A trustworthy artificial intelligence method dedicated to vehicle class driving and control

Abstract

The application belongs to the technical field of artificial intelligence, and discloses a trusted artificial intelligence method special for vehicle driving and control, which comprises the following steps: acquiring a video stream collected by a camera of a vehicle during driving of the vehicle; extracting feature information of each video frame of the video stream through a vehicle artificial intelligence perception network model, and predicting feature information of a next video frame based on the feature information of a current video frame; determining an online autonomous learning strategy of the vehicle artificial intelligence perception network model based on an error between the extracted feature information of the next video frame and the predicted feature information of the next video frame; and performing online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy, so as to obtain an online trained vehicle artificial intelligence perception network model. The application realizes online feedback learning and feedback training of the vehicle artificial intelligence perception network model, and improves the perception accuracy of the model.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and in particular relates to a trustworthy artificial intelligence method specifically for vehicle driving and control. Background Art

[0002] With the rapid development of artificial intelligence (AI) in recent years, it has gradually been applied to vehicles. Specifically, offline vehicle driving data can be collected and used as training data for model training, resulting in a pre-trained model. During vehicle operation, this pre-trained model generates predictions, which are then used to control the vehicle accordingly.

[0003] From the above description, it can be seen that in the related technology, a large amount of vehicle driving data is required to conduct offline training of the pre-trained model, resulting in a high training cost of the pre-trained model; and during the driving process of the vehicle, if a driving scenario that is not covered by the pre-trained model is encountered, the pre-trained model cannot make accurate predictions, resulting in a low prediction accuracy of the pre-trained model, which brings safety hazards to the vehicle. Summary of the Invention

[0004] The purpose of this invention is to enable online feedback and autonomous learning of a vehicle's artificial intelligence perception network model. After online training, the vehicle's artificial intelligence perception network model achieves higher perception accuracy, improving vehicle safety. Furthermore, online training eliminates the need for dedicated training data, reducing training costs. This online autonomous learning method improves the reliability and safety of intelligent perception and decision-making in unknown scenarios and operating conditions.

[0005] In a first aspect, an embodiment of the present invention provides a trusted artificial intelligence method specifically for vehicle driving and control, the method comprising:

[0006] During the driving of the vehicle, a video stream captured by a camera of the vehicle is obtained;

[0007] Extracting feature information of each video frame of the video stream through a vehicle artificial intelligence perception network model, and predicting feature information of a next video frame based on the feature information of a current video frame, where the current video frame is any video frame of the video stream;

[0008] Determining an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and an error between the predicted feature information of the next video frame;

[0009] Based on the online autonomous learning strategy, online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model to obtain the vehicle artificial intelligence perception network model after online training.

[0010] Optionally, extracting feature information of each video frame of the video stream by using a vehicle artificial intelligence perception network model includes:

[0011] Inputting the video stream into a vehicle artificial intelligence perception network model;

[0012] The vehicle artificial intelligence perception network model is used to extract feature information in multiple dimensions from each video frame.

[0013] Optionally, predicting feature information of a next video frame based on feature information of a current video frame includes:

[0014] For feature information of each dimension, feature information of the dimension of the next video frame is predicted based on the feature information of the dimension of the current video frame.

[0015] Optionally, determining an online autonomous learning strategy of the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and the error between the predicted feature information of the next video frame includes:

[0016] For feature information of each dimension, calculating an error between first feature information and second feature information, where the first feature information is the extracted feature information of the next video frame in the dimension, and the second feature information is the predicted feature information of the next video frame in the dimension;

[0017] An online autonomous learning strategy of the vehicle artificial intelligence perception network model is determined based on the calculated multiple errors.

[0018] Optionally, the method further includes:

[0019] Outputting multiple perception results through the vehicle artificial intelligence perception network model; each perception result is used to complete a perception task, and different perception results complete different perception tasks;

[0020] For each perception result, obtain the calibration result corresponding to the perception result from a pre-built database;

[0021] Calculating a loss function value of the vehicle artificial intelligence perception network model based on the perception results and the calibration results;

[0022] Based on the calculated loss function value, the online autonomous learning strategy of the vehicle artificial intelligence perception network model is adjusted to obtain an adjusted online autonomous learning strategy.

[0023] Optionally, performing online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain the vehicle artificial intelligence perception network model after online training includes:

[0024] Based on the adjusted online autonomous learning strategy, online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model to obtain a new online trained vehicle artificial intelligence perception network model.

[0025] In a second aspect, an embodiment of the present invention provides an artificial intelligence method specifically for vehicle-type driving controllability, the method comprising:

[0026] Obtaining a video stream captured by a camera of the vehicle during the vehicle's driving process;

[0027] Inputting the video stream into the online trained vehicle artificial intelligence perception network model to output a perception result through the online trained vehicle artificial intelligence perception network model;

[0028] The vehicle is controlled accordingly based on the perception result.

[0029] In a third aspect, an embodiment of the present invention provides an artificial intelligence device specifically for vehicle driving and control, the device comprising:

[0030] A video stream acquisition module is used to acquire the video stream captured by the vehicle's camera while the vehicle is traveling;

[0031] a feature information extraction and prediction module, configured to extract feature information of each video frame of the video stream through a vehicle artificial intelligence perception network model, and predict feature information of a next video frame based on the feature information of a current video frame, where the current video frame is any video frame of the video stream;

[0032] an online autonomous learning strategy determination module, configured to determine an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and an error between the predicted feature information of the next video frame;

[0033] The vehicle artificial intelligence perception network model training module is used to perform online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain the vehicle artificial intelligence perception network model after online training.

[0034] In a fourth aspect, an embodiment of the present invention provides an artificial intelligence device specifically for vehicle driving control, the device comprising:

[0035] A video stream acquisition module is used to acquire the video stream collected by the camera of the vehicle during the vehicle's driving process;

[0036] A perception result acquisition module, configured to input the video stream into the online trained vehicle artificial intelligence perception network model, so as to output a perception result through the online trained vehicle artificial intelligence perception network model;

[0037] A vehicle control module is used to control the vehicle accordingly based on the perception result.

[0038] In a fifth aspect, an embodiment of the present invention provides an electronic device, including:

[0039] at least one processor;

[0040] a memory for storing the at least one processor-executable instruction;

[0041] The at least one processor is configured to execute the instructions to implement the method described in any one of the first aspect or the second aspect.

[0042] In a sixth aspect, an embodiment of the present invention provides a computer-readable storage medium, which, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to execute the method described in any one of the first aspect or the second aspect.

[0043] In a seventh aspect, an embodiment of the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in either the first aspect or the second aspect.

[0044] The technical solution provided by an embodiment of the present invention is as follows: during vehicle driving, a video stream captured by a vehicle camera is obtained; the video stream is input into a vehicle artificial intelligence perception network model to be trained online, and the following steps are performed through the vehicle artificial intelligence perception network model: feature information of multiple dimensions of each video frame in the video stream is extracted; feature information of each video frame of the video stream is extracted through the vehicle artificial intelligence perception network model, and feature information of the next video frame is predicted based on the feature information of the current video frame; an online autonomous learning strategy of the vehicle artificial intelligence perception network model is determined based on the error between the extracted feature information of the next video frame and the predicted feature information of the next video frame; and online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain a vehicle artificial intelligence perception network model after online training.

[0045] As can be seen, in the embodiments of the present invention, the vehicle artificial intelligence perception network model can be trained online in real time while the vehicle is in motion, enabling online feedback and autonomous learning of the vehicle artificial intelligence perception network model. The online-trained vehicle artificial intelligence perception network model has higher perception accuracy, improving vehicle safety. Furthermore, online training does not require the collection of dedicated training data, reducing the training cost of the vehicle artificial intelligence perception network model. The online autonomous learning method of the present invention improves the reliability and safety of intelligent perception and decision-making in unknown scenarios and operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flowchart of a trusted artificial intelligence method specifically for vehicle driving and control provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of the process of online feedback learning and feedback training of a vehicle artificial intelligence perception network model provided by an embodiment of the present invention;

[0048] Figure 3 A flowchart of a trusted artificial intelligence method specifically for vehicle driving and control provided by an embodiment of the present invention;

[0049] Figure 4 A schematic diagram of an online enhancement process for a vehicle artificial intelligence perception network model provided by an embodiment of the present invention;

[0050] Figure 5 A flowchart of an artificial intelligence method specifically for vehicle-type driving control provided by an embodiment of the present invention;

[0051] Figure 6 A schematic diagram of the structure of an artificial intelligence device specifically for vehicle driving and control provided by an embodiment of the present invention;

[0052] Figure 7 A trusted artificial intelligence device specifically for vehicle driving and control provided by an embodiment of the present invention;

[0053] Figure 8 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be described in detail below through examples.

[0055] With the rapid development of artificial intelligence (AI) in recent years, it has gradually been applied to vehicles. However, existing AI technologies lack the ability to learn online or update in real time. As pre-trained models, they can only iterate based on pre-training data. This means that their knowledge and capabilities are trained based on data from before a specific point in time. Furthermore, during use, generative AI cannot actively learn or memorize new information, nor can it instantly update its knowledge base based on user interaction. This poses significant challenges to vehicle safety when encountering scenarios not covered in pre-training during driving.

[0056] As can be seen from the above description, the training cost of existing pre-trained models is enormous. They require a large amount of offline data for pre-training, and then the trained models are put into use. Furthermore, due to the extremely complex internal structure of the models, the training process is difficult to analyze, and the reliability of the results remains questionable. Furthermore, during driving, if a scenario not covered by pre-training is encountered, the pre-trained model's prediction accuracy is low. This can lead to incorrect vehicle control and serious consequences. Therefore, the reliability and safety of AI applications in vehicles are a significant challenge.

[0057] To address the aforementioned technical issues, embodiments of the present invention provide a vehicle AI capable of online learning. Compared to existing AI technologies, this invention reduces model complexity through online feedback learning, enabling a certain degree of internal analysis of the model to ensure safety and reliability that meet driving requirements. This will not only advance the development of AI theory but also promote the industrial application of intelligent technologies in the field of intelligent driving. It can also improve the reliability and safety of intelligent perception and decision-making in unknown scenarios and operating conditions.

[0058] The following is a detailed description of a trusted artificial intelligence method specifically for vehicle driving and control provided by an embodiment of the present invention.

[0059] like Figure 1 FIG. 1 is a flowchart of a trusted artificial intelligence method for vehicle driving and control provided by an embodiment of the present invention. The method may include the following steps:

[0060] S110 , while the vehicle is traveling, obtaining a video stream captured by a camera of the vehicle.

[0061] Specifically, the embodiments of the present invention use online learning technology to train the vehicle's artificial intelligence perception network model. Therefore, continuous video streams are the basis for the learning and perception of the vehicle's artificial intelligence perception network model. As the vehicle is driving, the vehicle's cameras capture real-time video streams, so the video streams captured by the vehicle's cameras can be obtained.

[0062] S120, extracting feature information of each video frame of the video stream through the vehicle artificial intelligence perception network model, and predicting feature information of the next video frame based on the feature information of the current video frame.

[0063] The current video frame is any video frame in the video stream.

[0064] As an implementation of an embodiment of the present invention, extracting feature information of each video frame of a video stream through a vehicle artificial intelligence perception network model may include the following steps, namely step a1 and step a2:

[0065] Step a1: Input the video stream into the vehicle artificial intelligence perception network model.

[0066] Step a2: Extract feature information of multiple dimensions of each video frame through the vehicle artificial intelligence perception network model.

[0067] Among them, before online training of the vehicle artificial intelligence perception network model, the vehicle artificial intelligence perception network model can be pre-trained, and the pre-training process is an offline training process. In order to improve the perception accuracy of the vehicle artificial intelligence perception network model obtained by pre-training, in an embodiment of the present invention, during the driving process of the vehicle, the video stream collected by the vehicle's camera is input into the pre-trained vehicle artificial intelligence perception network model to perform online training on the pre-trained vehicle artificial intelligence perception network model. Because the vehicle will encounter various new scenes during driving, by online training of the vehicle artificial intelligence perception network model, the vehicle artificial intelligence perception network model can be adapted to various new scenes, thereby improving the perception accuracy of the vehicle artificial intelligence perception network model. In addition, online training does not require the special collection of a large amount of training data, which can reduce the training cost of the vehicle artificial intelligence perception network model.

[0068] As an implementation of the embodiment of the present invention, extracting feature information of each video frame of a video stream through a vehicle artificial intelligence perception network model may include the following steps, namely step b1 and step b2:

[0069] Step b1: Input the video stream into the vehicle artificial intelligence perception network model.

[0070] Step b2: extract feature information of multiple dimensions of each video frame through the vehicle artificial intelligence perception network model.

[0071] Specifically, after obtaining the video stream, the video stream can be input into the vehicle artificial intelligence perception network model. After receiving the video stream, the vehicle artificial intelligence perception network model can extract feature information of multiple dimensions of the video stream, where the multiple dimensions may include two-dimensional, three-dimensional or four-dimensional, etc. The two-dimensional feature information may be plane feature information, the three-dimensional feature information may be spatial feature information, and the four-dimensional feature information may be time and space feature information.

[0072] Accordingly, in one embodiment, predicting feature information of the next video frame based on feature information of the current video frame may include the following step c1:

[0073] Step c1: For feature information of each dimension, predict feature information of the dimension of the next video frame based on the feature information of the dimension of the current video frame.

[0074] Specifically, assuming there is 2D feature information, 3D feature information, and 4D feature information, the 2D feature information of the next video frame can be predicted based on the 2D feature information of the current video frame, the 3D feature information of the next video frame can be predicted based on the 3D feature information of the current video frame, and the 4D feature information of the next video frame can be predicted based on the 4D feature information of the current video frame.

[0075] S130, determining an online autonomous learning strategy of the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and the error between the predicted feature information of the next video frame.

[0076] Specifically, after predicting the feature information of the next video frame, the error between the extracted feature information of the next video frame and the predicted feature information of the next video frame can be calculated, and based on the error between the two, the online autonomous learning strategy of the vehicle artificial intelligence perception network model can be determined.

[0077] As an implementation of an embodiment of the present invention, determining an online autonomous learning strategy for a vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and the error between the predicted feature information of the next video frame may include the following steps, namely step d1 and step d2:

[0078] Step d1: For each dimension of feature information, calculate the error between the first feature information and the second feature information.

[0079] The first feature information is the extracted feature information of the dimension of the next video frame, and the second feature information is the predicted feature information of the dimension of the next video frame.

[0080] Step d2: determining an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the calculated multiple errors.

[0081] Specifically, assuming that the extracted feature information has three dimensions, namely two-dimensional feature information, three-dimensional feature information, and four-dimensional feature information. Then, the error between the extracted two-dimensional feature information of the next video frame and the predicted two-dimensional feature information of the next video frame can be calculated. The error between the extracted three-dimensional feature information of the next video frame and the predicted three-dimensional feature information of the next video frame can also be calculated. The error between the extracted four-dimensional feature information of the next video frame and the predicted four-dimensional feature information of the next video frame can also be calculated. And based on the three calculated errors, the online autonomous learning strategy of the vehicle artificial intelligence perception network model is determined. Among them, the online autonomous learning strategy can be the model parameter adjustment information of the vehicle artificial intelligence perception network model.

[0082] S140, performing online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain the vehicle artificial intelligence perception network model after online training.

[0083] Specifically, after obtaining the online autonomous learning strategy of the vehicle artificial intelligence perception network model, the vehicle artificial intelligence perception network model can be subjected to online feedback learning and feedback training according to the online autonomous learning strategy. This process is the model parameter adjustment process of the vehicle artificial intelligence perception network model, thereby obtaining the vehicle artificial intelligence perception network model after online training. Assuming that the vehicle artificial intelligence perception network model needs to adjust three parameters a, b, and c, the online autonomous learning strategy can be: keep a unchanged, increase b, and decrease c. Then, based on this online autonomous learning strategy, a can be kept unchanged, b can be increased, and c can be decreased. The perception accuracy of the vehicle artificial intelligence perception network model after online training is higher, and online training does not require the special collection of training data, which reduces the training cost of the vehicle artificial intelligence perception network model.

[0084] The technical solution provided by an embodiment of the present invention is as follows: during vehicle driving, a video stream captured by a vehicle camera is obtained; the video stream is input into a vehicle artificial intelligence perception network model to be trained online, and the following steps are performed through the vehicle artificial intelligence perception network model: feature information of multiple dimensions of each video frame in the video stream is extracted; feature information of each video frame of the video stream is extracted through the vehicle artificial intelligence perception network model, and feature information of the next video frame is predicted based on the feature information of the current video frame; an online autonomous learning strategy of the vehicle artificial intelligence perception network model is determined based on the error between the extracted feature information of the next video frame and the predicted feature information of the next video frame; and online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain a vehicle artificial intelligence perception network model after online training.

[0085] As can be seen, in the embodiments of the present invention, the vehicle artificial intelligence perception network model can be trained online in real time while the vehicle is in motion, enabling online feedback and autonomous learning of the vehicle artificial intelligence perception network model. The online-trained vehicle artificial intelligence perception network model has higher perception accuracy, improving vehicle safety. Furthermore, online training does not require the collection of dedicated training data, reducing the training cost of the vehicle artificial intelligence perception network model. The online autonomous learning method of the present invention improves the reliability and safety of intelligent perception and decision-making in unknown scenarios and operating conditions.

[0086] Below Figure 2 Take this as an example to illustrate: Figure 2 The image space in is the image contained in each video frame in the video stream captured by the vehicle's camera, and the video input is to input the video stream into the vehicle's artificial intelligence perception network model. Figure 2 The feature network in is the above-mentioned vehicle artificial intelligence perception network model, which is used to extract the feature information of each video frame in the video stream. The feature space corresponding to time t-1 can be understood as the feature information of the current video frame, and the feature space corresponding to the predicted time t can be the feature information of the next video frame. The feature space corresponding to time t can be the real feature information of the current video frame extracted. By calculating the error between the predicted feature information of the current video frame and the real two-dimensional feature information, a gradient space can be constructed. The gradient space is an online autonomous learning strategy. Figure 2 The feedback network shown feeds back the online autonomous learning strategy to the feature network, and performs online feedback learning and feedback training on the vehicle artificial intelligence perception network model through the autonomous learning strategy to obtain the vehicle artificial intelligence perception network model after online training.

[0087] exist Figure 1 Based on the embodiment shown, in one embodiment, as Figure 3 As shown, the method may further include the steps of:

[0088] S310, outputting multiple perception results through the vehicle artificial intelligence perception network model.

[0089] Each perception result is used to complete a perception task, and different perception results complete different perception tasks.

[0090] Specifically, the vehicle artificial intelligence perception network model can form multiple perception task heads, that is, output multiple perception results, where the multiple perception results can include target recognition results, obstacle recognition results, motion feature perception results, operational space recognition results, semantic information results, lane line recognition results, traffic sign recognition results, etc., which are not specifically limited here.

[0091] S320: For each perception result, obtain a calibration result corresponding to the perception result from a pre-built database.

[0092] S330, calculating the loss function value of the vehicle artificial intelligence perception network model based on the perception results and calibration results.

[0093] Specifically, for each perception result, the calibration result corresponding to the perception result can be obtained from a pre-built database, and the loss function value of the vehicle artificial intelligence perception network model can be calculated based on the perception result and the calibration result to evaluate the perception result.

[0094] S340: Based on the calculated loss function value, the online autonomous learning strategy of the vehicle artificial intelligence perception network model is adjusted to obtain an adjusted online autonomous learning strategy.

[0095] Specifically, if the calculated loss function value is large, it means that the accuracy of the perception result is low. Therefore, based on the calculated loss function value, the online autonomous learning strategy of the vehicle artificial intelligence perception network model can be adjusted, that is, the model parameters of the vehicle artificial intelligence perception network model can be adjusted, so as to realize the online enhanced learning of the vehicle artificial intelligence perception network model.

[0096] by Figure 4 Take this as an example to illustrate: Figure 4 The image space in the image refers to the image contained in each video frame in the video stream captured by the vehicle's camera, and the video input refers to inputting the video stream into the vehicle's artificial intelligence perception network model. The feature information network is used to extract feature information from each video frame, and the feature space refers to the extracted feature information. The typical target recognition network is used to output the perception results. The typical knowledge base is a pre-constructed database. The loss function value of the vehicle's artificial intelligence perception network model is calculated based on the perception results and the calibration results in the typical knowledge base. If the loss function value is large, the online autonomous learning strategy of the vehicle's artificial intelligence perception network model is adjusted to obtain the adjusted online autonomous learning strategy.

[0097] exist Figure 4 Based on the illustrated embodiment, in one implementation, S140, performing online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on an online autonomous learning strategy to obtain an online trained vehicle artificial intelligence perception network model, may include the following steps:

[0098] S350, based on the adjusted online autonomous learning strategy, performs online feedback learning and feedback training on the vehicle artificial intelligence perception network model to obtain a new online trained vehicle artificial intelligence perception network model.

[0099] Specifically, the adjusted online autonomous learning strategy is obtained by integrating online feedback learning and online reinforcement learning. By performing feedback learning and feedback training on the vehicle artificial intelligence perception network model through the adjusted online autonomous learning strategy, the accuracy of the perception results of the vehicle artificial intelligence perception network model can be further improved.

[0100] It can be seen that in the embodiment of the present invention, during the movement of the vehicle, the vehicle artificial intelligence perception network model can be subjected to online feedback learning and online reinforcement learning in real time. The perception results output by the vehicle artificial intelligence perception network model after online training are more accurate, thereby improving the safety of the vehicle. In addition, online training does not require the special collection of training data, thereby reducing the training cost of the vehicle artificial intelligence perception network model.

[0101] The embodiment of the present invention also provides an artificial intelligence method specifically for vehicle driving control, such as Figure 5 As shown, the method may include the following steps:

[0102] S510, obtaining a video stream captured by a camera of the vehicle during the vehicle's driving process.

[0103] S520: Input the video stream into the online trained vehicle artificial intelligence perception network model to output a perception result through the online trained vehicle artificial intelligence perception network model.

[0104] S530: Control the vehicle accordingly based on the perception result.

[0105] Specifically, in actual applications, during the driving process of the vehicle, the video stream collected by the vehicle's camera is input into the vehicle artificial intelligence perception network model obtained through online training through the above embodiment. After receiving the video stream, the online trained vehicle artificial intelligence perception network model extracts the feature information of each video frame in the video stream, and outputs the perception result based on the feature information, and then controls the vehicle accordingly according to the perception result. For example, if the perception result is that there is an obstacle in front of the vehicle, then the vehicle can be controlled to avoid the obstacle. It is understandable that there are multiple perception results and multiple ways to control the vehicle, which will not be listed one by one here.

[0106] The technical solution of the embodiment of the present invention can perform online feedback learning and online reinforcement learning on the vehicle artificial intelligence perception network model in real time during the movement of the vehicle. The perception results output by the vehicle artificial intelligence perception network model after online training are more accurate, thereby improving the safety of the vehicle. In addition, online training does not require the special collection of training data, thereby reducing the training cost of the vehicle artificial intelligence perception network model.

[0107] The embodiment of the present invention also provides an artificial intelligence device 60 specifically for vehicle driving and control, such as Figure 6 As shown, the device includes:

[0108] The video stream acquisition module 610 is used to acquire the video stream captured by the camera of the vehicle while the vehicle is traveling;

[0109] a feature information extraction and prediction module 620 for extracting feature information of each video frame of the video stream using a vehicle artificial intelligence perception network model, and predicting feature information of a next video frame based on the feature information of a current video frame, where the current video frame is any video frame of the video stream;

[0110] an online autonomous learning strategy determination module 630 for determining an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and the error between the predicted feature information of the next video frame;

[0111] The model online training module 640 is used to perform online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain the vehicle artificial intelligence perception network model after online training.

[0112] The embodiment of the present invention provides an artificial intelligence device 70 specifically for vehicle driving control, such as Figure 7 As shown, the device includes:

[0113] The video stream acquisition module 710 is used to acquire the video stream collected by the camera of the vehicle during the vehicle's driving process;

[0114] A perception result acquisition module 720 is configured to input the video stream into the online trained vehicle artificial intelligence perception network model to output a perception result through the online trained vehicle artificial intelligence perception network model;

[0115] The vehicle control module 730 is used to control the vehicle accordingly based on the perception result.

[0116] The embodiment of the present invention further provides an electronic device 800, such as Figure 8 Shown, including:

[0117] at least one processor 801;

[0118] a memory 802 for storing the at least one processor-executable instruction;

[0119] The at least one processor is configured to execute the instructions to implement the above method.

[0120] In a sixth aspect, an embodiment of the present invention provides a computer-readable storage medium, which, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the above method.

[0121] In a seventh aspect, an embodiment of the present invention provides a computer program product, comprising a computer program, which implements the above method when executed by a processor.

[0122] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and purpose of the present invention.

Claims

1. A trustworthy artificial intelligence method specifically for vehicle driving and control, characterized in that: The method comprises: During the driving of the vehicle, a video stream captured by a camera of the vehicle is obtained; Extracting feature information of each video frame of the video stream through a vehicle artificial intelligence perception network model, and predicting feature information of a next video frame based on the feature information of a current video frame, where the current video frame is any video frame of the video stream; Determining an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and an error between the predicted feature information of the next video frame; Performing online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain an online trained vehicle artificial intelligence perception network model; Outputting multiple perception results through the vehicle artificial intelligence perception network model; each perception result is used to complete a perception task, and different perception results complete different perception tasks; For each perception result, obtain the calibration result corresponding to the perception result from a pre-built database; Calculating a loss function value of the vehicle artificial intelligence perception network model based on the perception results and the calibration results; Based on the calculated loss function value, adjusting the online autonomous learning strategy of the vehicle artificial intelligence perception network model to obtain an adjusted online autonomous learning strategy; Based on the adjusted online autonomous learning strategy, online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model to obtain a new online trained vehicle artificial intelligence perception network model.

2. The method according to claim 1, characterized in that The extracting feature information of each video frame of the video stream by using the vehicle artificial intelligence perception network model includes: Inputting the video stream into a vehicle artificial intelligence perception network model; The vehicle artificial intelligence perception network model is used to extract feature information in multiple dimensions from each video frame.

3. The method according to claim 2, characterized in that The predicting feature information of the next video frame based on feature information of the current video frame includes: For feature information of each dimension, feature information of the dimension of the next video frame is predicted based on the feature information of the dimension of the current video frame.

4. The method according to claim 2, characterized in that Determining an online autonomous learning strategy of the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and the error between the predicted feature information of the next video frame includes: For feature information of each dimension, calculating an error between first feature information and second feature information, where the first feature information is the extracted feature information of the next video frame in the dimension, and the second feature information is the predicted feature information of the next video frame in the dimension; An online autonomous learning strategy of the vehicle artificial intelligence perception network model is determined based on the calculated multiple errors.

5. An artificial intelligence method specifically for vehicle driving control, characterized in that: The method comprises: Obtaining a video stream captured by a camera of the vehicle during the vehicle's driving process; Inputting the video stream into the online trained vehicle artificial intelligence perception network model according to any one of claims 1 to 4, so as to output a plurality of perception results through the online trained vehicle artificial intelligence perception network model; The vehicle is controlled accordingly based on the multiple perception results.

6. A trusted artificial intelligence device specifically for vehicle driving and control, characterized in that: The device comprises: A video stream acquisition module is used to acquire the video stream captured by the vehicle's camera while the vehicle is traveling; a feature information extraction and prediction module, configured to extract feature information of each video frame of the video stream through a vehicle artificial intelligence perception network model, and predict feature information of a next video frame based on the feature information of a current video frame, where the current video frame is any video frame of the video stream; an online autonomous learning strategy determination module, configured to determine an online autonomous learning strategy for the vehicle artificial intelligence perception network model based on the extracted feature information of the next video frame and an error between the predicted feature information of the next video frame; A vehicle artificial intelligence perception network model training module is used to perform online feedback learning and feedback training on the vehicle artificial intelligence perception network model based on the online autonomous learning strategy to obtain an online trained vehicle artificial intelligence perception network model; The vehicle artificial intelligence perception network model training module is further used to output multiple perception results through the vehicle artificial intelligence perception network model; each perception result is used to complete a perception task, and different perception results complete different perception tasks; For each perception result, obtain the calibration result corresponding to the perception result from a pre-built database; Calculating a loss function value of the vehicle artificial intelligence perception network model based on the perception results and the calibration results; Based on the calculated loss function value, adjusting the online autonomous learning strategy of the vehicle artificial intelligence perception network model to obtain an adjusted online autonomous learning strategy; Based on the adjusted online autonomous learning strategy, online feedback learning and feedback training are performed on the vehicle artificial intelligence perception network model to obtain a new online trained vehicle artificial intelligence perception network model.

7. An artificial intelligence device specifically designed for vehicle driving control, characterized in that: The device comprises: A video stream acquisition module is used to acquire the video stream collected by the camera of the vehicle during the vehicle's driving process; a perception result acquisition module, configured to input the video stream into the online trained vehicle artificial intelligence perception network model according to any one of claims 1 to 4, so as to output a plurality of perception results through the online trained vehicle artificial intelligence perception network model; A vehicle control module is used to control the vehicle accordingly based on the multiple perception results.

8. An electronic device, characterized in that: include: at least one processor; a memory for storing the at least one processor-executable instruction; The at least one processor is configured to execute the instructions to implement the method according to any one of claims 1 to 4 or claim 5.

Images