Automatic driving test scene extraction method and device based on deep embedding clustering
By analyzing real traffic accident data using deep embedding clustering methods, pre-collision trajectory sequences are extracted and clustered, solving the problems of implantability and adaptability in the selection of autonomous driving test scenarios, and achieving more accurate scenario description and safety assessment.
Patent Information
- Application Number
- CN202310277416.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-21
- Publication Date
- 2026-02-10
- Estimated Expiration
- 2043-03-21
AI Technical Summary
Existing technologies have poor implantability and adaptability of autonomous driving test scenarios. The selection of scenarios mainly relies on the subjective understanding of safety experts and lacks objective, comprehensive and accurate scenario classification methods.
A deep embedding clustering method is adopted. By deeply deconstructing real traffic accident data, pre-collision trajectory sequences are extracted, dimensionality reduction is performed using a heap autoencoder, and clustering is performed using the k-means algorithm to obtain multi-class feature clusters. By backtracking scene information, scene descriptions for each type of typical accident are obtained.
It improves the implantability and adaptability of autonomous driving test scenarios, provides more accurate and efficient scenario descriptions, and enhances the safety assessment between autonomous vehicles and vulnerable road users.
Smart Images
Figure CN116432067B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of data processing technology, and in particular to a method for extracting autonomous driving test scenarios based on deep embedding clustering. Background Technology
[0002] Currently, the intelligentization and connectivity of automobiles are major trends in future automotive development, and scenario-based testing technology is an indispensable and crucial part of autonomous driving testing and evaluation. Selecting appropriate scenarios is key to the safety testing of autonomous vehicles, and real-world accident scenarios can serve as critical scenarios for testing the safety performance of autonomous vehicles. Generally, typical test scenarios are selected based on the subjective understanding of safety experts, and the static characteristics of collisions (such as geometric features and weather conditions) are limited.
[0003] It is evident that there is an urgent need for an implantable and adaptable method for extracting autonomous driving test scenarios based on deep embedding clustering. Summary of the Invention
[0004] In view of this, the present disclosure provides a method for extracting autonomous driving test scenarios based on deep embedding clustering, which at least partially solves the problems of poor implantability and adaptability in the prior art.
[0005] In a first aspect, embodiments of this disclosure provide a method for extracting autonomous driving test scenarios based on deep embedding clustering, including:
[0006] Step 1: Based on the testing requirements of intelligent vehicles, perform in-depth accident analysis on real traffic accidents to obtain accident data, reconstruct the collected accident data, and extract the pre-collision trajectory sequence of the accident participants.
[0007] Step 2: Combine the pre-collision trajectory sequence and static environment information to create a pre-collision trajectory matrix;
[0008] Step 3: Using the pre-collision trajectory matrix as input, perform dimensionality reduction using a heap autoencoder to obtain low-dimensional embedding features;
[0009] Step 4: Using the latent features as input variables, cluster them using the k-means algorithm to obtain multi-class feature clusters;
[0010] Step 5: Based on the labels in each feature cluster, backtrack the pre-collision trajectory matrix to obtain the scene information of each cluster and summarize the commonalities of each cluster to obtain the scene description corresponding to each type of typical accident.
[0011] According to a specific implementation of this disclosure, the pre-collision trajectory sequence includes static elements and dynamic elements. The dynamic elements include the longitudinal coordinates, lateral coordinates, speed, acceleration, orientation, steering, and relative position difference of the accident participants within each preset step length during a preset time period before the collision. The static elements include the vehicle types of the accident participants, lighting conditions, weather, road surface water conditions, road type, first collision point, number of lanes, and line-of-sight obstruction.
[0012] According to one specific implementation of an embodiment of this disclosure, the stacked autoencoder includes multiple stacked autoencoder layers, wherein each autoencoder layer includes an encoder and a decoder.
[0013] According to a specific implementation of an embodiment of this disclosure, step 3 specifically includes:
[0014] The encoder uses a preset function to map the pre-collision trajectory matrix from the input layer to the hidden layer, and the decoder uses a mapping function to map the hidden layer data to the output layer. By reconstructing the input samples, the encoder learns the embedding features and obtains low-dimensional embedding features.
[0015] According to a specific implementation of this disclosure, before step 4, the method further includes:
[0016] The appropriate K value is determined by calculating the sum of squares error and the average profile coefficient under different K values according to the elbow rule.
[0017] According to a specific implementation of this disclosure, the formula for calculating the sum of squares error is as follows:
[0018]
[0019] Where, dist(x,o) i ) are objects x and o i The distance between them, where K represents the number of clusters;
[0020] The formula for calculating the average profile coefficient is as follows:
[0021]
[0022] Where S(i) is the silhouette coefficient of i:
[0023]
[0024] Where i is an object in a cluster, a(i) is the average dissimilarity between i and other objects in the same cluster, and b(i) is the minimum average dissimilarity between i and objects in other clusters.
[0025] Secondly, embodiments of this disclosure also provide an electronic device, the electronic device comprising:
[0026] At least one processor; and,
[0027] The memory is communicatively connected to the at least one processor; wherein,
[0028] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the deep embedding clustering-based autonomous driving test scenario extraction method in the first aspect or any implementation thereof.
[0029] The autonomous driving test scenario extraction scheme based on deep embedding clustering in this embodiment includes: Step 1, according to the testing requirements of intelligent vehicles, performing deep accident deconstruction on real traffic accidents to obtain accident data, reconstructing the collected accident data, and extracting the pre-collision trajectory sequence of the accident participants; Step 2, combining the pre-collision trajectory sequence and static environment information to create a pre-collision trajectory matrix; Step 3, using the pre-collision trajectory matrix as input, performing dimensionality reduction processing using a heap autoencoder to obtain low-dimensional embedding features; Step 4, using the latent features as input variables, performing clustering using the k-means algorithm to obtain multiple feature clusters; Step 5, based on the labels in each feature cluster, backtracking the pre-collision trajectory matrix to obtain the scene information of each cluster and summarizing the scene commonalities of each cluster to obtain the scene description corresponding to each type of typical accident.
[0030] The beneficial effects of the embodiments of this disclosure are as follows: By using the deep embedding clustering algorithm based on the stacked autoencoder to solve the dimensionality problem, this method enriches the existing classification methods for autonomous driving safety test scenarios. These results provide the necessary information for constructing and evaluating the safety of test scenarios between autonomous vehicles and vulnerable road users. By using the entire pre-collision trajectory matrix as input, it will not have an adverse effect due to manually selecting variables as clustering variables. It provides a more accurate and efficient description than the prior art in each typical scenario, increasing the implantability and adaptability of typical scenarios in autonomous driving safety testing. Attached Figure Description
[0031] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0032] Figure 1A flowchart illustrating an autonomous driving test scenario extraction method based on deep embedding clustering, provided in an embodiment of this disclosure;
[0033] Figure 2 A schematic diagram of an electronic device provided in an embodiment of this disclosure. Detailed Implementation
[0034] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.
[0035] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.
[0036] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.
[0037] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0038] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.
[0039] This disclosure provides a method for extracting autonomous driving test scenarios based on deep embedding clustering, which can be applied to the analysis of autonomous driving scenarios in connected vehicle scenarios.
[0040] See Figure 1 This is a flowchart illustrating a method for extracting autonomous driving test scenarios based on deep embedding clustering, provided in an embodiment of this disclosure. Figure 1 As shown, the method mainly includes the following steps:
[0041] Step 1: Based on the testing requirements of intelligent vehicles, perform in-depth accident analysis on real traffic accidents to obtain accident data, reconstruct the collected accident data, and extract the pre-collision trajectory sequence of the accident participants.
[0042] Optionally, the pre-collision trajectory sequence includes static elements and dynamic elements. The dynamic elements include the longitudinal coordinates, lateral coordinates, speed, acceleration, orientation, steering, and relative position difference of the parties involved in the accident within each preset step length during a preset time period before the collision. The static elements include the vehicle types of the parties involved in the accident, lighting conditions, weather, road surface water conditions, road type, first collision point, number of lanes, and visibility obstruction.
[0043] In practical implementation, when testing intelligent vehicles, in-depth accident analysis of real traffic accidents can be performed according to actual testing needs, and relevant accident data can be collected and organized. The collected accident data is then reconstructed, and the pre-collision trajectories of the accident participants are extracted. In the extraction of typical autonomous driving test scenarios using deep embedding clustering, the pre-collision trajectory matrix contains the necessary elements for building simulation tests, divided into dynamic and static elements. Dynamic elements include: the state of the accident participants at every moment in the 10 seconds before the collision, specifically, with a step size of 0.01 seconds, the longitudinal coordinate, lateral coordinate, velocity, acceleration, orientation, steering, and relative position difference within each step. Static elements include the vehicle types of both parties involved in the accident, lighting conditions, weather, road surface water conditions, road type, first collision point, number of lanes, and line-of-sight obstruction. Of course, necessary elements can be added or removed according to actual needs.
[0044] Step 2: Combine the pre-collision trajectory sequence and static environment information to create a pre-collision trajectory matrix;
[0045] In practice, after obtaining the pre-collision trajectory sequence, a pre-collision trajectory matrix (PCM) can be created by combining the pre-collision trajectory sequence with static environment information.
[0046] Step 3: Using the pre-collision trajectory matrix as input, perform dimensionality reduction using a heap autoencoder to obtain low-dimensional embedding features;
[0047] Furthermore, the stacked autoencoder includes multiple stacked autoencoder layers, wherein each autoencoder layer includes an encoder and a decoder.
[0048] Furthermore, step 3 specifically includes:
[0049] The encoder uses a preset function to map the pre-collision trajectory matrix from the input layer to the hidden layer, and the decoder uses a mapping function to map the hidden layer data to the output layer. By reconstructing the input samples, the encoder learns the embedding features and obtains low-dimensional embedding features.
[0050] In practice, a heap autoencoder is used to reduce the dimensionality of the PCM, which includes:
[0051] The PCM matrix is used as the input variable, and dimensionality reduction is achieved through a stacked autoencoder. The stacked autoencoder consists of several stacked autoencoders (AEs). Each autoencoder minimizes the loss function between the input and output vectors to ensure that the embedding process extracts key features from the input data. The autoencoder comprises two sub-networks: encoder f. w and decoder g u The encoder uses the function f. w x =
[0052] [x1,x2,…,x n Mapping from the input layer to the hidden layer h = [h1, h2, ..., h2] k ], where k is the size of the hidden layer variable vector. The decoder uses the mapping function g u Map h to the output layer. Learn the embedded features h = f by reconstructing the input samples. w (x), that is, the loss function is:
[0053]
[0054] A stacked autoencoder is formed by connecting multiple autoencoder layers one after another. In a stacked autoencoder, the original input vector is sent to the first AE layer. After the first autoencoder is trained, its hidden feature output is sent to the next layer, and the same process is repeated until all AEs are trained. The hidden layer of the top-level autoencoder is the output of the stacked autoencoder, which can be further input into a clustering model for classification.
[0055] Step 4: Using the latent features as input variables, cluster them using the k-means algorithm to obtain multi-class feature clusters;
[0056] Based on the above embodiments, before step 4, the method further includes:
[0057] The appropriate K value is determined by calculating the sum of squares error and the average profile coefficient under different K values according to the elbow rule.
[0058] Furthermore, the formula for calculating the sum of squares error is as follows:
[0059]
[0060] Where, dist(x,o) i ) are objects x and o i The distance between them, where K represents the number of clusters;
[0061] The formula for calculating the average profile coefficient is as follows:
[0062]
[0063] Where S(i) is the silhouette coefficient of i:
[0064]
[0065] Where i is an object in a cluster, a(i) is the average dissimilarity between i and other objects in the same cluster, and b(i) is the minimum average dissimilarity between i and objects in other clusters.
[0066] In practice, the embedded features output by the heap autoencoder are clustered using the k-means algorithm, where:
[0067] The k-means algorithm is one of the simplest and most commonly used clustering algorithms. Given n samples x and K initialized cluster centers μ, k-means partitions the samples by minimizing the within-cluster mean squared error. The corresponding loss function is:
[0068]
[0069] Where s ij This indicates whether the i-th sample belongs to the j-th cluster; if it does, the value is 1, otherwise it is 0.
[0070] In the k-means clustering algorithm, choosing the number of clusters K is a major issue. A large K value may separate similar trajectories into different clusters; on the other hand, a small K value may merge some obviously different clusters. Based on the elbow rule, an appropriate K value was determined by calculating the sum of squared errors (SSE) and the average silhouette coefficient (ASC) for different K values. k For a specific cluster, the SSE value is defined as:
[0071]
[0072] Where dist(x,o)i ) are objects x and o i The distance between them. ASC is defined as:
[0073]
[0074] Where S(i) is the silhouette coefficient of i:
[0075]
[0076] Where i is an object in a cluster; a(i) is the average dissimilarity of i with other objects in the same cluster; and b(i) is the minimum average dissimilarity of i with objects in other clusters. The silhouette coefficient ranges from -1 to 1, with larger values indicating better results.
[0077] Step 5: Based on the labels in each feature cluster, backtrack the pre-collision trajectory matrix to obtain the scene information of each cluster and summarize the commonalities of each cluster to obtain the scene description corresponding to each type of typical accident.
[0078] In practice, the pre-collision trajectory matrix can be traced back based on the labels in each feature cluster to extract the common features of scene elements in each cluster. Specifically, dynamic information such as position, speed, and orientation at each moment of collision, and static information such as weather and lighting are extracted. The mode of each variable is taken as the common feature of typical scenes.
[0079] The autonomous driving test scenario extraction method provided in this embodiment solves the dimensionality problem by using a deep embedding clustering algorithm based on a heap autoencoder. This method enriches existing autonomous driving safety test scenario classification methods. These results provide necessary information for constructing and evaluating the safety of test scenarios between autonomous vehicles and vulnerable road users. By using the entire pre-collision trajectory matrix as input, it avoids adverse effects from manually selecting variables as clustering variables. It provides a more accurate and efficient description of each typical scenario than existing technologies, increasing the implantability and adaptability of typical scenarios in autonomous driving safety testing.
[0080] See Figure 2 This disclosure also provides an electronic device 20, which includes at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed, enables the at least one processor to perform the deep embedding clustering-based autonomous driving test scenario extraction method described in the foregoing method embodiments.
[0081] This disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the deep embedding clustering-based autonomous driving test scenario extraction method in the foregoing method embodiments.
[0082] This disclosure also provides a computer program product, which includes a computing program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions that, when executed by a computer, cause the computer to perform the autonomous driving test scenario extraction method based on deep embedding clustering in the foregoing method embodiments.
[0083] The following is for reference. Figure 2 The diagram illustrates a structural schematic of an electronic device 20 suitable for implementing embodiments of the present disclosure. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 2 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.
[0084] like Figure 2 As shown, the electronic device 20 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 202 or a program loaded from a storage device 208 into a random access memory (RAM) 203. The RAM 203 also stores various programs and data required for the operation of the electronic device 20. The processing unit 201, ROM 202, and RAM 203 are interconnected via a bus 204. An input / output (I / O) interface 205 is also connected to the bus 204.
[0085] Typically, the following devices can be connected to I / O interface 205: input devices 206 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 207 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 208 including, for example, magnetic tapes, hard disks, etc.; and communication devices 209. Communication device 209 allows electronic device 20 to communicate wirelessly or wiredly with other devices to exchange data. Although electronic device 20 with various devices is shown in the figure, it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed alternatively.
[0086] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 209, or installed from storage device 208, or installed from ROM 202. When the computer program is executed by processing device 201, it performs the functions defined in the methods of embodiments of this disclosure.
[0087] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.
[0088] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.
[0089] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, enable the electronic device to perform the relevant steps of the above-described method embodiments.
[0090] Alternatively, the aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, enable the electronic device to perform the relevant steps of the above method embodiments.
[0091] Computer program code for performing the operations of this disclosure can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, and conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0092] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0093] The units described in the embodiments of this disclosure can be implemented in software or in hardware.
[0094] It should be understood that the various parts of this disclosure can be implemented in hardware, software, firmware, or a combination thereof.
[0095] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.
Claims
1. A method for extracting autonomous driving test scenarios based on deep embedding clustering, characterized in that, include: Step 1: Based on the testing requirements of intelligent vehicles, perform in-depth accident analysis on real traffic accidents to obtain accident data, reconstruct the collected accident data, and extract the pre-collision trajectory sequence of the accident participants. The pre-collision trajectory sequence includes static elements and dynamic elements. The dynamic elements include the longitudinal coordinates, lateral coordinates, speed, acceleration, orientation, steering, and relative position difference of the accident participants within each preset step length during a preset time period before the collision. The static elements include the vehicle types of the two parties involved in the accident, lighting conditions, weather, road surface water conditions, road type, first collision point, number of lanes, and line of sight obstruction. Step 2: Combine the pre-collision trajectory sequence and static environment information to create a pre-collision trajectory matrix; Step 3: Using the pre-collision trajectory matrix as input, perform dimensionality reduction using a heap autoencoder to obtain low-dimensional embedding features; Step 4, use latent features as input variables, and... -The mean algorithm is used for clustering to obtain multiple feature clusters; Before step 4, the method further includes: The appropriate K value is determined by calculating the sum of squares error and the average profile coefficient under different K values according to the elbow rule. The formula for calculating the sum of squares error is as follows: ； in, It is an object and The distance between them Indicates the number of clusters; The formula for calculating the average profile coefficient is as follows: ； in yes Profile coefficients: ； in It is an object in a cluster. yes The average dissimilarity with other objects in the same cluster. yes The minimum average dissimilarity between objects in other clusters; Step 5: Based on the labels in each feature cluster, backtrack the pre-collision trajectory matrix to obtain the scene information of each cluster and summarize the commonalities of each cluster to obtain the scene description corresponding to each type of typical accident.
2. The method according to claim 1, characterized in that... The stacked autoencoder comprises multiple stacked autoencoder layers, wherein each autoencoder layer includes an encoder and a decoder.
3. The method according to claim 2, characterized in that... Step 3 specifically includes: The encoder uses a preset function to map the pre-collision trajectory matrix from the input layer to the hidden layer, and the decoder uses a mapping function to map the hidden layer data to the output layer. By reconstructing the input samples, the encoder learns the embedding features and obtains low-dimensional embedding features.
4. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the autonomous driving test scene extraction method based on deep embedding clustering as described in any one of claims 1-3.