Aircraft fleet takeoff scheduling method, device, equipment and storage medium

Abstract

The present application relates to a method, device, equipment and storage medium for scheduling the takeoff of a fleet of aircraft. The method includes: constructing a dispatch scheduling model; constructing a multi-island distribution model, and initializing the distribution model of each island; performing parallel iterative updates on the initialized distribution model of each island according to the local elite sample and the island global elite library, and when the first iteration termination condition is met, outputting the optimal fitness corresponding to the optimal sample of each island; communicating and fusing the distribution information of each island according to the pre-set multi-island information communication and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, and performing parallel iterative updates on the distribution parameters of each island after communication and fusion, and when the second iteration termination condition is met, outputting the optimal fleet dispatch scheduling plan. The use of this method can efficiently execute fleet takeoff scheduling tasks.

Description

Technical Field

[0001] The present application relates to the technical field of fleet scheduling, and in particular to a method, device, equipment and storage medium for scheduling takeoff of a fleet of aircraft. Background Art

[0002] After completing the maintenance and support operations, the fleet will taxi to each take-off position in turn to perform the departure mission. The departure mission involves the constraints of available warm-up positions and the collision avoidance safety of multiple aircraft taxiing. The sorting order directly affects the efficiency of the formation's mission execution, and it is necessary to reasonably plan the operation sequence and transfer path of each stage of the departure of multiple aircraft.

[0003] To address this issue, current research mainly focuses on the taxiing path planning from the parking position to the take-off position, but fails to consider the impact of the offshore platform environment, such as the constraints of available warm-up parking positions and the interference between the take-off position and the runway. Although there are many research results on multi-runway departure scheduling at civil aviation airports, there are many differences between this and the departure of aircraft from offshore platforms: (1) Civil aviation aircraft do not have warm-up self-inspection and can taxi directly from the parking position to the take-off runway, while some parking positions on offshore platforms, such as those in front of buildings, do not have warm-up conditions. Aircraft parked there need to be transferred to a warm-up position before they can perform the first warm-up self-inspection process before take-off; (2) Civil aviation airports have sufficient space and a clear route from the apron to the take-off runway, while the taxiing path of offshore platforms is a narrow open area, and real-time collision avoidance is required when multiple aircraft taxi; (3) Unlike the structure of multiple parallel take-off runways at civil aviation airports, there is interference between the take-off runways of offshore platforms and between the runways and take-off positions. For example, in the environment of the Kuznetsov offshore platform, the runway where the C3 take-off position is located interferes and overlaps with the runway where the C1 take-off position is located, which is a multi-type spatial resource interference problem and is more challenging than the departure scheduling problem at civil aviation airports. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, equipment and storage medium for scheduling takeoff of a fleet of aircraft in response to the above technical problems.

[0005] A method for scheduling takeoff of a fleet of aircraft, the method comprising:

[0006] Parsing the pre-set fleet dispatching task to obtain the aircraft number, parking stand number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers;

[0007] According to the aircraft number of the aircraft to be scheduled, the parking position number, the operation phase set, the multiple aircraft formations and the preset constraints, a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled is constructed with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function;

[0008] Constructing a multi-island distribution model and initializing the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model;

[0009] The cross entropy method is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library. When the termination condition of the first iteration is met, the optimal fitness corresponding to the optimal sample of each island is output; the local elite sample is the elite sample of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations;

[0010] According to the preset multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel using the cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output.

[0011] A takeoff scheduling device for a fleet of aircraft, the device comprising:

[0012] A task analysis module is used to analyze the preset fleet dispatch task to obtain the aircraft number, parking position number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers;

[0013] A dispatch scheduling model building module is used to build a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled based on the aircraft number of the aircraft to be scheduled, the parking position number, the operation stage set, the multiple aircraft formations and preset constraints, with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function;

[0014] A multi-island distribution model construction module is used to construct a multi-island distribution model and initialize the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model;

[0015] The distribution model update module is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library using the cross entropy method, and output the optimal fitness corresponding to the optimal sample of each island when the first iteration termination condition is met; the local elite samples are the elite samples of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations;

[0016] The scheduling plan output module is used to communicate and fuse the distribution information of each island according to the pre-set multi-island information communication and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, and to use the cross entropy method to parallel iteratively update the distribution parameters of each island after communication and fusion. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output.

[0017] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0018] Parsing the pre-set fleet dispatching task to obtain the aircraft number, parking stand number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers;

[0019] According to the aircraft number of the aircraft to be scheduled, the parking position number, the operation phase set, the multiple aircraft formations and the preset constraints, a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled is constructed with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function;

[0020] Constructing a multi-island distribution model and initializing the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model;

[0021] The cross entropy method is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library. When the termination condition of the first iteration is met, the optimal fitness corresponding to the optimal sample of each island is output; the local elite sample is the elite sample of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations;

[0022] According to the preset multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel using the cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0024] Parsing the pre-set fleet dispatching task to obtain the aircraft number, parking stand number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers;

[0025] According to the aircraft number of the aircraft to be scheduled, the parking position number, the operation phase set, the multiple aircraft formations and the preset constraints, a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled is constructed with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function;

[0026] Constructing a multi-island distribution model and initializing the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model;

[0027] The cross entropy method is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library. When the termination condition of the first iteration is met, the optimal fitness corresponding to the optimal sample of each island is output; the local elite sample is the elite sample of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations;

[0028] According to the preset multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel using the cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output.

[0029] The above-mentioned aircraft fleet takeoff scheduling method, device, computer equipment and storage medium analyze the pre-set fleet dispatching tasks, build a dispatching scheduling model for the dispatching priority and takeoff position allocation of the aircraft to be dispatched, and use the cross entropy optimization method based on multi-island distribution (Multi-island based Cross Entropy Optimization Method) to optimize the dispatching priority and takeoff position allocation of the aircraft to be dispatched. The multi-island distribution model is solved by using Multi-Island Entropy (MICE) to obtain the optimal fleet dispatch scheduling plan. The specific solution method is as follows: first, a multi-island distribution model is constructed and the distribution model of each island is initialized. Then, the model parameter update strategy is improved by introducing a guidance mechanism based on the global elite samples of the island, so that the optimization performance of the algorithm is more stable. Based on the improved model parameter update strategy, the multi-island distribution model is iteratively updated in parallel, so that each island gradually converges to a local optimal solution, and the optimal fitness of the optimal sample of each island is output. Finally, a multi-island information exchange and fusion mechanism is set according to the optimal fitness corresponding to each island, so that the distribution information of each island can be exchanged and fused to improve the diversity of the distribution of each island. The distribution models of each island after the exchange are iteratively updated in parallel until the termination condition of the second iteration is reached, and the optimal fleet dispatch scheduling plan is output. The optimal fleet dispatch scheduling plan is used for fleet takeoff scheduling, which can improve the efficiency of fleet takeoff and dispatch. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1This is an application scenario diagram of a method for scheduling takeoff of a fleet of aircraft in one embodiment;

[0031] Figure 2 A schematic diagram of the take-off and dispatching process of an offshore platform aircraft in one embodiment;

[0032] Figure 3 A schematic diagram of a change trend of a formation's dispatch efficiency in an embodiment;

[0033] Figure 4 A flowchart diagram of a flowchart diagram of a MICE algorithm in a specific embodiment;

[0034] Figure 5 This is a schematic diagram of the change trend of the objective function of fleet dispatch scheduling in one embodiment; wherein, Figure 5 (a) is a schematic diagram of the change trend of the fleet dispatch efficiency. Figure 5 (b) is a schematic diagram showing the trend of the time required for the completion of the sortie of the aircraft group. Figure 5 (c) is a schematic diagram of the taxiing time and change trend of the fleet;

[0035] Figure 6 A Gantt chart for scheduling the dispatching and departure operations of an optimal fleet in an embodiment;

[0036] Figure 7 A Gantt chart is used for scheduling takeoff positions in one embodiment;

[0037] Figure 8 is a structural block diagram of an aircraft fleet takeoff scheduling device in one embodiment;

[0038] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0040] In one embodiment, Figure 1 As shown, a method for scheduling takeoff of a fleet of aircraft is provided, comprising the following steps:

[0041] Step 102, parsing the preset fleet dispatching mission, obtaining the aircraft number, parking stand number, operation phase set and multiple aircraft formations of the aircraft to be dispatched.

[0042] An aircraft formation includes multiple aircraft numbers. Before the aircraft group is dispatched, the formation command post will divide the group into different formations to perform various tasks separately, and will specify the earliest ideal departure time and the latest departure time for each formation based on the progress of the overall task. Obviously, when the formation takes off later than the latest departure time, it may miss the best time to perform the flight mission. The parking stand number is obtained based on the parking stand type and the number of parking stands. The parking stand type numbers are A, B and C, which represent warm-up stands, waiting stands and take-off stands respectively.

[0043] The operation phase set includes multiple operation phases of the aircraft departure mission, such as Figure 2 The schematic diagram of the take-off and dispatch process of offshore platform aircraft is shown. After the fleet completes the maintenance and service support mission, the pilots receive the flight mission and board the aircraft, and then enter the dispatch and departure mission phase. This mission can be divided into four main operation phases:

[0044] (1) Warm-up self-inspection. According to the aircraft use requirements, the engine needs to be warmed up and self-inspected before takeoff, mainly to check whether the engine speed, temperature and other indicators meet the takeoff requirements under different working conditions. However, due to the huge tail flame heat flow formed by the nozzle during the warm-up process, not all parking spaces on the offshore platform can meet the warm-up requirements. For example, warming up the parking space in front of the building will cause ablation of the building. Therefore, during the initial deployment stage, the aircraft initially deployed in the non-warm-up parking space needs to be transferred to an empty warm-up parking space before the warm-up self-inspection operation can be carried out.

[0045] (2) Waiting behind the deflector. After the aircraft completes the warm-up self-check, it will determine the taxiing time according to the order of the assigned take-off positions: if the aircraft is the first to take off, it will ignore the waiting and taxi directly to the take-off position; if the previous aircraft has entered the take-off position, it can taxi to the waiting position behind the deflector and wait; otherwise, if the previous aircraft is still in the waiting position or there are other parking positions that interfere with the taxiing route, it is necessary to wait at the original warm-up parking position. This stage is actually a virtual process that does not involve any substantial work content (its working hours are 0). Since the take-off position corresponds to the waiting position one by one, that is, the selection of the take-off position is determined by this stage, the impact of this stage still needs to be considered.

[0046] (3) Take-off preparation. When the leading aircraft takes off and leaves the airport, and after a given cooling time interval for the deflector, the aircraft can taxi from the waiting position to the take-off position. If it is taxiing take-off mode, the rear wheels will stop in front of the stop block; if it is catapult take-off mode, the front wheels and the catapult slide block are connected through the towing rod. At the take-off position, the maintenance personnel will conduct a final inspection of the aircraft, and the pilot will also conduct a final self-inspection of the aircraft's control system, etc.

[0047] (4) Takeoff. The takeoff assistant checks the runway conditions based on the pilot and maintenance personnel’s feedback, gives a gesture indicating that the aircraft can take off, and the aircraft takes off and leaves the airport.

[0048] Step 104, based on the aircraft numbers, parking position numbers, operation phase sets, multiple aircraft formations and preset constraints of the aircraft to be scheduled, with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function, a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled is constructed.

[0049] The departure scheduling problem for aircraft (DSPA) can be expressed as arranging aircraft fleets for offshore platforms to be dispatched and take off. Under various constraints in the dispatch operation stage, the dispatch efficiency of each aircraft formation is maximized, the dispatch completion time of the aircraft fleet is minimized, and the taxiing time of the offshore platform is minimized by optimizing the allocation of warm-up positions and take-off positions of each aircraft and the operation sequence of each stage. The preset constraints include operation process sequence constraints, space resource occupation sequence constraints, waiting position and take-off position matching constraints, wake interval constraints, and take-off configuration constraints. If the parking position selection in each operation stage is regarded as processing equipment, the DSPA problem can be essentially classified as a hybrid flow shop scheduling problem containing multi-stage non-equivalent parallel machines. In addition to making decisions on the dispatch priority, the stage operation machine of the dispatch departure - the parking position will affect the subsequent operations, and the decision-making needs to expand the selection of parking positions.

[0050] The mathematical optimization model of DSPA is established with the following assumptions: (1) Once the operation at any stage and the transportation or taxiing process are started, they cannot be interrupted; (2) The influence of the wing deployment / folding state on the taxiing path is ignored, and the transfer parameters of the transfer path library are constant and are not affected by other transfer operations; (3) The interference of sudden factors such as task changes is not considered.

[0051] Step 106, construct a multi-island distribution model and initialize the distribution model of each island.

[0052] The distribution model of each island is a mutually independent truncated Gaussian distribution model.Before the multi-island distribution model is initialized, the parking positions and aircraft sequence priority decisions are integrated in a continuous coding form.

[0053] Step 108, using the cross entropy method to perform parallel iterative updates on the distribution model of each initialized island based on the local elite samples and the island global elite library. When the termination condition of the first iteration is met, the optimal fitness corresponding to the optimal sample of each island is output; the local elite sample is the elite sample of each island at each iteration, and the island global elite library is used to save the historical optimal samples of each island in previous iterations.

[0054] When updating the distribution model of each island in the multi-island distribution model, the cross entropy (CE) optimization method based on the continuous probability distribution model is introduced. The CE method is a stochastic optimization method based on adaptive importance sampling, which guides the search direction by continuously updating the parameters of the optimal probability distribution. This method has strong global optimization ability, simple structure, and fewer parameters. It has been promoted and applied in classic scheduling problems such as steelmaking-continuous casting production scheduling, vehicle path planning, resource-constrained project scheduling, and workshop scheduling. In the multi-level iteration process of the traditional CE method, the two core stages include: one is to randomly generate sample data from the sampling distribution model; the other is to update the parameters of the sampling part model based on the sample data so that the sampling model can generate better samples. The traditional CE method has a fast convergence speed when solving general optimization problems, but when applied to multi-modal complex constraint problems such as scheduling, it is easy to fall into local extreme values ​​and difficult to jump out, resulting in poor optimization stability. In this regard, the present invention has made an improved design based on the characteristics of the problem and adopted a multi-island based cross entropy optimization method (Multi-island based Cross Entropy, MICE) to improve the optimization stability.

[0055] Step 110, according to the pre-set multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel using the cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output.

[0056] The present invention adopts a multi-island distribution model for parallel optimization. In the parallel iteration, each island independently optimizes according to the iterative steps in the traditional CE algorithm. Each island may gradually converge to a local optimal solution. Therefore, after a certain number of cycles, the distribution information of each island is communicated and integrated to improve the diversity of the distribution of each island. The fleet dispatch scheduling plan includes a dispatch timing plan and a take-off position allocation plan.

[0057] In the above-mentioned aircraft fleet takeoff scheduling method, by analyzing the pre-set fleet dispatching tasks, a dispatching scheduling model for the dispatching priority and takeoff slot allocation of the aircraft to be dispatched is constructed, and the multi-island based cross entropy optimization method (Multi-island based Cross Entropy Optimization Method) is used to optimize the dispatching priority and takeoff slot allocation of the aircraft to be dispatched. The multi-island distribution model is solved by using the Multi-Island Entropy (MICE) algorithm to obtain the optimal fleet dispatch scheduling plan. The specific solution method is as follows: first, a multi-island distribution model is constructed and the distribution model of each island is initialized. Then, the model parameter update strategy is improved by introducing a guidance mechanism based on the global elite samples of the islands, so that the optimization performance of the algorithm is more balanced. Based on the improved model parameter update strategy, the multi-island distribution model is iteratively updated in parallel, so that each island gradually converges to a local optimal solution, and the optimal fitness of the optimal sample of each island is output. Finally, a multi-island information exchange and fusion mechanism is set according to the optimal fitness corresponding to each island, so that the distribution information of each island can be exchanged and fused to improve the diversity of the distribution of each island. The distribution models of each island after the exchange are iteratively updated in parallel until the termination condition of the second iteration is reached, and the optimal fleet dispatch scheduling plan is output. The optimal fleet dispatch scheduling plan is used for fleet takeoff scheduling, which can improve the efficiency of fleet takeoff and dispatch.

[0058] According to the requirements of the fleet sortie departure mission, three objective functions are constructed: fleet sortie efficiency, fleet sortie completion time, and fleet taxiing time. The three objective functions are in a lexicographic hierarchical relationship, that is, the priority is to maximize the fleet sortie efficiency, followed by minimizing the fleet sortie completion time, and finally minimizing the fleet taxiing time.

[0059] Maximize the fleet dispatch efficiency: According to the fleet dispatch efficiency and the fleet task weight of each aircraft formation at the time of dispatch, the maximum fleet dispatch efficiency is obtained as follows:

[0060]

[0061]

[0062] Among them, DE is the fleet dispatch efficiency, Z ib is the formation to which the i-th aircraft belongs b Ew ij is the decision variable, Ew i4 Indicates the end time of the fourth operation phase of machine i, EM b is the latest dispatch time of formation b, EA b The ideal time for the B formation to be deployed is b is the task weight coefficient of formation b, t b is the take-off time, b∈B, B is the aircraft formation set, μ b The formation dispatching efficiency of b formation, β bis the efficiency decreasing curvature control parameter. Figure 3 The schematic diagram of the change trend of the formation's dispatch effectiveness is shown in the figure. Figure 3 Indicates the formation parameters EA b =10min,EM b =40min and different β b The change trend of the formation dispatch effectiveness under the value conditions, β b The larger the value, the closer it is to linear decrease; on the contrary, b The smaller the value, the faster the performance decreases. b is the task weight coefficient of formation b, and ∑ b∈B we b =1, we b The larger the value, the higher the task weight. The formation takeoff priority can be determined by the formation mission weight.

[0063] Minimize the time to complete the fleet deployment: Considering that the fleet operation is carried out in a cycle according to a certain offshore platform cycle, due to the limited space on the offshore platform, the aircraft recovered in the next wave can only start approaching and landing after the current wave of aircraft is deployed. Therefore, the shorter the fleet deployment completion time, the sooner the next wave of aircraft can be recovered to perform the subsequent wave deployment support mission. The minimum fleet deployment completion time is expressed as:

[0064]

[0065] Minimize the taxiing time of the fleet: Since the aircraft taxis under the action of its own engine, its tail flame poses a certain threat to the support personnel and equipment of the offshore platform. Taxiing operation is an important source of danger for safety accidents. Therefore, considering the principle of dispatching nearby, the shorter the taxiing time of the fleet, the better. The minimized taxiing time of the fleet is expressed as;

[0066]

[0067] Among them, P ij is the parking position of the i-th aircraft before the j-th stage transfer, Indicates Sz ij The time is from parking position P ij Transfer to parking position P i(j+1) The transfer time is j = 1, and the transfer process is the taxiing process of the aircraft. For Sz ij The aircraft is at parking position P ij Transfer to parking position P i(j+1) The transfer path corresponds to the layout state of the interference parking position, Sz ij For the transfer process O ij The start time of the operation,

[0068]

[0069]

[0070] The constraints of the dispatch scheduling model include:

[0071] Operation process timing constraints: According to the transfer operation timing relationship, the support operation timing relationship, the connection relationship between the transfer operation and the support operation, and the deflector cooling reset time interval constraint of the current operation phase of the aircraft to be scheduled, the operation process timing constraints are obtained; the operation phase includes transfer operation and support operation; the transfer operation timing relationship is:

[0072]

[0073] Where I is the set of aircraft to be taken off, I = {1, 2, ..., Nf}, j∈J, J is the set of operation stages, J = {0, 1, 2, 3, 4}, 0 represents the initial deployment stage, 1 represents the warm-up self-check stage, 2 represents the waiting stage behind the deflector, 3 represents the take-off preparation stage, 4 represents the take-off dispatch stage, Ez ij is a decision variable, representing the end time Sz of the transfer of the jth operation stage of the i machine ij is a decision variable, which indicates the start time of the transfer of the jth operation stage of the i-th machine, ΔTz ij is the transit time of the aircraft to be taken off; ΔTz ij According to the parking space selection in the previous and next two stages:

[0074]

[0075] Transfer includes taxiing and mobilization. The difference between taxiing and mobilization lies in whether there is a traction vehicle towing. The transfer between the parking position and the take-off preparation position is taxiing, while the transfer between parking positions for warm-up operations that may be involved in the dispatch process is mobilization. The transfer time includes taxiing / mobilization time and untying / mooring time.

[0076] The timing relationship of the guarantee operation is:

[0077]

[0078] Among them, Ew ij is a decision variable, indicating the end time of the jth operation phase of machine i, Sw ij is a decision variable, indicating the start time of the jth operation phase of machine i, dw j To ensure working hours for stage j;

[0079] The connection between transshipment operations and support operations is:

[0080]

[0081]

[0082] The deflector plate cooling reset time interval constraint is:

[0083]

[0084] Yp ijejp is a decision variable, which indicates the priority of aircraft in different operation stages occupying the same parking position, Δtb is the cooling reset time of the deflector plate, P is the parking position set, including warm-up position, waiting position and take-off position, P j The parking space set of the jth operation phase.

[0085] Space resource occupation time sequence constraint: According to the multi-machine transportation collision avoidance constraint and the limited time sequence constraint of occupying the same space, the space resource occupation time sequence constraint is obtained;

[0086] The collision avoidance constraints for multi-aircraft transfer are as follows: Based on the transfer path library and the initial and target parking positions of the aircraft to be transferred, the feasible constraints for the transfer path are as follows:

[0087]

[0088] Among them, Road t (P ij ,P i(j+1) )for The abbreviation of represents the transfer path library index; according to the preset aircraft posture model, the coordinates of the vertices of the aircraft convex hull of the aircraft to be transferred under the reference point posture are solved to obtain the space area occupied by the aircraft convex hull. According to the space area occupied by the aircraft convex hull, the configuration of the aircraft to be transferred is obtained. According to the configurations of multiple aircraft to be transferred at the same time on the corresponding transfer path, the aircraft transfer collision avoidance constraint is obtained as follows:

[0089]

[0090]

[0091] in, is the configuration of the i-th aircraft.

[0092] Occupies the same space with limited timing constraints:

[0093]

[0094] Among them, Yr ijegr is a decision variable, indicating the priority of aircraft in different operation stages occupying the same take-off runway, g∈J, g≠j, R is the take-off runway set, and BM is the penalty term.

[0095] Matching constraints between waiting positions and take-off positions: According to the one-to-one correspondence between waiting positions and take-off positions on the same take-off runway, the matching constraints between waiting positions and take-off positions are obtained as follows:

[0096]

[0097] Among them, P ij is a decision variable, indicating the parking space occupied by the i-th machine in the j-th operation stage, The p waiting position corresponds to the take-off position of the next stage, p∈Ψ2,γ p ∈Ψ3,Ψ j is the set of available parking spaces for the jth operation phase.

[0098] Wake turbulence interval constraint: According to the wake turbulence time interval standard between consecutive takeoff aircraft, the wake turbulence interval constraint is obtained as follows:

[0099]

[0100] Among them, x ie is a decision variable, indicating the takeoff priority of the aircraft, δ i is the wake turbulence interval time after aircraft i takes off.

[0101] Takeoff configuration constraint: According to the takeoff configuration corresponding weight of the aircraft to be taken off and the takeoff weight upper limit of the takeoff position, the takeoff configuration constraint is obtained as follows:

[0102]

[0103] Among them, m i is the weight corresponding to the takeoff configuration of aircraft i, fm p is the upper limit of takeoff weight for takeoff position p;

[0104] Decision variable Yp ijegp , Yr ijegr 、x ie They are all Boolean 0-1 decision variables, namely:

[0105]

[0106]

[0107]

[0108]

[0109] The constraints involved in the process of the offshore platform fleet’s departure mainly include the following aspects: (1) Constraints on available warm-up parking spaces. As analyzed in the above process, the aircraft parked in front of the offshore platform building need to be transferred to other available warm-up parking spaces for warm-up operations. If all available warm-up parking spaces are occupied, they need to wait for an aircraft to finish warming up and taxi out before they can be transferred to the space. The constraints on available warm-up parking spaces are already included in the available parking space set Ψ j in the definition; (2) multi-aircraft transfer collision avoidance constraints; (3) parking stand layout blocking the taxiway, that is, placing aircraft on some parking stands will prevent other aircraft from taxiing smoothly into the take-off position; (4) parking stand / take-off stand layout and transfer process interference constraints on the take-off runway, which can be called runway intrusion in civil aviation surface management; (5) take-off wake interval constraints. Since the take-off runways of the offshore platform are too close, even if the aircraft take off from different take-off runways at different take-off positions, the wake vortices they generate will have a certain impact on the take-off of subsequent aircraft. Therefore, a certain wake interval time must be reserved between the aircraft launched from the offshore platform; (6) operation process constraints, that is, each aircraft must perform the departure operation in the same stage sequence; (7) take-off configuration constraints, mainly for ski-jump take-off, that is, the aircraft's mission configuration corresponds to the bomb load and fuel load, which requires it to take off at a matching take-off position. If the take-off runway is too short, it is impossible to ensure the safe launch of the aircraft with a large load configuration. Among the above constraints, constraints (2) to (4) can be classified as space resource type constraints. In addition, since the wings can be unfolded during the warm-up parking stage, or can be unfolded after reaching the waiting position or take-off position, the impact of the wingspan state on the taxiing path can be ignored. According to actual usage experience, the wings are generally unfolded in the take-off area.

[0110] In one embodiment, before randomly generating multiple individuals within the definition interval of each distribution model, it includes: using a continuous coding rule to encode the take-off position selected by each aircraft to be scheduled and the dispatch priority of each aircraft to be scheduled to obtain an individual corresponding to each aircraft to be scheduled; the integer part of the gene of any dimension in the individual represents the take-off position selected by each aircraft to be scheduled, and the decimal part of the gene represents the dispatch priority of each aircraft to be scheduled; the continuous coding rule also includes: for any dimensional gene in the individual, when the integer part of the gene is an odd number, the smaller the value of the decimal part of the gene, the higher the dispatch priority of the aircraft to be scheduled corresponding to the decimal part; when the integer part of the gene is an even number, the larger the value of the decimal part of the gene, the higher the dispatch priority of the aircraft to be scheduled corresponding to the decimal part.

[0111] In this embodiment, the encoded data is represented by x=[x i ] 1×Nf , where any dimension gene x i For the interval (1, NP c+1), whose integer part Int(x i ) represents the take-off position selected by the i-th aircraft, and the decimal part Dec(x i ) represents the take-off priority of the i-th aircraft. However, if Dec(x i ) is smaller, the higher the priority. When the value jumps from a corresponding takeoff position to another takeoff position, a priority step will occur. For example, let x i From 2.1 to 1.9, its takeoff position changes from C2 to C1, but its priority drops from the higher priority of C2 to the lower priority of C1. This step situation is obviously not conducive to the sampling and evaluation of continuous Gaussian distribution in cross entropy optimization. In view of this, the encoding form of this embodiment specifies any dimensional gene when its integer part Int(x i ) is an odd number, the decimal part Dec(x i ) is smaller, the higher the priority; and when Int(x i ) is an even number, the decimal part Dec(x i ) is larger, the higher the priority. Based on this priority form, the priorities of the head and tail of the aircraft in different take-off positions can be connected in series, so that gene x i In the interval (1, NP c +1) is a continuous and gradual change in priority to avoid sudden changes caused by changes in takeoff position selection. And after each scheduling decoding, the value of the decimal part will be corrected according to the ratio of the takeoff time to the entire operation stage, that is,

[0112]

[0113] Among them, ε is a very small real number to ensure that Dec(x i )∈(0,1),Int(x i ) represents the i Take an integer, mod(Int(x i ),2) means Int(x i ) divided by 2, (Int(x i ),2)=1 means Int(x i ) is an odd number, mod(Int(x i ),2)=0 means Int(x i ) is an even number.

[0114] The decoding process is the process of converting the priority arrangement of aircraft at each takeoff position into a specific dispatch scheduling plan. In order to build a decoding mechanism for dispatch scheduling, the scheduling process variables include the set to be scheduled, the set to be scheduled, and the record of completed stages. The set to be scheduled represents the set of aircraft waiting for transfer scheduling sorted by priority, and the set to be scheduled represents the set of aircraft that have been scheduled for takeoff and departure operations. The SL in the completed stage record of each aircraft i The value is the number of operation stages completed by aircraft i. The aircraft list includes aircraft numbers sorted by dispatch priority. The steps of obtaining the aircraft list include: performing isotropic processing on the decimal part of the coded data, specifically: setting the priority weight of aircraft i at an even-numbered takeoff position to 1-Dec(x i ), the priority weights of the aircraft at odd take-off positions remain unchanged, and then the aircraft numbers are sorted from small to large according to the priority weights to obtain the aircraft list.

[0115] The steps of dispatch scheduling decoding include:

[0116] S10: Initialize W g =π, SL=[0] 1×Nf ,index=1,where W g is the set to be scheduled, π is the aircraft list, S g It is the scheduled set, and SL is the completed stage record.

[0117] S20: When index>|W g |, enter S60.

[0118] S30: Select the aircraft number i to be dispatched * =W g (index) and its corresponding take-off position c * =Int(x i ), select the next stage of the aircraft to be scheduled according to the completed stage record, Switch (SL i +1);

[0119] Case 1: When the next stage of the aircraft to be dispatched is the warm-up self-test stage, determine i * Check whether the current parking position of the aircraft is a parking position that can be warmed up. If so, warm up the aircraft directly in place and Go to Case 2; otherwise, further determine whether there is an available warm-up parking position at the start time. If so, select from the available warm-up positions from the current parking position to the warm-up position and from the warm-up position to the take-off position c * The transfer time and the minimum corresponding warm-up position are executed, and the transfer and warm-up self-test is performed, and the Go to Case 2; otherwise, further determine whether the scheduled set is not empty. If so, select the earliest transfer time and the corresponding warm-up position from the scheduled aircraft, perform transfer and warm-up self-check after it taxis out, and set Go to Case 2. Otherwise, the machine cannot be warmed up at present. Set index = index + 1 and go to S20.

[0120] Case 2: When the next stage of the aircraft to be dispatched is the taxiing and waiting stage behind the deflector, first query the warm-up position To take-off position c * Interference parking position of the path library Find Is there any aircraft stopping, or have all the aircraft stopping been dispatched? If there is a stop but it is not dispatched, the taxiing condition is not met at present, so let index = index + 1 and go to S20; otherwise, find the earliest available time t1 of the path, and further query the taxiing path intrusion runway set in the same way. And find the earliest clearing time t2 of the runway from time t1 as the earliest taxiing start time, and use the delayed transfer collision avoidance mechanism to delay the collision-free taxiing start time t3. * Is the machine c * The take-off position has the highest priority. If it is, it will taxi directly to the take-off position and the waiting phase time is 0. Otherwise, it will taxi to the waiting position and set Proceed to Case 3.

[0121] Case 3: When the next stage of the aircraft to be dispatched is the take-off position preparation stage, query the take-off position c * The last aircraft launch time plus the deflector reset cooling time Δtb is used as the entry time, and the taxiing entry and take-off preparation operations are performed, and Proceed to Case 4.

[0122] Case 4: When the next stage of the aircraft to be dispatched is the take-off stage, query the previous aircraft release time plus the wake interval time as the earliest take-off time t1, and further search for the parking space clearance time of the interfering runway and the transfer time to leave the runway as the final take-off time t2, perform the take-off and departure operations, and let W g =W g -{i *},S g =S g ∪{i *}, set index = 1, and go to S20 to continue iterative scheduling.

[0123] S40: Complete the departure scheduling, output the departure timing plan and take-off slot allocation plan, and calculate the objective function value.

[0124] S50: The warm-up process is shifted to the right. After the dispatching sequence is allocated, considering the principle of starting as early as possible adopted by the dispatcher, that is, the work can be started as soon as the operating conditions are met, many aircraft will be warmed up at the beginning. However, some aircraft that taxi and dispatch later will maintain a long engine standby time after warming up, resulting in unnecessary fuel consumption. In this regard, the warm-up operation process is shifted to the right to the taxiing start time, that is, the aircraft can taxi out directly after warming up, avoiding unnecessary standby time after warming up.

[0125] S60: To facilitate the algorithm to compare individual fitness, the lexicographic order multi-objectives are combined into a single objective by assigning weights, and the individual fitness is defined as:

[0126]

[0127] In the formula, is the weight coefficient of the dispatch completion time, It is the weight coefficient of aircraft taxiing time. Commander can adjust the weight according to the deployment requirements.

[0128] In this embodiment, in Case 1, the earliest time of transfer out of the position is the earliest time of transfer out of the position by means of a tractor when transferring to other parking positions for warming up. In S50, after the warming up is completed, the machine slides to the deflector plate and waits, and the sliding start time is the time when the sliding starts in the sliding and waiting phases behind the deflector plate. If the machine needs to be transferred to another parking position for warming up, the transfer to the parking position behind the deflector plate after the warming up is called sliding.

[0129] The basic steps of the CE method based on the traditional Gaussian distribution sampling model include:

[0130] S1: Initialize Gaussian distribution parameters v0 = (μ0, σ0), where μ0 and σ0 are the mean and standard deviation of the initial distribution respectively, and set the number of iterations to k = 1.

[0131] S2: According to the current sampling distribution model Generate a set of N random samples {x 1,k ,x 2,k ,…,x N,k}, calculate the objective function value of the sample {f(x i,k )}, and sort them in ascending order according to the target value to obtain {f i}, find the ρ quantile b of the target value order statistic k =f [ρN] , and record the first ρN elite samples and the best sample

[0132] S3: Count the distribution parameters of the current elite samples respectively:

[0133]

[0134] S4: Calculate the new generation distribution parameters according to the smooth update rule, the smooth update rule is: Among them, 0≤α≤1 is the smoothing parameter.

[0135] S5: If the iteration termination condition is reached, the algorithm stops and outputs the optimal solution And the corresponding optimal index Otherwise, set k=k+1 and go to S2.

[0136] The traditional CE method has a fast convergence speed when solving general optimization problems, but when applied to multi-modal complex constraint problems such as scheduling, it is easy to fall into local extreme values ​​and difficult to jump out, resulting in weak optimization stability. In this regard, the present invention has made an improved design based on the characteristics of the problem and proposed a multi-island based cross entropy optimization method (Multi-island based Cross Entropy, MICE).

[0137] In one embodiment, the step of initializing the distribution model of each island includes: randomly generating multiple individuals within the defined interval of each distribution model, decoding the coded data corresponding to each individual using the preset dispatch scheduling decoding rules, and obtaining the solutions corresponding to the multiple individuals; using the K-means algorithm to divide the solutions corresponding to the multiple individuals into each island as samples, and initializing the distribution model of each island according to the samples on each island. In this embodiment, a multi-island distribution model structure is adopted, and the number of islands is N. island , the number of sampled populations on each island is Np, and each island is an independent truncated Gaussian distribution model, that is, the distribution of the sth island at the kth iteration can be expressed as The k-means clustering method is used to initialize the model parameters: First, randomly generate N in the defined interval island ×Np individuals, after decoding and correcting the individual codes, the solution set is divided into N island islands, the samples in each island are N s is the number of samples clustered into the sth island, then the distribution model parameters of any sth island can be initialized as:

[0138]

[0139] in, and represent the initial mean and initial standard deviation of the sth island distribution model respectively.

[0140] In one embodiment, the steps of parallel iteratively updating the distribution model of each island after initialization according to the local elite samples and the island global elite library using the cross entropy method include: randomly sampling according to the distribution model of the current iteration number of each island to obtain multiple individuals, using the dispatch scheduling decoding mechanism to calculate the solution of each individual to obtain multiple samples, and obtaining the elite sample and the optimal sample of each island according to the quantile of the order statistics of the multiple samples; selecting the elite sample of each island as the local elite sample to construct the island global elite library; the island global elite library includes the global elite sample; the global elite sample includes the historical optimal sample of each island in all iterations; according to the parameters of the distribution model constructed by the global elite sample of each island and the parameters of the distribution model constructed by the local sampling elite sample, parallel iteratively updating the distribution parameters of the distribution model of each island.

[0141] In this embodiment, in order to solve the problem that the number of elite samples sampled each time is too small and the optimization performance fluctuates, a guidance mechanism based on the island global elite sample is introduced. The ρNp elite samples selected in each iteration are defined as the sampled elite samples, which can also be regarded as local elite samples. A ρNp island global elite library is set to store the historical optimal samples obtained from previous searches, and the samples are different from each other. Then the parameters of the distribution model constructed by the sth island global elite sample can be expressed as Similarly, the parameters of the distribution model constructed by the local sampling elite sample are The calculation of parameters is based on the formula used to initialize the distribution model parameters. Based on the concepts of local sampling elite samples and island global elite samples, the improved smoothing update strategy of distribution parameters is:

[0142]

[0143] Where α1 and α2 are local smoothing parameters and global smoothing parameters respectively. Obviously, in the initial stage, Through this smooth update strategy, the historical optimal information is combined with the current sampling information to achieve a balance between algorithm exploration and development.

[0144] In one embodiment, according to a preset multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel and iteratively by using a cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output, including: randomly sampling and generating multiple individuals on each updated island, in any dimension of each individual, when the parameter random number is less than or equal to the preset crossover probability, roulette is played according to the optimal fitness of the remaining islands, and a communication island is randomly selected, and sampling is performed according to the distribution model of the communication island in the corresponding dimension to exchange and fuse the distribution information of the current island and the communication island; when the parameter random number is greater than the preset crossover probability, sampling is performed according to the distribution model of each island; according to the multiple new individuals after sampling, the distribution parameters of each island after exchange and fusion are updated in parallel and iteratively by using a cross entropy method, and when the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output; the optimal fleet dispatch scheduling plan includes an optimal dispatch timing plan and an optimal take-off position allocation plan.

[0145] In this embodiment, after a certain number of cycles ΔGs, the distribution information of each island is exchanged and integrated, and the optimal fitness of the sth island is The crossover probability is P c First, for any island s, Np individuals are randomly sampled, where each individual Sampling by dimension: For any dimension d, if the parameter random number rand(0,1)>P c , then according to the distribution model of island s Otherwise, the optimal fitness of the remaining islands is used to select Play roulette and randomly select another island s′, and follow the distribution model of the island in dimension d Secondly, for any island s, select ρNp elite samples and update the distribution model parameters according to the smooth update rule. The smooth update rule is:

[0146]

[0147] Among them, 0≤α≤1 is the smoothing parameter, In this embodiment, the parameter P c Control the AC fusion ratio, P c The larger the value, the greater the frequency of crossover among various populations and the stronger the degree of integration. At the same time, roulette-based island selection can ensure that islands with better optimization results can participate in more integration.

[0148] In a specific embodiment, Figure 4As shown in the figure, a flowchart of the MICE algorithm is provided. First, the constructed multi-island distribution model is initialized by the K-means algorithm. Second, the CE method based on the Gaussian distribution sampling model is used to perform parallel iterative updates on each island, wherein the update steps of each island distribution model include sampling populations based on the distribution model of the island, performing sortie scheduling decoding, updating global elite samples and local elite samples and their distribution parameters, and updating the sampling distribution model parameters using an improved smooth update strategy. When the condition mod(k,ΔGs)=0 is met, the distribution model of each island is updated again based on the multi-island communication fusion mechanism until the termination condition is met, and the optimal fleet sortie scheduling plan is output. mod(k,ΔGs)=0 means that an update is performed when k is an integer multiple of ΔGs, that is, an update is performed once the interval reaches ΔGs.

[0149] In one embodiment, Figure 5 As shown in Figure 1, a schematic diagram of the change trend of the objective function of fleet dispatch scheduling is provided, where: Figure 5 (a) is a schematic diagram of the change trend of the fleet dispatch efficiency. Figure 5 (b) is a schematic diagram showing the trend of the time required for the completion of the sortie of the aircraft group. Figure 5 (c) is a schematic diagram of the taxiing time and change trend of the fleet. The simulation parameters include: 16 aircraft are dispatched, and the offshore platform adopts the ski-jump takeoff mode. The 16 aircraft are arranged in parking positions A2~A13, B1~B4, and are numbered from 1 to 16 in the order of arrangement. Except for A3~A6, all of them are parking positions for warm-up. The take-off positions are C1~C3, and each take-off position corresponds to a waiting position and 1#~3# take-off runways. 1# and 3# take-off runways overlap, and the C1 take-off position and waiting position both interfere with the 3# take-off runway. The A12 parking position interferes with the C3 waiting position, which will affect the taxiing of C3. The fleet is divided into four formations according to the mission. The first formation is [1 6 11 12]; the second formation is [7 8 9 10]; the third formation is [13 14 4 5]; and the fourth formation is [2 315 16]. The task weights of the four formations are we = [0.3, 0.25, 0.25, 0.2], the ideal departure time is [10, 15, 20, 25] (min), and the latest departure time is [15, 20, 30, 30] (min). It is assumed that all aircraft configurations meet the constraints of each take-off position. The remaining operation time parameters include: the operation time of the four stages is dw = [180, 0, 60, 10] (s), the deflector cooling reset time Δtb = 30s, the wake interval time δ i = 10s, the tethering time is 30s. The weight of the scheduling target is The MICE algorithm parameters are set as follows: The number of islands is N island=4, the number of sampled populations on each island is Np=50, the elite sample quantile ρ=0.1, the smoothing parameters α1=0.5, α2=0.3, and the crossover probability P c =0.6, multi-island fusion interval generation ΔGs=30. The algorithm iteration termination condition takes the number of loop iterations as 200. After simulation optimization, the change trend of the objective function of the fleet dispatch scheduling is obtained, among which the dispatch efficiency DE converges to 1 after about 100 generations, that is, the dispatch of each formation meets the ideal departure time requirement. Figure 5 In the equation, evolution generation means evolutionary generation. Figure 5 In (b), makespan represents the time it takes for the fleet to be dispatched. When DE is optimal, the dispatch completion time C max It also continues to converge to 21 minutes. If the time from the first aircraft entering the takeoff position to the completion of the sortie is 4.21 minutes, the fleet will complete the sortie at an average frequency of 1.05 minutes per aircraft. The overall change of taxiing time is not large, and finally converges to 30.81 minutes, with an average taxiing time of 1.93 minutes per aircraft.

[0150] In one embodiment, Figure 6 As shown in the figure, a Gantt chart for optimal fleet departure operation scheduling is provided. Figure 7 As shown, a take-off position scheduling Gantt chart is provided. Figure 6 In the figure, the ordinate is the corresponding number of each aircraft, the abscissa is the dispatch timeline, and the "Jj" in the Gantt box represents the j-th stage support operation. The stages with too short waiting time are not marked in the figure. The gray Gantt box represents the transfer process, and the subscripts include the starting parking position and the target parking position of the transfer. Figure 7 In the figure, the vertical axis represents the number of each take-off position, and the "ij" in the Gantt box represents the jth stage operation of the i-th aircraft, which includes the take-off position preparation operation and the take-off sortie operation. Combining the two scheduling Gantt charts, it can be verified that the sortie operation sequence of the fleet can meet the various constraints of the scheduling model, and after the warm-up process is shifted to the right, it can ensure that the warm-up operation is more dispersed in time sequence, avoiding unnecessary startup waiting; Figure 7 In the above figure, the number of aircrafts that choose to take off from the three take-off positions is 3, 6 and 7 respectively. The reason why C1 is less selected is mainly because this take-off position interferes with the take-off operation of C3, and the distance from the aircraft arranged at the stern is farther than that of C3, so C3 becomes a better choice, and the C3 take-off position can only be used after the aircraft on the A12 parking position taxis out of the position. The scheduling results can verify the effectiveness of the method of the present invention.

[0151] It should be understood that although Figure 1 , Figure 2 and Figure 4The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 , Figure 2 and Figure 4 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0152] In one embodiment, Figure 8 As shown, a takeoff scheduling device for a fleet of aircraft is provided, comprising: a task parsing module 802, a dispatch scheduling model building module 804, a multi-island distribution model building module 806, a distribution model updating module 808 and a scheduling solution output module 810, wherein:

[0153] The task analysis module 802 is used to analyze the preset fleet dispatching task to obtain the aircraft number, parking position number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers;

[0154] The dispatch scheduling model building module 804 is used to build a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled based on the aircraft number, parking position number, operation stage set, multiple aircraft formations and preset constraints of the aircraft to be scheduled, with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxi time as the objective function;

[0155] The multi-island distribution model building module 806 is used to build a multi-island distribution model and initialize the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model;

[0156] The distribution model updating module 808 is used to perform parallel iterative updates on the distribution model of each island after the initialization according to the local elite samples and the island global elite library using the cross entropy method, and output the optimal fitness corresponding to the optimal sample of each island when the first iteration termination condition is met; the local elite samples are the elite samples of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in all iterations;

[0157] The scheduling scheme output module 810 is used to communicate and fuse the distribution information of each island according to the pre-set multi-island information communication and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, and to use the cross entropy method to parallel iteratively update the distribution parameters of each island after communication and fusion. When the second iteration termination condition is met, the optimal fleet dispatch scheduling scheme is output.

[0158] In one of the embodiments, the multi-island distribution model construction module 806 is also used to randomly generate multiple individuals within the defined interval of each distribution model, and use a preset dispatch scheduling decoding rule to decode the encoded data corresponding to each individual to obtain solutions corresponding to the multiple individuals; use the K-means algorithm to divide the solutions corresponding to the multiple individuals as samples into each island, and initialize the distribution model of each island according to the samples on each island.

[0159] In one of the embodiments, the multi-island distribution model construction module 806 is also used to use a continuous coding rule to encode the take-off position selected by each aircraft to be scheduled and the dispatch priority of each aircraft to be scheduled to obtain an individual corresponding to each aircraft to be scheduled; the integer part of the gene of any dimension in the individual represents the take-off position selected by each aircraft to be scheduled, and the decimal part of the gene represents the dispatch priority of each aircraft to be scheduled; the continuous coding rule is also used for any dimensional gene in the individual, when the integer part of the gene is an odd number, the smaller the decimal part value of the gene, the higher the dispatch priority of the aircraft to be scheduled corresponding to the decimal part; when the integer part of the gene is an even number, the larger the decimal part value of the gene, the higher the dispatch priority of the aircraft to be scheduled corresponding to the decimal part.

[0160] In one of the embodiments, the distribution model update module 808 is also used to randomly sample the distribution model of the current iteration number of each island to obtain multiple individuals, use the dispatch scheduling decoding mechanism to calculate the solution of each individual to obtain multiple samples, and obtain the elite sample and the optimal sample of each island according to the quantile of the order statistics of the multiple samples; select the elite sample of each island as the local elite sample to construct the island global elite library; the island global elite library includes the global elite sample; the global elite sample includes the historical optimal sample of each island in all iterations; according to the parameters of the distribution model constructed by the global elite sample of each island and the parameters of the distribution model constructed by the local sampling elite sample, the distribution parameters of the distribution model of each island are updated in parallel and iteratively.

[0161] In one of the embodiments, the scheduling scheme output module 810 is also used to randomly sample and generate multiple individuals on each updated island. In any dimension of each individual, when the parameter random number is less than or equal to the preset crossover probability, a roulette wheel is played according to the optimal fitness of the remaining islands, and a communication island is randomly selected. Sampling is performed according to the distribution model of the communication island in the corresponding dimension to communicate and fuse the distribution information of the current island with the communication island; when the parameter random number is greater than the preset crossover probability, sampling is performed according to the distribution model of each island; based on the multiple new individuals after sampling, the distribution parameters of each island after communication and fusion are updated in parallel using the cross entropy method, and when the second iteration termination condition is met, the optimal fleet dispatch scheduling scheme is output; the optimal fleet dispatch scheduling scheme includes the optimal dispatch timing scheme and the optimal take-off position allocation scheme.

[0162] In one embodiment, the dispatch scheduling model building module 804 is further used to obtain the operation process timing constraint according to the transfer operation timing relationship, the support operation timing relationship, the connection relationship between the transfer operation and the support operation, and the deflector cooling reset time interval constraint of the operation phase of the current aircraft to be scheduled; the operation phase includes the transfer operation and the support operation; the transfer operation timing relationship is:

[0163]

[0164] Where I is the set of aircraft to be taken off, I = {1, 2, ..., Nf}, j∈J, J is the set of operation stages, J = {0, 1, 2, 3, 4}, 0 represents the initial deployment stage, 1 represents the warm-up self-check stage, 2 represents the waiting stage behind the deflector, 3 represents the take-off preparation stage, 4 represents the take-off dispatch stage, Ez ij is a decision variable, representing the end time Sz of the transfer of the jth operation stage of the i machine ij is a decision variable, which indicates the start time of the transfer of the jth operation stage of the i-th machine, ΔTz ij is the transfer time of the aircraft to be taken off; the timing relationship of the guarantee operation is:

[0165]

[0166] Among them, Ew ij is a decision variable, indicating the end time of the jth operation phase of machine i, Sw ij is a decision variable, indicating the start time of the jth operation phase of machine i, dw j The working hours for the jth stage guarantee operation; the connection relationship between the transfer operation and the guarantee operation is:

[0167]

[0168]

[0169] The deflector plate cooling reset time interval constraint is:

[0170]

[0171] Yp ijejp is a decision variable, which indicates the priority of aircraft in different operation stages occupying the same parking position, Δtb is the cooling reset time of the deflector plate, P is the parking position set, including warm-up position, waiting position and take-off position, P j The parking space set of the jth operation stage; according to the multi-machine transfer collision avoidance constraint and the limited time sequence constraint of occupying the same space, the space resource occupation time sequence constraint is obtained; the limited time sequence constraint of occupying the same space is:

[0172]

[0173] Among them, Yr ijegr is a decision variable, which indicates the priority of aircraft in different operation stages occupying the same takeoff runway, g∈J, g≠j, R is the takeoff runway set, and BM is the penalty term. According to the one-to-one correspondence between the waiting position and the takeoff position on the same takeoff runway, the matching constraint between the waiting position and the takeoff position is obtained as follows:

[0174]

[0175] Among them, P ij is a decision variable, indicating the parking space occupied by the i-th machine in the j-th operation stage, The p waiting position corresponds to the take-off position of the next stage, p∈Ψ2,γ p ∈Ψ3, Ψ is the set of available parking spaces for the jth operation phase; according to the wake turbulence time interval standard between consecutive takeoff aircraft, the wake turbulence interval constraint is obtained as:

[0176]

[0177] Among them, x ie is a decision variable, indicating the takeoff priority of the aircraft, δ i is the wake interval time after aircraft i takes off; according to the takeoff configuration corresponding weight of the aircraft to be taken off and the takeoff weight upper limit of the takeoff position, the takeoff configuration constraint is obtained as follows:

[0178]

[0179] Among them, m i is the weight corresponding to the takeoff configuration of aircraft i, fm p is the upper limit of the takeoff weight of takeoff position p; according to the operation process timing constraints, space resource occupation timing constraints, waiting position and takeoff position matching constraints, wake interval constraints and takeoff configuration constraints, the constraints are obtained.

[0180] In one embodiment, the multi-island distribution model building module 806 is further used to initialize the to-be-scheduled set as a to-be-scheduled aircraft list, and initialize the scheduled set and the completed stage record; the to-be-scheduled aircraft list includes the aircraft numbers of the to-be-scheduled aircraft to be sorted by priority weights; the completed stage record is a matrix that records the completed stages of each to-be-scheduled aircraft; the aircraft numbers of the to-be-scheduled aircraft in the to-be-scheduled set are indexed, the take-off position corresponding to the to-be-scheduled aircraft is searched, and the next stage of the to-be-scheduled aircraft is selected according to the completed stage record. When the to-be-scheduled aircraft completes the take-off and dispatching stage, the completed stage record, the to-be-scheduled set and the scheduled set corresponding to the to-be-scheduled aircraft are updated; the index value is initialized, and the index value is updated according to the initialization. The index value of the index updates the aircraft number of the aircraft to be scheduled in the set to be scheduled, iterates to find the take-off position corresponding to the aircraft to be scheduled, selects the next stage of the aircraft to be scheduled according to the completed stage records, and when the preset iteration termination condition is met, outputs the dispatch timing plan and the take-off position allocation plan, and calculates the dispatch timing plan and the take-off position allocation plan to calculate the objective function value of the dispatch scheduling model, and obtains the individual fitness according to the objective function value; the next stage of selecting the aircraft to be scheduled according to the completed stage records includes: when the next stage of the aircraft to be scheduled is the warm-up self-inspection stage, the transfer and warm-up self-inspection operations of the aircraft to be scheduled are performed when the available warm-up parking position constraints are met, and the corresponding completed stage of the aircraft to be scheduled is updated. Complete the stage record; when the next stage of the aircraft to be scheduled is the taxiing and waiting stage after the deflector, query the transfer path library from the warm-up position to the take-off position of the aircraft to be scheduled, and when the feasible transfer path constraints are met, find the earliest available time of the path, determine the earliest taxiing start time according to the earliest runway clearance time, and use the delayed transfer collision avoidance mechanism to obtain the delayed taxiing start time. If the aircraft to be scheduled is the highest priority aircraft at the take-off position, the aircraft to be scheduled will perform taxiing and waiting operations after the deflector, and update the corresponding completed stage record of the aircraft to be scheduled; when the next stage of the aircraft to be scheduled is the take-off preparation stage, obtain the flight start time of the aircraft to be scheduled according to the release time of the previous aircraft at the take-off position and the deflector reset cooling time. When the aircraft enters the take-off position, the aircraft to be scheduled performs the take-off position preparation operation, and updates the corresponding completed stage record of the aircraft to be scheduled; when the next stage of the aircraft to be scheduled is the take-off and dispatch stage, the earliest available take-off time of the aircraft to be scheduled is obtained according to the release time of the previous aircraft at the take-off position and the wake turbulence interval time, and the parking position clearance time of the interfering runway is searched from the earliest available take-off time as the final available take-off time, the aircraft to be scheduled performs the take-off and dispatch operation, and updates the corresponding completed stage record, the set to be dispatched and the scheduled set of the aircraft to be scheduled; the dispatch scheduling decoding rules also include: according to the dispatch timing plan and the take-off position allocation plan, the operation time of the warm-up self-check stage is right-shifted to the taxi start time.

[0181] The specific definition of the aircraft fleet takeoff scheduling device can be found in the above definition of the aircraft fleet takeoff scheduling method, which will not be repeated here. Each module in the above-mentioned aircraft fleet takeoff scheduling device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0182] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for scheduling the takeoff of an aircraft fleet is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0183] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0184] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in the above embodiment when executing the computer program.

[0185] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in the above embodiment are implemented.

[0186] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0187] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0188] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for scheduling takeoff of a fleet of aircraft, characterized in that: The method comprises: Parsing the pre-set fleet dispatching task to obtain the aircraft number, parking stand number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers; According to the aircraft number of the aircraft to be scheduled, the parking position number, the operation phase set, the multiple aircraft formations and the preset constraints, a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled is constructed with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function; Constructing a multi-island distribution model and initializing the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model; The cross entropy method is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library. When the termination condition of the first iteration is met, the optimal fitness corresponding to the optimal sample of each island is output; the local elite sample is the elite sample of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations; According to the pre-set multi-island information exchange and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, the distribution information of each island is exchanged and fused, and the distribution parameters of each island after exchange and fusion are updated in parallel using the cross entropy method. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output; The method of communicating and fusing the distribution information of each island according to the preset multi-island information communication and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, and using the cross entropy method to perform parallel iterative updates on the distribution parameters of each island after communication and fusion, and outputting the optimal fleet dispatching scheduling plan when the second iteration termination condition is met includes: A plurality of individuals are randomly sampled on each updated island. In any dimension of each individual, when the parameter random number is less than or equal to the preset crossover probability, a roulette wheel is played according to the optimal fitness of the remaining islands, and a communication island is randomly selected. The communication island is sampled according to the distribution model of the communication island in the corresponding dimension, so as to communicate and merge the distribution information of the current island and the communication island. When the parameter random number is greater than the preset crossover probability, sampling is performed according to the distribution model of each of the islands; According to the multiple new individuals after sampling, the cross entropy method is used to parallelly iteratively update the distribution parameters of each island after communication fusion. When the second iteration termination condition is met, the optimal fleet dispatch scheduling plan is output; the optimal fleet dispatch scheduling plan includes the optimal dispatch timing plan and the optimal take-off position allocation plan.

2. The method according to claim 1, characterized in that The steps to initialize the distribution model for each island include: Randomly generate a plurality of individuals within a defined interval of each of the distribution models, and decode the coded data corresponding to each of the individuals using a preset dispatch scheduling decoding rule to obtain solutions corresponding to the plurality of individuals; The K-means algorithm is used to divide the solutions corresponding to the multiple individuals into various islands as samples, and the distribution model of each island is initialized according to the samples on each island.

3. The method according to claim 1, characterized in that Before randomly generating a plurality of individuals within the defined interval of each distribution model, the method comprises: The take-off position selected by each aircraft to be scheduled and the dispatch priority of each aircraft to be scheduled are encoded using a continuous encoding rule to obtain an individual corresponding to each aircraft to be scheduled; the integer part of the gene of any dimension in the individual represents the take-off position selected by each aircraft to be scheduled, and the decimal part of the gene represents the dispatch priority of each aircraft to be scheduled; The continuous encoding rules also include: For any dimension gene in the individual, when the integer part of the gene is an odd number, the smaller the decimal part of the gene is, the higher the dispatch priority of the aircraft to be dispatched corresponding to the decimal part is; When the integer part of the gene is an even number, the larger the decimal part of the gene is, the higher the dispatch priority of the to-be-scheduled aircraft corresponding to the decimal part is.

4. The method according to claim 1, characterized in that: The steps of parallel iteratively updating the distribution model of each island after initialization using the cross entropy method based on the local elite samples and the island global elite library include: Randomly sample according to the distribution model of the current iteration number of each island to obtain multiple individuals, use the dispatch scheduling decoding mechanism to calculate the solution of each individual to obtain multiple samples, and obtain the elite sample and the optimal sample of each island according to the quantile of the order statistics of the multiple samples; The elite samples of each island are selected as local elite samples to construct the global elite library of the island; the global elite library of the island includes the global elite samples; the global elite samples include the historical optimal samples of each island in all previous iterations; According to the parameters of the distribution model constructed by the global elite sample of each island and the parameters of the distribution model constructed by the local sampling elite sample, the distribution parameters of the distribution model of each island are updated in parallel and iteratively.

5. The method according to claim 1, characterized in that The steps of obtaining the constraint conditions include: According to the transfer operation timing relationship, the support operation timing relationship, the connection relationship between the transfer operation and the support operation, and the deflector cooling reset time interval constraint of the current operation phase of the aircraft to be scheduled, the operation process timing constraint is obtained; the operation phase includes the transfer operation and the support operation; the transfer operation timing relationship is: Where I is the set of aircraft to be taken off, I = {1, 2, ..., Nf}, j ∈ J, J is the set of operation stages, J = {0, 1, 2, 3, 4}, 0 represents the initial deployment stage, 1 represents the warm-up self-check stage, 2 represents the waiting stage after the deflector, 3 represents the take-off preparation stage, 4 represents the take-off dispatch stage, Ezij is a decision variable, which represents the end time of the transfer of the i-th operation stage, Szij is a decision variable, which represents the start time of the transfer of the i-th operation stage, ΔTz ij The transit time for the aircraft to be taken off; The timing relationship of the security operation is: Among them, Ewij is a decision variable, which indicates the end time of the j-th operation stage of the i-th machine, Swij is a decision variable, which indicates the start time of the j-th operation stage of the i-th machine, and dwj is the working hours of the j-th stage guarantee operation; The connection between the transfer operation and the support operation is as follows: The deflector plate cooling reset time interval constraint is: Ypijejp is a decision variable, which indicates the priority of aircraft in different operation stages occupying the same parking position, Δtb is the cooling reset time of the deflector plate, P is the parking position set, including warm-up position, waiting position and take-off position, and Pj is the parking position set of the jth operation stage; According to the multi-machine transportation collision avoidance constraint and the limited time sequence constraint of occupying the same space, the space resource occupation time sequence constraint is obtained; the limited time sequence constraint of occupying the same space is: in, Yrijegr is a decision variable, which indicates the priority of aircraft in different operation stages to occupy the same take-off runway, g∈J, g≠j, R is the take-off runway set, and BM is the penalty term; According to the one-to-one correspondence between the waiting position and the take-off position on the same take-off runway, the matching constraint between the waiting position and the take-off position is obtained as follows: Among them, Pij is a decision variable, which represents the parking space occupied by the i-th machine in the j-th operation stage. is the take-off position corresponding to the next stage of the p waiting position, p∈Ψ2, γp∈Ψ3, Ψ is the set of available parking positions for the jth operation stage; According to the wake turbulence time interval standard between consecutive takeoff aircraft, the wake turbulence interval constraint is obtained as follows: Among them, xie is a decision variable, indicating the takeoff priority of the aircraft, and δi is the wake turbulence interval time after the i aircraft takes off; According to the takeoff configuration corresponding weight of the aircraft to be taken off and the takeoff weight upper limit of the takeoff position, the takeoff configuration constraint is obtained as follows: Among them, mi is the weight corresponding to the takeoff configuration of aircraft i, and fmp is the upper limit of the takeoff weight of takeoff position p; Constraint conditions are obtained according to the operation process timing constraint, the space resource occupation timing constraint, the waiting position and take-off position matching constraint, the wake interval constraint and the take-off configuration constraint.

6. The method according to claim 2, characterized in that The dispatch scheduling decoding rules include: Initialize the set to be scheduled as a list of aircraft to be scheduled, and initialize the set to be scheduled and the record of completed stages; the list to be scheduled includes the aircraft numbers of the aircraft to be scheduled sorted by priority weights; the record of completed stages is a matrix recording the completed stages of each aircraft to be scheduled; Indexing the aircraft number of the aircraft to be scheduled in the set to be scheduled, searching for the take-off position corresponding to the aircraft to be scheduled, selecting the next stage of the aircraft to be scheduled according to the completed stage record, and updating the completed stage record, the set to be scheduled, and the scheduled set corresponding to the aircraft to be scheduled after the aircraft to be scheduled completes the take-off and dispatching stage; Initialize an index value, index the aircraft number of the aircraft to be scheduled in the updated set to be scheduled according to the initialized index value, iteratively search for the take-off position corresponding to the aircraft to be scheduled, select the next stage process of the aircraft to be scheduled according to the completed stage record, and when the preset iteration termination condition is met, output the dispatch timing plan and the take-off position allocation plan, and calculate the objective function value of the dispatch scheduling model calculated by the dispatch timing plan and the take-off position allocation plan, and obtain the individual fitness according to the objective function value; The next stage of selecting the aircraft to be scheduled according to the completed stage record includes: When the next stage of the aircraft to be scheduled is the warm-up self-check stage, the transfer and warm-up self-check operations of the aircraft to be scheduled are performed when the available warm-up parking space constraints are met, and the completed stage record corresponding to the aircraft to be scheduled is updated; When the next stage of the aircraft to be scheduled is the taxiing and waiting stage after the deflector, query the transfer path library from the warm-up position to the take-off position of the aircraft to be scheduled, and when the feasible transfer path constraints are met, find the earliest available time of the path, determine the earliest taxiing start time according to the earliest runway clearing time, and use the delayed transfer collision avoidance mechanism to obtain the delayed taxiing start time. If the aircraft to be scheduled is the highest priority aircraft at the take-off position, the aircraft to be scheduled will perform taxiing and waiting operations after the deflector, and update the corresponding completed stage record of the aircraft to be scheduled; When the next stage of the aircraft to be scheduled is the take-off preparation stage, the take-off entry time of the aircraft to be scheduled is obtained according to the take-off release time of the previous aircraft at the take-off position and the deflector reset cooling time, the aircraft to be scheduled performs the take-off preparation operation, and updates the completed stage record corresponding to the aircraft to be scheduled; When the next stage of the aircraft to be scheduled is the take-off stage, the earliest available take-off time of the aircraft to be scheduled is obtained according to the release time of the last aircraft at the take-off position and the wake turbulence interval time. From the earliest available take-off time, the parking position clearance time of the interfering runway is searched as the final available take-off time. The aircraft to be scheduled performs the take-off operation, and updates the completed stage record, the set to be scheduled, and the scheduled set corresponding to the aircraft to be scheduled; The dispatch scheduling decoding rules also include: According to the dispatching sequence plan and take-off position allocation plan, the operation time of the warm-up self-inspection phase is shifted right to the taxiing start time.

7. An aircraft fleet takeoff scheduling device, characterized in that: The device comprises: A task analysis module is used to analyze the preset fleet dispatch task to obtain the aircraft number, parking position number, operation phase set and multiple aircraft formations of the aircraft to be dispatched; the aircraft formation includes multiple aircraft numbers; A dispatch scheduling model building module is used to build a dispatch scheduling model for the dispatch priority and take-off position selection of the aircraft to be scheduled based on the aircraft number of the aircraft to be scheduled, the parking position number, the operation stage set, the multiple aircraft formations and preset constraints, with the maximum fleet dispatch efficiency, the minimum fleet dispatch completion time and the minimum fleet taxiing time as the objective function; A multi-island distribution model construction module is used to construct a multi-island distribution model and initialize the distribution model of each island; the distribution model of each island is a mutually independent truncated Gaussian distribution model; The distribution model update module is used to perform parallel iterative updates on the distribution model of each island after initialization based on the local elite samples and the island global elite library using the cross entropy method, and output the optimal fitness corresponding to the optimal sample of each island when the first iteration termination condition is met; the local elite samples are the elite samples of each island at each iteration, and the island global elite library is used to store the historical optimal samples of each island in previous iterations; A scheduling scheme output module is used to communicate and fuse the distribution information of each island according to a preset multi-island information communication and fusion mechanism and the optimal fitness corresponding to the optimal sample of each island, and to use a cross entropy method to perform parallel iterative updates on the distribution parameters of each island after communication and fusion, and output the optimal fleet dispatch scheduling scheme when the second iteration termination condition is met; The scheduling scheme output module is also used to randomly sample and generate multiple individuals on each updated island. In any dimension of each individual, when the parameter random number is less than or equal to the preset crossover probability, a roulette wheel is played according to the optimal fitness of the remaining islands, and a communication island is randomly selected. Sampling is performed according to the distribution model of the communication island in the corresponding dimension to communicate and fuse the distribution information of the current island with the communication island; when the parameter random number is greater than the preset crossover probability, sampling is performed according to the distribution model of each island; according to the multiple new individuals after sampling, the distribution parameters of each island after communication and fusion are updated in parallel using the cross entropy method, and when the second iteration termination condition is met, the optimal fleet dispatch scheduling scheme is output; the optimal fleet dispatch scheduling scheme includes an optimal dispatch timing scheme and an optimal take-off position allocation scheme.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images