CN115755134A - Vehicle positioning method and device based on Informer network and computer - Google Patents
Vehicle positioning method and device based on Informer network and computer Download PDFInfo
- Publication number
- CN115755134A CN115755134A CN202211296349.5A CN202211296349A CN115755134A CN 115755134 A CN115755134 A CN 115755134A CN 202211296349 A CN202211296349 A CN 202211296349A CN 115755134 A CN115755134 A CN 115755134A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- data
- informer
- traffic environment
- probability distribution
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000013528 artificial neural network Methods 0.000 claims abstract description 29
- 239000000203 mixture Substances 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000004458 analytical method Methods 0.000 claims abstract description 19
- 238000003062 neural network model Methods 0.000 claims abstract description 17
- 230000002159 abnormal effect Effects 0.000 claims abstract description 15
- 238000012163 sequencing technique Methods 0.000 claims abstract description 8
- 238000009826 distribution Methods 0.000 claims description 69
- 230000006870 function Effects 0.000 claims description 42
- 239000013598 vector Substances 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 description 11
- 238000004821 distillation Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention relates to a vehicle positioning method, a device and a computer based on an Informer network, wherein the method comprises the steps of judging whether the satellite positioning function of a vehicle is normal or not; clustering the traffic environment image data of the vehicle by using a Gaussian mixture model to obtain a plurality of traffic environment clustering data sets; sequencing the motion characteristic data of the vehicles under each type of traffic condition according to the running time sequence of the vehicles to obtain a motion characteristic time sequence data set; respectively training an Informer neural network model by utilizing motion characteristic time sequence data sets corresponding to various traffic conditions; when the satellite positioning function of the vehicle is abnormal, determining an Informer neural network prediction model by using a bulldozer distance analysis method; predicting the driving track of the vehicle by using an Informer neural network prediction model; the invention realizes the prediction of the vehicle running track by using the Informer network, and effectively solves the problem of accurately predicting the vehicle position under the condition of long-time failure of the GPS.
Description
Technical Field
The invention relates to the technical field of vehicle navigation positioning, in particular to a vehicle positioning method, a vehicle positioning device and a computer based on an Informer network.
Background
Vehicle navigation positioning technology as one of the key technologies of intelligent transportation systems, many Intelligent Transportation Systems (ITS) applications and Location Based Services (LBS) require the use of information about the location of a vehicle. Global Positioning System (GPS) kinematic relative positioning technology is widely used as a high-precision positioning method for vehicle position location. However, in the traditional GPS pure motion relative positioning technology, in some constrained urban road environments, such as a section of a tall building in forest, a section of a dense forest surrounding, a tunnel, and the like, when a satellite signal is blocked or the signal is unstable, the positioning capability cannot be accurately positioned or even lost, and reliable and accurate positioning cannot be provided for a vehicle, so that the vehicle track is lost.
Disclosure of Invention
In order to solve the technical problems that reliable positioning cannot be provided when satellite signals are blocked or unstable in the vehicle navigation positioning technology, the invention provides a vehicle positioning method, a vehicle positioning device and a computer based on an inform network.
The technical scheme for solving the technical problems is as follows:
judging whether the satellite positioning function of the vehicle is normal or not;
when the satellite positioning function of the vehicle is normal, clustering the traffic environment image data of the vehicle by using a Gaussian mixture model to obtain a multi-class traffic environment clustering data set; wherein each type of traffic environment clustering data set corresponds to one type of traffic condition;
sequencing a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of motion characteristic time sequence data sets; wherein each type of motion characteristic time series data set corresponds to one type of the traffic condition;
respectively training an Informer neural network model by utilizing the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models;
when the satellite positioning function of the vehicle is abnormal, comparing the current traffic environment image data of the vehicle with the traffic environment image data in the multi-class traffic environment clustering data sets by using a bulldozer distance analysis method, determining the current traffic condition class of the vehicle, and selecting a corresponding inform neural network prediction model according to the determined current traffic condition class of the vehicle;
and predicting the running track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model.
The beneficial effects of the invention are: the vehicle driving track prediction method based on the Informer network realizes the prediction of the vehicle driving track by using the Informer network, and effectively solves the problem of accurately predicting the vehicle position under the condition that the satellite positioning function is invalid for a long time. In the method for judging the traffic condition by combining the Gaussian mixture model and the KL divergence, a bulldozer distance analysis method is used for replacing the KL divergence, the problem that the KL divergence is asymmetric is solved, and the judgment effect is better than that of the KL divergence when measurement is not distributed in an overlapping mode. The driving track of the vehicle is predicted by applying the Informer network model, and a self-attention distillation technology, a probability sparse self-attention mechanism and a generative decoder of the Informer network model are used as core technologies of a basic network, so that the training speed and the reasoning speed are improved, the memory overhead of the network is reduced, and the prediction precision is improved.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, the method for clustering the traffic environment image data of the vehicle by using the Gaussian mixture model to obtain a multi-class traffic environment clustering data set comprises the following steps:
selecting a feature vector in the traffic environment image data;
calculating probability distribution of the feature vectors in the traffic environment image data of the vehicle by using a Gaussian mixture model;
and clustering all the traffic environment image data according to the corresponding probability distribution to obtain a plurality of types of traffic environment clustering data sets.
Further, calculating the probability distribution of the feature vector in the traffic environment image data of the vehicle by using a Gaussian mixture model, comprising the following steps:
using formulas
Calculating a probability distribution in traffic environment image data of the vehicle; wherein y represents the feature vector, p (y) represents the probability distribution, and N (y | μ |) k ,∑ k ) Represents the kth component, π, of the Gaussian mixture model k Representing a weight of each component of the gaussian mixture model.
Further, classifying all the traffic environment clustering data sets by using a bulldozer distance analysis method to obtain multiple types of traffic environment clustering data sets, and the method comprises the following steps:
comparing the current traffic environment image data of the vehicle with the traffic environment image data in the traffic environment cluster data sets by using a bulldozer distance analysis method to determine the current traffic condition category of the vehicle, comprising the following steps:
determining the current traffic condition category of the vehicle by utilizing a bulldozer distance analysis method according to the distance value between the target probability distribution and the pre-stored probability distribution; the target probability distribution is the probability distribution of the characteristic vectors in the current traffic environment image data of the vehicle, and the pre-stored probability distribution is the probability distribution of the characteristic vectors in the traffic environment image data in each type of the traffic environment cluster data sets.
Further, the method for determining the current traffic condition category of the vehicle by utilizing the bulldozer distance analysis method according to the distance value between the target probability distribution and the pre-stored probability distribution comprises the following steps:
calculating a distance value between the target probability distribution and a pre-stored probability distribution by using a bulldozer distance calculation formula; the calculation formula of the bulldozer distance is as follows,
will be minimum
The corresponding current traffic condition is classified as
Class;
and representing the traffic condition type corresponding to the prestored traffic environment image data of the vehicle.
The beneficial effect of adopting above-mentioned further technical scheme is that, use the KL divergence of modified bull-dozer distance analysis method to replace, solved the asymmetric problem of KL divergence, and when measuring no overlapping distribution, judge that the effect is superior to KL divergence, compare general bull-dozer distance and can shorten the solution time under the condition that improves and solve the precision.
Further, sorting a plurality of motion characteristic data of the vehicle corresponding to each type of the traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of motion characteristic time sequence data sets, comprising the following steps:
sorting a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of initial time sequence data sets;
and removing the empty data and the abnormal data in each type of the initial time sequence data set to obtain a plurality of types of the motion characteristic time sequence data sets.
Further, training an Informer neural network model by utilizing the motion characteristic time sequence data in each type of motion characteristic time sequence data set respectively to obtain a plurality of Informer neural network prediction models, and the method comprises the following steps:
performing data normalization processing on all the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of normalized data sets; wherein each of the normalized data sets corresponds to a type of the traffic condition.
Respectively training the Informer neural network models by using data in the plurality of normalized data sets to obtain a plurality of Informer neural network prediction models;
and verifying and/or testing the Informmer neural network prediction model by using partial data of the normalized data set.
Further, the motion characteristic data of the vehicle includes a travel time stamp, a heading angle, a pitch angle, a latitude, a longitude, an altitude, and a travel speed of the vehicle.
In order to solve the technical problem, the invention also provides a vehicle positioning device based on the Informer network, which has the following specific technical scheme:
an Informer network-based vehicle locating device, comprising:
the function judging module is used for judging whether the satellite positioning function of the vehicle is normal or not;
the model training module is used for clustering the traffic environment image data of the vehicle by using a Gaussian mixture model when the satellite positioning function of the vehicle is normal to obtain a multi-class traffic environment clustering data set; sequencing a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of motion characteristic time sequence data sets; respectively training an Informer neural network model by utilizing the motion characteristic time sequence data in each motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models; wherein each type of traffic environment clustering data set corresponds to a type of traffic condition;
the track judgment module is used for comparing the current traffic environment image data of the vehicle with the traffic environment image data in the traffic environment cluster data sets by utilizing a bulldozer distance analysis method when the satellite positioning function of the vehicle is abnormal, determining the current traffic condition category of the vehicle, and selecting a corresponding Informer neural network prediction model according to the determined current traffic condition category of the vehicle; and predicting the running track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model.
In order to solve the technical problem, the invention also provides a vehicle positioning device based on the Informer network, which has the following specific technical scheme:
a computer comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the vehicle positioning method based on the Informer network when executing the computer program.
Drawings
FIG. 1 is a block diagram of a vehicle positioning method based on an Informer network according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, the present embodiment provides a vehicle positioning method based on an Informer network, including the following steps:
s1, judging whether the satellite positioning function of a vehicle is normal or not; the determination of the satellite positioning function may be, among other things, determining whether a GPS signal is available.
S2, when the satellite positioning function of the vehicle is normal, acquiring traffic environment image data of the vehicle and motion characteristic data of the vehicle in real time; clustering the traffic environment image data of the vehicle by using a Gaussian mixture model to obtain a multi-class traffic environment clustering data set; wherein each type of traffic environment clustering data set corresponds to a type of traffic condition; the motion characteristic data of the vehicle includes at least a travel time stamp t, a heading angle ψ, a pitch angle θ, a latitude B, a longitude L, an altitude H, and a running speed v of the vehicle.
The method comprises the following steps of clustering traffic environment image data of a vehicle by using a Gaussian mixture model to obtain a multi-class traffic environment clustering data set, wherein the method comprises the following steps:
s20, selecting a characteristic vector in the traffic environment image data of the vehicle;
s21, calculating probability distribution of the feature vectors in the traffic environment image data of the vehicle by using a Gaussian mixture model; the method comprises the following specific steps:
calculating traffic environment image data of the feature vector in the vehicle by using a Gaussian mixture model
A probability distribution in the image data; wherein y represents the feature vector, p (y) represents the probability distribution, and N (y | μ |) k ,∑ k ) Represents the kth component, π, of the Gaussian mixture model k Representing a weight of each component of the gaussian mixture model.
And S22, clustering the traffic environment image data of all the vehicles according to the corresponding probability distribution to obtain a plurality of types of traffic environment clustering data sets.
And S3, sequencing the multiple motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain multiple types of motion characteristic time sequence data sets. The method comprises the following specific steps:
s30, performing data normalization processing on all the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of normalized data sets; wherein each of the normalized data sets corresponds to a type of the traffic condition.
The specific calculation formula for the normalization processing of the data is as follows:
wherein μ represents a mean value corresponding to each data in the time-series data set, σ represents a standard deviation corresponding to each data in the time-series data set, x represents each data in the time-series data set, and x represents a standard deviation corresponding to each data in the time-series data set * Representing the normalized numerical value of each data in the time sequence data set; the normalized values are included in the normalized data set.
S31, respectively training the Informer neural network model by using data in the plurality of normalized data sets to obtain a plurality of Informer neural network prediction models; specifically, 80% of data in the time sequence data set after normalization processing are respectively selected as training data to train the Informer neural network model, and a plurality of Informer neural network prediction models are obtained.
S32, verifying and/or testing the Informmer neural network prediction model by using partial data of the normalized data set. And selecting 10% of data in the time sequence data set after normalization as verification data, selecting the remaining 10% of data in the time sequence data set after normalization as test data, wherein the verification data is used for verifying the trained inform neural network model, and the test data is used for testing the test accuracy of the trained inform neural network model.
S4, respectively training an Informer neural network model by utilizing the motion characteristic time sequence data in each motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models;
after the data is processed in the last step, an input sequence x is set 1 ,x 1 ,...,x T The sequence is multidimensional data comprising a driving time stamp t, a heading angle psi, a pitch angle theta, a latitude B, a longitude L, an altitude H and a driving speed v of the vehicle, and the initial parameters are set as follows: encoder input with dimension 7, encoder input dimension 7, output dimension 7, model size 512, encoder layer number 2, decoder layer number 1, activation function gelu function,The learning rate was 0.0001 and the batch size of the input training data was 32, the model training process was as follows:
model input by filtered smoothed feature scalar
The local timestamp PE and the global timestamp SE; the conversion formula is:
in the formula: i ∈ { 1., L ∈ x α is a factor that balances the size between scalar mapping and local/global embedding.
In a formula corresponding to a feature scalar
A specific operation is to convert i-dimension to 512-dimension vectors by Conv 1D. The local timestamp adopts Positional embedding in a Transformer, and the calculation formula is as follows:
wherein d is model For the feature dimension of the input, the global timestamp uses a fully connected layer to map the input timestamp to 512-dimensional Embedding.
The specific method for generating the encoder comprises the following steps:
unifying the converted inputs
Inputting the data into an Encoder Encoder part of a model, firstly carrying out sparsity self-attention calculation on the data in an attention module, wherein each Key only concerns u main queries, Q is a Query vector, K is a Key vector and V is a value vectorThe calculation formula is as follows:
wherein,
is a sparse matrix of the same size as Q and which contains only the sparse metric M (Q) i And K) Top-u Query. One sampling factor c is added, setting u = clnLq. First, c × lnL keys are randomly sampled for each Query, and a sparsity score M (q) of each Query is calculated i ,K)。Q i ,K i ,V i I-th row of Q, K, V, respectively, d is Q i Of dimension (c), and L k ＝L q = L, sparsity metric M (q) i And K) is as follows:
and then, selecting N Query with the highest sparsity score, wherein the N Query is defaulted to c × lnL, only calculating dot product results of the N Query and Key, and the rest L-N queries are not calculated.
The output after sparse self-attention calculation has a redundant combination of V values, so that a distillation operation is required to give higher weight to the dominant feature having the main feature and generate a focused self-attention feature map at the next layer. In particular, the method is realized by four Convld convolutional layers and one maximum pooling layer.
After a combination of multiple sparsity self-attention layer calculations and distillation operations, the input of the Decoder of the Informer neural network model is obtained. For the Decoder, similar to the Decoder used by the Informer neural network model, the Decoder needs the following inputs in order for the algorithm to generate a long sequence of outputs:
wherein,
to input the original sequence of the Decoder,
to predict the sequence (filled in with 0 s) and then pass the sequence through a mask-based sparsity auto-attention layer, it prevents each location from focusing on future locations, thus avoiding auto-regression. And transmitting the output of the layer and the output of the Encoder to a multi-head attention layer, and outputting a result through one-time calculation. Finally, the final output is obtained through a full connection layer. And (3) performing Loss function Loss calculation on the predicted output and real value, wherein the Loss function adopts MSE, and the calculation formula is as follows:
wherein n is the number of samples, y i In order to be the real data,
is the prediction data. And continuously iterating until the training condition is terminated, and finally generating the required model. The training condition termination is specifically to reach the number of model iterations or trigger an early-stop mechanism because MSE does not decrease, and finally the Informmer prediction model with the minimum loss function is obtained.
And S5, when the satellite positioning function of the vehicle is abnormal, comparing the current traffic environment image data of the vehicle with the traffic environment image data in the multiple types of traffic environment clustering data sets by using a bulldozer distance analysis method, determining the current traffic condition category of the vehicle, and selecting a corresponding inform neural network prediction model according to the determined current traffic condition category of the vehicle.
The method comprises the following specific steps: determining the current traffic condition category of the vehicle by utilizing a bulldozer distance analysis method according to the distance value between the target probability distribution and the pre-stored probability distribution; the target probability distribution is the probability distribution of the characteristic vectors in the current traffic environment image data of the vehicle, and the pre-stored probability distribution is the probability distribution of the characteristic vectors in the traffic environment image data in each type of the traffic environment cluster data set.
The method for determining the current traffic condition category of the vehicle by utilizing the bulldozer distance analysis method according to the distance value between the target probability distribution and the pre-stored probability distribution comprises the following steps:
calculation formula of bulldozer distance
will be minimum
The corresponding current traffic condition is classified as
Class;
and representing the traffic condition type corresponding to the prestored traffic environment image data of the vehicle.
Are respectively as
And
a pre-stored probability distribution representing a class of low complexity traffic conditions,
representing a pre-stored probability distribution of the medium complexity traffic condition class,
representing a pre-stored probability distribution of high complexity traffic condition classes. When the traffic environment of the vehicle changes, the vehicle motion characteristic data collected in the period of time is packaged and identified, and is classified in the vehicle motion characteristic data set of the traffic condition corresponding to the vehicle motion characteristic data set.
And S6, predicting the running track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model.
The embodiment of the invention realizes the prediction of the vehicle running track by using the Informer network, and effectively solves the problem of accurately predicting the vehicle position under the condition that the satellite positioning function fails for a long time. In the method for judging the traffic condition by combining the Gaussian mixture model and the KL divergence, the KL divergence is replaced by an improved bulldozer distance analysis method, the problem of asymmetric KL divergence is solved, the judgment effect is better than that of the KL divergence when the measurement has no overlapped distribution, and compared with a common bulldozer distance method, the improved bulldozer distance method adopted by the embodiment of the invention can shorten the solving time under the condition of improving the solving precision. The driving track of the vehicle is predicted by applying the Informer network model, and a self-attention distillation technology, a probability sparse self-attention mechanism and a generator decoder of the Informer network model are used as core technologies of a basic network, so that the training speed and the reasoning speed are improved, the memory overhead of the network is reduced, and the prediction precision is improved.
Example 2
Based on embodiment 1, this embodiment provides a vehicle positioner based on Informer network, including:
the function judging module is used for judging whether the satellite positioning function of the vehicle is normal or not;
the model training module is used for clustering the traffic environment image data of the vehicle by using a Gaussian mixture model when the satellite positioning function of the vehicle is normal to obtain a multi-class traffic environment clustering data set; sequencing a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of motion characteristic time sequence data sets; respectively training an Informer neural network model by utilizing the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models; wherein each type of traffic environment clustering data set corresponds to a type of traffic condition;
the track judgment module is used for comparing the current traffic environment image data of the vehicle with the traffic environment image data in the traffic environment cluster data sets by utilizing a bulldozer distance analysis method when the satellite positioning function of the vehicle is abnormal, determining the current traffic condition category of the vehicle, and selecting a corresponding Informer neural network prediction model according to the determined current traffic condition category of the vehicle; and predicting the driving track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model. In this embodiment, the function determining module, the model training module and the trajectory determining module may be computer function system modules, or may be specific computer hardware with corresponding functions or specific computers with corresponding functions.
The embodiment of the invention realizes the prediction of the vehicle running track by using the Informer network, and effectively solves the problem of accurately predicting the vehicle position under the condition that the GPS fails for a long time; meanwhile, different traffic conditions can be classified by adopting a GMM and IMP-Wasserstein distance method, the purpose is to distinguish trained models according to different traffic conditions, and when the GPS is unavailable, the accuracy can be greatly improved by selecting corresponding models for prediction according to the current traffic environment of the vehicle. Wherein GMM represents a Gaussian mixture model, and Wassertein Distance is Wassertein Distance, also called bulldozer Distance, namely Earth Mover's Distance, EMD. IMP-Wasserstein is a modified bulldozer distance, and a specific calculation formula of IMP-Wasserstein is shown in a bulldozer distance calculation formula in example 1. By using a self-attention distillation technology, a probability sparse self-attention mechanism and a generative decoder of an Informer network as core technologies of a basic network, the training speed and the reasoning speed are improved, the memory overhead of the network is reduced, and the prediction precision is improved. The invention can be used for assisting the positioning of the vehicle and improving the positioning capability of the vehicle in a complex environment.
Example 3
Based on embodiment 1, the present embodiment provides a computer, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the vehicle positioning method based on the Informer network in embodiment 1 when executing the computer program. The memory in the embodiment may be a computer internal memory, a computer external storage device, a storage hard disk, a mobile storage device, a cloud memory, and the like; the vehicle positioning method based on the Informer network is realized by using a computer program, so that the operation efficiency is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A vehicle positioning method based on an Informer network is characterized by comprising the following steps:
judging whether the satellite positioning function of the vehicle is normal or not;
when the satellite positioning function of the vehicle is normal, clustering the traffic environment image data of the vehicle by using a Gaussian mixture model to obtain a multi-class traffic environment clustering data set; wherein each type of traffic environment clustering data set corresponds to one type of traffic condition;
sequencing a plurality of motion characteristic data of a plurality of vehicles corresponding to each type of traffic condition according to the running time sequence of the vehicles to obtain a multi-type motion characteristic time sequence data set; wherein each type of motion characteristic time series data set corresponds to one type of the traffic condition;
training an Informer neural network model by respectively utilizing the motion characteristic time sequence data in each motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models;
when the satellite positioning function of the vehicle is abnormal, comparing the current traffic environment image data of the vehicle with the traffic environment image data in the multi-class traffic environment clustering data sets by using a bulldozer distance analysis method, determining the current traffic condition class of the vehicle, and selecting a corresponding inform neural network prediction model according to the determined current traffic condition class of the vehicle;
and predicting the driving track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model.
2. The vehicle positioning method based on the Informer network as claimed in claim 1, wherein a gaussian mixture model is used to cluster the traffic environment image data of the vehicle to obtain a plurality of types of traffic environment cluster data sets, comprising the steps of:
selecting a feature vector in the traffic environment image data;
calculating the probability distribution of the feature vectors in the traffic environment image data by utilizing a Gaussian mixture model;
and clustering all the traffic environment image data according to the corresponding probability distribution to obtain a plurality of types of traffic environment clustering data sets.
3. The Informer network-based vehicle positioning method as claimed in claim 2, wherein the probability distribution of the feature vectors in the traffic environment image data of the vehicle is calculated using a gaussian mixture model, comprising the steps of:
using formulas
Calculating a probability distribution in traffic environment image data of the vehicle; wherein y represents the feature vector, p (y) represents the probability distribution, and N (y | μ |) k ,∑ k ) Represents the kth component, π, of the Gaussian mixture model k Representing a weight of each component of the gaussian mixture model.
4. The method as claimed in claim 2, wherein the step of comparing the current traffic environment image data of the vehicle with the traffic environment image data in the plurality of types of traffic environment cluster data sets by using a bulldozer distance analysis method to determine the current traffic condition category of the vehicle comprises the steps of:
determining the current traffic condition category of the vehicle by utilizing a bulldozer distance analysis method according to the distance value between the target probability distribution and the pre-stored probability distribution; the target probability distribution is the probability distribution of the characteristic vectors in the current traffic environment image data of the vehicle, and the pre-stored probability distribution is the probability distribution of the characteristic vectors in the traffic environment image data in each type of the traffic environment cluster data sets.
5. The Informer network-based vehicle locating method according to claim 4, wherein said vehicle's current traffic condition category is determined according to the distance value between the target probability distribution and the pre-stored probability distribution by using a bulldozer distance analysis method, comprising the steps of:
calculating a distance value between the target probability distribution and a pre-stored probability distribution by using a bulldozer distance calculation formula; the calculation formula of the bulldozer distance is as follows,
where s represents the target probability distribution for the current traffic situation, s (x) and
is defined in a complex plane R n The probability distribution of (a) to (b),
is R n ×R n The combined distribution of the upper and lower beams,
is s and
set of all joint distributions γ combined, s and
is that
Sample x and sample y are obtained by sampling (x, y) to y from the joint distribution gamma,
is a kernel density function for x, y, t =1/α, α is 1, p is 2 for all cases where t > 0,
represents s and
the distance values between the corresponding pre-stored probability distributions;
will be at a minimum
The corresponding current traffic condition is classified as
Class;
and representing the traffic condition type corresponding to the prestored traffic environment image data of the vehicle.
6. The Informmer network-based vehicle positioning method as claimed in claim 1, wherein a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition are sorted according to a running time sequence of the vehicle to obtain a plurality of types of motion characteristic time-series data sets, comprising the following steps of:
sequencing a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of initial time sequence data sets;
and removing the empty data and the abnormal data in each type of the initial time sequence data set to obtain a plurality of types of the motion characteristic time sequence data sets.
7. The Inform network-based vehicle positioning method as claimed in claim 6, wherein the Inform neural network model is trained by using the motion feature time-series data in each type of the motion feature time-series data set to obtain a plurality of Inform neural network prediction models, comprising the steps of:
performing data normalization processing on all the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of normalized data sets; wherein each said normalized data set corresponds to a type of said traffic condition;
respectively training the Informer neural network model by using data in the plurality of normalized data sets to obtain a plurality of Informer neural network prediction models;
and verifying and/or testing the Informmer neural network prediction model by using partial data of the normalized data set.
8. The Informer network-based vehicle locating method as claimed in claim 1, wherein said vehicle motion characteristic data includes a vehicle travel time stamp, a heading angle, a pitch angle, a latitude, a longitude, an altitude, and a driving speed.
9. A vehicle positioning device based on an Informer network is characterized by comprising:
the function judging module is used for judging whether the satellite positioning function of the vehicle is normal or not;
the model training module is used for clustering the traffic environment image data of the vehicle by using a Gaussian mixture model when the satellite positioning function of the vehicle is normal to obtain a multi-class traffic environment clustering data set; sequencing a plurality of motion characteristic data of the vehicle corresponding to each type of traffic condition according to the running time sequence of the vehicle to obtain a plurality of types of motion characteristic time sequence data sets; respectively training an Informer neural network model by utilizing the motion characteristic time sequence data in each type of motion characteristic time sequence data set to obtain a plurality of Informer neural network prediction models; wherein each type of traffic environment clustering data set corresponds to a type of traffic condition;
the track judgment module is used for comparing the current traffic environment image data of the vehicle with the traffic environment image data in the traffic environment cluster data sets by utilizing a bulldozer distance analysis method when the satellite positioning function of the vehicle is abnormal, determining the current traffic condition category of the vehicle, and selecting a corresponding Informer neural network prediction model according to the determined current traffic condition category of the vehicle; and predicting the running track of the vehicle when the satellite positioning function is abnormal by using the selected Informer neural network prediction model.
10. A computer, characterized in that it comprises a memory, which stores a computer program, and a processor, which when executing said computer program, carries out the steps of the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211296349.5A CN115755134A (en) | 2022-10-21 | 2022-10-21 | Vehicle positioning method and device based on Informer network and computer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211296349.5A CN115755134A (en) | 2022-10-21 | 2022-10-21 | Vehicle positioning method and device based on Informer network and computer |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115755134A true CN115755134A (en) | 2023-03-07 |
Family
ID=85352663
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211296349.5A Pending CN115755134A (en) | 2022-10-21 | 2022-10-21 | Vehicle positioning method and device based on Informer network and computer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115755134A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116819581A (en) * | 2023-08-29 | 2023-09-29 | 北京交通大学 | Autonomous satellite positioning precision prediction method and device based on Informir |
| CN117807413A (en) * | 2023-08-08 | 2024-04-02 | 长安大学 | Vehicle lane change track prediction method based on random forest and improved Informier model |
-
2022
- 2022-10-21 CN CN202211296349.5A patent/CN115755134A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117807413A (en) * | 2023-08-08 | 2024-04-02 | 长安大学 | Vehicle lane change track prediction method based on random forest and improved Informier model |
| CN116819581A (en) * | 2023-08-29 | 2023-09-29 | 北京交通大学 | Autonomous satellite positioning precision prediction method and device based on Informir |
| CN116819581B (en) * | 2023-08-29 | 2023-11-21 | 北京交通大学 | Autonomous satellite positioning precision prediction method and device based on Informir |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115755134A (en) | Vehicle positioning method and device based on Informer network and computer | |
| CN107451619A (en) | A kind of small target detecting method that confrontation network is generated based on perception | |
| CN108460427A (en) | A kind of disaggregated model training method, device and sorting technique and device | |
| CN114202120A (en) | Urban traffic travel time prediction method aiming at multi-source heterogeneous data | |
| CN113505537A (en) | Building energy consumption detection method and device, computer equipment and storage medium | |
| CN113932801A (en) | Crowdsourcing-based real-time matching updating method for auxiliary driving map | |
| CN110795522B (en) | Method and device for predicting track position of mobile user | |
| CN111932612B (en) | Intelligent vehicle vision positioning method and device based on second-order hidden Markov model | |
| CN115730637A (en) | Multi-mode vehicle trajectory prediction model training method and device and trajectory prediction method | |
| CN111190211A (en) | GPS failure position prediction positioning method | |
| CN113159403A (en) | Method and device for predicting pedestrian track at intersection | |
| CN113406623A (en) | Target identification method, device and medium based on radar high-resolution range profile | |
| CN113587944A (en) | Quasi-real-time vehicle driving route generation method, system and equipment | |
| CN110135451B (en) | Flight path clustering method based on distance from point to line segment set | |
| US20240262385A1 (en) | Spatio-temporal pose/object database | |
| Zeng et al. | On the importance of contextual information for building reliable automated driver identification systems | |
| CN116543603B (en) | Flight path completion prediction method and device considering airspace situation and local optimization | |
| CN114595770B (en) | Long time sequence prediction method for ship track | |
| CN117749658A (en) | Fault prediction method, network operation and maintenance management platform, electronic equipment and medium | |
| CN113204246A (en) | Unmanned aerial vehicle running state detection method | |
| CN115293297B (en) | Method for predicting track of ship driven by intention | |
| CN117933492B (en) | Ship track long-term prediction method based on space-time feature fusion | |
| CN111832154B (en) | Identification test method for space decision consistency under large-scale data drive | |
| CN112686325B (en) | Underwater target search scheme evaluation decision method based on gray scale envelope | |
| CN112712197B (en) | Data processing method, device, equipment and storage medium |
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 |