed if nly prbabilistic mdels are t be estimated. Randm Flights Destinatin Airprt Mdel Extractin In this case, we will btain the cmpnents f several prbability distributin functins (estimated frm the data) t statistically mdel the flight arrival t each airprt. We will estimate a set f cncurrent arrival t airprt prcesses. We will assume Pissn prcesses, which can be parameterized just by the average number f flights per time unit. As it is nt a 2.A.1-4

statinary prcess, we will calculate this per hur, and there culd be different distributins fr: Different days f the week. Special days, t be marked by the user. Different Origins: as defined in 0. Different Aircraft Types: We will have an equivalence (based n a wake vrtex turbulence, engine cunt, engine type and ICAO aircraft descriptin) f aircraft types frm the aircraft table, s that we can aggregate equivalent aircraft. Thse three aspects are crrelated. The apprach used t perfrm this estimatin is: Perfrm a segmentatin f flights by destinatin airprt, rigin and aircraft type. Perfrm a segmentatin f days by day f week and special day s table. Each f thse segments is what we will call a day type. Fr each destinatin airprt, rigin airprt cell, aircraft type and day type, we will calculate the average number f arrival flights per hur. This is the nly parameter needed fr the generatin f the arrival traffic mdel. Entry Pints Assignatin This functin will take the infrmatin f all flights and find deterministic patterns abut the TMA entry pints f the flights with a determined rigin and destinatin airprt. Please remember that rigins are gruped using the airprt grid. If there is mre than ne ptential entry pint fr a given pair f rigin and destinatin airprt, we will btain a discrete prbability distributin functin and its assciate parameters (percentages related t each entry pint) t mdel the TMA entry pint assignatin. S there will be a table with: Origin airprt cell Destinatin Airprt Aircraft Type Class Entry pint list: Fr each entry pint Identificatin Percentage f entrances There will be a percentage threshld, s that if nly a very small amunt f flights d nt fllw the deterministic pattern we will assume the assignatin is deterministic and the mdel will nt take int accunt thse nn-deterministic flights, treating them as utliers. STAR Assignatin This functin will take the infrmatin f all flights and find deterministic patterns abut the STAR used by the flights with a determined entry pint and destinatin airprt. If there are mre than ne ptential STAR fr a given pair entry pintdestinatin airprt, we will btain a discrete prbability distributin functin and its assciate parameters (percentages related t each STAR) t mdel the STAR assignatin. S there will be a table with: Entry pint Destinatin Airprt Aircraft Type Class STAR list: Fr each STAR Identificatin Percentage f entrances There will be anther utlier percentage threshld fr this assignatin. Undefined & Unknwn STAR Assignatin Mdel Extractin Smetimes, the flights culd fllw a rute nt cnsidered by the STAR definitin. If this is the case, we must define a pseud-star fr thse rutes, and perfrm an analysis similar t the STAR extractin but taking int accunt thse pseud-stars, in additin t the nminal nes. S there will be a table with: Entry pint Destinatin Airprt Aircraft Type Class STAR list: Fr each STAR Identificatin (pseud-star identificatin, prvided by the analyzer) Percentage f entrances In additin, the pseud-star must be prvided as an utput and used by the traffic generatr as if it were a nminal STAR. 2.A.1-5

This will be prvided in an independent file t be appended t the peratinal cntext mdel f the traffic generatr. Initial Arrival Flight Data Calculatin Mdel Nw we cnsider the way t estimate mdel fr the arrival flight initial cnditins. Please nte each kind f flight (arrival, departure, dmestic and verfly) will have its wn initial data calculatin prcedure, as different data is needed fr the injectin f each kind f flight in a simulatin. We will estimate the cmpnents f several prbability distributin functin (estimated frm the data) and its crrespnding parameters: Cruise velcity (crrespnding t entry pint fix). There will be a distributin different fr each rigin, destinatin airprt within the TMA, and aircraft type. Entry pint altitude (crrespnding t entry fix pint). There will be a distributin different fr each rigin, destinatin airprt within the TMA, and aircraft type. In rder t calculate the time f injectin at simulatin, we will mdel it as the subtractin f several terms: ETA, which will be btained at generatin time frm the deterministic pattern f frm the randm arrival pattern. Landing delay, which is the time difference between the ETA and the actual landing time. This is an additive term. There will be a distributin estimated per rigin airprt cell, destinatin airprt, entry pint and aircraft type. Duratin f flight inside the TMA (difference between entry fix pint and landing time, which is last fix time). This is a subtractive term. It shuld be necessary t perfrm a segmentatin f this item per STAR, and per aircraft type, and btain a distributin fr each f thse sets. Duratin f flight prir t entrance in TMA. This is anther subtractive term. This will be a user-defined time cnstant. N calculatin related t this term will be perfrmed during analysis. Detailed Functinality f Synthetic Traffic Generatr The traffic generatr will use a traffic pattern descriptin file and prduce a set f synthetic flights cnsistent with that descriptin. Traffic pattern descriptin file cntains a list f deterministic and prbabilistic patterns (with respect t several different aspects such as flight peridicity, rute r significant pints), and this methd will parse the descriptin f such a pattern and prduce a flight r set f flights cnsistent with this pattern. The whle list f synthetic flights is btained as an aggregatin f the flights generated based n all thse patterns. Figure 4. Synthetic Traffic Generatr. As the mdels used t generate arrival, departure, dmestic and verfly flights are different, and the data prvided fr each f thse patterns are different, generatin is divided fr thse same types f flights. In additin, there is a traffic validatin functin which will test, bth prir t generatin and after generatin, that peratinal limits (maximum number f peratins per hur in an airprt, ) are nt surpassed. There shuld be sme means remving ptential cnflicts prir t entrance in TMA. 2.A.1-6

Generatin will be perfrmed fr a cntiguus time interval, which culd be: A set f hurs A cmplete day (f predefined type with respect t traffic patterns). A set f cnsecutive days (f predefined types with respect t traffic patterns), up t a cmplete year. Initial Traffic Validatin This functin will check if the requested traffic pattern is within certain peratinal limits, defined by the user. Adding all traffic patterns in each peratinal cntext element, we will check the fllwing limits: Airprt number f peratins per hur. Entry pint number f peratins per hur. Exit pint number f peratins per hur. TMA traffic per hur. In this case, we will sum all traffic dmestic t TMA. This will be cmpared with a threshld. Entry Pints check. This functin checks that all Entry pint identifiers in the traffic patterns are described in the Operatinal Cntext input file. If any f thse threshlds is exceeded, r there is an unknwn entry pint, we will stp generatin, and infrm abut the prblematic traffic pattern. Arrival Flights This part f the generatr is in charge f creating arrival flights accrding t the arrival flight traffic patterns. There will be a similar functin fr departure, dmestic and verfly flights. We will detail nly this kind f flights prcessing see Figure 5). Figure 5. Synthetic Traffic Generatr fr Arrivals This is perfrmed in a set f steps where we cnsecutively refine the infrmatin regarding individual flights r grups f flights. In the fllwing diagrams the (D?) symbl stands fr a test where we are wrking with a flight r grup f flights either deterministic r nt with respect t the next step in the refinement. Next we will detail each f thse functins in the fllwing subsectins. Deterministic Destinatin Airprt Selectin This functin will select and generate the deterministic pattern f incming flights. Using the infrmatin f the flight mdel traffic pattern file, btained by the traffic analyzer, the deterministic pattern f flights with an ETA and destinatin airprt will be selected. This selectin will include the repetitin patterns f flights in the fllwing cases: Every day f the week. Several days f the week. One day f the week. Special days marked by the user S first the system will need t decide if the given day is f any f thse previus types, and perfrm an aggregatin f the deterministic traffic patterns f all the sets the current day belngs t. The infrmatin generated by the deterministic destinatin airprt selectin will be a list f flights. Each f thse flights will have: Flight Identificatin Origin: defined using the grid defined in the analyzer. Destinatin airprt ETA t the airprt Aircraft Type Class: Identifier f the aircraft type class, generated using the discrete prbability functin f aircraft type classes estimated in the analysis phase as a cmbinatin f aircraft s Descriptin, Engine Type, Engine Cunt and WTC. Randm Flights Destinatin Airprt Mdel Generatin In this case, we will generate the randm patterns f the flight arrival t each airprt using the Pissn prcesses in the analysis phase. We will perfrm a nested lp ver destinatin airprts, 2.A.1-7

rigin airprt cells (using the airprt grid), aircraft type and day type, and fr each hur we will generate the number f flights with all thse cnditins fllwing the previusly defined arrival prcess. Each flight will have: Flight Identificatin (Synthetic) Origin. Destinatin airprt. ETA t the airprt (btained using the Pissn prcess). Aircraft Type Class. Deterministic Entry Pint Assignatin This functin will select and generate the infrmatin abut the deterministic TMA entry pints f the flights with a determined rigin, destinatin airprt and Aircraft Type Class. S, fr a flight whse rigin and destinatin airprt are related t a deterministic entry pint (ne with 100% ccurrence accrding t flight pattern entrance), we will cmplete the flight with the crrespnding deterministic entry pint. Therefre, nw, fr each flight we will have: Flight Identificatin (Synthetic) Origin. Destinatin airprt. ETA t the airprt. Aircraft Type Class. Entry pint. Prbabilistic Entry Pints Assignatin Mdel Extractin In the case f flights with a given rigin, destinatin airprt and Aircraft Type Class that use several TMA entry pints, we will use a discrete prbability distributin functin and its assciate parameters (percentages related t each entry pint) t generate the TMA entry pint assignatin. Therefre, nw, fr each flight we will als have: Flight Identificatin (Synthetic) Origin. Destinatin airprt. ETA t the airprt. Aircraft Type Class. Entry pint. Deterministic STAR Assignatin This functin will select and generate the deterministic patterns abut the STAR used by the flights with a determined entry pint, destinatin airprt and Aircraft Type Class. S, fr a flight whse entry pint and destinatin airprt are related t a deterministic STAR (ne with 100% ccurrence accrding t flight pattern), we will cmplete the flight with the crrespnding deterministic STAR. Therefre, nw, fr each flight we will have: Flight Identificatin (Synthetic) Origin. Destinatin airprt. ETA t the airprt. Aircraft Type Class. Entry pint. STAR. Please nte that with respect t this pint undefined STARS are equivalent t defined STARS. Prbabilistic STAR Assignatin Mdel Extractin In the case f flights with a given entry pint, destinatin airprt and Aircraft Type Class that use several STARs, we will use a discrete prbability distributin functin and its assciate parameters (percentages related t each STAR) t generate the STAR assignatin. Therefre, nw, fr each flight we will have: Flight Identificatin (Synthetic) Origin. Destinatin airprt. ETA t the airprt. Aircraft Type Class. Entry pint. STAR. Please nte that with respect t this pint undefined STARS are equivalent t defined STARS. Initial Arrival Flight Data Calculatin Mdel Nw we cnsider the way t generate values fr the arrival flight initial cnditins. Please nte each kind f flight (departure, dmestic, verfly) will have its wn initializatin prcedure. We will generate the values f a prbability distributin functin (estimated frm the data): 2.A.1-8

Aircraft Type. Identifier f the type, generated using the discrete prbability functin f aircraft types estimated in the analysis phase fr this aircraft type class identifier. Cruise velcity (crrespnding t entry fix pint ). We will generate a value frm the analyzer prvided distributin fr each rigin, destinatin airprt within the TMA, and aircraft type Entry pint altitude (crrespnding t entry fix pint ). We will generate a value frm the analyzer prvided distributin fr each rigin, destinatin airprt within the TMA, and aircraft type. In rder t generate the time f injectin at simulatin, we will subtract several terms. Flight initializatin will be perfrmed t ensure arrival t entry pint at the predefined time: (ETA+Landing Delay- Duratin f flight inside the TMA): ETA, which will be generated frm the deterministic pattern f frm the randm arrival pattern. Landing delay, which is the time difference between the ETA and the actual landing time. This is an additive term. We will use the adequate distributin estimated fr rigin airprt cell, destinatin airprt, and aircraft type. We will use the distributin prvided by analyzer. Duratin f flight inside the TMA (difference between entry pint fix and landing time, which is last fix time). This is a subtractive term. We will use the adequate distributin estimated fr STAR and aircraft type. We will use the distributin prvided by analyzer. Duratin f flight prir t entrance in TMA. This is a subtractive term. This will be a user-defined time cnstant. We will assume a flight at cnstant height, heading and speed, cnverging twards the TMA, injected at the previusly defined time instant. The heading will be calculated assuming the flight came directly frm the center f the rigin cell t the entry pint fllwing a cnstant heading trajectry. We will calculate the initial psitin fr the injectin assuming this flight pattern and predicting back in time entry pint psitin by duratin f flight prir t entrance in TMA term. In additin, In rder t be able t perfrm accurate predictins f aircraft trajectry, we must calculate aircraft weight at initial psitin and perfrmance mdel t be used. It is btained thrugh the fllwing apprximate algrithm. Frm a set f simplified tables with weights and perfrmances fr different heights (Cnsumptin per minute fr different flight segments) we culd btain reasnable weights fr the aircraft. The defined methd is based n the explitatin f BADA[5] data. Aircraft weight M initial is calculated as: M initial = M dest (h,d) +ROF Cruise (cruise_altitude)* T exc +M min + r * M pyld where: M dest (h,d) is the estimated mass f the fuel needed t fly t the destinatin airprt. h is the entry pint altitude d is the distance travelled inside the TMA (frm the entry pint t the airprt) ROF cruise,lw (h) is the cruise rate f fuel cnsumptin f height h, with lw weight. ROF descend (h) is the cruise rate f fuel cnsumptin f height h cruise_altitude is the en rute cruise altitude. T exc is a sample f a nrmal randm variable taking int accunt remaining fuel fr 100 minutes and ptential increases 2.A.1-9

r decreases f cnsumptin alng flight. It will have a mean f 100 and a variance given in additinal flight mdel definitin file M 100 is the estimated fuel needed fr 100 minutes f verflying (ROF Cruise (cruise_altitude)*(100) M mim is the minimum mass f the aircraft as defined in BADA. M pyld is the paylad mass f the aircraft as defined in BADA. r is a sample f a unifrm randm variable frm Min_Ocupatin t Max_Ocupatin (parameters defined in additinal flight mdel definitin file). M dest (h,d) is the fuel cnsumptin needed t arrive frm injectin pint t airprt, calculated using the next algrithm (based n a table similar t that f PTF BADA files): First we shall cmpute the time needed t descent frm FL i+1 t FL i, using a vertical speed f ROCD i, t be T desc (i). T desc (i) = FL FL i+1 i *100 ROCD i Then we can cmpute the fuel spent t descend frm height h t airprt level (assumed t be zer), as: FL( h) M ( h) = ROF ( i)* T ( i) desc _ nm i= descend desc If we assume we are descending at the nminal true airspeed VTAS descend (i) between FL i+1 and FL i,the traversed distance during the cmplete nminal descend will be: h d ( h) = VTAS ( i)* T ( i) desc _ nm i= descend desc If we assume we are prir t descend we are at level flight at the nminal true airspeed, and T prev is the duratin f flight prir t entrance in TMA (user defined time cnstant), the cmplete weight will be: M dest (h,d) = ROF Cruise,lw (h) * T prev + d d (h) desc _ nm + M VTAS cruise (h) desc _ nm (h) Entrance Cnflict Remval After aggregating arrival and verfly flights, there shuld be a test t see if there is a ptential cnflict at any entry pint. If this is the case, there shuld be a mdificatin in injectin times f cnflicting aircraft, s that they are separated. Cnflict detectin will be perfrmed assuming a minimum time difference at entry pint. Cnflict reslutin will be perfrmed increasing time injectin f the last aircraft t arrive t the airprt until there is n cnflict. Please nte this culd induce a new cnflict in the entry pint, s there shuld be a recursive slutin fr this prblem btaining cnflict free traffic. Final Traffic Validatin This functin will recheck if the generated traffic pattern is within the peratinal limits defined by the user. We will allw fr a 10% additinal guard, s that even nt cmpletely cmpliant traffic is allwed. The fllwing limits will be checked against threshlds. Airprt number f peratins per hur. Entry pint number f peratins per hur. Exit pint number f peratins per hur. TMA traffic per hur. If any f thse threshlds is exceeded, we will nt save generated flights, and infrm abut the prblematic traffic pattern. Otherwise, we will save all the flights in the generated flights file, rdered by injectin time. 2.A.1-10

Traffic Analyzer and Generatr HMI Next we will summarize the input and utput HMIs f bth Traffic Analyzer and Traffic Generatr. Special emphasis is given t utput HMIs, due t the fact that they are the means t be used by the peratr t either understand the traffic situatin r t validate the generated scenari. Specific day wrst hur. Specific day. Traffic Analyzer Input HMI This interface allws fr the selectin f XML files including traffic data, peratinal cntext definitin, aircraft type and aircraft type class definitin, and t edit a user preferences file r select a preciusly generated user preferences file. Once all thse files are prvided, it has a buttn t invke traffic analysis. Traffic Pattern Analyzer Output HMI Next we will cnsider the presentatin f the results f the analysis mdule. There are means fr the representatin f aggregated data and pattern related data. Aggregated Traffic Data The analysis mdule will represent different views f traffic. There are tw result presentatin mdes: Traffic Map: The user select a time interval t present the results, and a time increment, and the sftware shws, using a clr cde, the number f aircraft passing thrugh each leg in each time interval (aggregated frm all flight types and adding deterministic flights and the mean number f randm flights), as shwn in Figure 6. Several different representatins may be dne in this mde: Nrmal days average hur. Nrmal days average day. Nrmal days average week. Nrmal days wrst hur. Nrmal days wrst day. Special days average hur. Special days average day. Specific hur. Specific day average hur. Figure 6. Example f Traffic Map Number f flights plt: The user selects a scenari element (SID, STAR, entry pint, exit pint, airprt, etc) and a plt f the evlutin f number f aircraft versus time (per hur) is shwn (see Figure 7). The user will be able t define ne f the fllwing time intervals Average day (24 hurs): It will perfrm the average f the desired traffic fr all days, separated by hur. Average Week (7 days x 24 hurs): It will perfrm the average f the desired traffic fr all weeks, separated by hur. It will be represented using seven number f flights plts, ne per day in the week. 2.A.1-11

Wrst day (24 hurs): It will find the highest f the desired traffic fr all days, separated by hur. There shuld be a methd t select an hur and btain details f the wrst day. Wrst Week (7 days x 24 hurs): It will find the highest f the desired traffic fr all days, separated by hur. It will be represented using seven number f flights plts, ne per day in the week. There shuld be a methd t select an hur and day and btain details f the wrst week. Special days (24 hurs): It will find the highest f the desired traffic fr all days, separated by hur. Specific day (24 hurs): It will shw the traffic fr this day, separated by hur. Specific week (7 days x 24 hurs): It will shw the traffic fr this week, separated by hur. It will be represented using seven number f flights plts, ne per day in the week. Figure 7. Example f Number f Flights Plt fr Specific Day The user will be able t aggregate/discriminate data ver the fllwing set f items: Type f flight: Arrival, Departure, Dmestic, Overfly, r All Data (aggregated) TMA element: TMA, SID, STAR, Dmestic rute, Overfly rute, Entry Pint, Exit Pint, Leg, Waypint, Origin Airprt, Destinatin Airprt, Origin Cell, Destinatin Cell. Type f pattern: Deterministic, Randm (we will use the mean number f flights in the randm pattern), Ttal (aggregatin f deterministic and randm) Deterministic Pattern Data The system will allw fr search f a specific flight number, and will prvide a means t shw its pattern data: Origin and Destinatin: Airprts/Airprt cells depending n the type f flight ETA/Departure Time/Entrance time mdel: Expected Time f Arrival t the airprt/departure frm airprt/entrance t TMA time mdel depending n the type f flight Repetitin Pattern: ne f the fur cases (Every day f the week, Several days f the week (specify which nes), ne day f the week (specifying which ne), Special days) Aircraft Types: Set f pairs (Designatr f an aircraft type f this class, Percentage f entrances fr assciated Aircraft Type Class) Additinal Traffic Data Additinal data frm the analysis must be visible t the user. It will be a methd t shw in a tabular manner data in the traffic pattern file: Arrival Pattern: Entry pint assignatin. STAR Assignatin. Cruise velcity distributins Entry pint distributins Landing delay mdel distributins Departure Pattern: Exit pint assignatin. SID Assignatin. Departure delay mdel distributins. 2.A.1-12

Dmestic pattern Rute Assignatin. Departure delay mdel distributins. Overfly pattern Entry Pint Assignatin Exit Pint Assignatin Overfly rute Assignatin Cruise velcity distributins Entry pint distributins Synthetic Traffic Generatr Input HMI This interface allws fr the selectin f XML files including traffic pattern data, additinal mdel data, peratinal cntext definitin, aircraft type and aircraft type class definitin, and t edit a user preferences file r select a previusly generated user preferences file. Once all thse files are prvided, it has a buttn t invke traffic generatin. Synthetic Traffic Generatr Output HMI Next we will cnsider the presentatin f the results f the generatr mdule. Aggregated Traffic Data The generatin mdule will represent different views f traffic, by aggregating all flights. All data is derived frm generated flights, nt frm generatr inputs. There are tw result presentatin mdes: Traffic Map: The user select a time interval t present the results, and a time increment, and the sftware shws, using a clr cde, the number f aircraft passing thrugh each leg in each time interval (aggregated frm all flight types and adding deterministic flights and the mean number f randm flights. The representatin is quite similar t that f the analysis, but with slightly different cases. Number f flights plt: The user selects a scenari element (SID, STAR, entry pint, exit pint, airprt, etc) and a plt f the evlutin f number f aircraft versus time (per hur) is shwn, as in the analysis, but calculated by cunting individual generated flights. The user will be able t define ne f the fllwing time intervals Average day (24 hurs): It will perfrm the average f the desired traffic fr all days, separated by hur. Average Week (7 days x 24 hurs): It will perfrm the average f the desired traffic fr all weeks, separated by hur. It will be represented using seven number f flights plts, ne per day in the week. Specific day (24 hurs): It will shw the traffic fr this day, separated by hur. Specific week (7 days x 24 hurs): It will shw the traffic fr this week, separated by hur. It will be represented using seven number f flights plts, ne per day in the week. The user will be able t aggregate/discriminate data ver the fllwing set f items: Type f flight: Arrival, Departure, Dmestic, Overfly, r All Data (aggregated) TMA element: TMA, SID, STAR, Dmestic rute, Overfly rute, Entry Pint, Exit Pint, Leg, Waypint, Origin Airprt, Destinatin Airprt, Origin Cell, Destinatin Cell. Detailed Traffic Data There will be a means t prvide, in a tabular manner, the data regarding the generated flights fr a given time interval. The prvided data will be different fr the different kinds f flights, and cmprises all the infrmatin needed fr flight simulatin. Cnclusins and Future Wrk In this paper we have summarized the design f a prttype traffic analysis and generatin tl. These systems are currently being validated and will serve initially as the basis fr traffic injectin in ATLANTIDA prject. Future develpments f this tl include: 2.A.1-13

Validatin and refinement f traffic mdels with experts (cntrllers, ATM engineers, ). Special emphasis shuld be given t the search f ptential crrelatins amng distributins f different flight aspects. Cmpletin f prttype HMI develpment. Inclusin f cmpany related fields, as cmpany preferences will have a clear impact n final trajectry. Adaptatin f prttype t INDRA ATM cmmercial prducts. Research n ptential means f explitatin f the tl. References [1] Integra-TSG Web page [2] AENA-PITOT Web page [3] R. Oaks, S. Liu, D. Zhu, M. Pagline, W. J. Hughes. Methdlgy fr the generatin air traffic scenaris based n recrded traffic data. DASC 2003. Indianaplis. [4] ATLANTIDA Web page [5] BADA Eurcntrl Web page 28th Digital Avinics Systems Cnference Octber 25-29, 2009 2.A.1-14