CN108764560B - Aircraft scene trajectory prediction method based on long-short term memory neural network - Google Patents
Aircraft scene trajectory prediction method based on long-short term memory neural network Download PDFInfo
- Publication number
- CN108764560B CN108764560B CN201810495952.3A CN201810495952A CN108764560B CN 108764560 B CN108764560 B CN 108764560B CN 201810495952 A CN201810495952 A CN 201810495952A CN 108764560 B CN108764560 B CN 108764560B
- Authority
- CN
- China
- Prior art keywords
- neural network
- track
- scene
- data
- prediction
- 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
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 16
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 10
- 238000003062 neural network model Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000001174 ascending effect Effects 0.000 claims description 2
- 230000033001 locomotion Effects 0.000 abstract description 10
- 230000000694 effects Effects 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 abstract description 4
- 230000007774 longterm Effects 0.000 abstract description 4
- 238000001514 detection method Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 8
- 230000029305 taxis Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 208000025174 PANDAS Diseases 0.000 description 1
- 208000021155 Paediatric autoimmune neuropsychiatric disorders associated with streptococcal infection Diseases 0.000 description 1
- 240000000220 Panda oleosa Species 0.000 description 1
- 235000016496 Panda oleosa Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G06Q50/40—
Abstract
The invention provides an aircraft scene track prediction method based on a long-short term memory neural network, which combines an LSTM neural network and a polynomial fitting method to implement a track prediction technology, can theoretically predict the position of any time within 60 seconds of a long period by setting an incremental sampling period, but predicts the position of an overlong period, and training data generated by preprocessing has low quality, so that the prediction precision is too low, and the method has no practical effect on the detection of scene sliding conflict, so that the prediction of the position of any time within 30 seconds of a medium and long term is relatively proper. The invention can implicitly simulate the scene motion state of the aircraft according to the context of the track sequence by means of the characteristic of historical memorability of the LSTM neural network, can be used for predicting the positions of airport taxiways and aircrafts on runways in future time periods, avoids the scene sliding conflict of the aircrafts, lays a cushion for real-time path planning and ensures the safe and efficient operation of the airports.
Description
Technical Field
The invention relates to a prediction technology of airport aircraft scene tracks.
Background
With the high-speed continuous development of the global air transportation industry, airport scene traffic is increasingly busy, particularly in the airport scene traffic control process, resources such as scene space, time, manpower and the like cannot be fully utilized, the potential safety hazard of taxi conflict still exists in the initially planned aircraft taxi path and the operation process, and the operation safety and efficiency of the airport scene are directly influenced. The geographical position of the aircraft in the future time period can be judged by utilizing a track prediction technology, the time for the aircraft to reach a key intersection and the leaving time can also be predicted, the traffic jam on the scene is reduced, the sliding conflict is avoided, the total sliding time of the aircraft is shortened, the high-speed operation of an airport is ensured, and the service quality is improved.
At present, the traditional trajectory prediction method based on a dynamics model, a Kalman filtering algorithm, a hidden Markov model and the like depends on aircraft performance parameters which are difficult to obtain in the actual motion process and is influenced by scene control, the operation intention of an aircraft driver and weather, and the accuracy of the trajectory prediction is influenced by the established aircraft dynamics model. In addition, the multidimensional state vector space in the historical data set has large calculation amount, cannot store excessive historical information, is not suitable for medium and long term prediction, and influences the real-time performance of the track prediction.
Because of strong correlation and dependence between the position sequences of the aircraft taxiing motion, implicit relation between the position sequences and the characteristic sequence value can be analyzed by means of a time sequence method, and the position of the future moment can be predicted according to the historical position sequences of the aircraft taxiing motion. And Long-Short Term Memory neural Networks (LSTM) are important instruments for processing time series problems in the field of artificial intelligence. However, the LSTM can only predict the position at some discrete time in the future, which is determined by the sampling period of the training data, and cannot predict the position at any time in the middle or long term in the future.
Disclosure of Invention
The invention provides an aircraft scene trajectory prediction method based on an LSTM neural network with an attenuation window, which can realize medium and long term prediction.
The technical scheme adopted by the invention for solving the technical problems is that the aircraft scene track prediction method based on the long-short term memory neural network comprises the following steps:
step 1, acquiring a historical taxi data set of the aircraft, wherein the historical taxi data set comprises longitude, latitude and speed data, and setting an incremental sampling period to preprocess a taxi data sequence and divide the taxi data sequence into training data and test data;
step 3, inputting test data to an LSTM neural network model under different sampling periods to obtain predicted values, after reverse normalization, adding a basic item during first-order difference processing to obtain a track predicted position, wherein the track predicted position consists of longitude and latitude;
and 4, forming a track prediction sequence by the track prediction positions under different sampling periods according to the ascending sequence of the sampling periods, calculating the relative distance of the scene (azimuth angle can be further calculated for the curve sliding track prediction) with the first track point in the track prediction sequence as a base point and the other track points respectively, processing the relative distance of the scene and time by adopting a polynomial fitting method, obtaining a track prediction curve fitting equation, and obtaining the relative distance of the scene at any time in medium and long periods.
The LSTM neural network is adopted alone to predict the positions of some discrete moments, and the discrete moments are determined by the sampling period of the training data; and the position at any time in the future can be predicted by adopting a polynomial fitting method alone, but the error is too large. According to the invention, an LSTM neural network and a polynomial fitting method are combined to implement a trajectory prediction technology, and the position of any time within 60 seconds of a long period can be theoretically predicted by setting an incremental sampling period, but the position of an overlong period is predicted, the quality of training data generated by preprocessing is low, so that the prediction precision is too low, and the method has no practical effect on the detection of scene sliding conflict, so that the position of any time within 30 seconds of a medium-long period is relatively suitable for prediction.
The method has the advantages that by means of the characteristic of historical memorability of the LSTM neural network, the scene motion state of the aircraft can be simulated implicitly according to the context of the track sequence, the method can be used for predicting the positions of aircraft on taxiways and runways of airports in future time periods, avoiding the scene sliding conflict of the aircraft, laying a cushion for real-time path planning and ensuring safe and efficient operation of the airports.
Drawings
FIG. 1 is a flow chart of an embodiment.
Fig. 2 is a flowchart of a supervised learning sequence conversion operation.
Fig. 3 is a schematic diagram of changes in the motion state of an aircraft scene.
Fig. 4 is a polynomial fit graph.
Detailed Description
As shown in fig. 1, a certain airport in China is taken as an example, and the implementation process mainly comprises the following steps:
step 1, acquiring a historical taxis data set (longitude, latitude and speed) of the aircraft, carrying out pretreatment of equidistant sampling, first-order difference, normalization and supervised learning sequence conversion on a taxis data sequence, and dividing the taxis data sequence into training data and test data:
step 1.1, selecting historical track data on a certain straight taxiway according to historical sliding data, and utilizing the following formula:
Tk＝k·T，k∈N* (1)
where k is the sampling factor, N*And T is the period of the scene monitoring radar for acquiring the track data. Equidistant sampling yields n sequences of trajectory points, denoted as:
[(x1,y1,v1),(x2,y2,v2),…,(xn,yn,vn)]
(xt,yt,vt) Respectively representing longitude, latitude and speed of a track point at the t-th moment, wherein t is 1,2, …, n;
performing difference on basic items of adjacent moments in the track point sequence, namely performing first-order difference processing;
x′t＝xt+1-xt,y′t＝yt+1-yt,v′t＝vt+1-vt (2)
new n-1 sequences of trajectory points are obtained, denoted as:
[(x′1,y′1,v′1),(x'2,y'2,v'2),…,(x'n-1,y'n-1,v'n-1)]
step 1.2, using sklern tool in machine learning library, normalizing track point Sequence after first order difference to scale to range of [ -1,1] by using minmaxscale () function, and storing output parameter scaler of function, then converting normalized data into supervised learning Sequence by using pandas tool in machine learning library, the operation flow is shown in fig. 2, n _ vars represents characteristic number of Input data, n _ in represents number of Input Sequence, n _ out represents number of predicted Sequence, using DataFrame () function to initialize two data objects of Input Sequence and Forecast Sequence, Shift () function respectively circularly constructs Input Sequence data and predicted Sequence label data, and using Concat () function to splice final supervised learning track point Sequence data, namely training data set, the sample format of which is shown in table 1.
TABLE 1 supervised learning sequence data Format
| xt-3 | yt-3 | vt-3 | xt-2 | yt-2 | vt-2 | xt-1 | yt-1 | vt-1 | xt | yt | vt | xt+1 | yt+1 | vt+1 |
| x1 | y1 | v1 | x2 | y2 | v2 | x3 | y3 | v3 | x4 | y4 | v4 | x5 | y5 | v5 |
| x2 | y2 | v2 | x3 | y3 | v3 | x4 | y4 | v4 | x5 | y5 | v5 | x6 | y6 | v6 |
| x3 | y3 | v3 | x4 | y4 | v4 | x5 | y5 | v5 | x6 | y6 | v6 | x7 | y7 | v7 |
| x4 | y4 | v4 | x5 | y5 | v5 | x6 | y6 | v6 | x7 | y7 | v7 | x8 | y8 | v8 |
The data set is divided into a training data set and a testing data set according to the proportion of 80% and 20%, the training data set is used for training the model, and the testing data set is used for evaluating the model.
step 2.1 introduce long-short term memory neural network LSTM, using input gate itForgetting door ftAnd an output gate otThe method controls the information addition, the information discard and the information forgetting, and has the following specific functions:
input door it: the extent to which the control information is added to the storage unit;
forget door ft: controlling the output of the storage unit at the previous moment and the input of the current moment, and transmitting the output and the input of the storage unit at the current moment to the information discarding degree in the storage unit at the current moment;
output gate ot: controlling the information in the storage unit at the current moment to be transmitted to the current hidden state htThe degree of information discard.
At time t, the parameter update relationship of forward propagation and backward propagation is as follows, wherein the symbol is defined as shown in table 2.
it＝σ(Wi·[ht-1,Xt]+bi)
ft＝σ(Wf·[ht-1,Xt]+bf) (3)
ot＝σ(Wo·[ht-1,Xt]+bo)
ct＝ft*ct-1+it*tanh(Wc·[ht-1,Xt]+bc) (4)
ht＝ot*tanh(ct)
TABLE 2 LSTM network architecture basic parameters
Step 2.2 analysis of the variation of the state of motion of the aircraft on the scene, as shown in FIG. 3, according to the memorable nature of the LSTM neural network, for example, during the time period t2～t3In case of internal position prediction, t1～t2The data flow generated in the time period can cause a false image disturbance of an accelerated motion state to the motion state at the current moment, so that the prediction error is increased, and the precision of model training is influenced; considering the introduction of an attenuation memory mechanism, the influence of historical data on the motion state of the current moment is weakened on the data flow in the hidden layer of the LSTM neural network according to exponential attenuation sliding along a memory window of a training set, specifically, an attenuation coefficient lambda epsilon (0, 1) is introduced into the hidden layer, the value is suggested to be 0.9, and the attenuation coefficient lambda serves as a unit state in the hidden layer:
where t represents the most recent data time within the decaying memory sliding window, ctIndicating the cell state at the current time t and W the length of the decaying memory sliding window.
Step 2.3, building an LSTM training model by adopting a deep learning framework Keras, inputting training data, retaining an iterative weight matrix in the training process by using a ModelCheckpoint () function, and configuring network parameters, wherein the parameters are shown in a table 3:
TABLE 3 LSTM network parameter configuration
Step 3, inputting test data to an LSTM neural network model under different sampling periods to obtain a predicted value, after inverse normalization, adding a basic item during first-order difference processing to obtain a track predicted position, wherein the track predicted position consists of longitude and latitude:
step 3.1, inputting the test data into the training model to obtain a predicted value
Obtained after reverse normalization with a stored scaler
Adding the basic term (x) in the first order difference processingl,yl) Obtaining a predicted position:
and circularly executing the steps 2.3 and 3.1 until a prediction period T is obtained through a Root Mean Square Error (RMSE) evaluation model1Then, all the predicted positions reaching the root mean square error threshold are averaged to obtain the predicted point
Step 3.2, the sampling period is updated in an increasing mode, the steps 1,2 and 3.1 are repeated, and different sampling periods T are obtainedkLower track prediction point
k is 1,2, …, P represents the total number of sampling periods set.
step 4.1 obtaining a trajectory prediction sequence under an incremental sampling period according to step 3
Selecting a sampling period T1Track sequence points of
As a base point, and calculating the track sequence point under other sampling periods by using the following formula
Scene relative distance s, k between longitude and latitude is 2,3, …, P:
where d is the median of the data calculation, r is 6371km, and is the radius of the earth, and the experimental data are shown in table 4.
TABLE 4 relative distance of scene at different times
Step 4.2, processing the functional relation between the relative distance s and the time t of the scene by adopting a polynomial fitting method, and solving the following m-degree polynomial:
sm(t)＝a0+a1t+…+amtm (8)
wherein, aiAnd i ∈ m is a polynomial coefficient, and the table 4 data is subjected to polynomial solution by using a sklern tool library in python to obtain a cubic, quartic and quintic multi-form fitting curve, which is shown in fig. 4. The figure shows that the fitting effect of the fifth power is better, and the polynomial function is solved as follows:
s(t)＝0.00401×t5-0.172×t4+2.53×t3-14.2×t2+34.6×t-15.6 (9)
the relative distance of the field aircraft at any time t in the linear taxiway can be obtained through the formula, and since the example is linear taxiway track prediction, the azimuth angle does not need to be calculated; if the curve sliding track is predicted, the relative distance and the azimuth angle of the scene can be simultaneously calculated according to the longitude and the latitude of the two track prediction sequence points, and the scene sliding conflict can be more favorably detected.
Theoretically, the position of any time within 60 seconds of a long period can be predicted by setting the incremental sampling factor k, but the position of the excessively long period is predicted, the training data generated by preprocessing has low quality, so that the prediction precision is too low, and the detection of scene sliding conflict has no practical effect, so that the position of any time within a medium-long period (within 30 seconds) is relatively suitable for prediction.
Claims (5)
1. The aircraft scene trajectory prediction method based on the long-short term memory neural network is characterized by comprising the following steps of:
step 1, acquiring a historical taxi data set of the aircraft, wherein the historical taxi data set comprises longitude, latitude and speed data, and setting an incremental sampling period to preprocess a taxi data sequence and divide the taxi data sequence into training data and test data;
step 2, constructing a long-short term memory neural network LSTM neural network model, inputting training data, configuring network parameters and finishing the training of the model;
step 3, inputting test data to an LSTM neural network model under different sampling periods to obtain predicted values, after reverse normalization, adding a basic item during first-order difference processing to obtain a track predicted position, wherein the track predicted position consists of longitude and latitude;
and 4, forming a track prediction sequence by the track prediction positions under different sampling periods according to the ascending sequence of the sampling periods, taking the first track point in the track prediction sequence as a base point, respectively calculating with the other track points to obtain the relative distance of the scene, processing the relative distance of the scene and the time by adopting a polynomial fitting method to obtain a track prediction curve fitting equation, and obtaining the relative distance of the scene at any time in the medium-long term.
2. The method of claim 1, wherein an attenuation coefficient λ e (0, 1) is introduced into the hidden layer of the LSTM neural network model, and the unit states acting in the hidden layer:
where t represents the most recent data time within the sliding window, ctIndicating the state of the cell at the current time t, and W indicating the slip adjusted according to the changing characteristics of the trajectory dataThe length of the moving window.
3. The method of claim 2, wherein λ attenuation coefficient is 0.9 and W is 10.
4. The method of claim 1, wherein the preprocessing in step 1 includes equidistant sampling, first order difference, normalization and supervised learning sequence conversion;
and 3, inputting the test data to the LSTM neural network model to obtain a predicted value, performing inverse normalization on the predicted value, and adding a basic item during first-order difference processing to obtain a track predicted position.
5. The method of claim 1, wherein the step 4 polynomial fit is a quintic fit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810495952.3A CN108764560B (en) | 2018-05-22 | 2018-05-22 | Aircraft scene trajectory prediction method based on long-short term memory neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810495952.3A CN108764560B (en) | 2018-05-22 | 2018-05-22 | Aircraft scene trajectory prediction method based on long-short term memory neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764560A CN108764560A (en) | 2018-11-06 |
| CN108764560B true CN108764560B (en) | 2022-03-15 |
Family
ID=64007537
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810495952.3A Active CN108764560B (en) | 2018-05-22 | 2018-05-22 | Aircraft scene trajectory prediction method based on long-short term memory neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764560B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109814101B (en) * | 2019-01-25 | 2020-11-20 | 北京邮电大学 | Method and device for predicting position of aircraft |
| CN109769285B (en) * | 2019-02-18 | 2021-05-28 | 西安交通大学 | Routing method for communication between vehicles based on position prediction |
| CN110232169B (en) * | 2019-05-09 | 2022-01-04 | 北京航空航天大学 | Track denoising method based on bidirectional long-time and short-time memory model and Kalman filtering |
| CN110334167B (en) * | 2019-06-26 | 2023-06-09 | 浪潮软件集团有限公司 | Position deviation early warning method and device based on neural network track |
| CN110443448B (en) * | 2019-07-01 | 2022-03-29 | 华中科技大学 | Bidirectional LSTM-based airplane position classification prediction method and system |
| CN110471055B (en) * | 2019-07-08 | 2021-03-30 | 岭澳核电有限公司 | Method and device for predicting flight path, readable storage medium and terminal equipment |
| CN111310965A (en) * | 2019-10-30 | 2020-06-19 | 北京航空航天大学 | Aircraft track prediction method based on LSTM network |
| CN110826788A (en) * | 2019-10-30 | 2020-02-21 | 南京智慧航空研究院有限公司 | Airport scene variable slide-out time prediction method based on big data deep learning |
| CN111027692A (en) * | 2019-11-04 | 2020-04-17 | 中国电子科技集团公司第十四研究所 | Target motion situation prediction method and device |
| CN110795522B (en) * | 2019-11-06 | 2023-01-06 | 中国人民解放军战略支援部队信息工程大学 | Method and device for predicting track position of mobile user |
| CN112329974B (en) * | 2020-09-03 | 2024-02-27 | 中国人民公安大学 | LSTM-RNN-based civil aviation security event behavior subject identification and prediction method and system |
| CN111965639B (en) * | 2020-09-09 | 2024-04-09 | 扬州船用电子仪器研究所(中国船舶重工集团公司第七二三研究所) | Radar signal analysis method based on bidirectional long-short-term memory neural network |
| CN112327903B (en) * | 2020-09-15 | 2021-09-17 | 南京航空航天大学 | Aircraft trajectory generation method based on deep mixed density network |
| CN114187783B (en) * | 2021-12-06 | 2023-10-31 | 中国民航大学 | Method for analyzing and predicting potential conflict in airport flight area |
| CN114298196B (en) * | 2021-12-22 | 2022-06-17 | 中国民航科学技术研究院 | Automatic judgment method for operation modes of multiple runways of airport |
| CN117076839B (en) * | 2023-10-17 | 2023-12-26 | 中国民用航空总局第二研究所 | Airport aircraft track dynamic prediction method based on dual incremental neural network |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060155432A1 (en) * | 2005-01-07 | 2006-07-13 | United Technologies Corporation | Methods and systems for monitoring atmospheric conditions, predicting turbulent atmospheric conditions and optimizing flight paths of aircraft |
| CN106571068A (en) * | 2015-01-07 | 2017-04-19 | 江苏理工学院 | Method for predicting trajectory of aircraft based on 4D flight path operation |
| WO2016156236A1 (en) * | 2015-03-31 | 2016-10-06 | Sony Corporation | Method and electronic device |
| CN106096767A (en) * | 2016-06-07 | 2016-11-09 | 中国科学院自动化研究所 | A kind of link travel time prediction method based on LSTM |
| CN107492113B (en) * | 2017-06-01 | 2019-11-05 | 南京行者易智能交通科技有限公司 | A kind of moving object in video sequences position prediction model training method, position predicting method and trajectory predictions method |
-
2018
- 2018-05-22 CN CN201810495952.3A patent/CN108764560B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764560A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764560B (en) | Aircraft scene trajectory prediction method based on long-short term memory neural network | |
| CN114048889B (en) | Aircraft trajectory prediction method based on long-term and short-term memory network | |
| KR20210036845A (en) | Semi-supervised variational autoencoder for indoor localization | |
| CN112862171B (en) | Flight arrival time prediction method based on space-time neural network | |
| Zhao et al. | Entrotaxis-Jump as a hybrid search algorithm for seeking an unknown emission source in a large-scale area with road network constraint | |
| JP6663606B2 (en) | Unmanned aerial vehicle position estimation method and system | |
| Candela et al. | Planetary robotic exploration driven by science hypotheses for geologic mapping | |
| CN115564114A (en) | Short-term prediction method and system for airspace carbon emission based on graph neural network | |
| Yang et al. | LSTM-based deep learning model for civil aircraft position and attitude prediction approach | |
| CN114842681A (en) | Airport scene flight path prediction method based on multi-head attention mechanism | |
| CN116468186A (en) | Flight delay time prediction method, electronic equipment and storage medium | |
| Bastas et al. | Data driven aircraft trajectory prediction with deep imitation learning | |
| Han et al. | Short-term 4D trajectory prediction based on LSTM neural network | |
| CN117217089A (en) | Track prediction method based on two-way long-short-term memory network | |
| Choudhury et al. | The planner ensemble: Motion planning by executing diverse algorithms | |
| CN113076686A (en) | Aircraft trajectory prediction method based on social long-term and short-term memory network | |
| US11487301B2 (en) | Method and device for generating an optimum vertical trajectory intended to be followed by an aircraft | |
| CN110779526B (en) | Path planning method, device and storage medium | |
| Graas et al. | Quantifying accuracy and uncertainty in data-driven flight trajectory predictions with gaussian process regression | |
| Neloy et al. | Alpha-N-V2: Shortest path finder automated delivery robot with obstacle detection and avoiding system | |
| CN115630566A (en) | Data assimilation method and system based on deep learning and dynamic constraint | |
| He et al. | Research on an Obstacle Avoidance Method for UAV | |
| CN112766300A (en) | Aviation big data preprocessing technology | |
| Li et al. | Exploiting Unlabeled RSSI Fingerprints in Multi-Building and Multi-Floor Indoor Localization through Deep Semi-Supervised Learning Based on Mean Teacher | |
| CN115826013B (en) | Beidou satellite positioning method based on light reinforcement learning under urban multipath environment |
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 |