A low-altitude flight intelligent scheduling method and system

Abstract

The application discloses a kind of low-altitude flight intelligent scheduling method and system, it is related to flight scheduling technical field, its technical solution main point is including: obtaining the flight information of current to be scheduled aircraft, current to be scheduled aircraft is multiple;According to flight information, constraint condition and objective function are structured;According to constraint condition and objective function, lagrange optimization function is obtained;Based on bayesian optimization algorithm and lagrange relaxation method, lagrange optimization function is solved, and the scheduling scheme of current to be scheduled aircraft is obtained, the application is fused bayesian optimization algorithm in lagrange relaxation method to realize the dynamic update of multiplier, when facing flight delay or airspace adjustment, the technical problem that lagrange relaxation method cannot adjust scheduling scheme in time, it is difficult to meet the practical demand of complex low-altitude flight scheduling.

Description

Technical Field

[0001] The present invention relates to the technical field of flight scheduling, and more particularly to a low-altitude flight intelligent scheduling method and system. Background Art

[0002] Currently, the number of low-altitude flights continues to grow, covering various types such as eVTOL (electric vertical take-off and landing aircraft), drones, helicopters, etc. Without a scientific and reasonable scheduling strategy, a large number of flight delays will lead to economic losses. However, the scheduling solutions provided by existing technologies are difficult to meet the real-time dynamic scheduling needs to cope with dynamic interference such as sudden changes in wind speed and new flight applications. Therefore, the existing technology has shortcomings. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a low-altitude flight intelligent scheduling method and system, which solves the problem that the existing technology is difficult to meet the real-time dynamic scheduling needs by introducing a Bayesian optimization algorithm.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] The present invention provides a low-altitude flight intelligent scheduling method, comprising:

[0006] Obtaining flight information of a currently scheduled aircraft, where there are multiple currently scheduled aircraft;

[0007] Constructing constraints and objective functions based on the flight information;

[0008] Obtaining a Lagrangian optimization function according to the constraints and the objective function;

[0009] The Lagrangian optimization function is solved based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain a scheduling plan for the current aircraft to be scheduled.

[0010] As a further improvement of the present invention, the method of solving the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain the scheduling plan for the current aircraft to be scheduled includes:

[0011] Constructing a sample set according to the Lagrangian optimization function;

[0012] performing a first iterative operation based on the sample set, the first iterative operation comprising: calculating a subgradient based on a current multiplier, determining whether the subgradient satisfies a preset termination condition, and if not, obtaining a first function value based on the current multiplier and a subgradient method, obtaining a second function value based on the current sample set and a Bayesian optimization algorithm, updating the current sample set based on the first function value and the second function value, selecting a maximum value between the first function value and the second function value, updating the current multiplier until the preset termination condition is satisfied, and outputting the current function value;

[0013] A scheduling solution for the current aircraft to be scheduled is obtained according to the current function value.

[0014] As a further improvement of the present invention, constructing a sample set according to the Lagrangian optimization function includes:

[0015] A second iterative operation is performed according to the Lagrangian optimization function. The second iterative operation includes obtaining a current scheduling scheme and a current subgradient according to a current multiplier and the Lagrangian optimization function, updating the current multiplier according to the current subgradient to obtain a first multiplier and a first function value, and placing the first multiplier and the first function value as samples into a sample set until the number of samples in the sample set reaches a preset number.

[0016] As a further improvement of the present invention, obtaining the second function value according to the current sample set and the Bayesian optimization algorithm includes:

[0017] Constructing a Gaussian model based on the current sample set;

[0018] Based on the expectation strategy and the Gaussian model, a second multiplier is obtained;

[0019] The second function value is obtained according to the second multiplier and the Lagrangian optimization function.

[0020] As a further improvement of the present invention, the second multiplier is obtained based on the expectation strategy and the Gaussian model, including:

[0021] Obtaining a predicted probability distribution of the second multiplier according to the Gaussian model and the kernel function;

[0022] Obtaining a predicted value of the second multiplier according to the predicted probability distribution;

[0023] Obtaining an optimal function value according to the current sample set;

[0024] The second multiplier is obtained according to the predicted value and the optimal function value.

[0025] As a further improvement of the present invention, the low-altitude flight intelligent scheduling method further includes:

[0026] Sending instructions to each of the aircraft to be scheduled according to the scheduling plan, so that each of the aircraft to be scheduled flies according to the scheduling plan;

[0027] Determining the flight position of each of the aircraft to be scheduled and the newly added aircraft;

[0028] The scheduling plan is updated according to the flight position and the newly added aircraft.

[0029] As a further improvement of the present invention, updating the scheduling scheme according to the flight position includes:

[0030] Obtaining a predicted landing time point according to the flight position;

[0031] Determining a delayed aircraft based on the predicted landing time point and the landing time point in the scheduling plan;

[0032] updating the constraint conditions and the objective function according to the predicted landing time point;

[0033] The scheduling scheme is updated according to the updated objective function and constraints.

[0034] As a further improvement of the present invention, updating the scheduling plan according to the newly added aircraft includes:

[0035] Obtaining the take-off time of the newly added aircraft;

[0036] updating the aircraft to be scheduled according to the take-off time and the scheduling plan;

[0037] updating the constraint conditions and the objective function according to the aircraft to be scheduled;

[0038] The scheduling scheme is updated according to the updated objective function and constraints.

[0039] As a further improvement of the present invention, updating the scheduling scheme according to the updated objective function and constraint conditions includes:

[0040] Obtaining an update function according to the updated objective function and the constraint conditions;

[0041] The update function is solved according to the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain an updated scheduling solution.

[0042] The present invention provides a low-altitude flight intelligent scheduling system, comprising:

[0043] A collection module, configured to obtain flight information of a plurality of aircraft to be scheduled;

[0044] A construction module, constructing constraint conditions and objective functions according to the flight information;

[0045] A calculation module, which obtains a Lagrangian optimization function according to the constraint conditions and the objective function;

[0046] The scheduling module solves the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain a scheduling plan for the current aircraft to be scheduled.

[0047] By integrating the Bayesian optimization algorithm into the traditional Lagrangian relaxation method, the present invention achieves dynamic updating of the multiplier, solving the problems of the traditional Lagrangian relaxation method in the face of flight delays or airspace adjustments, such as the difficulty in adjusting the multiplier parameters and the slow convergence speed. This enables the final scheduling solution to be applicable to situations where the number of flights increases sharply and the scheduling scenarios are complex. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flowchart of the steps of a low-altitude flight intelligent scheduling method of the present invention;

[0049] Figure 2 is a flow chart of the iterative steps in the present invention;

[0050] Figure 3 Schematic diagram of the traversal sequence in the present invention;

[0051] Figure 4 is a flow chart of the data assimilation step in the present invention;

[0052] Figure 5 Schematic diagram of flight conflict detection in the present invention;

[0053] Figure 6 This is a scheduling framework diagram in the present invention. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations of the technical solution of the present invention.

[0055] The term "and / or" in the following text simply describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Additionally, the character " / " generally indicates an "or" relationship between the related objects.

[0056] like Figure 1 As shown, the embodiment of the present application provides a low-altitude flight intelligent scheduling method, including:

[0057] Get the flight information of the aircraft currently to be scheduled. There are multiple aircraft currently to be scheduled.

[0058] Construct constraints and objective functions based on flight information;

[0059] According to the constraints and the objective function, the Lagrangian optimization function is obtained;

[0060] The Lagrangian optimization function is solved based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain the scheduling plan for the current aircraft to be scheduled.

[0061] Among them, none of the aircraft currently to be scheduled have taken off, and the take-off and landing points of the aircraft currently to be scheduled are the same or different. The flight information of the aircraft currently to be scheduled includes a preset take-off time point, a preset landing time point, and a safety interval time, etc. The safety interval time between two aircraft is calculated by obtaining the minimum allowable time interval between the take-off time points of the two aircraft and the minimum allowable time interval between the landing time points of the two aircraft, and the minimum value between the two minimum time intervals is used as the safety interval time between the two aircraft to ensure flight safety. The specific value is determined according to the type of aircraft.

[0062] Specifically, this embodiment sets the objective function based on the goal of minimizing the delay cost. First, the delay time of each aircraft is calculated as:

[0063] ;

[0064] in, Indicates the The delay time of each aircraft, Indicates that after scheduling The predicted landing time of each aircraft, Indicates the first flight in the initial flight information. The preset landing time of each aircraft.

[0065] According to the loss cost of the aircraft per unit time, the delay cost during the delay time is calculated as:

[0066] ;

[0067] in, Indicates the The delay cost of an aircraft, Indicates the The types of aircraft can be divided into VTOL (electric vertical take-off and landing aircraft), drones and helicopters, etc. This embodiment does not limit this. Indicates the The loss cost of an aircraft per unit time.

[0068] After obtaining the delay time and delay cost of each aircraft, the average delay cost of the aircraft to be scheduled can be calculated. for:

[0069] ;

[0070] in, is the total number of aircraft currently to be scheduled, Represents and Related functions for calculating the The delay cost of an aircraft, but this embodiment There is no restriction on the specific expression of Can be equal to the above , which can also be equal to other formulas.

[0071] According to the average delay cost, the objective function can be obtained for:

[0072] ;

[0073] in, is the weight coefficient, is the maximum value of the delay time of each aircraft, that is, .

[0074] Furthermore, to ensure the feasibility of the scheduling scheme, this embodiment sets the following constraints. First, for each aircraft, the flight time after scheduling should be greater than or equal to the flight time before scheduling. The flight time includes the take-off time and the landing time point, that is, it should meet the following requirements:

[0075] ;

[0076] in, Indicates that after scheduling The take-off time of each aircraft, Indicates the first flight in the initial flight information. The preset takeoff time of each aircraft.

[0077] Furthermore, for any two aircraft with the same takeoff point, the time difference between their takeoffs should not be less than the safety interval, and for any two aircraft with the same landing point, the time difference between their landings should not be less than the safety interval, that is:

[0078] ;

[0079] ;

[0080] in, Indicates that after scheduling The predicted landing time of each aircraft, Indicates that after scheduling The take-off time of each aircraft, Indicates the aircraft and the The safety distance between the aircraft , and at this time aircraft and the The take-off and landing points of the two aircraft are the same.

[0081] Then construct conflict constraints based on the initial flight information :

[0082] ;

[0083] in, Indicates the number of combinations. For example, if there are three aircraft to be dispatched, three different combinations can be obtained by randomly selecting two different aircraft and grouping them together. , Represents the index, used to traverse all combinations. is a binary variable, for Two aircraft in a combination and , when they clash, ,on the contrary 0, where conflict refers to the take-off time between two aircraft with the same take-off point being less than the safety interval or the landing time being less than the safety interval. is a large constant, which makes the algorithm finitely satisfy the conflict constraints and then satisfy other constraints. For example, , but this embodiment does not limit this, and those skilled in the art can set other values ​​according to actual conditions.

[0084] According to the objective function and the above three constraints, the Lagrangian optimization function is obtained as follows:

[0085] ;

[0086] in, For about The piecewise linear function of A multiplier vector, each element in the multiplier vector corresponds to a constraint condition, is the total number of constraints, represents transpose, for The elements, It is by The linear function of the constraint transformation is in the form of , represents the scheduling scheme, where Indicates the landing time of the first aircraft. Indicates the take-off time of the first aircraft, and so on. A variable coefficient matrix, Middle and First Constraint-related elements, No. The threshold constant of the constraint condition is used to quantify the strictness of the constraint. The ultimate goal of this embodiment is to solve the Lagrangian optimization function to obtain the final scheduling solution. , represents the objective function under the current flight information, With the above The meaning of the expression is essentially the same. In optimization problems, when the dependency relationship in the objective function is clear, it can be omitted. Directly written as , in order to further highlight the constraints on Dependence, writing , Represents the linear function under the current flight information, for The feasible region of The range of values ​​that can be taken and the basic conditions that can be met, for example, in this embodiment Contains information such as flight departure time, It stipulates the legal scope of information.

[0087] This embodiment sets constraints so that the final scheduling plan meets the setting of the safety interval time, thereby achieving the feasibility of the scheduling plan. This embodiment sets an objective function based on the delay time, and sets a weight coefficient to balance the overall delay cost and the individual maximum delay cost, so as to avoid sacrificing individual flights to achieve the overall optimality and ensure the fairness of flight services. Then, a Lagrangian optimization function is constructed based on the objective function and the constraints. The constraints are converted into penalty terms by introducing Lagrangian multipliers to avoid falling into the local optimal situation. When constructing the Lagrangian optimization function, the constraints are converted into linear functions to facilitate subsequent solution based on the duality theory. Finally, the Lagrangian optimization function is solved by the Bayesian optimization algorithm and the Lagrangian relaxation method, which solves the traditional problems of difficult multiplier adjustment and slow convergence speed when facing flight delays or airspace adjustments, so that the final scheduling plan can be applied to situations where the number of flights has increased sharply and the scheduling scenarios are complex.

[0088] Furthermore, this embodiment provides a step of solving the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain a scheduling plan for each aircraft to be scheduled, including:

[0089] Construct a sample set based on the Lagrangian optimization function;

[0090] Performing a first iterative operation based on the sample set, the first iterative operation including: calculating a subgradient based on a current multiplier, determining whether the subgradient satisfies a preset termination condition, and if not, obtaining a first function value based on the current multiplier and a subgradient method, obtaining a second function value based on the current sample set and a Bayesian optimization algorithm, updating the current sample set based on the first function value and the second function value, selecting a maximum value between the first function value and the second function value, updating the current multiplier until the preset termination condition is satisfied, and outputting the current function value;

[0091] According to the current function value, the scheduling plan for each aircraft to be scheduled is obtained.

[0092] Specifically, such as Figure 2 As shown, first randomly initialize a multiplier vector , and then Substituting into the Lagrangian optimization function, we get , then solve Get the current scheduling plan for:

[0093] ;

[0094] in, Indicates smallest Value, solving the Lagrangian optimization function when the multiplier vector is known is a conventional technical means for those skilled in the art. Those skilled in the art can choose linear programming or quadratic programming to solve it, and this embodiment does not limit this.

[0095] Then and Add to sample set In this case, there is only A sample. Then calculate exist Subgradient at ,in, is the constraint coefficient matrix, the number of rows is the total number of constraints, and the number of columns is equal to the number of elements contained in the scheduling scheme. No. Behavior ,Right now , is a constant term vector whose dimension is equal to the total number of constraints, for The elements, namely .

[0096] Then according to the current subgradient Update the current multiplier vector , get the updated multiplier vector, which is the first multiplier , ,in represents the Euclidean norm, Represents the objective function value of the current feasible solution. A feasible solution is a solution that satisfies all constraints. When no feasible solution appears at the beginning of the iteration, a feasible solution that satisfies all constraints can be constructed to obtain , can be updated in subsequent iterations , that is, in subsequent iterations, whenever a solution is obtained, it is necessary to determine whether it is a feasible solution. If so, the objective function of the solution is calculated and compared with the current Compare the two and take the minimum value as the updated , used for subsequent calculations.

[0097] Afterwards Substituting into the Lagrangian optimization function, we get , That is The corresponding first function value, then The samples are placed in the current sample set, and the above steps of updating the multiplier vector and calculating the function value are repeated until the number of samples in the sample set reaches a preset number. This embodiment does not impose a limit on the preset number.

[0098] Assume that it is repeated After the number of samples in the sample set reaches the preset number, the following steps are performed for each subsequent iteration. As an example of iteration, the previous steps are the same as above. After being put into the sample set, a Gaussian model is trained based on the samples in the current sample set:

[0099] ;

[0100] in, Indicates that in the known sample set When , the probability distribution of the Lagrangian optimization function value is, represents a Gaussian process, is the Lagrangian optimization function value, which is the output variable of the Gaussian function. for The mean function can be set to a constant such as , or a linear function, which is not limited in this embodiment. is the kernel function, and Represents any two different first multipliers in the sample set. The kernel function is used to describe the covariance between the function values ​​corresponding to the two first multipliers. Those skilled in the art can select different kernel function types according to actual scenarios, and this embodiment does not limit this.

[0101] Based on the Gaussian model, the predicted probability distribution of the second multiplier can be obtained for:

[0102] ;

[0103] in, represents a normal distribution, is the covariance matrix corresponding to the sample set, and its dimension is , is the number of samples in the current sample set, Rank The elements of the column are , For the The first multiplier corresponding to the samples, For the The first multiplier corresponding to the samples, is the covariance vector of the sample set and the second multiplier, and its dimension is 1, its The elements are , is the second multiplier, is the vector of function values ​​in the sample set, that is , is the autocovariance of the second multiplier, Represents the prior mean of the Gaussian process, which can also be set to a constant such as , or a linear function, which is not limited in this embodiment.

[0104] Finally, based on the predicted probability distribution, the expectation strategy is used for exploration, and the second multiplier is obtained as:

[0105] ;

[0106] in is the predicted mean, also known as the acquisition function, , is the minimum value of all function values ​​in the current sample set, Indicates smallest , Express expectations.

[0107] Then Substituting into the Lagrangian optimization function, we get , The second multiplier The corresponding second function value, and Add to the current sample set and then compare and ,Will and The multiplier corresponding to the minimum value in is used as the current multiplier in the next iteration process to continue the iteration step until the preset termination condition is met, where the preset termination condition is that the calculated subgradient is zero. At this time, the function value corresponding to the subgradient is output, and the current scheduling plan corresponding to the function value is used as the scheduling plan to be executed.

[0108] Among them, as the iterative steps proceed, the current scheduling plan will continue to change, that is, the take-off time and landing time of the aircraft corresponding to each scheduling plan are different, and the conflict constraints are obtained based on the current scheduling plan. That is, when a new scheduling plan is obtained, the conflict constraints need to be updated, and the Lagrangian optimization function is updated, and the updated Lagrangian optimization function is used in subsequent steps.

[0109] This embodiment performs calculations based on a Bayesian optimization algorithm, i.e., explores the second multiplier based on a Gaussian process and an expectation strategy, is independent of the gradient direction, and can dynamically optimize the multiplier through probability distribution to find the global optimal solution. This solves the problem that the solution obtained by the traditional Lagrangian relaxation method may not meet the feasibility requirements of the original problem, and that the multiplier update depends on experience, parameter adjustment is difficult, and the convergence speed is slow or even oscillates. In addition, this embodiment solves the problem of a large duality gap between the Lagrangian dual solution and the original problem, which leads to a decrease in solution quality, so that the final execution scheduling solution can meet the practical needs of complex low-altitude flight scheduling.

[0110] Furthermore, after obtaining the scheduling plan, the scheduling plan can be evaluated. Specifically, the evaluation can be based on indicators such as the number of conflicts, average delay time, flow rate, control workload, and non-linear coefficient. Specifically:

[0111] Number of conflicts for:

[0112] ;

[0113] in, Indicates the aircraft after scheduling and Minimum distance separation during flight, Indicates the preset distance interval, represents the decision function, that is, if , then the aircraft is judged and Conflict, at this time If there is no conflict, .

[0114] Average delay time for:

[0115] ;

[0116] in Indicates the The delay time of each aircraft is the delay time between the scheduled landing time and the corresponding landing time in the initial flight information. Indicates the number of aircraft that experienced delays.

[0117] According to the location of each take-off point and landing point, the flight range is divided into multiple flight areas. Traffic volume in the flight area for:

[0118] ;

[0119] in Indicates the The aircraft passed The number of times in each area, the flow That is The total number of aircraft passing through the area.

[0120] Regulating workloads for:

[0121] ;

[0122] in Indicates the conflict type is The severity of the conflict is The number of aircraft, Represents the conflict adjustment factor, which indicates the difficulty of adjusting conflicts of different severity. It represents the conflict severity coefficient, which indicates the inherent difficulty of adjusting different conflict types. This embodiment does not limit the conflict type, conflict severity, conflict adjustment factor and conflict severity coefficient, and those skilled in the art can set them by themselves.

[0123] Non-linearity coefficient for each aircraft It can be expressed as:

[0124] ;

[0125] in, The flight path of the aircraft involved in the scheduling plan The length of the flight segment, is the number of flight segments, The route distance between the starting and ending points.

[0126] By comparing the values ​​of these indicators with the preset data, a specific evaluation result can be obtained. Then, based on the specific evaluation results, one can choose to execute the scheduling plan or readjust the parameters (such as the preset quantity) and recalculate the scheduling plan. The indicators provided in this embodiment are only examples, and this embodiment does not limit the values ​​of the indicators. Those skilled in the art can set them on their own or combine other evaluation indicators for evaluation based on this example.

[0127] Based on the above analysis, it can be summarized that the steps of constructing a sample set based on the Lagrangian optimization function provided in this embodiment include:

[0128] A second iterative operation is performed according to the Lagrangian optimization function. The second iterative operation includes obtaining a current scheduling scheme and a current subgradient according to the current multiplier and the Lagrangian optimization function, updating the current multiplier according to the current subgradient, obtaining a first multiplier and a first function value, and placing the first multiplier and the first function value as samples into a sample set until the number of samples in the sample set reaches a preset number.

[0129] The step of obtaining the second function value according to the current sample set and the Bayesian optimization algorithm provided in this embodiment includes:

[0130] Construct a Gaussian model based on the current sample set;

[0131] Based on the expectation strategy and Gaussian model, the second multiplier is obtained;

[0132] A second function value is obtained according to the second multiplier and the Lagrangian optimization function.

[0133] The steps of obtaining the second multiplier based on the expectation strategy and the Gaussian model provided in this embodiment include:

[0134] According to the Gaussian model and kernel function, the predicted probability distribution of the second multiplier is obtained;

[0135] Obtain the predicted value of the second multiplier according to the predicted probability distribution;

[0136] Get the optimal function value based on the current sample set;

[0137] The second multiplier is obtained according to the predicted value and the optimal function value.

[0138] Furthermore, the above-mentioned low-altitude flight intelligent scheduling method further includes:

[0139] Sending instructions to each aircraft to be scheduled according to the scheduling plan, so that each aircraft to be scheduled flies according to the scheduling plan;

[0140] Determine the flight position of each aircraft to be scheduled and the newly added aircraft;

[0141] Update the scheduling plan based on flight locations and newly added aircraft.

[0142] Specifically, after obtaining the scheduling plan, each aircraft is scheduled to take off according to the take-off time of each aircraft in the scheduling plan. However, during the flight of the aircraft, emergencies may occur. For example, a certain aircraft is affected by environmental factors, and the landing time point is delayed. The delayed landing time point and the landing time point of the subsequent aircraft are less than the safety time interval, or there is a new aircraft at this time, and the initial flight information of the new aircraft conflicts with the current scheduling plan. The conflict includes that the take-off time or landing time of the new aircraft and one of the aircraft currently to be scheduled is less than the safety time interval. For these situations, the scheduling plan needs to be updated.

[0143] Furthermore, this embodiment provides a step of updating a scheduling plan based on a flight position, including:

[0144] Get the predicted landing time based on the flight position;

[0145] Determine the delayed aircraft based on the predicted landing time and the landing time in the scheduling plan;

[0146] Update the constraints and objective function based on the predicted landing time point;

[0147] Update the scheduling plan based on the updated objective function and constraints.

[0148] Specifically, it is first necessary to detect the aircraft that has taken off in real time and obtain the data transmitted back by the aircraft at the current moment, such as the current position and speed, and then input the data transmitted back at the current moment into the preset scheduling model to obtain the state parameters of each future moment. The state parameters include speed. Then, the output results of the scheduling model are input into the observation model to obtain the predicted position of each future moment, and then obtain the predicted landing time point corresponding to the current moment. Among them, predicting the state parameters and predicted landing time points of each future moment based on the data transmitted back in real time is a technical means commonly used by technical personnel in this field. Therefore, this embodiment does not limit the specific formulas of the scheduling model and the observation model. Technical personnel in this field can select appropriate model formulas as needed.

[0149] After executing the above steps for each aircraft that has already taken off, delayed aircraft are identified based on each predicted landing time at the current moment (i.e., aircraft whose predicted landing time is later than the landing time in the schedule). Next, aircraft with the same takeoff or landing point as the delayed aircraft are identified, and aircraft with landing times after the delayed aircraft are selected. These aircraft are used as the updated aircraft to be scheduled, and the schedule is updated accordingly. This embodiment assumes that there is only one delayed aircraft, and the steps for updating the schedule are similar to those for obtaining the schedule described above. It should be noted that the initial flight information at this point is determined based on the previously determined schedule. That is, if there are aircraft that have already taken off among the updated aircraft to be scheduled at the current moment, the takeoff times of these aircraft are known in the initial flight information and are specific values, so they do not need to participate in the subsequent iterative update steps. Furthermore, compared to the previous steps, a constraint is added: the updated landing time of each aircraft to be scheduled must be at least a safe time interval away from the predicted landing time of the delayed aircraft. Since the flight information and constraints of the aircraft to be scheduled change, the objective function will also change accordingly, but the overall solution steps remain unchanged and will not be described in detail in this embodiment.

[0150] Furthermore, in addition to adding a constraint condition, this embodiment provides another method that can ensure that the updated landing time point of each aircraft to be scheduled and the predicted landing time point of the delayed aircraft are not less than the safety time interval without adding a constraint condition. Specifically, after executing the above steps to obtain a scheduling plan without adding a constraint condition, the updated landing time point of each aircraft to be scheduled is obtained, and each landing time point is sorted according to the landing order. Then, the predicted landing time point of the delayed aircraft is placed at the first place in the sequence. Then, the intervals between two adjacent time points are compared in turn to see if they are not less than the safety time interval. If so, no adjustment is made. If not, Figure 3 As shown, Figure 3 Each box in the sequence represents a time point. The landing time point of the latter of the two adjacent time points is updated so that the interval between them is equal to the safety time interval. Then, when performing the next comparison, the updated landing time point needs to be used for comparison until the sequence is traversed. Based on the landing time point of each updated aircraft to be scheduled and the take-off time point in the scheduling plan obtained by executing the above steps, the final updated scheduling plan corresponding to the current moment is obtained, and the flight mission of each updated aircraft to be scheduled is executed according to the scheduling plan.

[0151] Moreover, at each subsequent moment, it is necessary to perform steps to determine the delayed aircraft and subsequently update the scheduling plan, wherein, in the subsequent update step, the data input to the scheduling model is not the data transmitted in real time by the aircraft. For example, after receiving the data transmitted in real time by the aircraft, it is necessary to obtain the predicted data of the scheduling model and the observation model for the aircraft at that moment in the previous moment, and then perform data assimilation on the real-time transmitted data and the predicted data to obtain the data input to the scheduling model at the current moment, and then obtain each predicted landing time point at the current moment based on the output result of the scheduling model, wherein the data input to the scheduling model at the current moment includes position and speed, etc. Data assimilation is a method of fusing the real-time transmitted data with the predicted data to obtain more accurate data. For example, a weighted or Kalman filtering method can be used, which is not limited in this embodiment.

[0152] Further, such as Figure 4 As shown, since the preset scheduling model is usually obtained based on historical data, it cannot fully adapt to the current scenario to obtain accurate output results. Based on this, in this embodiment, the real aircraft system is first detected to obtain the data transmitted back by the aircraft in real time, and the data is assimilated with the predicted data of the aircraft at the previous moment by the scheduling model and the observation model. The assimilated data can better reflect the actual status of the aircraft at the current moment, and after each data assimilation, it is necessary to correct the parameters in the scheduling model based on the deviation between the current data after assimilation and the predicted state parameters of the scheduling model at the previous moment, and then input the assimilated data into the scheduling model after adjusting the parameters, and then proceed to the subsequent steps.

[0153] This embodiment provides two solutions to aircraft delays. The first adds constraints, allowing the algorithm to search within a solution space that satisfies all constraints, ultimately obtaining a globally optimal solution. This approach is more suitable for scenarios with sufficient computing power. The other solution provided by this embodiment does not add constraints, but instead determines the final scheduling solution by traversing a sequence. This solves the problem of increasing difficulty and slowing down the solution speed as constraints are added, making it more suitable for scheduling scenarios with high real-time requirements. Furthermore, this embodiment adjusts the data input to the scheduling model through data assimilation, allowing for correction of predicted data using real-time data. When errors in real-time data are caused by aircraft failures, the predicted data is integrated to reduce the impact of data errors on the scheduling solution. Ultimately, the integrated data is obtained, leading to a more optimal scheduling solution.

[0154] Furthermore, the embodiment provides a step of updating the scheduling plan based on the newly added aircraft, including:

[0155] Get the takeoff time of the newly added aircraft;

[0156] Update the aircraft to be scheduled according to the take-off time and scheduling plan;

[0157] Update the constraints and objective function according to the aircraft to be scheduled;

[0158] Update the scheduling plan based on the updated objective function and constraints.

[0159] Specifically, when a new aircraft needs to be scheduled and the initial flight information of the aircraft conflicts with the current flight information, the scheduling plan needs to be updated. Specifically, Figure 5 As shown, the current flight information includes an existing path, existing flight information, and a prohibited intersection schedule. The existing path includes the flight path of the current aircraft to be scheduled, and the existing flight information includes the takeoff and landing times of the current aircraft to be scheduled. The prohibited intersection schedule records the time period when the current aircraft to be scheduled occupies a certain flight area and prohibits other aircraft from entering the flight area. The flight information of the newly added aircraft includes the takeoff time, landing time, takeoff point, landing point, and the time period when the aircraft occupies a certain flight area and prohibits other aircraft from entering the flight area. Based on this, conflicts between the initial flight information of the aircraft and the current flight information specifically include overlapping flight paths at a certain time, takeoff time or landing time less than the safety time interval, and overlapping time periods occupying the same flight area. If there is a conflict, the scheduling plan and prohibited intersection schedule are updated through recalculation. If there is no conflict, the initial flight information of the newly added aircraft is directly added to the current scheduling plan and prohibited intersection schedule, completing the update of the current scheduling plan and prohibited intersection schedule.

[0160] Specifically, when there is a conflict, the initial flight information of the newly added aircraft is first obtained, including the take-off time, landing time, take-off point and landing point, and then the flight status of the currently scheduled aircraft is determined, from which the aircraft with the same take-off point or landing point as the newly added aircraft and which have not yet completed their flight are determined. The aircraft that have not yet completed their flight specifically refers to aircraft that have taken off but whose scheduled landing time is later than the current time and the time difference with the current time is greater than the preset time window, and aircraft that have not yet taken off.

[0161] After that, the aircrafts that have not completed the flight and the new aircraft are collectively regarded as updated aircrafts to be dispatched, and are recalculated based on the above steps of obtaining the scheduling scheme to obtain an updated scheduling scheme, wherein the initial flight information of the aircrafts that have not completed the flight is obtained according to the previously obtained scheduling scheme, and since the initial flight information changes, the constraint condition and the objective function need to be updated, and the Lagrange optimization function is updated according to the updated constraint condition and the objective function, and the updated Lagrange optimization function is denoted as an update function, and the above steps of obtaining the scheduling scheme by the Bayesian optimization are executed again.

[0162] The prior art usually uses a static scheduling method to schedule the new aircraft when the new aircraft causes a flight conflict, that is, the flight information of the aircrafts to be dispatched is kept unchanged, and only the flight information of the new aircraft is adjusted. However, the embodiment adopts a dynamic scheduling method, that is, a preset time window is determined, and only the aircrafts whose landing time after scheduling is later than the current time and the time difference between the landing time after scheduling and the current time is greater than the preset time window are rescheduled, and the aircrafts whose landing time after scheduling is later than the current time and the time difference between the landing time after scheduling and the current time is less than or equal to the preset time window (that is, the aircrafts that will land) do not need to be rescheduled, so that the number of updated aircrafts to be dispatched is reduced, the calculation complexity is reduced, the aircrafts that will land are avoided to be scheduled, and flight safety is ensured.

[0163] According to the above analysis, it can be summarized that the step of updating the scheduling scheme according to the updated objective function and constraint condition provided by the embodiment includes:

[0164] obtaining an update function according to the updated objective function and constraint condition;

[0165] solving the update function according to the Bayesian optimization algorithm and the Lagrange relaxation method to obtain an updated scheduling scheme.

[0166] Further, the embodiment of the application provides a low-altitude flight intelligent scheduling system, which includes:

[0167] an acquisition module configured to acquire flight information of current aircrafts to be dispatched, the current aircrafts to be dispatched being a plurality of aircrafts;

[0168] a construction module configured to construct a constraint condition and an objective function according to the flight information;

[0169] a calculation module configured to obtain a Lagrange optimization function according to the constraint condition and the objective function;

[0170] a scheduling module configured to solve the Lagrange optimization function based on a Bayesian optimization algorithm and a Lagrange relaxation method to obtain a scheduling scheme of each aircraft to be dispatched.

[0171] Specifically, as shown in Figure 6 According to the system provided by the embodiment, a complete scheduling framework can be obtained. First, resource plans, emergencies and digital twin information are obtained, input into a flight scheduling model, and a final flight schedule, i.e., a scheduling scheme, is obtained. Finally, the scheduling scheme is executed and evaluated. The flight scheduling model is equivalent to the low-altitude flight intelligent scheduling system provided by the embodiment, which is used to execute the low-altitude flight intelligent scheduling method provided by the embodiment.

[0172] The low-altitude flight intelligent scheduling method and system provided by the embodiment solves the complex conflict constraint problem in low-altitude flight scheduling by fusing the Lagrange multiplier adaptive low-altitude flight scheduling method based on Bayesian optimization and relaxing the conflict constraint into a Lagrange multiplier penalty term through the Lagrange relaxation technique, and converting the original problem into an optimization problem with a penalty term. Specifically, the embodiment first constructs a constraint condition for the conflict relationship in the low-altitude flight information, relaxes the conflict constraint by using the Lagrange multiplier, constructs a Lagrange optimization function, and the function includes a flight delay cost term and a multiplier penalty term. Subsequently, a Bayesian optimization framework is adopted, the objective function in the Lagrange multiplier space is modeled by using a Gaussian process, and the expected improvement acquisition function is used to guide the intelligent selection and update of the multiplier parameters, breaking the traditional single update strategy based on the sub-gradient, realizing the adaptive adjustment of the multiplier, effectively improving the optimization efficiency and convergence speed, and constantly narrowing the conflict range by alternately updating the Lagrange multiplier and the flight information, gradually approaching the optimal scheduling scheme, balancing global search and local exploration, reducing the risk of falling into local optimum, and being applicable to complex low-altitude flight environments with multiple conflict constraints. It can minimize scheduling conflicts while optimizing flight delay time and improving flight operation efficiency.

[0173] Those skilled in the art will appreciate that embodiments of the application can be provided as methods, systems, or computer program products. Therefore, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0174] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0175] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0176] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A low-altitude flight intelligent scheduling method, characterized in that: include: Obtaining flight information of a currently scheduled aircraft, where there are multiple currently scheduled aircraft; Constructing constraints and objective functions based on the flight information; Obtaining a Lagrangian optimization function according to the constraints and the objective function; Solving the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain a scheduling plan for the current aircraft to be scheduled; The method of solving the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain the scheduling plan for the current aircraft to be scheduled includes: Constructing a sample set according to the Lagrangian optimization function; performing a first iterative operation based on the sample set, the first iterative operation comprising: calculating a subgradient based on a current multiplier, determining whether the subgradient satisfies a preset termination condition, and if not, obtaining a first function value based on the current multiplier and a subgradient method, obtaining a second function value based on the current sample set and a Bayesian optimization algorithm, updating the current sample set based on the first function value and the second function value, selecting a maximum value between the first function value and the second function value, updating the current multiplier until the preset termination condition is satisfied, and outputting the current function value; Obtaining a scheduling plan for the current aircraft to be scheduled according to the current function value; The obtaining of the second function value according to the current sample set and the Bayesian optimization algorithm includes: Constructing a Gaussian model based on the current sample set; Based on the expectation strategy and the Gaussian model, a second multiplier is obtained; Obtaining the second function value according to the second multiplier and the Lagrangian optimization function; The second multiplier is obtained based on the expectation strategy and the Gaussian model, including: Obtaining a predicted probability distribution of the second multiplier according to the Gaussian model and the kernel function; Obtaining a predicted value of the second multiplier according to the predicted probability distribution; Obtaining an optimal function value according to the current sample set; The second multiplier is obtained according to the predicted value and the optimal function value.

2. A low-altitude flight intelligent scheduling method according to claim 1, characterized in that: The constructing of a sample set according to the Lagrangian optimization function includes: A second iterative operation is performed according to the Lagrangian optimization function. The second iterative operation includes obtaining a current scheduling scheme and a current subgradient according to a current multiplier and the Lagrangian optimization function, updating the current multiplier according to the current subgradient to obtain a first multiplier and a first function value, and placing the first multiplier and the first function value into a sample set until the number of samples in the sample set reaches a preset number.

3. The low-altitude flight intelligent scheduling method according to claim 1, characterized in that: The method further comprises: Sending instructions to each of the aircraft to be scheduled according to the scheduling plan, so that each of the aircraft to be scheduled flies according to the scheduling plan; Determining the flight position of each of the aircraft to be scheduled and the newly added aircraft; The scheduling plan is updated according to the flight position and the newly added aircraft.

4. A low-altitude flight intelligent scheduling method according to claim 3, characterized in that: Updating the scheduling plan according to the flight position, including: Obtaining a predicted landing time point according to the flight position; Determining a delayed aircraft based on the predicted landing time point and the landing time point in the scheduling plan; updating the constraint conditions and the objective function according to the predicted landing time point; The scheduling scheme is updated according to the updated objective function and constraints.

5. The low-altitude flight intelligent scheduling method according to claim 3, characterized in that: Updating the scheduling plan based on the newly added aircraft includes: Obtaining the take-off time of the newly added aircraft; updating the aircraft to be scheduled according to the take-off time and the scheduling plan; updating the constraint conditions and the objective function according to the aircraft to be scheduled; The scheduling scheme is updated according to the updated objective function and constraints.

6. A low-altitude flight intelligent scheduling method according to claim 4 or 5, characterized in that: The updating of the scheduling scheme according to the updated objective function and constraint conditions includes: Obtaining an update function according to the updated objective function and the constraint conditions; The update function is solved according to the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain an updated scheduling solution.

7. A low-altitude flight intelligent dispatching system, characterized in that: include: A collection module, configured to obtain flight information of a plurality of aircraft to be scheduled; A construction module, constructing constraint conditions and objective functions according to the flight information; A calculation module, which obtains a Lagrangian optimization function according to the constraint conditions and the objective function; A scheduling module, which solves the Lagrangian optimization function based on a Bayesian optimization algorithm and a Lagrangian relaxation method to obtain a scheduling plan for the current aircraft to be scheduled; The method of solving the Lagrangian optimization function based on the Bayesian optimization algorithm and the Lagrangian relaxation method to obtain the scheduling plan for the current aircraft to be scheduled includes: Constructing a sample set according to the Lagrangian optimization function; performing a first iterative operation based on the sample set, the first iterative operation comprising: calculating a subgradient based on a current multiplier, determining whether the subgradient satisfies a preset termination condition, and if not, obtaining a first function value based on the current multiplier and a subgradient method, obtaining a second function value based on the current sample set and a Bayesian optimization algorithm, updating the current sample set based on the first function value and the second function value, selecting a maximum value between the first function value and the second function value, updating the current multiplier until the preset termination condition is satisfied, and outputting the current function value; Obtaining a scheduling plan for the current aircraft to be scheduled according to the current function value; The obtaining of the second function value according to the current sample set and the Bayesian optimization algorithm includes: Constructing a Gaussian model based on the current sample set; Based on the expectation strategy and the Gaussian model, a second multiplier is obtained; Obtaining the second function value according to the second multiplier and the Lagrangian optimization function; The second multiplier is obtained based on the expectation strategy and the Gaussian model, including: Obtaining a predicted probability distribution of the second multiplier according to the Gaussian model and the kernel function; Obtaining a predicted value of the second multiplier according to the predicted probability distribution; Obtaining an optimal function value according to the current sample set; The second multiplier is obtained according to the predicted value and the optimal function value.

Images