Information processing method, device and storage medium

Abstract

Embodiments of the present invention provide an information processing method, device, and storage medium. The method includes: obtaining a second sample set based on a sample extraction strategy and a first sample set; clustering the second sample set according to a lane change clustering strategy to obtain at least two clustering information items; comparing the at least two clustering information items with at least two first data items obtained during actual vehicle driving; and obtaining a first feedback sample based on the comparison result; locating, based on the first feedback sample, failure scenarios that are inconsistent with road tests of actual vehicle driving; and modifying an existing vehicle lane change model based on the failure scenarios located by the first feedback sample, and performing lane change selection based on the modified vehicle lane change model. The use of embodiments of the present invention can improve lane change accuracy.

Description

[0001] This application is a divisional application of the Chinese patent application with the application date of February 21, 2019, application number 201910130653.4, and invention name “An information processing method, device and storage medium”. Technical Field

[0002] The present invention relates to the field of information processing technology, and in particular to an information processing method, device and storage medium. Background Art

[0003] One application scenario of information processing is in autonomous driving of vehicles. In order to realize the motion planning of automatic driving of the vehicle, it is necessary to evaluate the trajectory of the vehicle's motion. Vehicle lane changing is an important part of the motion trajectory. Related technologies cannot automatically locate failure scenarios that are inconsistent with actual road tests, that is, they cannot effectively locate the problem (bad case) and pick it out. If the problem cannot be located, the vehicle's lane change decision will be inaccurate, and the lane change accuracy will decrease when the lane change selection is made based on the vehicle's lane change decision, which will ultimately make it difficult to ensure the feasibility and safety of the lane change results during actual driving. Summary of the Invention

[0004] An embodiment of the present invention provides an information processing method to solve one or more technical problems in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides an information processing method, the method comprising:

[0006] Obtain a second sample set according to the sample extraction strategy and the first sample set;

[0007] performing clustering processing on the second sample set according to the lane change clustering strategy to obtain at least two clustering information, comparing the at least two clustering information with at least two first data obtained during actual driving of the vehicle, and obtaining a first feedback sample based on the comparison result;

[0008] Locating, based on the first feedback sample, a failure scenario that is inconsistent with an actual road test of the vehicle;

[0009] According to the failure scenario located by the first feedback sample, the existing vehicle lane change model is modified, and the lane change selection is performed according to the modified vehicle lane change model.

[0010] In one embodiment, obtaining the second sample set according to the sample extraction strategy and the first sample set includes:

[0011] Obtaining a neural network constituting the existing vehicle lane change model, wherein the neural network includes an input layer, an intermediate layer, and an output layer;

[0012] Extracting all samples corresponding to the input layer and the output layer from the first sample set as first subsamples;

[0013] extracting a portion of samples corresponding to the middle layer from the first sample set as second subsamples;

[0014] The second sample set is obtained according to the first subsample and the second subsample.

[0015] In one embodiment, comparing the at least two clustering information with at least two first data obtained during actual driving of the vehicle, and obtaining a first feedback sample based on the comparison result, includes:

[0016] Querying target data that does not match each clustering information from the at least two first data;

[0017] The target data is used as a first feedback sample.

[0018] In one embodiment, the method further comprises:

[0019] Clustering the second sample set according to the lane change clustering strategy, if no clustering information is obtained, the current scene is a boundary scene in actual driving of the vehicle;

[0020] Taking the boundary scene as a second feedback sample;

[0021] The second feedback sample is added to the first feedback sample.

[0022] In one embodiment, the method further comprises:

[0023] Inputting the first feedback sample into the existing vehicle lane change model, and calculating and obtaining a probability distribution of the feedback sample;

[0024] Obtaining feedback entropy according to the probability distribution of the feedback samples;

[0025] Whether the vehicle meets the requirements of a road test for actual driving is determined based on the feedback entropy.

[0026] In one embodiment, the method further comprises:

[0027] Adding the first feedback sample to the second sample set to obtain a third sample set;

[0028] performing clustering processing on the third sample set according to the lane change clustering strategy to obtain a first subclass for characterizing a left lane change, a second subclass for characterizing a right lane change, and a third subclass for characterizing straight driving;

[0029] According to the first subclass, the second subclass and the third subclass, corresponding subnetworks in the neural network constituting the existing vehicle lane change model are adjusted.

[0030] In a second aspect, an embodiment of the present invention provides an information processing device, the device comprising:

[0031] a sample processing unit, configured to obtain a second sample set according to the sample extraction strategy and the first sample set;

[0032] a first feedback sample processing unit, configured to perform clustering processing on the second sample set according to the lane change clustering strategy, and when obtaining at least two pieces of clustering information, compare the at least two pieces of clustering information with at least two pieces of first data obtained during actual driving of the vehicle, and obtain a first feedback sample based on the comparison result;

[0033] a positioning unit, configured to locate, based on the first feedback sample, a failure scenario that is inconsistent with a road test of an actual vehicle driving;

[0034] The lane change selection unit is configured to modify an existing vehicle lane change model according to a failure scenario located by the first feedback sample, and perform lane change selection according to the modified vehicle lane change model.

[0035] In one embodiment, the sample processing unit is further configured to:

[0036] Obtaining a neural network constituting the existing vehicle lane change model, wherein the neural network includes an input layer, an intermediate layer, and an output layer;

[0037] Extracting all samples corresponding to the input layer and the output layer from the first sample set as first subsamples;

[0038] extracting a portion of samples corresponding to the middle layer from the first sample set as second subsamples;

[0039] The second sample set is obtained according to the first subsample and the second subsample.

[0040] In one embodiment, the first feedback sample processing unit is further configured to:

[0041] Querying target data that does not match each clustering information from the at least two first data;

[0042] The target data is used as a first feedback sample.

[0043] In one embodiment, the device further comprises:

[0044] a clustering processing unit, configured to perform clustering processing on the second sample set according to the lane change clustering strategy, and if no clustering information is obtained, the current scene is a boundary scene in actual driving of the vehicle;

[0045] a second feedback sample processing unit, configured to use the boundary scene as a second feedback sample;

[0046] The first sample adding unit is configured to add the second feedback sample to the first feedback sample.

[0047] In one embodiment, the device further comprises:

[0048] a first computing unit, configured to input the first feedback sample into the existing vehicle lane change model and calculate a probability distribution of the feedback sample;

[0049] A second operation unit, configured to obtain feedback entropy according to the probability distribution of the feedback samples;

[0050] A judgment unit is used to judge whether the feedback entropy meets the road test of actual driving of the vehicle.

[0051] In one embodiment, the device further comprises:

[0052] a second sample adding unit, configured to add the first feedback sample to the second sample set to obtain a third sample set;

[0053] a subclass processing unit, configured to perform clustering processing on the third sample set according to the lane change clustering strategy to obtain a first subclass for representing a left lane change, a second subclass for representing a right lane change, and a third subclass for representing straight driving;

[0054] An adjustment unit is configured to adjust corresponding sub-networks in a neural network constituting the existing vehicle lane change model according to the first sub-class, the second sub-class, and the third sub-class.

[0055] In a third aspect, an embodiment of the present invention provides an information processing device, the functions of which can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions.

[0056] In one possible design, the apparatus includes a processor and a memory, the memory being configured to store a program that supports the apparatus in executing any of the aforementioned information processing methods, and the processor being configured to execute the program stored in the memory. The apparatus may also include a communication interface for communicating with other devices or a communication network.

[0057] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium for storing computer software instructions used by an information processing device, which includes a program for executing any of the above-mentioned information processing methods.

[0058] One of the above technical solutions has the following advantages or beneficial effects:

[0059] In an embodiment of the present invention, a second sample set is obtained based on a sample extraction strategy and a first sample set, and the second sample set is clustered according to a lane change clustering strategy. When at least two clustering information are obtained, the at least two clustering information are compared with at least two first data obtained during the actual driving of the vehicle, and a first feedback sample is obtained based on the comparison result. Based on the first feedback sample, a failure scenario that is inconsistent with the road test of the actual driving of the vehicle is located. Based on the failure scenario located by the first feedback sample, the existing vehicle lane change model is corrected, and lane change selection is performed based on the corrected vehicle lane change model. Since the clustering information obtained by the clustering process is compared with the data during the actual driving of the vehicle, a first feedback sample (bad example) can be obtained. Based on the first feedback sample, a failure scenario that is inconsistent with the road test of the actual driving of the vehicle can be located, and the problem (bad example) can be found and picked out. By correcting the existing vehicle lane change model through the located failure scenario and performing lane change selection based on the corrected vehicle lane change model, the accuracy of the vehicle lane change can be improved.

[0060] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar components or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed herein and should not be construed as limiting the scope of the invention.

[0062] Figure 1 A flowchart of an information processing method according to an embodiment of the present invention is shown.

[0063] Figure 2 A schematic diagram illustrating sample screening according to an embodiment of the present invention.

[0064] Figure 3 A flowchart of an information processing method according to an embodiment of the present invention is shown.

[0065] Figure 4A flowchart of an information processing method according to an embodiment of the present invention is shown.

[0066] Figure 5 A flowchart of an information processing method according to an embodiment of the present invention is shown.

[0067] Figure 6 A structural block diagram of an information processing device according to an embodiment of the present invention is shown.

[0068] Figure 7 A structural block diagram of an information processing device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0069] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0070] In the related art, lane change decision models based on machine learning and deep learning (such as relationship models and simulation models) will inevitably have failure scenarios in actual road tests, because both machine learning and deep learning cannot effectively identify bad cases. If a manual labeling method is used, it is applicable to takeover problems (such as scenarios where vehicles collide with each other, are blocked by obstacles, or vehicles are stagnant), and is also suitable for scenarios where identifiable behaviors can be clearly defined by rules (such as pressing solid lines, etc.). In addition, the lane change decision model cannot know whether the bad case is due to scene differences or insufficient feature information expression resulting in learning deviations. By adopting the following embodiment of the present invention, scene differences and bad cases can be automatically discovered by optimizing the lane change decision model to form feedback samples, and then automatically locate failure scenarios that do not conform to actual road tests based on the feedback samples, and continue to improve and enhance the accuracy of lane changes using the model based on the failure scenarios.

[0071] Figure 1 1 is a flow chart showing an information processing method according to an embodiment of the present invention. Figure 1 As shown, the process includes:

[0072] Step 101: Obtain a second sample set according to a sample extraction strategy and a first sample set.

[0073] Step 102: Cluster the second sample set according to the lane change clustering strategy. When at least two clustering information are obtained, the at least two clustering information are compared with at least two first data obtained during actual driving of the vehicle, and a first feedback sample is obtained based on the comparison result.

[0074] Step 103: Based on the first feedback sample, locate the failure scenario that is inconsistent with the actual road test of the vehicle.

[0075] Step 104: Modify the existing vehicle lane change model based on the failure scenario located by the first feedback sample, and perform lane change selection based on the modified vehicle lane change model.

[0076] In one embodiment, considering that the first scene information obtained during human driving can only reflect the real-world human scene, while the second scene information obtained during machine-simulated driving can only reflect dynamic simulation results, neither represents real-world driving and cannot 100% cover all data, it is necessary to fuse the two. The first sample set is composed of the first scene information collected during human driving and the second scene information collected during machine-simulated driving. The first sample set formed by fusing the first and second scene information may be inaccurate because perception in simulations can be inherently inaccurate. For example, the perception module in the simulation may have inherent anomalies, such as noise, which can affect perception and lead to inaccurate perception. For example, in simulations, obstacles can be visible from a distance, but in the real world, distant obstacles are often invisible to the vehicle's sensors. This part of the scene is inconsistent with the actual driving scene and needs to be filtered out. In other words, any inaccurate perception needs to be filtered out to match the actual driving scene. After filtering out these inaccurate perceptions, the first sample set obtained will be relatively accurate.

[0077] The first scene information is an image capture of the actual driving conditions of the vehicle in the lane. The first scene information obtained based on the capture results is information obtained by capturing the vehicle driving in any lane on a frame-by-frame basis. Specifically, in addition to the environmental information surrounding the vehicle, the vehicle's driving route, and obstacle information (static or dynamic obstacle information), the information also includes the driving behavior of the professional driver. Feature extraction of the driving behavior during the driving process can obtain feature information, such as acceleration, speed, speed limit, corresponding timestamp, and location information. Feature information such as acceleration, speed, speed limit, corresponding timestamp, and location information is identified and labeled to obtain label information such as left lane change, right lane change, or straight driving. The first scene information includes at least the environmental information surrounding the vehicle, the vehicle's driving route, obstacle information (static or dynamic obstacle information), and feature information and label information obtained by analyzing the driving behavior of the professional driver. The speed limit here means that if the current vehicle (such as the main vehicle) finds that there are other vehicles (such as the vehicle in front) on the current vehicle's route, if the main vehicle's speed is 80 miles per hour, in order to avoid collision, the speed limit needs to be reduced to 50 miles per hour to increase the distance between the main vehicle and the vehicle in front, ensure the driving safety of both parties, and avoid rear-end collisions.

[0078] For the second scene information, the driving condition of the vehicle on the driving lane is simulated and then image capture is performed. The second scene information is obtained according to the capture results, which at least includes the environmental information around the vehicle, the vehicle's driving route, and obstacle information (static or dynamic obstacle information). It is necessary to utilize the feature information and label information obtained by analyzing the driving behavior of professional drivers. Therefore, the feature information and label information are also covered in the second scene information. The simulation is based on this information to simulate the driving behavior of professional drivers to see what data can be output. Since one is real driving and the other is a simulation of real driving, there are certain differences between the two.

[0079] The first sample set obtained after filtering out the above-mentioned samples that do not match the actual driving scene, even if it is relatively accurate, may still deviate from the actual driving scene. For example, when there is a failure scene but the failure scene cannot be located in time, it is easy to deviate. This failure scene needs to be found in time. First, the second sample set is obtained based on the sample extraction strategy and the above-mentioned first sample set. Specifically, a neural network constituting an existing vehicle lane change model is obtained, and the neural network includes an input layer, an intermediate layer, and an output layer. Figure 2 As shown, Figure 2 Figure 1 is a schematic diagram of sample screening. All samples corresponding to the input layer 121 and the output layer 123 of the first sample set 11 are extracted as first subsamples 132. A portion of the samples corresponding to the intermediate layer 122 of the first sample set 11 are extracted as second subsamples 131. A second sample set 14 is obtained based on the first subsamples 132 and the second subsamples 131. The second sample set is clustered according to the lane change clustering strategy. When at least two clusters are obtained, the at least two clusters are compared with at least two first data points obtained during actual vehicle driving. A first feedback sample is obtained based on the comparison result. Based on the first feedback sample, failure scenarios that are inconsistent with the actual road driving test are located. Based on the failure scenarios located by the first feedback sample, the existing vehicle lane change model is modified, and lane change selection is performed based on the modified vehicle lane change model. According to the embodiments of the present invention, since failure scenarios that are inconsistent with actual road testing can be automatically located based on the feedback samples, the accuracy of lane change using the model can be further improved and enhanced based on the failure scenarios.

[0080] Figure 3 1 is a flow chart showing an information processing method according to an embodiment of the present invention. Figure 3 As shown, the process includes:

[0081] Step 201: Obtain a second sample set according to a sample extraction strategy and the first sample set.

[0082] Step 202: Cluster the second sample set according to the lane change clustering strategy. When at least two clustering information are obtained, the at least two clustering information are compared with at least two first data obtained during actual vehicle driving. Target data that does not match each clustering information is retrieved from the at least two first data, and the target data is used as the first feedback sample.

[0083] In one example, based on the second sample set, (offline) sub-category clustering is performed within the three categories of going straight, changing lanes to the left, and changing lanes to obtain clustering results. Specific clustering methods can include K-means or hierarchical clustering. The samples in the second sample set can also be numbered, and an associated index is established between the sample number, input feature layer, key intermediate layer output, output probability layer, lane direction, and sub-category number.

[0084] In one example, clustering information (e.g., offline clustering information) is compared with first data from actual vehicle travel (e.g., online data) to determine the affiliation of data in the second sample set with existing categories, as well as new, anomalous data that does not belong to any subclass. The purpose of comparing clustering information (e.g., offline clustering information) with first data from actual vehicle travel (e.g., online data) is to determine whether a particular scenario conforms to the clustering information obtained through clustering. This comparison includes at least two decision branches: first, if it conforms to the clustering information, it is a good example, which is not used as feedback; second, if it does not conform to the clustering information, it is a bad case, which is used as feedback. By comparing the bad case with the good example and the feature space index, the online prediction component of the clustering algorithm can distinguish whether the sample type does not exist, is a boundary scenario, or other situations. This bad case is used as feedback and retrained together with the previously existing data in the second sample set.

[0085] Step 203: Based on the first feedback sample, locate the failure scenario that is inconsistent with the actual road test of the vehicle.

[0086] Step 204: Modify the existing vehicle lane change model based on the failure scenario located by the first feedback sample, and perform lane change selection based on the modified vehicle lane change model.

[0087] In one example, Figure 2As shown, all samples corresponding to the input layer 121 and the output layer 123 in the first sample set 11 are extracted as the first subsample 132, and some samples corresponding to the intermediate layer 122 in the first sample set 11 are extracted as the second subsample 131. The second sample set 14 is obtained based on the first subsample 132 and the second subsample 131. After clustering processing is performed on the second sample set, multiple clustering information is compared with multiple first data obtained during actual vehicle driving. The multiple first data also include data that does not match the clustering information. The data that matches the clustering information is used as the target data, and the target data is used as the first feedback sample. Based on the first feedback sample, a failure scenario that does not match the road test of actual vehicle driving is located. Based on the failure scenario located by the first feedback sample, the existing vehicle lane change model is corrected, and lane change selection is performed based on the corrected vehicle lane change model.

[0088] Figure 4 1 is a flow chart showing an information processing method according to an embodiment of the present invention. Figure 4 As shown, the process includes:

[0089] Step 301: Obtain a second sample set according to a sample extraction strategy and the first sample set.

[0090] Step 302: Cluster the second sample set according to the lane change clustering strategy. When at least two clustering information are obtained, the at least two clustering information are compared with at least two first data obtained during actual vehicle driving. Target data that does not match each clustering information is retrieved from the at least two first data, and the target data is used as the first feedback sample.

[0091] Step 303: Cluster the second sample set according to the lane change clustering strategy. If no clustering information is obtained, the current scene is a boundary scene in actual driving of the vehicle, and the boundary scene is used as the second feedback sample.

[0092] Step 304: Locate failure scenarios that are inconsistent with the actual road test of the vehicle based on the first feedback sample and the second feedback sample.

[0093] In the aforementioned example of clustering and obtaining more than one feedback sample, (offline) sub-category clustering is performed based on the second sample set within the three categories of going straight, changing lanes to the left, and changing lanes to obtain clustering results. The specific clustering method can be K-means or hierarchical clustering. The samples in the second sample set can also be numbered, and an associated index is established between the sample number, the input feature layer, the output of the key intermediate layer, the output probability layer, the lane direction, and the sub-category number. The clustering information (e.g., offline clustering information) is compared with the first data from the actual vehicle driving (e.g., online data) to determine the affiliation of the data in the second sample set with the existing categories, as well as any anomalous new data that does not belong to any sub-category. The purpose of comparing the clustering information (e.g., offline clustering information) with the first data from the actual vehicle driving (e.g., online data) is to determine whether a particular scenario conforms to the clustering information obtained from the clustering. This includes at least two decision branches: first, if it conforms to the clustering information, it is a good case and no feedback is provided; second, if it does not conform to the clustering information, it is a bad case and feedback is provided. Furthermore, a special case of a bad case is a situation where there is a scene but no clustering information, usually corresponding to a boundary scene. For this case, it will also be used as feedback. By comparing the bad case and the good case with the feature space index, the online prediction part of the clustering algorithm is used to distinguish whether the sample type does not exist or is a boundary scene or other situation. The bad case and / or the special case of the bad case are all used as feedback samples, and the feedback samples are input into the second sample set. Since the data in the second sample set cannot locate the failure scene and cannot fully cover the real driving scene, the feedback samples of the second sample set are input for data repair and completion as new training data, and the new training data is used for model optimization (for example, for the existing vehicle lane change model). A more accurate processing result can be output, so that the lane change accuracy according to the optimized vehicle lane change model will be greatly improved.

[0094] Step 305: Modify the existing vehicle lane change model based on the failure scenario located by the first feedback sample and the second feedback sample, and perform lane change selection based on the modified vehicle lane change model.

[0095] Figure 5 1 is a flow chart showing an information processing method according to an embodiment of the present invention. Figure 5 As shown, the process includes:

[0096] Step 401: Obtain a second sample set according to a sample extraction strategy and the first sample set.

[0097] Step 402: Cluster the second sample set according to the lane change clustering strategy. When at least two clustering information are obtained, the at least two clustering information are compared with at least two first data obtained during actual vehicle driving. Target data that does not match each clustering information is retrieved from the at least two first data, and the target data is used as the first feedback sample.

[0098] Step 403: Based on the first feedback sample, locate the failure scenario that is inconsistent with the actual road test of the vehicle.

[0099] Step 404: Modify the existing vehicle lane change model based on the failure scenario located by the first feedback sample, and perform lane change selection based on the modified vehicle lane change model.

[0100] Step 405: Input the first feedback sample into the existing vehicle lane change model, and calculate to obtain the probability distribution of the feedback sample.

[0101] Step 406: Obtain feedback entropy based on the probability distribution of the feedback samples, and determine whether the feedback entropy conforms to the road test of actual vehicle driving based on the feedback entropy.

[0102] In one example, an existing vehicle lane change model can serve as a basic decision model. The first feedback sample is input into the decision model and optimized to improve its performance. Specifically, the first feedback sample is input into the decision model, and a probability distribution of the feedback sample is calculated. Based on the probability distribution of the feedback sample, feedback entropy is calculated. A higher entropy indicates a better example.

[0103] In one embodiment, different subnetwork structures can also be adjusted through different clustering information obtained through clustering. Specifically, the first feedback sample is added to the second sample set to obtain a third sample set. The third sample set is clustered according to the lane change clustering strategy to obtain a first subclass for characterizing left lane changes, a second subclass for characterizing right lane changes, and a third subclass for characterizing straight driving. According to the first subclass, the second subclass, and the third subclass, the corresponding subnetworks in the neural network constituting the existing vehicle lane change model are adjusted. The same neural network and data are shared by all lane models (relationship models, simulation models, decision models, etc.).

[0104] Figure 6A structural block diagram of an information processing device is shown, which includes: a sample processing unit 21, which is used to obtain a second sample set based on a sample extraction strategy and a first sample set; a first feedback sample processing unit 22, which is used to cluster the second sample set according to a lane change clustering strategy, and when at least two clustering information are obtained, compare the at least two clustering information with at least two first data obtained during actual vehicle driving, and obtain a first feedback sample based on the comparison result; a positioning unit 23, which is used to locate a failure scenario that is inconsistent with a road test of actual vehicle driving based on the first feedback sample; and a lane change selection unit 24, which is used to modify an existing vehicle lane change model based on the failure scenario located by the first feedback sample, and perform lane change selection based on the modified vehicle lane change model.

[0105] In one embodiment, the sample processing unit is further used to: obtain a neural network constituting the existing vehicle lane change model, the neural network including an input layer, an intermediate layer, and an output layer; extract all samples corresponding to the input layer and the output layer from the first sample set as first subsamples; extract some samples corresponding to the intermediate layer from the first sample set as second subsamples; and obtain the second sample set based on the first subsamples and the second subsamples.

[0106] In one embodiment, the first feedback sample processing unit is further configured to: query target data that does not match each clustering information from the at least two first data; and use the target data as the first feedback sample.

[0107] In one embodiment, the device further includes: a clustering processing unit, configured to cluster the second sample set according to a lane change clustering strategy; if no clustering information is obtained, the current scene is a boundary scene in actual vehicle driving; a second feedback sample processing unit, configured to use the boundary scene as a second feedback sample; and a first sample adding unit, configured to add the second feedback sample to the first feedback sample.

[0108] In one embodiment, the device further includes: a first computing unit, configured to input the first feedback sample into the existing vehicle lane change model to calculate a probability distribution of the feedback sample; a second computing unit, configured to obtain feedback entropy based on the probability distribution of the feedback sample; and a judgment unit, configured to judge whether the feedback entropy conforms to a road test for actual vehicle driving based on the feedback entropy.

[0109] In one embodiment, the device further includes: a second sample adding unit, configured to add the first feedback sample into the second sample set to obtain a third sample set; a subclass processing unit, configured to cluster the third sample set according to a lane change clustering strategy to obtain a first subclass for characterizing a left lane change, a second subclass for characterizing a right lane change, and a third subclass for characterizing straight driving; and an adjustment unit, configured to adjust each corresponding subnetwork in the neural network constituting the existing vehicle lane change model according to the first subclass, the second subclass, and the third subclass.

[0110] The functions of the modules in the devices of the embodiments of the present invention can be found in the corresponding descriptions of the above methods, which will not be repeated here.

[0111] Figure 7 FIG. 1 shows a structural block diagram of an information processing device according to an embodiment of the present invention. Figure 7 As shown, the apparatus includes: a memory 910 and a processor 920. The memory 910 stores a computer program that can be executed on the processor 920. When the processor 920 executes the computer program, the autonomous driving method in the above embodiment is implemented. The number of the memory 910 and the processor 920 can be one or more.

[0112] The device also includes a communication interface 930 for communicating with external devices and performing data exchange transmission.

[0113] The memory 910 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0114] If the memory 910, processor 920, and communication interface 930 are implemented independently, the memory 910, processor 920, and communication interface 930 can be connected to each other via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0115] Optionally, in a specific implementation, if the memory 910, the processor 920 and the communication interface 930 are integrated on a chip, the memory 910, the processor 920 and the communication interface 930 can communicate with each other through an internal interface.

[0116] An embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the program is executed by a processor, any of the methods described in the above embodiments is implemented.

[0117] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless otherwise inconsistent.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0119] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0120] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0121] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0122] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0123] Furthermore, the functional units in the various embodiments of the present invention may be integrated into a single processing module, each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules. If the integrated modules are implemented as software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a magnetic disk, or an optical disk, etc.

[0124] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various modifications and substitutions within the technical scope disclosed in the present invention, and such modifications and substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. An information processing method, characterized in that: The method comprises: Using a neural network of an existing vehicle lane change model, a second sample set is obtained according to a sample extraction strategy and the first sample set; performing clustering processing on the second sample set according to the lane change clustering strategy to obtain at least two clustering information, querying target data that does not match each clustering information from at least two first data obtained during actual driving of the vehicle, and using the target data as the first feedback sample; Locating, based on the first feedback sample, a failure scenario that is inconsistent with an actual road test of the vehicle; Modifying an existing vehicle lane change model based on the failure scenario located by the first feedback sample, and performing lane change selection based on the modified vehicle lane change model; The method further comprises: Inputting the first feedback sample into the existing vehicle lane change model, and calculating and obtaining a probability distribution of the feedback sample; Obtaining feedback entropy according to the probability distribution of the feedback samples; Whether the vehicle meets the requirements of a road test for actual driving is determined based on the feedback entropy.

2. The method according to claim 1, characterized in that The neural network includes an input layer, an intermediate layer and an output layer; The neural network using the existing vehicle lane change model obtains the second sample set according to the sample extraction strategy and the first sample set, including: Extracting all samples corresponding to the input layer and the output layer from the first sample set as first subsamples; extracting a portion of samples corresponding to the middle layer from the first sample set as second subsamples; The second sample set is obtained according to the first subsample and the second subsample.

3. The method according to claim 1, characterized in that The method further comprises: Clustering the second sample set according to the lane change clustering strategy, if no clustering information is obtained, the current scene is a boundary scene in actual driving of the vehicle; Taking the boundary scene as a second feedback sample; The second feedback sample is added to the first feedback sample.

4. The method according to claim 1, wherein The method further comprises: Adding the first feedback sample to the second sample set to obtain a third sample set; performing clustering processing on the third sample set according to the lane change clustering strategy to obtain a first subclass for characterizing a left lane change, a second subclass for characterizing a right lane change, and a third subclass for characterizing straight driving; According to the first subclass, the second subclass and the third subclass, corresponding subnetworks in the neural network constituting the existing vehicle lane change model are adjusted.

5. An information processing device, characterized in that The device comprises: a sample processing unit, configured to use a neural network of an existing vehicle lane change model to obtain a second sample set based on a sample extraction strategy and the first sample set; a first feedback sample processing unit configured to cluster the second sample set according to the lane change clustering strategy, and when at least two clustering information are obtained, query the at least two first data obtained during actual driving of the vehicle for target data that does not match each clustering information, and use the target data as the first feedback sample; a positioning unit, configured to locate, based on the first feedback sample, a failure scenario that is inconsistent with a road test of an actual vehicle driving; a lane change selection unit, configured to modify an existing vehicle lane change model based on a failure scenario located by the first feedback sample, and perform lane change selection based on the modified vehicle lane change model; The device further comprises: a first computing unit, configured to input the first feedback sample into the existing vehicle lane change model and calculate a probability distribution of the feedback sample; A second operation unit, configured to obtain feedback entropy according to the probability distribution of the feedback samples; A judgment unit is used to judge whether the feedback entropy meets the road test of actual driving of the vehicle.

6. The device according to claim 5, characterized in that The neural network includes an input layer, an intermediate layer and an output layer; The neural network using the existing vehicle lane change model obtains the second sample set according to the sample extraction strategy and the first sample set, including: Extracting all samples corresponding to the input layer and the output layer from the first sample set as first subsamples; extracting a portion of samples corresponding to the middle layer from the first sample set as second subsamples; The second sample set is obtained according to the first subsample and the second subsample.

7. The device according to claim 5, characterized in that The device further comprises: a clustering processing unit, configured to perform clustering processing on the second sample set according to the lane change clustering strategy, and if no clustering information is obtained, the current scene is a boundary scene in actual driving of the vehicle; a second feedback sample processing unit, configured to use the boundary scene as a second feedback sample; The first sample adding unit is configured to add the second feedback sample to the first feedback sample.

8. The device according to claim 5, characterized in that The device further comprises: a second sample adding unit, configured to add the first feedback sample to the second sample set to obtain a third sample set; a subclass processing unit, configured to perform clustering processing on the third sample set according to the lane change clustering strategy to obtain a first subclass for representing a left lane change, a second subclass for representing a right lane change, and a third subclass for representing straight driving; An adjustment unit is configured to adjust corresponding sub-networks in a neural network constituting the existing vehicle lane change model according to the first sub-class, the second sub-class, and the third sub-class.

9. An information processing device, characterized in that The device comprises: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method according to any one of claims 1 to 4.

10. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images