Aircraft push-out and taxiing collaborative decision-making method

Abstract

The invention relates to the technical field of aircraft scheduling management, in particular to an aircraft push-out and taxiing collaborative decision-making method, which constructs a cosine curve push-out rate control module, a push-out and taxiing collaborative decision-making model construction module and an adaptive algorithm design module. Searching an optimal queuing threshold value on multiple runways; secondly, multiple actual operation constraint conditions are emphatically considered, and a collaborative decision-making model with the dual optimization objectives of minimizing the total push-out time deviation duration and the total take-off time deviation duration of the aircraft and minimizing the total departure penalty cost and the oil consumption cost is constructed; and then fusing a Markov chain algorithm of continuous time, a genetic simulated annealing algorithm and a Q-learning reinforcement learning algorithm, scientifically arranging a departure aircraft scheduling sequence, deducing time, a sliding path and target runway distribution through an algorithm solving model, and finally improving the operation efficiency of the multi-runway airport in a mixed operation mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aircraft dispatch management, and particularly relates to an aircraft push-out and taxiing collaborative decision-making method. BACKGROUND

[0002] Airport aircraft apron dispatch is a complex and crucial process, and its main goal is to ensure the safety of aircraft entering and leaving the port and to maximize the efficiency of airport operation. The entire dispatch process involves multiple operating stages from the parking space to the runway, and requires airport managers to accurately plan, predict and avoid possible emergencies. At the system architecture level, the parking space system, the taxiway network system and the runway system constitute the three core modules of aircraft dispatch, and each subsystem maintains functional independence while working collaboratively to ensure the safety and efficiency of aircraft ground operation.

[0003] The air transportation system is susceptible to weather conditions, air traffic control restrictions, airport facilities and management levels, and other factors, and the resulting flight delays, delay accumulation and even flight cancellations are significantly random and unpredictable. The sudden changes in aircraft operation time can disrupt the established path planning scheme, causing runway entrance congestion and a chain of delay effects, thereby significantly increasing the complexity of aircraft dispatch. In recent years, scholars have increasingly conducted in-depth research on airport ground resource optimization and dispatch, particularly on the optimization of key resources such as runways, taxiways and parking spaces. On the one hand, for the dispatch of runway and taxiway network node resources, when the aircraft push-out time changes, its taxi path changes, thereby causing dynamic changes in the entire apron operation situation. On the other hand, the differences in aircraft taxi time and push-out delay duration further change the apron network state, directly affecting the controller's allocation decisions for subsequent aircraft taxi path nodes and runway resources.

[0004] However, under the traditional operating mode, there is a lack of effective coordination mechanism among each control unit, making it difficult to achieve global optimal decision-making. Firstly, most current departure control methods focus on the collaborative optimization of arrival and departure runway sequencing, but the overall coordination mechanism of push-out control and path planning is relatively insufficient. Secondly, in the process of aircraft taxi path planning, existing researches are mostly limited to push-out time slot allocation problems, without considering push-out rate control as a key constraint condition. Thirdly, previous researches are mostly based on the assumption of single-runway airport or fixed runway operation mode, which is difficult to adapt to the actual needs of future multi-runway airport dynamic operation. SUMMARY

[0005] The purpose of the present application is to provide an aircraft push-out and taxiing collaborative decision-making method, which reasonably transfers the taxiway waiting period to the parking space waiting stage, optimizes the taxi path and runway allocation scheme, and effectively reduces fuel consumption.

[0006] To achieve the above object, the application provides an aircraft push-out and taxi coordination decision method, comprising the following steps:

[0007] Step 1: constructing a field network structure diagram, abstracting the airport field structure into a field network structure diagram;

[0008] Step 2: constructing a dynamic push-out rate control model, and then determining the optimal queuing threshold of the taxiway queue;

[0009] Step 3: constructing a departure scheduling coordination decision model with the dual optimization objectives of minimizing the total push-out time deviation and the total take-off time deviation, minimizing the total departure penalty cost and the fuel consumption cost;

[0010] Step 4: obtaining the optimal solution of the model by using the adaptive algorithm of the fusion of the continuous-time Markov chain algorithm, the genetic simulated annealing algorithm and the Q-learning reinforcement learning algorithm.

[0011] Optionally, in the execution process of Step 2, a cosine curve push-out rate control strategy is adopted, the multi-runway scene and the taxiway waiting condition are comprehensively considered, the planned push-out time table of the whole day is taken as the input, and the dynamic push-out rate control model is constructed with the dual optimization objectives of minimizing the parking space waiting penalty cost and the taxiway waiting fuel consumption cost, and then the optimal queuing threshold of the taxiway queue is determined.

[0012] Optionally, the definition of the cosine curve push-out rate control function in Step 2 is as follows:

[0013]

[0014] Wherein, λ n is the actual push-out rate; λ is the applied push-out rate; n is the current taxiway queuing length; the taxiway queue length threshold N is an unknown value, which is a to-be-determined target.

[0015] Optionally, the objective function of the minimum push-out cost of the dynamic push-out rate control model is as follows:

[0016]

[0017] Wherein, m=1,2,...∈S out is the departure aircraft index; r=1,2,...∈R is the runway set; c fule.2 represents the fuel consumption oil price of the taxiway waiting, and the unit is RMB / min; t m.h is the parking space waiting time, and the unit is min; t m.wait represents the taxiway waiting time, and the unit is min.

[0018] Optionally, the objective function built in step 3 includes a total deviation time function and a total departure cost function, which are respectively expressed as follows:

[0019]

[0020] Wherein, S m is the actual push-off time; K m is the planned push-off time; T m.f is the actual take-off time; T m.plantakeoff is the planned take-off time; c fule.1 is the fuel cost per minute during taxiing, RMB / min; T m.1 is the time of reaching the runway; and ξ is the stand penalty coefficient.

[0021] Optionally, in step 4, a hierarchical optimization architecture is adopted, the outer optimization module realizes global optimization of the aircraft departure scheduling scheme by using an improved genetic simulated annealing hybrid algorithm; meanwhile, a continuous-time Markov chain algorithm is introduced to search for the lowest cost under different taxiway queue length thresholds by using an iterative loop, so as to solve the optimal taxiway queue threshold; in addition, a Q-learning algorithm is introduced to dynamically adjust the crossover probability P c and the mutation probability P m .

[0022] Optionally, the execution process of the adaptive algorithm includes the following steps:

[0023] Step 1: initialize the population, randomly generate a certain number of chromosomes, generate an initial scheme, i.e., the initial order of the aircraft conforms to the FCFS principle, generate the initial push-off time and the initial path; initialize the crossover probability P c and the mutation probability P m ; initialize the temperature; initialize the Q table, set all values in the Q table to 0; set the iteration counter to 0; initialize the queue threshold;

[0024] Step 2: determine whether the final iteration number is reached? If “yes”, jump to step 14; if “no”, proceed to step 3;

[0025] Step 3: obtain the chromosome code;

[0026] Step 4: initialize / update the threshold N;

[0027] Step 5: read the aircraft m departure scheme, including the start scheduling time, the push-off time, the path and the target runway;

[0028] Step 6: determine whether the current queue length is less than the queue threshold when the aircraft m is pushed off? If “yes”, proceed to step 7; if “no”, it is infeasible, jump to step 9;

[0029] Step 7: judge F < (1-n / N)?

[0030] If yes, go to step 8; if no, it is not feasible, jump to step 9;

[0031] Step 8: judge whether the current aircraft index is less than the last aircraft index m < M? If yes, update the aircraft index to m+1, calculate the objective function value, and jump to step 5; otherwise, go to step 9;

[0032] Step 9: judge the relationship between the team leader threshold and the threshold limit N < N max ? If yes, jump to step 4 and update the taxiway team leader threshold to N+1; if no, continue step 10;

[0033] Step 10: compare all objective function values and output the optimal solution, i.e. end the inner loop;

[0034] Step 11: evaluate using the fitness function based on the total departure cost of the aircraft, and use the tournament selection strategy to select parent individuals from the current population according to the fitness;

[0035] Step 12: the agent gets the current time step t and state s t ; according to certain rules, the selected parent individuals are crossed, recombined, and mutated, and the scheduling sequences and path adjustments of part of the aircrafts in the two individuals are exchanged;

[0036] Step 13: simulated annealing replacement, update the population; the state is changed to s t+1 , the strategy function is updated, the agent selects the next action a t+1 and updates the Q table, and jumps to step 2;

[0037] Step 14: output the optimal solution, i.e. the individual with the highest fitness.

[0038] The application provides an aircraft pushback and taxi coordination decision-making method, which takes a departure aircraft as a research object, is characterized by fusing pushback rate control, and focuses on integrating airport aircraft pushback control and path planning processes; and is specifically based on a cosine-based dynamic pushback control (CPC) module, a pushback and taxi coordination decision-making model construction module, and an adaptive algorithm design module; the CPC module is based on a Markov state transition model and explores the optimal queuing threshold on multiple runways; the pushback and taxi coordination decision-making model construction module focuses on actual operation constraint conditions such as pushback rate, taxi conflict risk, fuel consumption, and taxiway queuing conditions, and constructs a coordination decision-making model with dual optimization objectives of minimizing the deviation length of the total pushback time and the takeoff time and minimizing the total departure penalty cost and fuel consumption; then, the Markov chain algorithm in continuous time, the genetic simulated annealing algorithm, and the Q-learning reinforcement learning algorithm are fused, the model is solved through the algorithms to scientifically arrange the departure aircraft scheduling sequence, the pushback time, the taxi path, and the target runway allocation, and finally the operation efficiency of a multi-runway airport in a mixed operation mode is improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0040] Figure 1 is a schematic diagram of an airport surface network structure in a specific embodiment of an aircraft pushback and taxi coordination decision-making method of the present application.

[0041] Figure 2 is a characteristic curve diagram of a pushback rate of an aircraft pushback and taxi coordination decision-making method of the present application.

[0042] Figure 3 is a step flowchart of an adaptive algorithm of an aircraft pushback and taxi coordination decision-making method of the present application.

[0043] Figure 4 is a schematic diagram of departure demand distribution of a specific embodiment of the present application.

[0044] Figure 5 is a schematic diagram of simulation results of each scheme of a specific embodiment of the present application.

[0045] Figure 6 is a schematic diagram of a queuing length comparison of each scheme of a specific embodiment of the present application.

[0046] Figure 7 is a sliding time comparison schematic diagram of each scheme of the specific embodiment of the present application.

[0047] Figure 8 is a total departure time comparison schematic diagram of each scheme of the specific embodiment of the present application.

[0048] Figure 9 is a sliding fuel consumption cost comparison schematic diagram of each scheme of the specific embodiment of the present application. DETAILED DESCRIPTION

[0049] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0050] The present application provides an aircraft push-out and taxiing collaborative decision-making method, comprising the following steps:

[0051] S1: constructing a field network structure diagram, abstracting the airport field structure into a field network structure diagram;

[0052] S2: constructing a dynamic push-out rate control model, and then determining the optimal queuing threshold of the taxiway queue;

[0053] S3: constructing a departure dispatch collaborative decision-making model with the dual optimization objectives of minimizing the total push-out time deviation and the total take-off time deviation, and minimizing the total departure penalty cost and the fuel consumption cost;

[0054] S4: obtaining the optimal solution of the model by fusing the Markov chain algorithm in continuous time, the genetic simulated annealing algorithm, and the Q-learning reinforcement learning algorithm.

[0055] The following is further described in combination with the execution steps and specific embodiments:

[0056] Step S1: constructing a field network structure diagram, abstracting the airport field structure into a field network structure diagram;

[0057] The ground taxiing system of Beijing International Airport is constructed as a field simulation structure model. The field taxiing network diagram is as follows Figure 1As shown, the taxiway network is composed of gate entry / exit points, taxiway intersection nodes and runway nodes. The present application analyzes the taxiway network of a field with two parallel runways and taxiways, including 132 gate points, 2 runways and 484 taxiway nodes. The first runway is 3200m x 50m, 4E level, and the code is 36L / 18R; the second runway is 3800m x 60m, 4E level, and the code is 36R / 18L.

[0058] In step S2, a Cosine-based Dynamic Pushback Control (CPC) strategy is adopted. The multi-runway scenario and taxiway waiting situation are comprehensively considered. The all-day planned pushback schedule is taken as the input. The minimum gate waiting penalty cost and taxiway waiting fuel consumption cost are taken as the dual optimization objectives. A dynamic pushback control model is constructed to determine the optimal taxiway queue length threshold.

[0059] Specifically, the following steps are included:

[0060] 2-1) The definition of the Cosine-based Dynamic Pushback Control function is as follows:

[0061]

[0062] Wherein, λ n is the actual pushback rate; λ is the application pushback rate; n is the current taxiway queue length; the taxiway queue length threshold N is an unknown value, which is a to-be-determined target.

[0063] The core objective of the CPC strategy is to determine the optimal control rate and set the corresponding control implementation threshold, considering the specific N value constraint applicable to different scale airports. The characteristic curve of the pushback rate is as shown in Figure 2 When the number of aircraft in the taxiway queue has not reached the control value N, the probability of allowing the aircraft requesting to take off to leave is represented as λ[0.5cos(πn / N)+0.5]. Once the number of aircraft in the taxiway queue exceeds the threshold N, no other aircraft is allowed to leave.

[0064] 2-2) Construct the objective function:

[0065]

[0066] Equation (2) is to minimize the pushback cost; m=1,2,...∈S out is the departure aircraft index; r=1,2,...∈R is the runway set; c fule.2 represents the taxiway waiting fuel consumption oil price, RMB / min; t m.ht is the maximum stand-by time of the stand, min; t m.wait t is the stand waiting time, min. According to the current operation standard of Beijing Capital International Airport, if the aircraft can complete the gate push-out operation within the time limit of 30 minutes of the expected take-off time, it meets the "on-time" operation standard. Based on this, the maximum stand waiting time threshold is defined as 30 minutes, that is, the aircraft must complete the stand push-out operation within the time limit of 30 minutes after the planned push-out time. When the stand waiting time is small, the operator does not directly implement economic penalty measures, but guides the aircraft to stay in the stand for a reasonable time limit through the operation management system, while effectively avoiding long-term occupation of stand resources. Therefore, the present application introduces a stand penalty function wherein the fuel cost function and the exponential function are equal at the 30-minute time node by mathematical modeling method, so as to accurately deduce the stand penalty coefficient ξ = [ln(c fule.2 ·t+1)] / t.

[0067] 2-3) Constraint condition:

[0068]

[0069] Formula (3) is the push-out rate control.

[0070] 0≤n≤N≤N ideal (4)

[0071] Formula (4) is the taxiway queue length constraint, the maximum taxiway queue length cannot be greater than the control threshold N, and the upper limit of the value of N is N deal It is set to 30.

[0072] t m.h ≤30 (5)

[0073] Formula (5) is the push-out deviation time constraint.

[0074] The process of step S3 is the push-out and taxiway cooperative decision-making model construction process,

[0075] The existing pushback rate control model uses the predicted taxi time as input and does not consider the dynamic taxi time changes. Therefore, on the basis of the pushback rate control model, a departure scheduling collaborative decision model is established by considering the aircraft taxi path on the taxiway system, taxi conflict and taxi time. The scheduling model takes the pushback rate control as a constraint, guides the aircraft to stay at the parking stand for a period of time to reduce the taxiway queue waiting time, further reduces the overall taxi time by optimizing the path, so as to reduce the fuel consumption cost of taxiing. However, staying at the parking stand will inevitably affect the actual pushback time of the aircraft, and further affect the path and target runway of the aircraft. Therefore, a departure scheduling collaborative decision model is constructed with the dual optimization objectives of minimizing the total pushback time deviation and takeoff time deviation of the aircraft and minimizing the total departure penalty and fuel consumption cost.

[0076] 3-1) Construction of objective function:

[0077]

[0078] Equation (6) is the total deviation time function; equation (7) is the total departure cost function. Wherein, S m is the actual pushback time; K m is the planned application pushback time; T m.f is the actual takeoff time; T m.plantakeoff is the planned takeoff time; c fule.1 is the fuel consumption cost per minute during taxiing, RMB / min; T m.1 is the time to reach the runway; and ξ is the parking stand penalty coefficient.

[0079] 3-2) Constraint condition:

[0080]

[0081] Equation (8) is the pushback rate control constraint.

[0082] S m =K m +t m.h (9)

[0083]

[0084] T m.f =T m.1 +t m.wait (11)

[0085] Equations (9)-(11) are continuous time constraints, which respectively represent the actual pushback time of the aircraft, the time when the aircraft enters the taxiway waiting queue and the actual takeoff time. Wherein, is the taxiing distance from node i to node j, m; V is the taxiing speed of the aircraft, m / s; D is the path node set, i, j ∈ D, i ≠ j.

[0086]

[0087] Equations (12)-(13) are the push-out sequence constraints, which limit the deviation between the planned take-off time and the actual take-off time of the aircraft to be controlled within a reasonable interval of 15 minutes (the allowable time difference η is set to 15 min). Wherein, g is the maximum allowable time difference of the push-out time; η is the maximum allowable time difference of the take-off time.

[0088] E msj +E smj = 1 (14)

[0089]

[0090] Equations (14)-(16) are the taxi node safety interval constraints, which strictly check the timing relationship of the aircraft passing through the nodes to avoid potential conflicts. For the case of two consecutive aircrafts arriving at the same node or sharing the same taxiway, the time difference must meet the minimum safety interval standard: for a specific node, the subsequent arriving aircraft and the previous aircraft must maintain a Δt time interval; for the taxiway scenario: (1) if the aircraft m, s pass in the same direction, the node passing sequence consistency must be maintained, that is, m must pass through the starting node first, then it must also pass through the end node first; (2) if the aircraft m, s pass in opposite directions, the one that passes through the starting node of the road segment first must also pass through the end node first. Wherein, T j is the time of arrival at node j ∈ P.

[0091]

[0092] Equation (17) formulates the runway exclusivity constraint, which stipulates that only a single aircraft is allowed to occupy the runway during a single runway operation period, and the arrival time of the two aircrafts entering the runway must be at least μ service time apart. Wherein, μ is the runway service rate; X is a very large number.

[0093]

[0094] Equation (18) explicitly specifies the runway uniqueness constraint, which requires that each departing aircraft must be assigned to a unique runway and prohibits multiple runways from being assigned simultaneously.

[0095]

[0096]

[0097] Formulas (19) and (20) set the sequence exclusivity constraint to ensure that a single aircraft can only occupy a single sequence position and eliminate sequence overlap. mk The dispatch sequence for aircraft; The queue sequence for aircraft at the end of the taxiway for runway r.

[0098]

[0099] Formulas (21)-(25) are binary variables. When two aircraft are scheduled to be launched consecutively, K ms =1 means s∈S out Rank m∈S out After that, otherwise K ms =0; when two aircraft pass through the taxiing node consecutively, E msi =1 means s∈S arrives at node i after m∈S, otherwise E msi =0; when two consecutive aircraft arrive at runway r, H ms =1 means s∈S out Rank m∈S out After that, otherwise H ms =0; for an aircraft, R mr =1 means the aircraft takes off from runway r, otherwise R mr =0; if aircraft m taxis from node i to node j, then Otherwise

[0100] Step S4 is an adaptive algorithm design process. In the present application, a combination of a continuous-time Markov chain algorithm, a genetic simulated annealing algorithm and a Q-learning reinforcement learning algorithm is used to propose an adaptive genetic simulated annealing algorithm based on a pushback control strategy (Control Pushback-based Genetic Adaptive Simulated Annealing with Reinforcement Learning, CPC-GASS-RL), referred to as CGR algorithm. The CGR adopts a hierarchical optimization architecture: the outer optimization module realizes global optimization of the aircraft departure scheduling scheme by using an improved genetic simulated annealing hybrid algorithm. The module particularly considers three key time parameters: the estimated pushback time, the taxiing duration period and the takeoff time, to ensure that the generated initial scheme meets various operating constraints. The inner pushback control module introduces a real-time dynamic evaluation mechanism, continuously monitors the queuing state of the taxiway system, and verifies the feasibility of the initial scheme. When the scheme does not meet the constraints, the module will start the adaptive adjustment program, and implement a two-dimensional correction for the aircraft that does not meet the constraints: (1) time dimension, recalculate the optimal pushback time within the allowed time flexibility range (2) space dimension, re-plan the optimal taxiing path based on the current field operating state.

[0101] In the genetic simulated annealing hybrid algorithm, a fitness function based on the total pushback time deviation of the aircraft, the total takeoff time deviation and the total penalty and fuel consumption cost (collectively referred to as departure cost) is used, which can accurately select individuals with excellent performance. The formula is as follows:

[0102]

[0103] In the CGR algorithm, in order to adapt to the dynamics of the field, a continuous-time Markov chain algorithm is introduced to search for the lowest cost under different taxiway queuing length thresholds by using an iterative loop, to solve the optimal taxiway queuing threshold. This process embeds the changing state in the genetic simulated annealing algorithm in an interactive form, thereby affecting the planning of the key time points and path scheme in the algorithm, calculating its fitness, and then finding the optimal solution within the period under this condition. During the algorithm running process, the effect of the genetic algorithm is mainly affected by the crossover probability P c and the mutation probability P m . With the increase of the number of iterations, the reinforcement learning process is activated, and the agent can continuously optimize P c and P m through past experience and current learning state. Therefore, by introducing the Q-learning algorithm to dynamically adjust P c and P m , the performance of the genetic algorithm can be effectively improved. First, the agent obtains the current time step t and state wt , take corresponding action a in the preset policy function t , through genetic, repair and simulated annealing replacement process, the algorithm changes from the current state to w t+1 , at the same time, the agent obtains feedback from the change of algorithm performance, updates the policy function according to the feedback, and then selects the next action a t+1 . The agent updates the Q table according to the state of the current population, historical data and prediction of future state.

[0104] The algorithm performs the following steps:

[0105] Step 1: initialization. Initialize the population, randomly generate a certain number of chromosomes, generate the initial scheme, that is, the initial order of the aircraft conforms to the FCFS principle, generate the initial push-out time and the initial path; initialize the crossover probability P c and the mutation probability P m ; initialize the temperature; initialize the Q table, set all values in the Q table to 0; set the iteration counter to 0; initialize the queue threshold.

[0106] Step 2: determine whether the final iteration number is reached? If yes, jump to step 14; if no, go to step 3, enter the interaction stage.

[0107] Step 3: get the chromosome code.

[0108] Step 4: initialize / update threshold N.

[0109] Step 5: read the aircraft m departure scheme, including the start scheduling time, push-out time, path and target runway.

[0110] Step 6: determine whether the current queue length is less than the queue threshold when the aircraft m is pushed out? If yes, go to step 7; if no, it is infeasible, jump to step 9.

[0111] Step 7: determine whether F < (1-n / N)? If yes, go to step 8; if no, it is infeasible, jump to step 9.

[0112] Step 8: determine whether the current aircraft index is less than the last aircraft index m < M? If yes, update the aircraft index to m+1, calculate the objective function value, and jump to step 5; otherwise, go to step 9.

[0113] Step 9: determine whether the queue length threshold and the threshold limit are related N < N max ? If yes, jump to step 4 and update the taxiway queue length threshold to N+1; if no, continue step 10;

[0114] Step 10: compare all objective function values and output the optimal scheme (end inner loop).

[0115] Step 11: Fitness evaluation. The fitness function based on the total departure cost of the aircraft is used for evaluation, and the selection strategy tournament selection is used to select the parent individuals from the current population according to the fitness.

[0116] Step 12: The agent obtains the current time step t and state s t ; the selected parent individuals are subjected to crossover, recombination, and mutation operations according to certain rules, and the scheduling sequences and adjustment paths of part of the aircrafts in the two individuals are exchanged.

[0117] Step 13: Simulated annealing replacement, update the population. The state is changed to s t+1 , the strategy function is updated, the agent selects the next action a t+1 , and the Q table is updated. Go to step 2.

[0118] Step 14: Output the optimal solution, i.e. the individual with the highest fitness.

[0119] Further, the application also carries out simulation comparison experiments, and specifically selects the performance data of an air transportation system of a certain year of Beijing Capital International Airport as the data source, which contains flight number, arrival and departure state, aircraft model, planned / actual push time, taxi time, planned / actual takeoff time, planned / actual landing time, etc. Among them, the aircraft data from November to the next January is preliminarily analyzed, the flight plan of this period is consistent with the half-year data, and the daily average departure demand obtained accordingly ensures the reliability of the research results, and the departure demand distribution is as shown in Figure 4 .

[0120] The simulation experiments are all completed in MATLAB R2024 software. The experimental settings are as follows:

[0121] (1) Flight schedule. Including the flight number, arrival and departure state, planned push time, planned takeoff time, and planned landing time of 756 aircrafts.

[0122] (2) Parameter setting of the aircraft. Since the takeoff and landing aircrafts of the Capital Airport are mainly composed of heavy and medium aircrafts, the safety time interval Δt is set to 1 min; the taxi speed V is 10 m / s; the maximum stand waiting time is 30 min; the runway port release interval is a key factor for determining the takeoff time of the departure aircraft, and the release interval of the departure aircrafts queuing on the runway is generally the runway service time μ = 2 min; the fuel cost consumed per minute during taxiing c fule = 120 yuan, the fuel cost consumed per minute during taxiing c fule.2 = 60 yuan / kg; the stand waiting penalty coefficient ξ = 0.245.

[0123] (3) Algorithm parameter setting. The maximum evolution generation L = 200; the population size O = 100; the minimum mutation probability P = 0.01, the maximum mutation probability P = 0.21; the minimum crossover probability P = 0.4, the maximum crossover probability P = 0.9; the initial temperature T = 100; the termination temperature T = 0.0001; the attenuation coefficient a = 0.98. p m2 m1 c2 c1

[0124] (4) Experimental scheme setting. The effectiveness of the model and algorithm is verified by the comparative experimental method. The core of the push-out strategy is to plan the push-out time of the aircraft according to the waiting time of the parking stand, the taxi time and the taxiway waiting queue situation; the core of the taxi strategy is to avoid the conflict of the aircraft in the taxi road network, and to select the target runway by minimizing the fuel consumption cost. In order to explore the influence of the scheduling optimization method on the optimization result and ensure the continuity of the push-out operation, the sequencing, assignment and path scheme of 284 departures in the peak period (12:40-17:00) are optimized. Four groups of comparative schemes are set, and the scheme description is shown in Table 1. Case 1 is the baseline scheme, that is, the simulation is run according to the planned schedule and the planned path; Case 2 implements the push-out control and runway sequencing scheduling method, that is, the push-out strategy; Case 3 implements the path optimization and runway assignment scheduling method, that is, the taxi strategy; Case 4 implements the push-out control and runway sequencing, path optimization and runway assignment scheduling method, that is, the combination of the scheduling optimization methods of Case 2 and Case 3. Among them, CASE4 is the optimal scheduling method of the application.

[0125] Table 1 Experimental scheme

[0126] Scheme Description CASE 1 No push-out strategy implemented, no coasting strategy implemented CASE 2 Push-out strategy implemented, no coasting strategy implemented CASE 3 No push-out strategy implemented, coasting strategy implemented CASE 4 Push-out strategy implemented, coasting strategy implemented

[0127] Table 2 presents the simulation results of the four schemes, including five indexes of 1. the maximum taxiway queue length, 2. the total fuel consumption cost (RMB), 3. the average taxi time (min), 4. the average taxi completion time (min), and 5. the average taxiway waiting time (min). Due to the difference in technical complexity, the results of the four schemes are different. The research content CASE4 of the application is the optimal scheme.

[0128] Table 2 Comparison of simulation results of schemes

[0129] Scheme 1 2 3 4 5 CASE 1 41 2388.86 18.07 21.75 3.68 CASE 2 35 2352.91 18.07 21.15 3.08 CASE 3 18 2127.11 16.62 18.83 2.22 CASE 4 19 1969.98 15.61 17.22 1.61

[0130] As Figure 5 ​​​​​​​As shown, the overall data trend is that the index value gradually decreases with the increase of technical complexity (from CASE1 to CASE4) (for ease of display, the unit of fuel consumption cost in the figure is 102RMB). Figures 6 to 8 The queuing length, taxi time and total departure time of each aircraft in the four groups of schemes are shown. From Figure 6 and Figure 7 Overall, there is an obvious regularity in all schemes, that is, the queuing length and taxi time in the peak period of arrival and departure demand are significantly higher than those in the low peak period, which shows that the peak period needs more push-out control. From Figure 6 the comparison of the queuing length results, the increase of the push-out control strategy significantly improves the queuing of the aircraft on the taxiway; from Figure 7 the comparison of the taxi time results, the taxi time range gradually decreases. Taking CASE1 as an example, the maximum queuing length is 41, the average taxiway waiting time is 3.68min, and the average taxiway waiting fuel consumption is 220.56RMB. Compared with CASE1, the maximum queuing length of CASE2 is 35, the average taxiway waiting time is reduced by 16.11%, and the average taxiway waiting fuel consumption is 185.02RMB, which is reduced by 16.11%. As Figure 8 shown, the implementation of the push-out strategy converts part of the taxiway waiting time into stand waiting time, thereby reducing the fuel consumption cost, verifying the effectiveness of the push-out strategy. Correspondingly, the stand waiting time of CASE2 is no longer 0. Therefore, the benefit of taxiway waiting time is at the cost of increasing stand waiting time. The above analysis shows that compared with no control, the CPC method of the application has higher economic benefit.

[0131] As Figure 8 and Figure 9 shown, by comparing the total departure time and taxi fuel consumption cost of the four groups of schemes, the effectiveness of the taxi strategy is verified. The average taxiway waiting time of CASE1 is 3.68min, the average completion taxi time is 21.75min, and the average taxi fuel consumption cost is 2388.86RMB. Compared with CASE1, the average taxiway waiting time of CASE3 is reduced by 39.74%, the average completion taxi time is reduced by 13.39%, and the average taxi fuel consumption cost is reduced by 10.96%. The implementation of the taxi strategy optimizes the taxi path and target runway allocation, reduces the total completion taxi time, and thereby reduces the fuel consumption cost, verifying the effectiveness of the taxi strategy.

[0132] Overall, in the non-optimized method CASE1 scheme, part of the aircraft needs to wait for a very long time on the taxiway, greatly increasing the departure cost; compared with CASE1, the maximum queue length of the push-out and taxi coordination decision method CASE4 is greatly reduced to 19, the average taxiway waiting time is 1.61 min, which is reduced by 56.24%, the average taxi completion time is 17.22 min, which is reduced by 20.81%, and the average taxi fuel consumption cost is 1969.98 RMB, which is reduced by 17.53%. It shows that under the action of the push-out and taxi coordination decision method, the fuel consumption is greatly reduced, and the operation efficiency of the multi-runway airport in the mixed operation mode is improved.

[0133] Tables 3 and 4 show part of the optimal scheduling results. In Table 3, the aircraft serial number represents the numerical number of each aircraft, the planned push-out time, the actual push-out time, the planned take-off time, the time take-off time indicates the time deviation and the duration of the departure operation, and the total departure fuel consumption cost directly indicates the surface operation cost. In Table 4, the path and target runway indicate the running state of the aircraft at each node.

[0134] Table 3 optimal scheduling results

[0135]

[0136]

[0137] Table 4 path and target runway

[0138]

[0139]

[0140] To sum up, compared with the prior art, the beneficial effects of the present application are:

[0141] 1. The push-out and taxi coordination decision model is constructed, mainly considering the push-out rate control, the taxi conflict, the taxiway queue situation constraint, and taking the minimum total push-out time deviation and the total take-off time deviation and the minimum total departure penalty and fuel consumption cost as the optimization target.

[0142] 2. A push-out and taxi coordination method is proposed. The method includes a push-out strategy and a taxi strategy. The push-out strategy plans the push-out time of the aircraft according to the stand waiting time, taxi time and taxiway waiting queue situation. The taxi strategy avoids the conflict of the aircraft in the taxi network, and selects the target runway by minimizing the fuel consumption cost. In terms of time, the push-out time of the departing aircraft is controlled by considering the taxiway waiting situation at the runway port, so as to transfer the taxiway waiting time to the stand waiting time as much as possible, reduce the taxiway waiting time, and in turn reduce the fuel consumption cost. In terms of space, the taxi path is planned by considering the conflict of the taxi nodes and the runway occupation situation, the overall taxi time is optimized, and the overall fuel consumption cost is reduced.

[0143] 3. A genetic simulated annealing algorithm with nested Markov state transition is proposed, and a reinforcement learning mechanism is introduced in the algorithm running process to accelerate the convergence of the algorithm.

[0144] The above only discloses one or more preferred embodiments of the present application, of course cannot limit the scope of the right of the present application, the person skilled in the art can understand that the all or part of the processes of the above-mentioned embodiments are implemented, and the equivalent changes made according to the claims of the present application still belong to the scope covered by the present application.

Claims

1. A collaborative decision-making method for aircraft pushback and taxiing, characterized in that: The following steps are involved: Step 1: Construct a scene network structure diagram and abstract the airport scene structure into a scene network structure diagram; Step 2: Build a dynamic push-out rate control model to determine the optimal queue threshold for the taxiway queue; Step 3: Construct a departure scheduling collaborative decision-making model with the dual optimization objectives of minimizing the total pushback time deviation and total takeoff time deviation of aircraft, and minimizing the total departure penalty cost and fuel consumption cost; Step 4: Integrate the continuous-time Markov chain algorithm, genetic simulated annealing algorithm and Q-learning reinforcement learning algorithm to obtain the adaptive algorithm to solve the optimal solution of the model.

2. The aircraft pushback and taxiing collaborative decision-making method according to claim 1, wherein: During the execution of step 2, a cosine curve-based pushback rate control strategy is adopted, comprehensively considering the multi-runway scenario and taxiway waiting conditions. The planned pushback schedule for the entire day is used as input, and the dual optimization objectives of minimizing the penalty cost of waiting at the parking stand and the fuel consumption cost of waiting on the taxiway are taken. A dynamic pushback rate control model is constructed to determine the optimal queuing threshold of the taxiway queue.

3. The aircraft pushback and taxiing collaborative decision-making method according to claim 2, wherein: The definition of the cosine curve push-out rate control function in step 2 is as follows: Among them, λ n is the actual push-out rate; λ is the push-out application rate; n is the current taxiway queue length; the taxiway queue length threshold N is an unknown value and is a target to be determined.

4. The aircraft pushback and taxiing collaborative decision-making method according to claim 3, wherein: The objective function of minimizing the launch cost of the dynamic launch rate control model is: Where m=1,2,...∈S out is the index of the departing aircraft; r=1,2,...∈R is the runway set; c fule.2 Indicates the fuel consumption price during taxiing and waiting, in RMB / min; t m.h is the parking space waiting time, in min; t m.wait Indicates the taxiway waiting time in minutes.

5. The aircraft pushback and taxiing collaborative decision-making method according to claim 4, wherein: The objective function constructed in step 3 includes the total deviation time function and the total departure cost function, which are expressed as follows: Among them, S m is the actual launch time; K m Time to launch the plan application; T m.f is the actual take-off time; T m.plantakeoff is the planned take-off time; c fule.1 is the fuel cost consumed per minute during taxiing, RMB / min; T m.1 is the time of arrival at the runway; ξ is the parking stand penalty coefficient.

6. The aircraft pushback and taxiing collaborative decision-making method according to claim 5, wherein: In step 4, a hierarchical optimization architecture is adopted. The outer optimization module realizes the global optimization of the aircraft departure scheduling plan through an improved genetic simulated annealing hybrid algorithm. At the same time, a continuous-time Markov chain algorithm is introduced to use iterative loops to search for the minimum cost under different taxiway queue length thresholds to solve the optimal taxiway queue threshold. In addition, the Q-learning algorithm is introduced to dynamically adjust the crossover probability P. c and mutation probability P m .

7. The aircraft pushback and taxiing collaborative decision-making method according to claim 6, wherein: The execution process of the adaptive algorithm includes the following steps: Step 1: Initialize the population, randomly generate a certain number of chromosomes, generate an initial plan, that is, the initial order of the aircraft follows the FCFS principle, generate the initial launch time and initial path; initialize the crossover probability P c and mutation probability P m ; Initialization temperature; Initialize the Q table, set all values ​​in the Q table to 0; set the iteration counter to 0; initialize the queue threshold; Step 2: Determine whether the final number of iterations has been reached. If yes, go to step 14; if no, go to step 3. Step 3: Obtain chromosome code; Step 4: Initialize / update threshold N; Step 5: Read the departure plan of aircraft m, including the start scheduling time, pushback time, path and target runway; Step 6: Determine whether the current queue length is less than the queue threshold when aircraft m is pushed out. If "yes", proceed to step 7; if "no", it is not feasible and jump to step 9. Step 7: Determine whether F < (1-n / N)? If "yes", proceed to step 8; if "no", it is not feasible and jump to step 9; Step 8: Determine whether the current aircraft index is less than the last aircraft index m<M. If so, update the aircraft index to m+1, calculate the objective function value, and jump to step 5; otherwise, proceed to step 9. Step 9: Determine the relationship between the captain threshold and the threshold limit N<N max If "yes", jump to step 4 and update the taxiway length threshold to N+1; if "no", continue to step 10; Step 10: Compare all objective function values ​​and output the optimal solution, ending the inner loop; Step 11: Use the fitness function based on the total departure cost of the aircraft for evaluation, adopt the selection strategy tournament selection, and select the parent individual from the current population based on the fitness; Step 12: The agent obtains the current time step t and state s t Perform crossover, recombination, and mutation operations on the selected parent individuals according to certain rules, and exchange the scheduling sequences and adjustment paths of some aircraft in the two individuals; Step 13: Simulated annealing replacement, update the population; the state changes to s t+1 , update the policy function, and the agent chooses the next action a t+1 And update the Q table and jump to step 2; Step 14: Output the optimal solution, that is, the individual with the highest fitness.

Images