Method and system for propagation delay-aware extraction based on spectral traffic pattern prototype library

Abstract

This invention discloses a method and system for propagation delay perception extraction based on a spectrum traffic pattern prototype library. The method includes: acquiring historical spatiotemporal observation data; filling the starting positions of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data; performing sliding windowing along the time dimension on the filled historical spatiotemporal observations to obtain a local time window observation subsequence for each time step; performing Fast Fourier Transform and normalization on the local time window observation subsequences to obtain the normalized current local spectrum representation; and performing a pattern search based on an attention mechanism in a pre-constructed spectrum traffic pattern prototype library to obtain the propagation delay perception representation. This invention can be flexibly integrated into various deep learning-based spatiotemporal sequence modeling networks, exhibiting good versatility and portability.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method, system, storage medium, and electronic device for propagation delay perception extraction based on a spectrum traffic pattern prototype library. Background Technology

[0002] In urban road networks, traffic conditions between adjacent or upstream / downstream road nodes often exhibit propagation delays: congestion events occurring upstream take a certain amount of time to spread to downstream nodes. This propagation delay is closely related to the road topology and historical traffic patterns, and exhibits a certain degree of periodicity and repeatability. For nodes... and The delayed coupling relationship between them can be formally described as:

[0003] in Indicates delayed transmission. This is the amplitude attenuation coefficient.

[0004] Existing deep learning-based spatiotemporal sequence modeling methods typically extract features solely from the raw time-domain observations of nodes, implicitly assuming that the spatial correlation between nodes is instantaneous and ignoring the propagation delay effect of traffic conditions between nodes. Furthermore, directly performing delay pattern matching or clustering based on raw time-domain observations is susceptible to random noise: two road segments with the same propagation delay may exhibit significant differences in waveform amplitude due to varying traffic peak values, making it difficult for time-domain shape similarity methods to correctly classify them into the same delay category. Summary of the Invention

[0005] This invention provides a propagation delay perception extraction method, system, storage medium, and electronic device based on a spectrum traffic pattern prototype library. It can be flexibly used in various deep learning-based spatiotemporal sequence modeling networks to improve the downstream network's ability to model spatiotemporal correlations with propagation delay characteristics.

[0006] This invention provides a propagation delay perception extraction method based on a spectrum traffic pattern prototype library, comprising: Acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step; Perform a Fast Fourier Transform and normalization on the observed subsequence of the local time window to obtain the normalized local spectral representation; A pattern search based on an attention mechanism is performed in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay perception representation.

[0007] Furthermore, according to the above-mentioned propagation delay perception extraction method based on the spectrum traffic pattern prototype library, zero vectors are filled at the starting position of the time dimension of the historical spatiotemporal observation data to obtain the filled historical spatiotemporal observation data. A sliding window is then applied along the time dimension to the filled historical spatiotemporal observations to obtain a local time window observation subsequence for each time step, including: The length of the padding at the starting position of the time dimension of the historical spatiotemporal observation data is After finding the zero vector, with a sliding step of 1 and a window length of... The sliding window method is used to select the local time window observation subsequence corresponding to the current time step. ,when The corresponding position is filled with a zero vector.

[0008] Furthermore, according to the above-mentioned propagation delay sensing extraction method based on the spectrum traffic pattern prototype library, the local time window observation subsequence is subjected to fast Fourier transform and normalization to obtain the current local spectrum representation, including: Perform a Fast Fourier Transform on the observation subsequence of the local time window to obtain the transformation result; The real and imaginary parts of the transformation result are concatenated and flattened to obtain the current local spectrum representation; The current local spectrum representation is normalized to obtain the normalized current local spectrum representation.

[0009] Furthermore, according to the above-described propagation delay perception extraction method based on a spectrum traffic pattern prototype library, the construction process of the spectrum traffic pattern prototype library includes: Historical spatiotemporal observation data of the target road network are obtained, and a set of candidate time subsequences is obtained by sliding windowing along the time dimension. For each candidate time subsequence in the candidate time subsequence set, perform Fast Fourier Transform and normalization to obtain a normalized candidate spectral feature vector. Calculate the cluster center of the normalized candidate spectral feature vector and use the cluster center as the spectral traffic pattern prototype library of the target road network.

[0010] Furthermore, according to the above-mentioned propagation delay perception extraction method based on the spectrum traffic pattern prototype library, the normalization method is L2 norm normalization: The current local spectrum representation or the candidate spectrum feature vector is normalized using its own L2 norm, i.e. ,in This represents the current local spectrum or candidate spectrum feature vector before normalization. This is the normalized representation of the current local spectrum or the candidate spectrum feature vector.

[0011] Furthermore, according to the above-mentioned propagation delay perception extraction method based on a spectrum traffic pattern prototype library, a pattern search based on an attention mechanism is performed in the pre-constructed spectrum traffic pattern prototype library to obtain the propagation delay perception representation, including: The current local spectrum representation is mapped through a projection layer to obtain a query vector; Each spectrum traffic pattern prototype in the preset frequency domain traffic pattern prototype library is mapped through two linear projection layers to obtain key vectors and value vectors. Calculate the pattern retrieval weight based on the query vector and the key vector; Based on the pattern retrieval weights, the prototype value vectors corresponding to each spectrum traffic pattern prototype are weighted and summed to obtain the propagation delay perception representation.

[0012] Furthermore, according to the above-mentioned propagation delay perception extraction method based on the spectrum traffic pattern prototype library, the pattern retrieval weight is calculated by the following formula:

[0013] in, For query vector, For key vectors, The projected feature dimension. For activation function, Retrieve weights for the model; Based on the pattern retrieval weights, the prototype value vectors corresponding to each spectrum traffic pattern prototype are weighted and summed, as expressed by the following formula:

[0014] in, This refers to the prototype value vector, i.e., the value vector, corresponding to each of the aforementioned spectrum traffic mode prototypes. For nodes At any moment The propagation delay perception representation.

[0015] This invention also provides a propagation delay perception extraction system based on a spectrum traffic pattern prototype library, comprising: The acquisition and preprocessing module is used to acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step. The feature extraction module is used to perform fast Fourier transform and normalization on the local time window observation subsequence to obtain the normalized current local spectrum representation; The pattern retrieval module is used to perform attention-based pattern search in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay-aware representation.

[0016] The present invention also provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute any of the above-described propagation delay perception extraction methods based on a spectrum traffic pattern prototype library.

[0017] The present invention also provides an electronic device, including a processor and a memory, wherein the processor is electrically connected to the memory, the memory is used to store instructions and data, and the processor is used in the steps of the propagation delay perception extraction method based on the spectrum traffic pattern prototype library described above.

[0018] This invention provides a method, system, storage medium, and electronic device for propagation delay perception extraction based on a spectrum traffic pattern prototype library. In the offline phase, this invention extracts and clusters recurring frequency domain patterns from historical data to construct a spectrum traffic pattern prototype library that represents historical delays. Compared to directly learning fixed correlation weights between nodes, this method can more fully explore the repeatable propagation patterns inherent in historical data. In the online phase, attention retrieval of the prototype library is performed using the local spectrum representation of each node at the current moment. This allows the propagation delay perception representation to adaptively retrieve the most relevant pattern information from the historical propagation pattern library based on the spectral characteristics of the current node, exhibiting strong adaptability. By expressing patterns using frequency domain features rather than raw time domain values, the interference of absolute observational value differences between different nodes and time periods, as well as random noise, on pattern matching is reduced, making the extracted propagation delay perception more stably characterize the oscillation rhythm and propagation patterns inherent in the observation sequence. Furthermore, the propagation delay perception extracted by this invention is a general feature representation, independent of specific downstream network structures and specific feature fusion methods. It can be flexibly integrated into various deep learning-based spatiotemporal sequence modeling networks, exhibiting good versatility and portability. Attached Figure Description

[0019] The technical solution and other beneficial effects of the present invention will become apparent from the following detailed description of specific embodiments of the invention, in conjunction with the accompanying drawings.

[0020] Figure 1 A flowchart of a propagation delay perception extraction method based on a spectrum traffic pattern prototype library provided in an embodiment of the present invention.

[0021] Figure 2This is a flowchart illustrating the propagation delay perception extraction based on a spectrum traffic pattern prototype library, as described in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the propagation delay perception extraction system based on a spectrum traffic pattern prototype library provided in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention provides a method, system, storage medium, and electronic device for propagation delay perception extraction based on a spectrum traffic pattern prototype library. The propagation delay perception extraction system based on a spectrum traffic pattern prototype library provided by this invention can be integrated into an electronic device, such as a terminal or server. The terminal can include a tablet computer, laptop computer, personal computer (PC), microprocessor box, or other devices.

[0026] Please see Figure 1 , Figure 1 The flowchart illustrates a propagation delay perception extraction method based on a spectrum traffic pattern prototype library, provided in an embodiment of the present invention. This method, applied in electronic devices, includes the following steps: S1. Obtain historical spatiotemporal observation data. Fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data. Perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step.

[0027] First, obtain the instance spatiotemporal observation data, denoted as... .

[0028] Step S1 involves filling the historical spatiotemporal observation data with zero vectors at the beginning of the time dimension to obtain the filled historical spatiotemporal observation data. Then, a sliding window is applied along the time dimension to obtain the local time window observation subsequence for each time step. This includes the following steps: Fill the starting position of the time dimension of historical spatiotemporal observation data with a length of [missing information]. After finding the zero vector, with a sliding step of 1 and a window length of... The sliding window method is used to select the local time window observation subsequence corresponding to the current time step. ,when The corresponding position is filled with a zero vector.

[0029] Specifically, let the preset local window length be... ,exist The starting position of the time dimension is filled with a length of 100 mm. The zero vector, then along the time dimension with a sliding step of 1 and a window length of... The sliding window method is used to obtain the local time window observation subsequence corresponding to each time step:

[0030] when When the zero vector is filled, the corresponding position is filled with a zero vector.

[0031] Since the window only contains the current time step and historical observations before it, and does not contain future information, the causality of the propagation delay perception extraction process is guaranteed.

[0032] This invention employs Discrete Fourier Transform (DFT) to extract the spectral features of local window observation subsequences for the following reasons: First, traffic flow at different nodes and time periods often exhibits significant differences in absolute values ​​(e.g., different road classes have different capacities). These differences in absolute flow levels can mask the similarities in the rhythm of change and trends of the subsequences. The spectral features obtained through Fourier transform and normalization primarily characterize the oscillation rhythm and relative fluctuation structure within the subsequence, weakening the interference of absolute value differences. This allows local windows with similar propagation rhythms but different absolute flow levels to still be identified as similar traffic patterns. Second, the DFT transforms the spectral features of a length w... The time-domain subsequence is mapped to a limited number of frequency-domain coefficients, which is equivalent to a compact representation of the original signal in the frequency domain. This can suppress the interference of random noise in the original observation data on the pattern matching results and improve the stability of the extracted spectral features. Thirdly, this invention retains both the real and imaginary parts when splicing the frequency-domain coefficients, that is, it retains the amplitude and phase information of each frequency component. Therefore, the spectral features can not only characterize the periodic fluctuation amplitude contained in the subsequence, but also distinguish different propagation patterns such as rising and falling edges, which is beneficial for distinguishing propagation modes with different directions and rhythms. In summary, using discrete Fourier transform to extract spectral features is more conducive to characterizing the propagation modes contained in the local window of spatiotemporal observation that are independent of the absolute numerical level, thereby improving the accuracy of subsequent pattern clustering and pattern retrieval.

[0033] S2 performs a Fast Fourier Transform and normalization on the local time window observation subsequence to obtain the normalized current local spectrum representation.

[0034] Specifically, step S2 includes the following steps: S21, perform a fast Fourier transform on the local time window observation subsequence to obtain the transform result.

[0035] This can be expressed by the following formula: ) in, For a local time window observation subsequence, FFT is the Fast Fourier Transform. For the transformation result S22, after concatenating the real and imaginary parts of the transformation result, flatten it to obtain the current local spectrum representation.

[0036] This can be expressed by the following formula:

[0037] in, For the real part, The virtual part, For splicing operations, This represents the current local spectrum.

[0038] S23, normalize the current local spectrum representation to obtain the normalized current local spectrum representation.

[0039] The normalization method used is L2 norm normalization.

[0040] Normalize the current local spectrum representation using the L2 norm of the current local spectrum representation or the candidate spectrum feature vector itself, i.e. ,in This is the current local spectrum representation before normalization. This is the normalized representation of the current local spectrum.

[0041] S3 performs an attention-based pattern search in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay perception representation.

[0042] The construction process of the spectrum traffic mode prototype library includes: Historical spatiotemporal observation data of the target road network are obtained, and a set of candidate time subsequences is obtained by sliding windowing along the time dimension.

[0043] Specifically, the number of days used for extracting candidate subsequences is set to... The number of observation time steps per day is determined by the sampling time interval (for example, when the sampling interval is 300 seconds, the number of observation time steps per day is 288). Therefore, the total number of time steps corresponding to the candidate extraction time range is... = × Daily observation time steps. Within this time frame, for each node. Along the time dimension with window length A sliding window is used with a sliding step size of 1 to obtain all candidate time subsequences corresponding to the node. The candidate time subsequences obtained by all nodes within this time range are summarized to form a candidate time subsequence set.

[0044] For each candidate time subsequence in the candidate time subsequence set, perform Fast Fourier Transform and normalization to obtain the normalized candidate spectral feature vector. Calculate the cluster center of the normalized candidate spectral feature vector and use the cluster center as the spectral traffic pattern prototype library of the target road network.

[0045] The process of Fast Fourier Transform and Normalization is the same as in step S2.

[0046] Specifically, the cluster centers of the normalized candidate spectral feature vectors are calculated, and these cluster centers are used as a prototype library of spectral traffic patterns for the target road network, including: For the normalized candidate spectral feature vector set ,in This represents the number of candidate spectral eigenvectors. The dimension of a single spectral feature vector is represented by... Clustering is performed using the mean clustering algorithm, with a preset number of clusters (i.e., the number of spectrum traffic pattern prototypes). After clustering, we get Cluster centers:

[0047] The aforementioned cluster centers are considered as a prototype library of spectral traffic patterns for this road network. This prototype library was built offline for pattern retrieval in the subsequent online phase.

[0048] In one specific embodiment, for historical traffic flow data of a city's road network, the extraction period can be set to 14 days or 21 days, and the window length... It can be set to 6, the number of spectrum traffic mode prototypes. The value can be set to 16, and the sampling time interval can be set to 300 seconds (i.e., 5 minutes). The above parameters are only preferred embodiments, and those skilled in the art can adjust them according to the actual scenario. They do not constitute a limitation on the scope of protection of the present invention.

[0049] In one embodiment, step S3 includes the following steps: S31, the current local spectrum representation is mapped through the projection layer to obtain the query vector.

[0050] S32, map each spectrum traffic pattern prototype in the preset frequency domain traffic pattern prototype library through two linear projection layers to obtain key vectors and value vectors.

[0051] S33, calculates the pattern retrieval weight based on the query vector and key vector.

[0052] S34. Based on the pattern retrieval weight, the prototype value vectors corresponding to the prototypes of each spectrum traffic pattern are weighted and summed to obtain the propagation delay perception representation.

[0053] Figure 2 This is a flowchart of the propagation delay perception extraction based on a spectrum traffic pattern prototype library, as described in an embodiment of the present invention. Figure 2 As shown, specifically, the current local spectrum is represented Mapped via linear projection layer The query vector is obtained; the spectrum traffic mode prototype library is used. The prototype traffic modes of each spectrum in the data are projected through two independent linear projection layers. and The mapping yields the key vector and value vector corresponding to the prototype, respectively:

[0054] in, For query vector, For key vectors, For value vectors, For the prototype of traffic modes of various spectrums, Parameters.

[0055] Calculate query vector With prototype key vector The scaled dot product, after being normalized by Softmax, yields the pattern retrieval weights:

[0056] Where d is the dimension of the projected feature.

[0057] And retrieve the prototype value vector according to the pattern retrieval weight. Perform a weighted summation to obtain the nodes. At any moment Propagation delay perception representation :

[0058] This propagation delay perception representation incorporates historical spectral traffic pattern information most relevant to the current local observation window. It is then extended to... After [number] nodes, the propagation delay-aware representation is obtained. Due to the time-shifting nature of the Fourier transform, the delay in the time domain signal corresponds to a linear shift in the phase of each component in the frequency domain. Therefore, the phase structure information retained in the spectral features implicitly carries the propagation delay pattern corresponding to the local observation window. By matching the propagation delay perception representation with the attention retrieval of the historical spectral pattern prototype library, the resulting propagation delay perception representation implicitly encodes the soft correspondence between the current node and the historical propagation delay pattern category, thereby endowing the downstream network with the ability to perceive propagation delay. This is a general feature representation, and the present invention does not limit its specific downstream applications. In practical applications, propagation delay sensing... Depending on the downstream network structure, the downstream spatiotemporal sequence modeling network can be accessed in various ways. For example, it can be added to the key vector or query vector of the self-attention mechanism in the downstream network to enhance the calculation of the association strength between nodes; it can be concatenated with the node features extracted from the downstream network along the feature dimension and then fused through a linear layer; or it can be weighted and fused with the node features of the downstream network after calculating the gating weights through the Sigmoid gating mechanism. None of the above access methods change the effectiveness of the propagation delay perception extraction process described in steps 1 to 3 of this invention. The specific access method can be determined according to the actual structure of the downstream network and application requirements, and this invention does not limit it.

[0059] Regarding the training method for pattern prototypes: The spectrum traffic pattern prototype P can be set as parameters initialized by the cluster centers in the above steps and kept fixed during the network training phase (i.e., only used as prior knowledge obtained offline and not involved in gradient updates), or it can be set as learnable parameters initialized by the cluster centers and further fine-tuned and updated during network training with gradient backpropagation. Neither of the above two settings affects the effectiveness of the online retrieval process in step S3 of this invention. The specific choice can be determined based on factors such as the scale of training data and computing resources in the actual application scenario.

[0060] The propagation delay perception extraction method proposed in this invention can be flexibly embedded as an independent functional module into the feature processing flow of any deep learning-based spatiotemporal sequence modeling network. It does not rely on specific temporal feature extraction methods or limit specific downstream fusion methods, requiring only the input of the original observation sequence of the node to be processed, thus exhibiting good versatility and portability. In one embodiment applied to traffic flow prediction, the propagation delay perception obtained by the method described in this invention can be used as an auxiliary feature in the downstream temporal prediction network. A loss function is constructed by combining it with real traffic flow data, and all learnable parameters, including spectral feature projection parameters and pattern prototype parameters, are jointly trained using an optimization algorithm. Finally, a model capable of outputting traffic flow prediction results for road nodes at multiple future time steps is obtained.

[0061] Based on the method described in the above embodiments, this embodiment will further describe the propagation delay perception extraction system based on the spectrum traffic pattern prototype library. The propagation delay perception extraction system based on the spectrum traffic pattern prototype library can be implemented as an independent entity or integrated into an electronic device. The electronic device can be a terminal, server, or other device. The terminal can include a tablet computer, a laptop computer, a personal computer (PC), a microprocessor box, or other devices.

[0062] Please see Figure 3 , Figure 3 This invention specifically describes a propagation delay perception extraction system based on a spectrum traffic pattern prototype library, applied in electronic devices. The system may include: The acquisition and preprocessing module is used to acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step. The feature extraction module is used to perform fast Fourier transform and normalization on the local time window observation subsequence to obtain the normalized current local spectrum representation; The pattern retrieval module is used to perform attention-based pattern search in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay-aware representation.

[0063] In specific implementation, the above modules and / or units can be implemented as independent entities, or they can be arbitrarily combined and implemented as the same or several entities. For the specific implementation of the above modules and / or units, please refer to the previous method embodiments. For the specific beneficial effects that can be achieved, please also refer to the beneficial effects in the previous method embodiments, which will not be repeated here.

[0064] In addition, this embodiment of the invention also provides an electronic device, which may be a computer, tablet computer, or other similar device. This electronic device can implement the steps of any embodiment of the propagation delay perception extraction method based on a spectrum traffic pattern prototype library provided in this embodiment of the invention. Therefore, it can achieve the beneficial effects that any propagation delay perception extraction method based on a spectrum traffic pattern prototype library provided in this embodiment of the invention can achieve, as detailed in the preceding embodiments, and will not be repeated here.

[0065] Figure 4 A specific structural block diagram of an electronic device provided in an embodiment of the present invention is shown. This electronic device can be used to implement the propagation delay perception extraction method based on a spectrum traffic pattern prototype library provided in the above embodiments. The electronic device 500 can be a terminal, server, or other device. The terminal can include a tablet computer, laptop computer, personal computer (PC), microprocessor box, or other devices.

[0066] The memory 520 can be used to store software programs and modules, such as the program instructions / modules corresponding to those in the above embodiments. The processor 580 executes various functional applications and data processing by running the software programs and modules stored in the memory 520. The memory 520 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 520 may further include memory remotely located relative to the processor 580, and these remote memories can be connected to the electronic device 500 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0067] The input unit 530 can be used to receive input numeric or character information, and to generate a keyboard and mouse related to user settings and function control.

[0068] Display unit 540 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces, which can be composed of graphics, text, icons, video, and any combination thereof. Display unit 540 may include display panel 541, which may optionally be configured in the form of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), or other similar forms.

[0069] Electronic device 500, through transmission module 570 (e.g., Wi-Fi module), can help users receive requests, send information, etc., providing users with wireless broadband internet access. Although transmission module 570 is shown in the figure, it is understood that it is not an essential component of electronic device 500 and can be omitted as needed without changing the essence of the invention.

[0070] The processor 580 is the control center of the electronic device 500. It connects to various parts of the phone via various interfaces and lines, and performs various functions and processes data of the electronic device 500 by running or executing software programs and / or modules stored in the memory 520, and by calling data stored in the memory 520, thereby providing overall monitoring of the electronic device. Optionally, the processor 580 may include one or more processing cores; in some embodiments, the processor 580 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor 580.

[0071] Electronic device 500 also includes a power supply 590 (such as a battery) that supplies power to various components. In some embodiments, the power supply may be logically connected to processor 580 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 590 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0072] Although not shown, the electronic device 500 also includes cameras (such as front-facing cameras and rear-facing cameras), Bluetooth modules, etc., which will not be described in detail here. Specifically, in this embodiment, the display unit of the electronic device is a touch screen display, and the mobile terminal also includes a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors. One or more programs contain instructions for performing the following operations: Acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step; Perform a Fast Fourier Transform and normalization on the observed subsequence of the local time window to obtain the normalized local spectral representation; A pattern search based on an attention mechanism is performed in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay perception representation.

[0073] In practice, the above modules can be implemented as independent entities or combined in any way to be implemented as the same or several entities. For the specific implementation of the above modules, please refer to the previous method implementation examples, which will not be repeated here.

[0074] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. Therefore, embodiments of the present invention provide a storage medium storing multiple instructions that can be loaded by a processor to execute the steps of any embodiment of the propagation delay perception extraction method based on a spectrum traffic pattern prototype library provided by the present invention.

[0075] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0076] Since the instructions stored in the storage medium can execute the steps in any embodiment of the propagation delay perception extraction method based on the spectrum traffic pattern prototype library provided in the embodiments of the present invention, the beneficial effects that any propagation delay perception extraction method based on the spectrum traffic pattern prototype library provided in the embodiments of the present invention can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0077] The foregoing has provided a detailed description of a propagation delay perception extraction method, system, storage medium, and electronic device based on a spectrum traffic pattern prototype library provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for propagation delay perception extraction based on a spectrum traffic pattern prototype library, characterized in that, The method includes: Acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step; Perform a Fast Fourier Transform and normalization on the observed subsequence of the local time window to obtain the normalized local spectral representation; A pattern search based on an attention mechanism is performed in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay perception representation.

2. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 1, characterized in that, The historical spatiotemporal observation data is padded with zero vectors at the beginning of the time dimension to obtain padded historical spatiotemporal observation data. A sliding window is then applied along the time dimension to obtain a local time window observation subsequence for each time step, including: The length of the padding at the starting position of the time dimension of the historical spatiotemporal observation data is After finding the zero vector, with a sliding step of 1 and a window length of... The sliding window method is used to select the local time window observation subsequence corresponding to the current time step. ,when The corresponding position is filled with a zero vector.

3. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 1, characterized in that, Performing a Fast Fourier Transform and normalization on the observed subsequences within the local time window yields the current local spectral representation, including: Perform a Fast Fourier Transform on the observation subsequence of the local time window to obtain the transformation result; The real and imaginary parts of the transformation result are concatenated and flattened to obtain the current local spectrum representation; The current local spectrum representation is normalized to obtain the normalized current local spectrum representation.

4. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 1, characterized in that, The construction process of the spectrum traffic mode prototype library includes: Historical spatiotemporal observation data of the target road network are obtained, and a set of candidate time subsequences is obtained by sliding windowing along the time dimension. For each candidate time subsequence in the candidate time subsequence set, perform Fast Fourier Transform and normalization to obtain a normalized candidate spectral feature vector. Calculate the cluster center of the normalized candidate spectral feature vector and use the cluster center as the spectral traffic pattern prototype library of the target road network.

5. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 1 or 4, characterized in that, The normalization method is L2 norm normalization: The current local spectrum representation or the candidate spectrum feature vector is normalized using its own L2 norm, i.e. ,in This represents the current local spectrum or candidate spectrum feature vector before normalization. This is the normalized representation of the current local spectrum or the candidate spectrum feature vector.

6. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 5, characterized in that, A pattern search based on an attention mechanism is performed in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay-aware representation, including: The current local spectrum representation is mapped through a projection layer to obtain a query vector; Each spectrum traffic pattern prototype in the preset frequency domain traffic pattern prototype library is mapped through two linear projection layers to obtain key vectors and value vectors. Calculate the pattern retrieval weight based on the query vector and the key vector; Based on the pattern retrieval weights, the prototype value vectors corresponding to each spectrum traffic pattern prototype are weighted and summed to obtain the propagation delay perception representation.

7. The propagation delay perception extraction method based on a spectrum traffic pattern prototype library according to claim 6, characterized in that, The pattern retrieval weight is calculated using the following formula: in, For query vector, For key vectors, The projected feature dimension. For activation function, Retrieve weights for the model; Based on the pattern retrieval weights, the prototype value vectors corresponding to each spectrum traffic pattern prototype are weighted and summed, as expressed by the following formula: in, This refers to the prototype value vector, i.e., the value vector, corresponding to each of the aforementioned spectrum traffic mode prototypes. For nodes At any moment The propagation delay perception representation.

8. A propagation delay perception extraction system based on a spectrum traffic pattern prototype library, characterized in that, include: The acquisition and preprocessing module is used to acquire historical spatiotemporal observation data, fill the starting position of the time dimension of the historical spatiotemporal observation data with zero vectors to obtain the filled historical spatiotemporal observation data, and perform sliding windowing on the filled historical spatiotemporal observations along the time dimension to obtain the local time window observation subsequence for each time step. The feature extraction module is used to perform fast Fourier transform and normalization on the local time window observation subsequence to obtain the normalized current local spectrum representation; The pattern retrieval module is used to perform attention-based pattern search in a pre-built spectrum traffic pattern prototype library to obtain a propagation delay-aware representation.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to execute the propagation delay perception extraction method based on a spectrum traffic pattern prototype library as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The method includes a processor and a memory, the processor being electrically connected to the memory, the memory being used to store instructions and data, and the processor being used to execute the steps in the propagation delay perception extraction method based on a spectrum traffic pattern prototype library as described in any one of claims 1 to 7.

Images