CN109147396B - Method and device for allocating airport parking positions - Google Patents
Method and device for allocating airport parking positions Download PDFInfo
- Publication number
- CN109147396B CN109147396B CN201810969491.9A CN201810969491A CN109147396B CN 109147396 B CN109147396 B CN 109147396B CN 201810969491 A CN201810969491 A CN 201810969491A CN 109147396 B CN109147396 B CN 109147396B
- Authority
- CN
- China
- Prior art keywords
- flight
- probability
- sample
- time
- state matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G5/00—Traffic control systems for aircraft, e.g. air-traffic control [ATC]
- G08G5/0043—Traffic management of multiple aircrafts from the ground
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G5/00—Traffic control systems for aircraft, e.g. air-traffic control [ATC]
- G08G5/0073—Surveillance aids
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The embodiment of the invention provides a method and a device for allocating airport parking positions, wherein the method comprises the following steps: constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment; inputting the environment state matrix into a probability distribution model, and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flights to each stand; and allocating the stand for the flight according to the probability map. According to the embodiment of the invention, the airport environment state matrix is constructed according to the flight to be distributed and the state information of the airport environment, so that the modeling accuracy is improved, and the safety of the parking lot distribution is improved; when the status information of the flights to be distributed and the airport environment changes, the corresponding parking lot distribution scheme can be obtained only by reconstructing the airport environment status matrix, and the management is easy; the environment state matrix is directly input into the probability distribution model, and the probability distribution model is used for distributing flights in real time, so that the operation time is saved, and the problem efficiency of airplane allocation is improved.
Description
Technical Field
The embodiment of the invention relates to the field of airport operation optimization, in particular to an airport parking space allocation method and device.
Background
In recent years, air transportation has been rapidly developed. However, the development of air transportation also brings great challenges to civil aviation operation management, and the air traffic pressure borne by an airport as the starting point of traffic is continuously increased, so that the operation efficiency is affected, and the development of air transportation becomes one of the main causes of flight delay. In the airport operation process, the allocation of the stand is one of the core resources, namely the starting point of an departing flight and the end point of an approaching flight, and the allocation result of the stand directly influences the allocation scheme of personnel and materials. Therefore, the optimal allocation of the parking spaces plays an important role in guaranteeing the safety and the high-efficiency operation of the airport.
In the actual operation of airport stand allocation, the allocation of stands is generally divided into two stages, namely pre-allocation and dynamic allocation.
In the pre-allocation stage, the prior art algorithm pre-allocates the flight according to the known flight plan, flight level resources, operation constraint conditions, and the like. However, in the distribution method in the prior art, various actual operation constraints need to be abstracted in the optimization problem modeling, and since the constraint conditions in the prior art are difficult to fully cover all the actual operation constraints, the distribution scheme in the prior art has great potential safety hazards. In addition, when the airport allocation management rules are modified, the airport allocation method in the prior art needs to re-model the constraint conditions of the optimization problem, needs re-intervention of an algorithm developer, and is not easy to manage in a later period.
In the dynamic allocation stage, the flight time variation problem caused by abnormal situations such as flight delay needs to be adjusted in time. The dynamic allocation problem has high randomness because of more outburst situations in airport operation and frequent change of flight time, but the requirements on the robustness and the real-time performance of the algorithm in the prior art are higher. However, the distribution method in the prior art is basically based on the pre-distribution of flights, so the calculation time is long, and the real-time requirement of dynamic distribution is difficult to meet.
Aiming at the problems of potential safety hazard, difficult management and long calculation time in the prior art, a new technical scheme is urgently needed to be provided to overcome the problems.
Disclosure of Invention
Embodiments of the present invention provide a method and apparatus for assigning airport stands that overcomes, or at least partially solves, the above-mentioned problems.
In a first aspect, an embodiment of the present invention provides a method for allocating airport parking spaces, including: constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment; inputting the environment state matrix into a probability distribution model, and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flights to each stand; and allocating the stand for the flight according to the probability map.
In a second aspect, an embodiment of the present invention provides an apparatus for allocating airport parking spaces, including: the modeling module is used for constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment; the processing module is used for inputting the environment state matrix into the probability distribution model to obtain a probability map output by the probability distribution model, and the probability map is used for indicating the probability of distributing the flights to each stand; and the distribution module is used for distributing the stand for the flight according to the probability map.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the steps of the method for allocating airport stands provided in the first aspect.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the method for allocating airport stands provided in the first aspect.
According to the embodiment of the invention, the airport environment state matrix is constructed according to the flight to be distributed and the state information of the airport environment, so that the modeling accuracy is improved, and the safety of the parking space distribution is improved; when the status information of the flights to be distributed and the airport environment changes, the corresponding parking lot distribution scheme can be obtained only by reconstructing the airport environment status matrix, and the management is easy; the environment state matrix is directly input into the probability distribution model, and the probability distribution model is used for distributing the flights in real time, so that the operation time is saved, and the problem efficiency of airplane allocation is improved.
Drawings
Fig. 1 is a schematic flow chart of an allocation method of airport parking spaces according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a process for training a policy network model according to an embodiment of the present invention;
FIG. 3 is a schematic view of an episode simulation process according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an airport stand allocation device provided by an embodiment of the present invention;
fig. 5 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The allocation of airport parking spaces is the core business of airport production scheduling. Unreasonable airport stand allocation schemes can cause flight delay and congestion, reduce passenger satisfaction, affect normal operation of related departments, and even cause accidents. Therefore, the parking spaces of the airport need to be reasonably distributed so as to ensure that passengers can conveniently get on and off flights, take luggage, transfer and get in and out of the port, and ensure that various operations on the ground of the airport are smoothly carried out while the parking spaces of the airport are reasonably utilized.
Fig. 1 is a schematic flow chart of an allocation method of airport parking spaces according to an embodiment of the present invention.
As shown in fig. 1, the method includes:
102, inputting the environment state matrix into a probability distribution model to obtain a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flights to each stand;
and 103, allocating an aircraft stop for the flight according to the probability map.
Specifically, in step 101, the status of the airport environment includes the resource occupancy status of the airport stand, the stand allocation rule, the predicted time of flight entering the stand, the predicted time of flight exiting the stand, and the like. And constructing an environment state matrix corresponding to the flight to be distributed according to the actual occupation state of the flight stand to be distributed, the current stand distribution rule, the estimated time of entering the stand of the flight corresponding to the flight to be distributed listed in the flight schedule and the estimated time of leaving the stand of the flight.
After obtaining the probability map of the flight to assign to each stand at the airport, the system may select the corresponding stand according to the probability map as the actual stand of the flight, step 103. For example, the probability map for flight A describes the probability that flight A is assigned to each stand, and the system may select the stand with the highest probability as the actual stand for flight A.
According to the embodiment of the invention, the airport environment state matrix is constructed according to the flight to be distributed and the state information of the airport environment, so that the modeling accuracy is improved, and the safety of the parking space distribution is improved; when the status information of the flights to be distributed and the airport environment changes, the corresponding parking lot distribution scheme can be obtained only by reconstructing the airport environment status matrix, and the management is easy; the environment state matrix is directly input into the probability distribution model, and the probability distribution model is used for distributing the flights in real time, so that the operation time is saved, and the problem efficiency of airplane allocation is improved.
On the basis of the above embodiment, as an alternative embodiment, the status information of the airport environment includes: the method comprises the following steps of (1) parking space resource occupation state information, parking space distribution rules, the estimated time of a flight entering a parking space and the estimated time of the flight exiting the parking space; correspondingly, according to the status information of the flights to be distributed and the airport environment, an environment status matrix is constructed, which comprises the following steps: according to the occupation state of the parking space resources, a parking state matrix is constructed; constructing a constraint vector of the flight according to the stop allocation rule; constructing a flight state matrix according to the halt state matrix, the constraint vector, the estimated time of the flight entering the halt station and the estimated time of the flight exiting the halt station; and constructing an environment state matrix according to the shutdown state matrix and the flight state matrix.
Specifically, the current time t is taken as the 0 th time step, and then the occupancy state of the airport parking space resource from the 0 th time step to the mth time step, that is, the current time t to the time t + mxΔ t, is used to represent the occupancy state of the airport parking space resource observable at the time t. Where Δ t represents a time step, i.e., a preset duration, e.g., 5 minutes for one time step, and M represents the number of time steps. For example, if the current time is 10:00, T is five minutes, and M is 10, then time T to T + M × Δ T represents time 10:00 to time 10: 50. And establishing a corresponding stand-off state matrix according to the stand-off resource occupation state. For example, if the number of airport stands is N, then the stand state matrix at time t is the (M +1) XN matrix At。
Wherein, aij,tThe condition that the jth stand is occupied at the t + i time step is observed at the t time step, and if the jth stand is occupied at the t + i time step, aij,t1, otherwise aij,t＝0。
And constructing a rule constraint vector of the flight according to the stop allocation rule. For example, flight v has a regular constraint vector of Pv。
Pv＝(pv1,pv2,...,pvN) (2)
Wherein p isvnIndicating whether the flight v can stop at stop n, if sovn1, otherwise pvn0, wherein N is 1,2, …, N.
Obtaining the predicted arrival time of the flight v from the flight information of the flight v in the flight schedule, which is the tth after t1A time step, i.e. time t + t1X Δ t, then the occupancy matrix is
Constraint vector P according to the rules of flight vvAnd occupancy matrix
Deriving a flight v available stop vector Cvt。
Suppose the predicted time of flight arrival and stop is t after t1At each time step, the estimated time of flight leaving the stand is tth after t2One time step, then flight status matrix B for flight vvt。
Wherein the content of the first and second substances,m,t＝Cvt,m∈[t1,t2]，
if the flight v can be allocated to the stand n, and the downtime of the flight v on the stand n is t after t1A time step until t th after t2A time step, then bmn,t＝1,m∈[t1,t2]，
The shutdown status matrix A according to flight vtAnd the state matrix BvtConstructing the environment state matrix s of the flight vt，
According to the embodiment of the invention, the airport environment state matrix is constructed according to the parking space resource occupation state, the parking space allocation rule and the estimated time of the flight entering the parking space and exiting the parking space, the corresponding constraint vector can be established according to different parking space allocation principles, the modeling accuracy is improved, and the parking space allocation safety is improved.
On the basis of the above embodiment, as an alternative embodiment, before inputting the environment state matrix to the probability distribution model, the method includes: constructing a strategy network model; performing plot simulation based on a policy network model, wherein each plot simulation starts from the 0 th time step, sequentially adds 1 time step, simulates flights of each simulation time in flight schedule samples, and obtains a state sample corresponding to each simulation time; and updating the policy network model by adopting the state samples, and obtaining a probability distribution model after the updating is finished.
Specifically, setting structural parameters of a policy network model, including the number of layers of a neural network, the number of neurons in each layer and the like, calling a Theano deep learning library or other libraries capable of realizing the neural network to obtain the policy network model, and initializing coefficients theta such as weight, bias and the like of the policy network by using random numbers. The embodiment of the present invention does not limit the specific type of the policy network model, for example, the policy network model may be a fully-connected neural network. And sequentially adding a time step from the 0 th time step to the maximum time step number T, wherein the specific duration of the time step can be set by itself, for example, the time step can be set to 5 minutes, or can be set to 1 minute, and the smaller the time step is, the larger the matrix dimension required for observing the occupancy state of the parking space resource with the same time length is. And simulating the flight of each simulation time in the flight schedule samples to obtain a state sample corresponding to each simulation time. And training the strategy network model by using the state sample obtained by simulation through a certain method, continuously updating the strategy network coefficient theta, and finally obtaining the probability distribution model.
Fig. 2 is a schematic diagram of a policy network model training process according to an embodiment of the present invention.
As shown in fig. 2, includes:
step 201: initializing strategy network model training parameters, wherein the parameters comprise flight schedule sample number J, training iteration round number I, parallel simulation plot number K of each flight schedule sample in each round of training, and simulation maximum time step number T of each plot;
step 202: initializing information of the stand, such as the number and the position of the stands, basic information of airplane types allowed to stop at the stands and the like;
step 203: reading stand allocation rules, e.g., other more complex rules such as which airline flights each stand allows to stop on;
step 204: initializing an environment state matrix S of the flight according to the initialized stop position information and the stop position distribution principletThe size of (d);
step 205: reading a flight schedule sample;
step 206: judging whether the number of flight schedule samples is larger than or equal to J or not; if the number of the acquired flight schedule samples is judged to be larger than or equal to J, executing step 209; if the number of the flight schedule samples is judged to be less than J, executing step 207;
step 207: randomly selecting a flight schedule from the flight schedule samples;
step 208: superposing a random fluctuation moment for each flight in the selected flight schedule to generate a new flight schedule, adding the new flight schedule into a flight schedule sample, and executing step 206;
step 209: generating a rule constraint vector for each flight in the flight schedule sample according to the flight information, the stop information and the distribution rule information;
step 210: setting structural parameters of a strategy network model, calling a Theano deep learning library or other libraries capable of realizing a neural network to obtain the strategy network model, and initializing coefficients theta such as weight, bias and the like of the strategy network by using random numbers;
step 211: initializing a strategy network training loop variable, i is 1, j is 1, and k is 1;
step 212: starting the strategy network model training of the ith round;
step 213: selecting a jth flight schedule sample for training;
step 214: performing k episode simulations according to the selected flight schedule sample, wherein each episode simulation starts from the 0 th time step and sequentially adds 1 time step to obtain a corresponding state sample, and the episode simulation mode can be sequential simulation or parallel simulation;
step 215: updating the policy network coefficient theta by adopting the state sample;
step 216: judging whether J is smaller than J, if J is judged to be smaller than J, J is J +1, executing step 213, and if J is judged to be not smaller than J, executing step 217;
step 217: judging whether I is smaller than I, if so, executing step 212, and if not, executing step 218;
step 218: and storing and updating the rough network coefficient theta after all rounds of strategy network training are finished, and obtaining a probability distribution model.
According to the embodiment of the invention, the corresponding state sample is obtained by performing plot simulation on the flight schedule sample, and the strategy network model is trained by using the state sample, so that the probability distribution model is finally obtained.
On the basis of the above embodiment, as an alternative embodiment, the state sample includes: an environmental state matrix sample, an action sequence sample, and an immediate reward sample; correspondingly, on the basis of the policy network model, sequentially adding 1 time step from the 0 th time step, simulating the flight at each simulation time in the flight schedule sample, and obtaining a state sample corresponding to each simulation time, including: starting from the 0 th time step, sequentially adding 1 time step, and constructing an environment state matrix sample of the flight corresponding to each simulation time; if the fact that the flight exists at the real time corresponding to the simulation time is judged and known, inputting the environment state matrix sample into the strategy network model, outputting the sample to obtain a corresponding sample probability map, and distributing the stop positions of the flight according to the sample probability map to obtain a distribution result; obtaining an action sequence sample and an immediate reward sample of the flight corresponding to the simulation moment according to the distribution result; and if the flight does not exist at the real time corresponding to the simulation time, setting the action sequence sample and the immediate reward sample of the flight corresponding to the simulation time to be 0.
Specifically, the number of flight schedule samples is J, and the number of episode simulations is K, then K episode simulations are performed on the J flight schedules, respectively, to obtain corresponding state samples. And (3) sequentially adding 1 time step from the 0 th time step to the maximum time step number T for each plot simulation of each flight schedule sample. For example, for the k-th episode simulation of the jth flight schedule, 1 time step is added in sequence from the 0 th time step to the maximum time step number T, and each time step corresponds to one simulation time. Fig. 3 is a schematic view of an episode simulation process according to an embodiment of the present invention. As shown in fig. 3, in the ith round of policy network model training, performing the kth episode simulation on the jth flight schedule includes:
step 301: setting t to be 0 and constructing an environment state matrix S corresponding to the 0 th time0 kSpecifically, according to the parking space resource occupation state of the 0 th time step, a corresponding parking state matrix A is established according to the formula (1)0 kInitializing flight state matrix B of 0 th time step by using all 0 data0 kAnd constructing an environment state matrix S corresponding to the 0 th time step according to a formula (6)0 k；
Step 302: starting the simulation of the t time step;
step 303: judging whether a flight arrives at a real time corresponding to the simulation time corresponding to the t-th time step, if so, executing a step 304, and if not, executing a step 308;
step 304: construction of St kWill St kInputting the data into a policy network model to obtain a flight v in a jth flight schedule sample1A probability map of (a);
step 305: adding security filter to obtain filtered probability map, and specifically, constructing the second one according to formula (4)Available parking space vector C for t time stepst kAccording to Ct kFinding out the unavailable parking positions of the parking positions, and setting the probability corresponding to the corresponding parking positions in the probability map as 0 to obtain a corresponding probability map;
step 306: the flights v1 are distributed according to the filtered flight probability graph, and the action sequence a of the stand is obtained according to the distribution resultt kAnd immediate reward rt kStep 309 is executed;
step 308: setting at k＝0，rt k＝0；
Step 309: store St k、at kAnd rt k；
Step 310: judging whether T is greater than T, wherein T is a preset maximum time step number, if yes, executing step 311, if T is not greater than T, and if T is greater than T, ending;
step 311: updating the parking space resource occupation state A according to the distribution result of the flight parking spacest k；
Step 312: reading information of a next different flight v2 from the flight schedule;
step 314 is based on the updated parking space resource occupancy state At kConstructing the environment state matrix S corresponding to the flight v2t kStep 302 is performed.
According to the embodiment of the invention, corresponding state samples are obtained by performing plot simulation on flight schedule samples, and the samples are provided for the training of the strategy network model.
On the basis of the above embodiment, as an optional embodiment, training the policy network model by using the state samples includes: and training the strategy network model by adopting a state sample through a gradient descent method.
Specifically, the status samples correspond to the episode simulation of the flight schedule samples, each flight schedule obtains one corresponding status sample through one episode simulation, and K corresponding status samples are obtained through K episode simulations. Similarly, a round of plot simulation of J flight schedule samples corresponds to J × K state samples, and plot simulation of I training iteration rounds is performed, so that J × K × I corresponding state samples are obtained.
Then, in a round of strategy network training, an episode simulation is performed on one flight schedule in the J flight schedule samples, and T + 1S are obtainedt、atAnd rt，t∈[0,T]. The discount sum V of the immediate award corresponding to the T-th time step in the scenario simulation is obtained by the following formula (7)tThe policy network coefficient θ is updated by substituting the equation (8). And sequentially carrying out J multiplied by K times of updating, reserving the strategy network coefficient theta updated at the last time, obtaining a corresponding probability distribution model, and taking the probability distribution model as an initial value of the strategy network coefficient in the next round of strategy network training.
Where alpha is the step size of the update coefficient theta,
is the gradient direction; gamma is the discount rate for immediate rewards, rtIs an immediate reward at time t; vtIs the discount sum of the immediate rewards at and all times after time t, and θ is the policy network coefficient.
According to the embodiment of the invention, the strategy network model is trained by adopting the state sample through a gradient descent method, so that the probability distribution model is obtained.
On the basis of the above embodiment, as an alternative embodiment, before allocating the stand of the flight according to the probability map, the method includes: constructing an available stop position vector of the flight according to the stop state matrix, the constraint vector, the estimated time of the flight entering the stop position and the estimated time of the flight exiting the stop position; and setting the corresponding probability of the unavailable stand in the probability map to be 0 according to the available stand vector.
Specifically, the shutdown status matrix A according to flight vtEstimated time of flight arrival at the station (tth after tth)1The time corresponding to each time step) and the expected time of flight leaving the stand (the tth after t)2The time corresponding to each time step), according to the formula (3), obtaining the occupancy matrix of the flight arrival and stop time
Occupancy matrix for the time when a flight enters a stop
Constraint vector P according to flight vvAnd occupancy matrix
And
according to the formula (4), obtaining the available stop position vector C of the flight vvt. Wherein available parking bit vectors if cvt,nAnd 0, the nth stand is not available. And setting the distribution probability of the aircraft stop corresponding to 0 in the available aircraft stop vector to be 0.
According to the embodiment of the invention, the available stop position vector of the flight is constructed according to the stop state matrix and the constraint vector, and the distribution probability of the stop position corresponding to 0 in the available stop position vector is set as 0, so that the flight is not distributed to the unavailable stop position, and the safety of airport stop position distribution is improved.
On the basis of the above embodiment, as an alternative embodiment, the allocating the flight stops according to the probability map includes: according to the probability map, selecting the stand corresponding to the maximum probability in the probability map to be allocated to the flight; or, according to the probability map, selecting the positions corresponding to the corresponding probabilities in the probability map to be allocated to the flights by adopting a roulette method.
Specifically, after the probability map of the flight is obtained, the system allocates the actual stand of the flight to be allocated according to the probability map. And the system selects the aircraft stop corresponding to the maximum probability value in the probability map to be allocated to the flight. For example, the maximum probability value in the probability map of the flight corresponds to n stands, and the system selects n stands as the actual stands of the flight. Or after obtaining the probability map corresponding to the flight, the system automatically removes the aircraft stop corresponding to the probability value 0, selects the remaining probability values by adopting a roulette method, and takes the aircraft stop corresponding to the selected probability value as the actual aircraft stop of the flight. Or the system firstly screens the probability which is larger than a certain value, such as 50 percent, then selects the probability value from the screened probability values by a roulette method, and takes the aircraft stop corresponding to the selected probability value as the actual aircraft stop of the flight. The manner in which the actual stops of flights are selected based on the probability map is not particularly limited herein.
The embodiment of the invention directly selects the stand corresponding to the maximum probability value in the probability map as the actual stand of the flight or selects the stand corresponding to the corresponding probability value in the probability map as the actual stand of the flight by adopting a roulette method, thereby providing various choices for the allocation mode of the stand of the airport.
Fig. 4 is a schematic structural diagram of an airport stand allocation device according to an embodiment of the present invention. As shown in fig. 4, the apparatus includes: a modeling module 401, a processing module 402 and an assignment module 403. The modeling module 401 is configured to construct an environment state matrix according to the state information of the flight to be distributed and the airport environment; the processing module 402 is configured to input the environment state matrix into a probability distribution model, and obtain a probability map output by the probability distribution model, where the probability map is used to indicate a probability of allocating a flight to each stand; the allocation module 403 is configured to allocate a flight stop for the flight according to the probability map.
Specifically, the airport state of the flight includes the airport stand resource occupation state, the stand allocation rule, the predicted time when the flight enters the stand, the predicted time when the flight exits the stand and the like. The modeling module 401 constructs an environment state matrix corresponding to the flight to be distributed according to the actual occupation state of the flight stop to be distributed, the current stop distribution rule, the estimated time of the flight entering the stop corresponding to the flight to be distributed listed in the flight schedule and the estimated time of the flight exiting the stop.
And constructing a probability distribution model, for example, training a neural network to obtain the probability distribution model. The probability distribution model is used for providing a distribution scheme of flight stops according to the environment state matrix of the flight. The processing module 402 inputs the environment state matrix of the flight into the probability distribution model, and the probability distribution model outputs a probability map corresponding to the flight. The probability map corresponding to a flight describes the probability that the flight is assigned to each stand at the airport. For example, assuming that an airport has 100 stops, the probability map for flight A bears 100 corresponding probabilities that flight A is assigned to the 100 stops.
Having derived a probability map of the assignment of flights to each stand at the airport, the assignment module 403 may select the corresponding stand from the probability map as the actual stand for the flight. For example, the probability map of flight a describes the probability that flight a is assigned to each stand, and the crew selects the stand with the highest probability as the actual stand of flight a.
According to the embodiment of the invention, the modeling module 401 constructs the airport environment state matrix according to the flight to be distributed and the state information of the airport environment, so that the modeling accuracy is improved, and the safety of the parking space distribution is improved; when the status information of the flight to be distributed and the airport environment changes, the modeling module 401 can obtain a corresponding parking space distribution scheme only by reconstructing the airport environment status matrix, and the management is easy; the processing module 402 directly inputs the environment state matrix into the probability distribution model, and real-time flight distribution is performed by using the probability distribution model, so that the operation time is saved, and the problem efficiency of airplane allocation is improved.
On the basis of the above embodiment, as an alternative embodiment, the status information of the airport environment includes: the method comprises the following steps of (1) parking space resource occupation state information, parking space distribution rules, the estimated time of a flight entering a parking space and the estimated time of the flight exiting the parking space; correspondingly, according to the status information of the flights to be distributed and the airport environment, an environment status matrix is constructed, which comprises the following steps: according to the occupation state of the parking space resources, a parking state matrix is constructed; constructing a constraint vector of the flight according to the stop allocation rule; constructing a flight state matrix according to the halt state matrix, the constraint vector, the estimated time of the flight entering the halt station and the estimated time of the flight exiting the halt station; and constructing an environment state matrix according to the shutdown state matrix and the flight state matrix.
On the basis of the above embodiment, as an alternative embodiment, before inputting the environment state matrix to the probability distribution model, the method includes: constructing a strategy network model; based on a policy network model, sequentially adding 1 time step from the 0 th time step, and performing plot simulation on flights at each simulation time in the flight schedule samples to obtain a state sample corresponding to each simulation time; and training the strategy network model by adopting the state sample, and obtaining a probability distribution model after the training is finished.
On the basis of the above embodiment, as an alternative embodiment, the state sample includes: an environmental state matrix sample, an action sequence sample, and an immediate reward sample; correspondingly, on the basis of the policy network model, sequentially adding 1 time step from the 0 th time step, simulating the flight at each simulation time in the flight schedule sample, and obtaining a state sample corresponding to each simulation time, including: starting from the 0 th time step, sequentially adding 1 time step, and constructing an environment state matrix sample of the flight corresponding to each simulation time; if the fact that the flight exists at the real time corresponding to the simulation time is judged and known, inputting the environment state matrix sample into the strategy network model, outputting the sample to obtain a corresponding sample probability map, and distributing the stop positions of the flight according to the sample probability map to obtain a distribution result; obtaining an action sequence sample and an immediate reward sample of the flight corresponding to the simulation moment according to the distribution result; and if the flight does not exist at the real time corresponding to the simulation time, setting the action sequence sample and the immediate reward sample of the flight corresponding to the simulation time to be 0.
On the basis of the above embodiment, as an optional embodiment, training the policy network model by using the state samples includes: and training the strategy network model by adopting a state sample through a gradient descent method.
On the basis of the above embodiment, as an alternative embodiment, before allocating the stand of the flight according to the probability map, the method includes: constructing an available stop position vector of the flight according to the stop state matrix, the constraint vector, the estimated time of the flight entering the stop position and the estimated time of the flight exiting the stop position; and setting the corresponding probability of the unavailable stand in the probability map to be 0 according to the available stand vector.
On the basis of the above embodiment, as an alternative embodiment, the allocating the flight stops according to the probability map includes: according to the probability map, selecting the stand corresponding to the maximum probability in the probability map to be allocated to the flight; or, according to the probability map, selecting the positions corresponding to the corresponding probabilities in the probability map to be allocated to the flights by adopting a roulette method.
Fig. 5 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention. As shown in fig. 5, the electronic device may include: a processor (processor)510, a communication Interface (Communications Interface)520, a memory (memory)530, and a bus 540, wherein the processor 510, the communication Interface 520, and the memory 530 communicate with each other via the bus 540. The communication interface 540 can be used for the transmission of information between the allocation device of the airport stand and the electronic device. Processor 510 may call logic instructions in memory 530 to perform the following method: constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment; inputting the environment state matrix into a probability distribution model, and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flights to each stand; and allocating the stand for the flight according to the probability map.
Furthermore, the logic instructions in the memory 530 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
An embodiment of the present invention provides a non-transitory computer-readable storage medium storing computer instructions, which cause a computer to execute the method for allocating airport stands provided in the above embodiment, for example, the method includes: constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment; inputting the environment state matrix into a probability distribution model, and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flights to each stand; and allocating the stand for the flight according to the probability map.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (8)
1. A method for allocating airport parking spaces is characterized by comprising the following steps:
constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment;
inputting the environment state matrix into a probability distribution model, and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flight to each stand;
allocating a stand for the flight according to the probability map;
the status information of the airport environment includes: the method comprises the following steps of (1) parking space resource occupation state information, parking space distribution rules, the estimated time of a flight entering a parking space and the estimated time of the flight exiting the parking space;
correspondingly, the constructing an environment state matrix according to the state information of the flight to be distributed and the airport environment comprises the following steps:
according to the parking space resource occupation state, a parking state matrix is constructed; constructing a constraint vector of the flight according to the stop allocation rule;
constructing a flight state matrix according to the shutdown state matrix, the constraint vector, the predicted time of the flight entering the stand and the predicted time of the flight exiting the stand;
constructing the environment state matrix according to the shutdown state matrix and the flight state matrix;
before inputting the environment state matrix to a probability distribution model, the method comprises the following steps:
constructing a strategy network model;
performing episode simulation based on the strategy network model, wherein each episode simulation starts from the 0 th time step, sequentially adds 1 time step, and simulates flights at each simulation time in flight schedule samples to obtain a state sample corresponding to each simulation time;
updating the strategy network model by adopting the state sample, and obtaining the probability distribution model after the updating is finished;
the state sample, comprising: an environmental state matrix sample, an action sequence sample, and an immediate reward sample.
2. The method of claim 1, wherein the step of simulating the flight at each simulation time in the flight schedule samples by sequentially adding 1 time step from the 0 th time step based on the policy network model to obtain the status sample corresponding to each simulation time comprises:
sequentially adding 1 time step from the 0 th time step to construct an environment state matrix sample of the flight corresponding to each simulation time;
if the fact that the flight exists at the real time corresponding to the simulation time is judged and known, inputting the environment state matrix sample to the strategy network model, outputting the environment state matrix sample to obtain a corresponding sample probability graph, and distributing the stop positions of the flight according to the sample probability graph to obtain a distribution result; obtaining an action sequence sample and an immediate reward sample of the flight corresponding to the simulation moment according to the distribution result;
and if the fact that no flight exists at the real time corresponding to the simulation time is judged and obtained, setting the action sequence sample and the immediate reward sample of the flight corresponding to the simulation time to be 0.
3. The method of assigning airport stands of claim 1, wherein said updating the policy network model with the status samples comprises: and updating the strategy network model by adopting the state sample through a gradient descent method.
4. The method of assigning airport stands according to claim 1, further comprising, prior to said assigning stands to flights according to the probability map:
constructing an available stand vector of the flight according to the stop state matrix, the constraint vector, the predicted time of the flight entering the stand and the predicted time of the flight exiting the stand;
and setting the corresponding probability of the unavailable stand in the probability map to be 0 according to the available stand vector.
5. The method of assigning airport stands of claim 1, wherein the assigning stands to flights according to the probability map comprises:
according to the probability map, selecting a stand corresponding to the maximum probability in the probability map to be allocated to the flight; alternatively, the first and second electrodes may be,
and selecting the aircraft stop corresponding to the corresponding probability in the probability map to be allocated to the flight by adopting a roulette method according to the probability map.
6. An airport stand allocation apparatus, comprising:
the modeling module is used for constructing an environment state matrix according to the state information of the flights to be distributed and the airport environment;
the processing module is used for inputting the environment state matrix into a probability distribution model and obtaining a probability map output by the probability distribution model, wherein the probability map is used for indicating the probability of distributing the flight to each stand;
the distribution module is used for distributing the stand for the flight according to the probability map;
the status information of the airport environment includes: the method comprises the following steps of (1) parking space resource occupation state information, parking space distribution rules, the estimated time of a flight entering a parking space and the estimated time of the flight exiting the parking space;
correspondingly, the constructing an environment state matrix according to the state information of the flight to be distributed and the airport environment comprises the following steps:
according to the parking space resource occupation state, a parking state matrix is constructed; constructing a constraint vector of the flight according to the stop allocation rule;
constructing a flight state matrix according to the shutdown state matrix, the constraint vector, the predicted time of the flight entering the stand and the predicted time of the flight exiting the stand;
constructing the environment state matrix according to the shutdown state matrix and the flight state matrix;
before inputting the environment state matrix to a probability distribution model, the method comprises the following steps:
constructing a strategy network model;
performing episode simulation based on the strategy network model, wherein each episode simulation starts from the 0 th time step, sequentially adds 1 time step, and simulates flights at each simulation time in flight schedule samples to obtain a state sample corresponding to each simulation time;
updating the strategy network model by adopting the state sample, and obtaining the probability distribution model after the updating is finished;
the state sample, comprising: an environmental state matrix sample, an action sequence sample, and an immediate reward sample.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the method of assigning airport stands according to any one of claims 1 to 5.
8. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for allocating airport stands of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810969491.9A CN109147396B (en) | 2018-08-23 | 2018-08-23 | Method and device for allocating airport parking positions |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810969491.9A CN109147396B (en) | 2018-08-23 | 2018-08-23 | Method and device for allocating airport parking positions |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109147396A CN109147396A (en) | 2019-01-04 |
| CN109147396B true CN109147396B (en) | 2020-10-30 |
Family
ID=64827657
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810969491.9A Active CN109147396B (en) | 2018-08-23 | 2018-08-23 | Method and device for allocating airport parking positions |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109147396B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3696094A1 (en) * | 2019-02-12 | 2020-08-19 | General Electric Company | Aircraft stand recovery optimization |
| CN109872064B (en) * | 2019-02-12 | 2021-07-30 | 民航成都信息技术有限公司 | Airport resource allocation method, device, equipment and medium |
| CN111627255B (en) * | 2019-02-27 | 2023-03-14 | 阿里巴巴集团控股有限公司 | Information processing method, device and system |
| CN109948844B (en) * | 2019-03-15 | 2022-05-20 | 民航成都信息技术有限公司 | Optimization method, device, equipment and medium for shutdown position distribution robustness |
| CN110288857B (en) * | 2019-06-26 | 2021-10-01 | 中国民航大学 | Airport parking lot fast scheduling method based on time-space dimension decomposition |
| CN110443448B (en) * | 2019-07-01 | 2022-03-29 | 华中科技大学 | Bidirectional LSTM-based airplane position classification prediction method and system |
| CN112785097A (en) * | 2019-11-04 | 2021-05-11 | 顺丰科技有限公司 | Parking space allocation method and device, storage medium and computer equipment |
| TWI724755B (en) * | 2020-01-17 | 2021-04-11 | 桃園國際機場股份有限公司 | Flight operation and transportation system |
| CN113095543B (en) * | 2021-03-01 | 2024-01-12 | 北京工业大学 | Distribution method and system for airport stand and taxiway |
| CN116993137B (en) * | 2023-09-28 | 2023-12-05 | 民航成都信息技术有限公司 | Method and device for determining stand, electronic equipment and medium |
| CN117933490A (en) * | 2024-03-14 | 2024-04-26 | 中国民航大学 | Airport scene dragging scheduling optimization method, electronic equipment and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104361769B (en) * | 2014-10-27 | 2017-07-11 | 广州市中南民航空管通信网络科技有限公司 | A kind of flight data processing method and field prison front end processor |
| CN104751681B (en) * | 2015-03-09 | 2017-05-03 | 西安理工大学 | Statistical learning model based gate position allocation method |
| CN106447215A (en) * | 2016-10-12 | 2017-02-22 | 合肥飞友网络科技有限公司 | Method for pre-distributing aircraft berths in airport |
| CN107085976B (en) * | 2017-04-21 | 2018-03-30 | 民航成都信息技术有限公司 | The time-bounded dynamic constrained method in airliner station level ground aircraft gate |
| CN107578159A (en) * | 2017-08-29 | 2018-01-12 | 飞友科技有限公司 | It is a kind of to adapt to the abnormal seat in the plane auto-allocation method of flight itineraries |
-
2018
- 2018-08-23 CN CN201810969491.9A patent/CN109147396B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109147396A (en) | 2019-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109147396B (en) | Method and device for allocating airport parking positions | |
| Rzevski et al. | Managing complexity | |
| US20190332944A1 (en) | Training Method, Apparatus, and Chip for Neural Network Model | |
| JP6122621B2 (en) | Simulation and visualization of project planning and management | |
| Dijk et al. | The recoverable robust stand allocation problem: a GRU airport case study | |
| CN112036676A (en) | Intelligent on-demand management of ride sharing in a transport system | |
| CN111192090A (en) | Seat allocation method and device for flight, storage medium and electronic equipment | |
| US8504402B1 (en) | Schedule optimization using market modeling | |
| US10719639B2 (en) | Massively accelerated Bayesian machine | |
| Tahir et al. | An improved integral column generation algorithm using machine learning for aircrew pairing | |
| Atkin et al. | A metaheuristic approach to aircraft departure scheduling at London Heathrow airport | |
| van der Weide et al. | Robust long-term aircraft heavy maintenance check scheduling optimization under uncertainty | |
| CN106779240A (en) | The Forecasting Methodology and system of civil aviaton's market macroscopic view index | |
| EP3306538A1 (en) | Total-ordering in process planning | |
| US9153138B1 (en) | Agent-based airfield conflict resolution | |
| Zhang et al. | A heuristic maintenance scheduling framework for a military aircraft fleet under limited maintenance capacities | |
| US10313457B2 (en) | Collaborative filtering in directed graph | |
| US10417651B2 (en) | Fuel consumption predictions using associative memories | |
| CN115660446A (en) | Intelligent generation method, device and system for air traffic control plan | |
| Wang | A dynamic fault tree approach for time-dependent logical modeling of autonomous flight systems | |
| Stasak et al. | Business process modeling linguistic approach–Problems of business strategy design | |
| CN113095543A (en) | Distribution method and system for airport parking space and taxiways | |
| Goggins | Stochastic modeling for airlift mobility | |
| de Freitas Cunha et al. | On the impact of MDP design for reinforcement learning agents in resource management | |
| Ukai et al. | An Aircraft Deployment Prediction Model Using Machine Learning Techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |