Vehicle risk model training method and device, electronic equipment and computer medium

Abstract

The invention provides a vehicle risk model training method and device, electronic equipment and a computer medium, and the method comprises the steps: obtaining the driving data and environment perception data of an automatic driving vehicle, and the driving data comprises the actual driving data and the planned driving data generated by an initial risk training model; taking data, different from the planned driving data, in the actual driving data as manual intervention driving data; determining first sensing data matched with the manual intervention driving data in the environment sensing data, and determining risk level information matched with the manual intervention driving data according to the first sensing data; and training an initial risk training model based on the risk level information, the first perception data, the planning driving data and the manual intervention driving data. The driving safety of the automatic driving vehicle can be improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and particularly to a method, device, electronic device, and computer medium for training a vehicle risk model. Background Art

[0002] With the continuous development of the field of intelligent driving technology, the market share of vehicles with autonomous driving mode or assisted driving mode is also increasing continuously, and more and more users purchase vehicles with autonomous driving mode or assisted driving mode.

[0003] When the vehicle is in the autonomous driving mode or assisted driving mode, the vehicle can obtain environmental information through sensors and control the vehicle's driving according to the environmental information, reducing the driver's operation and improving the driver's driving experience. However, at present, the development of the field of intelligent driving technology is not yet mature enough, and there are problems with driving insecurity in the driving of autonomous vehicles. Summary of the Invention

[0004] The present invention provides a method, device, electronic device, and computer medium for training a vehicle risk model, which can improve the driving safety of autonomous vehicles.

[0005] The technical solution of the present invention: A method for training a vehicle risk model includes: obtaining the driving data and environmental perception data of the autonomous vehicle, where the driving data includes actual driving data and planned driving data generated by an initial risk training model; taking the data in the actual driving data that is different from the planned driving data as manually intervened driving data; determining first perception data in the environmental perception data that matches the manually intervened driving data, and determining risk level information that matches the manually intervened driving data according to the first perception data; training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data.

[0006] The vehicle risk model training method of this application, based on the planned driving data generated by the initial risk training model, the real-time obtained environmental perception data, and the actual driving data, can accurately obtain the manually intervened driving data of the vehicle from the autonomous driving mode to the manually intervened state. Then, determine the first perception data that matches the manually intervened driving data, and the first perception data is the environmental perception data of the autonomous vehicle in the manually intervened state. Then, match the risk level according to the first perception data to determine the risk level of this person intervening to control the autonomous vehicle. Finally, train the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data to make the planned driving data trained by the initial risk training model more accurate.

[0007] In some embodiments, the risk level information includes emergency risks and non-emergency risks; before training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data, the method further includes: detecting whether the risk level information is an emergency risk; the training of the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data includes: when it is detected that the risk level information is an emergency risk, training the initial risk training model based on the first perception data matching the emergency risk, the planned driving data, and the manually intervened driving data matching the emergency risk.

[0008] In some embodiments, the risk level information includes emergency risks and non-emergency risks; before training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data, the method further includes: detecting whether the risk level information is an emergency risk; the training of the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data includes: when it is detected that the risk level information is a non-emergency risk, determining the risk priority of the risk level according to the first perception data matching the non-emergency risk; based on the risk priority, sorting the manually intervened driving data matching the non-emergency risk and adding it to the training task list; training the initial risk training model based on the training task list.

[0009] In some embodiments, the non-emergency risks include medium risks and low risks; the risk priorities include a first priority and a second priority, and the level of the first priority is higher than the level of the second priority, and the vehicle risk model preferentially trains the manually intervened driving data corresponding to the first priority; the determining the risk priority of the risk level according to the first perception data matching the non-emergency risk when it is detected that the risk level information is a non-emergency risk includes: if it is detected that the risk level information is a medium risk, determining the risk priority of the medium risk to be the first priority; if it is detected that the risk level information is a low risk, determining the risk priority of the low risk to be the second priority.

[0010] In some embodiments, the training of the initial risk training model based on the training task list includes: extracting the manually intervened driving data matching the first priority from the training task list; training the initial risk training model every first preset time based on the first perception data matching the first priority, the planned driving data, and the manually intervened driving data matching the first priority.

[0011] In some embodiments, it further includes: extracting the manual intervention driving data matching the second priority from the training task list; training the initial risk training model every second preset time based on the first perception data matching the second priority, the planned driving data, and the manual intervention driving data matching the second priority, wherein the duration of the second preset time is greater than the duration of the first preset time.

[0012] In some embodiments, obtaining the driving data and environmental perception data of the autonomous driving vehicle, wherein the driving data includes actual driving data and planned driving data generated by the initial risk training model, includes: obtaining the planned itinerary and environmental perception data of the autonomous driving vehicle; obtaining the actual driving data of the autonomous driving vehicle based on the environmental perception data; and obtaining the planned driving data generated by the initial risk training model based on the planned itinerary and the initial risk training model.

[0013] An embodiment of the present application further provides a vehicle risk model training device, including: an acquisition module, configured to acquire the driving data and environmental perception data of the autonomous driving vehicle, wherein the driving data includes actual driving data and planned driving data generated by the initial risk training model; an extraction module, configured to use the data different from the planned driving data in the actual driving data as the manual intervention driving data; a determination module, configured to determine the first perception data matching the manual intervention driving data in the environmental perception data, and determine the risk level information matching the manual intervention driving data according to the first perception data; and a training module, configured to train the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manual intervention driving data.

[0014] An embodiment of the present application further provides an electronic device, the electronic device includes a processor and a memory, the memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the above-mentioned vehicle risk model training method.

[0015] An embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, and when the computer instructions run on an electronic device, the electronic device is enabled to execute the above-mentioned vehicle risk model training method.

[0016] Compared with the prior art, the above vehicle risk model training method, device, electronic device, and computer-readable storage medium. In the driving process of an autonomous vehicle, the vehicle risk model training method obtains manual intervention driving data. And performs risk level classification on the manual intervention driving data. When the risk level information of the manual intervention driving data is an emergency risk, an initial risk training model is immediately trained based on the first perception data, planned driving data, and manual intervention driving data. To improve the driving safety of the autonomous vehicle. When the risk level information of the manual intervention driving data is a non-emergency risk, the manual intervention driving data is sorted by priority, and the sorted manual intervention driving data is added to the training task list. Then, at preset time intervals, different priority manual intervention driving data is extracted from the training task list. Finally, an initial risk training model is trained based on the first perception data, planned driving data, and manual intervention driving data. When the risk level information corresponding to the manual intervention driving data is not high, the initial risk training model is trained at preset time intervals to reduce the cost of model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flowchart of the steps of the vehicle risk model training method according to an embodiment of the present application.

[0018] Figure 2 is a schematic structural diagram of the vehicle risk model training device according to an embodiment of the present application.

[0019] Figure 3 is a schematic structural diagram of the electronic device according to an embodiment of the present application.

[0020] MAIN ELEMENT SYMBOL DESCRIPTION

[0021] Electronic device 100

[0022] Memory 20

[0023] Processor 30

[0024] Computer program 40

[0025] Vehicle risk model training device 200

[0026] Obtaining module 210

[0027] Extracting module 220

[0028] Determining module 230

[0029] Training module 240 DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] In order to more clearly understand the above objects, features and advantages of the present application, the present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0031] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application. The described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs. The terms used in the description of the present application herein are only for the purpose of describing specific embodiments, and are not intended to limit the present application.

[0033] Furthermore, it should be noted that in this document, the terms "comprises", "comprising" or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising such element.

[0034] In the embodiments of the present application, the term "at least one" means one or more, and "a plurality" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The terms "first", "second", "third", "fourth", etc. (if any) in the description, claims and drawings of the present application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0035] In the embodiments of the present application, words such as "exemplary" or "for example" are used to mean as an example, illustration or explanation. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific manner.

[0036] The vehicle risk model training method of the present application can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a processor, a microprogrammed control unit (MCU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc. The electronic device can be a portable electronic device (such as a mobile phone, a tablet computer), a personal computer, a server, etc.

[0037] Figure 1 It is a step flowchart of an embodiment of the vehicle risk model training method of the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted. This vehicle risk model training method is applied to a vehicle risk model training system. This vehicle risk model training system can be installed in an autonomous vehicle.

[0038] To facilitate the understanding of the technical solutions described in the present application, some terms in the present application are explained as follows:

[0039] In the embodiments of the present application, an autonomous vehicle (Autonomous vehicles), also known as a driverless car, a computer-driven car, or a wheeled mobile robot, is an intelligent vehicle that realizes driverless through a computer device. In practical applications, an autonomous vehicle relies on the collaborative cooperation of artificial intelligence, visual computing, radar, monitoring devices, and global positioning devices, enabling the computer device to automatically and safely operate a motor vehicle without any active operation by humans.

[0040] Refer to Figure 1 As shown, the vehicle risk model training method may include the following steps.

[0041] S100. Obtain the driving data and environmental perception data of the autonomous vehicle, where the driving data includes the actual driving data and the planned driving data generated by the initial risk training model.

[0042] In some embodiments, a data collection device is installed in an autonomous vehicle. The data collection device can be an in-vehicle sensor, radar, camera, Global Positioning System (GPS), etc. The in-vehicle sensor is mainly used to detect short-range obstacles around the vehicle, such as distance perception when parking. Radar mainly detects the position and speed of surrounding objects by emitting radio waves and measuring their reflections. The camera is used to capture images on the road to help identify other vehicles, pedestrians, road signs, traffic signals, etc. GPS mainly obtains the real-time positioning information of the vehicle and plans the driving route for the vehicle. In other embodiments, the autonomous vehicle may further include other types of data collection devices, and the present application does not limit the types included in the data collection device.

[0043] In some embodiments, the data collection device is used to collect environmental perception data within a preset range centered on the autonomous vehicle. Among them, the environmental perception data includes dynamic environmental perception data and static environmental perception data. The dynamic environmental perception data can be the movement information of dynamic obstacles, traffic signal information, etc. The static environmental perception data can be the width of the road, the position information of static obstacles, etc. The preset range can be 1 km or 1.5 km, and the present application does not limit the specific value of the preset range.

[0044] For example, the data collection device can collect environmental perception data within a radius of 1 km centered on the autonomous vehicle. The environmental perception data includes the driving state of other vehicles, the lane information of the road, the position information of road signs, the information of traffic command devices, etc.

[0045] In some embodiments, the data collection device is also used to obtain the departure position and target position of the autonomous vehicle. Based on the departure position and target position, a planned itinerary is planned by GPS. In this embodiment, the vehicle risk model training system obtains the planned itinerary of the autonomous vehicle and the environmental perception data of the autonomous vehicle. Based on the environmental perception data, the actual driving data of the autonomous vehicle is obtained. The actual driving data refers to the actual driving path information of the autonomous vehicle. Based on the planned itinerary and the initial risk training model, the planned driving data generated by the initial risk training model is obtained.

[0046] For example, the starting position of the driver is Location A, and the target position is Location B. Based on Location A and Location B, the vehicle risk model training system can obtain the planned itinerary between Location A and Location B. In this embodiment, the driver drives the autonomous vehicle to start along the planned itinerary from Location A to Location B. During the driving process, the vehicle risk model training system obtains environmental perception data in real time, and based on the real-time obtained environmental perception data, obtains the actual driving data of the autonomous vehicle. At the same time, based on the planned itinerary and the initial risk training model, the planned driving data generated by the initial risk training model is obtained.

[0047] S200. Use the data in the actual driving data that is different from the planned driving data as the manually intervened driving data.

[0048] In some embodiments, the actual driving data may include the actual driving path and actual braking information. The vehicle risk model training system can obtain the actual driving path and actual braking information of the autonomous vehicle in real time. Among them, the actual braking information includes actual steering wheel rotation information, actual throttle information, actual braking information, etc. In other embodiments, the actual braking information may further include emergency braking information and other braking information other than the emergency braking information. The present application does not limit the content of the actual braking information.

[0049] Specifically, the planned driving data may include the planned driving path and planned braking information. The planned braking information includes planned steering wheel rotation information, planned throttle information, and planned braking information. The vehicle risk model training system compares the actual driving path with the planned driving path, and the actual braking information with the planned braking information. The different data between the actual driving data and the planned driving data, as well as the different data between the actual braking data and the planned braking data are jointly used as the manually intervened driving data.

[0050] For example, when the autonomous vehicle is driving on the first road, the first road has a first turn. When the autonomous vehicle is approaching the first turn, based on the planned driving data, the autonomous vehicle does not automatically change direction at the first turn. After the driver discovers this problem, the driver manually turns the steering wheel in time, so that the autonomous vehicle safely turns through the first turn. Based on this, the vehicle risk model training system compares the planned steering wheel rotation information and the actual steering wheel rotation information of the vehicle, and the planned steering wheel rotation information and the actual steering wheel rotation information are different. At the same time, the vehicle risk model training system compares the actual driving path with the planned driving path and finds them different. Therefore, the different data between the actual steering wheel rotation information and the planned steering wheel rotation information, and the different data between the actual driving path and the planned driving path are jointly used as the manually intervened driving data. For example, the manually intervened driving data is the rotation angle, rotation direction of the steering wheel under manual intervention at the first turn of the first road, and the actual driving position information of the autonomous vehicle, etc.

[0051] In some embodiments, the actual driving data, planned driving data, environment perception data, and manual intervention driving data are stored in a preset database, so as to facilitate subsequent training of the initial risk training model based on the data in the preset database.

[0052] S300. Determine first perception data that matches the manual intervention driving data from the environment perception data, and determine risk level information that matches the manual intervention driving data according to the first perception data.

[0053] In this embodiment, the risk level information may include emergency risks and non-emergency risks. Among them, non-emergency risks include medium risks, low risks, and no risks. Determine the first perception data that matches the manual intervention data from the dynamic environment perception data and the static environment perception data. The first perception data can facilitate the vehicle risk model training system to analyze the planned driving data and the manual intervention driving data, so as to determine the risk level information that matches the manual intervention driving data.

[0054] For example, in step S200, based on the dynamic environment perception data and the static environment perception data at the first turning point. Among them, the dynamic environment perception data may include the driving information of other vehicles and traffic signal information, etc.; the static environment perception data includes the turning angle and turning position information of the first turning, etc. The vehicle risk model training system analyzes the planned driving data (i.e., the driving data of the autonomous vehicle without turning the direction at the first turning point) and the manual intervention driving data (when the steering wheel of the autonomous vehicle is turned manually at the first turning point, the actual driving data of the autonomous vehicle), and obtains that if the vehicle is not manually controlled, the vehicle will hit the road railing and cause damage to the vehicle. At this time, the first perception data is the position information of the first turning of the first road. Therefore, based on the first perception data, determine that the risk level information of this manual intervention driving data is medium risk.

[0055] In one embodiment, it is assumed that the autonomous driving vehicle is traveling on the second road at a relatively fast speed. At the same time, there is a pothole at a certain position on the second road. When the autonomous driving vehicle is about to reach the pothole, based on the planned driving data, the autonomous driving vehicle does not reduce the speed at the pothole to smoothly pass the pothole. After the driver discovers the problem, he promptly steps on the foot brake, so that the autonomous driving vehicle smoothly passes the pothole. Based on this, the vehicle risk model training system can obtain the manual intervention driving data of the vehicle on the second road. The manual intervention driving data is the pedaling stroke of the foot brake under manual intervention at the pothole on the second road, the driving speed of the autonomous driving vehicle, etc. The planned driving data (i.e., the driving data when the autonomous driving vehicle does not reduce the speed at the pothole) and the manual intervention driving data (the actual driving data of the autonomous driving vehicle when the foot brake of the autonomous driving vehicle is stepped on at the pothole) are analyzed to obtain that if the human does not intervene to control the vehicle, the vehicle will collide at the pothole. At this time, the first perception data is the location information of the pothole on the second road. Therefore, based on the first perception data, it is determined that the risk level information of the manual intervention driving data is low risk.

[0056] In another embodiment, it is assumed that the autonomous driving vehicle is traveling on a third road, and the road light at a certain intersection on the third road is about to change from red to green. When the autonomous driving vehicle is about to reach the intersection, the autonomous driving vehicle reduces its speed based on the planned driving data. After the driver discovers the problem, he steps on the accelerator, so that the autonomous driving vehicle passes the intersection when the road light just changes from red to green. Based on this, the vehicle risk model training system can obtain the manual intervention driving data of the vehicle on the third road. The manual intervention driving data is the pedaling stroke of the accelerator under manual intervention at the road light of the third road. The planned driving data (i.e., the driving data when the autonomous driving vehicle reduces its speed at the intersection) and the manual intervention driving data (the actual driving data of the autonomous driving vehicle when the accelerator of the autonomous driving vehicle is stepped on at the intersection) are analyzed to obtain that if the human does not intervene to control the vehicle, the vehicle will stop first and wait until the road light has turned green before passing through the intersection to reduce the driving risk. At this time, the first perception data is the dynamic transformation information of the road light of the third road. Therefore, based on the first perception data, it is determined that the risk level information of this manual intervention in the driving data is risk-free.

[0057] In yet another embodiment, it is assumed that an autonomous vehicle is traveling on a fourth road. At the same time, there is also an obstacle vehicle traveling on the fourth road, and the obstacle vehicle is in front of the vehicle. When the obstacle vehicle brakes suddenly during driving, the vehicle does not take emergency braking measures based on the planned driving path. After the driver discovers this problem, emergency braking measures are taken in a timely manner, enabling the autonomous vehicle to safely bypass the obstacle vehicle. Based on this, the vehicle risk model training system can obtain the manual intervention driving data of the vehicle on the fourth road. The manual intervention driving data includes the time of manual intervention, the angle of the steering wheel rotation, etc. when bypassing the obstacle vehicle on the fourth road. By analyzing the planned driving data (i.e., the driving speed and vehicle position information of the autonomous vehicle when the obstacle vehicle brakes suddenly) and the manual intervention driving data (the actual driving data of the autonomous vehicle under emergency braking), it is obtained that if the vehicle is not manually controlled, the vehicle will hit the obstacle vehicle, resulting in a driving accident and causing injuries to the occupants in the vehicle. At this time, the first perception data is the position information of the obstacle vehicle on the fourth road and the dynamic information of the sudden braking. Therefore, based on the first perception data, the risk level information of this manual intervention driving data is determined to be an emergency risk.

[0058] It can be understood that when the risk level information of the manual intervention driving data is determined to be an emergency risk, it proves that based on the planned driving data, if the vehicle is not manually controlled, the occupants in the vehicle will be injured or the vehicle will cause injuries to pedestrians. When the risk level information of the manual intervention driving data is determined to be a medium risk, it proves that based on the planned driving data, if the vehicle is not manually controlled, the vehicle will have a serious impact, resulting in a relatively high maintenance cost for the subsequent vehicle. When the risk level information of the manual intervention driving data is determined to be a low risk, it proves that based on the planned driving data, if the vehicle is not manually controlled, the vehicle will have a minor impact or collision, which has no effect on the driving of the vehicle and the maintenance cost of the vehicle is relatively low. When the risk level information of the manual intervention driving data is determined to be a no risk, it proves that based on the planned driving data, if the vehicle is not manually controlled, the vehicle will not have any driving accidents and will not incur any maintenance costs.

[0059] S400. Train the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manual intervention driving data.

[0060] In some embodiments, it is detected whether the risk level information is an emergency risk. If it is detected that the risk level information is an emergency risk, an initial risk training model is trained based on the first perception data matched with the emergency risk, the planned driving data, and the manual intervention driving data matched with the emergency risk. For example, the risk level information corresponding to the manual intervention driving data when the autonomous vehicle is driving on the fourth road is an emergency risk. In order to avoid the autonomous vehicle being unable to take emergency braking measures in a timely manner when encountering such emergencies in the future, thereby improving driving safety and protecting personnel from harm. After the vehicle risk model training system manually intervenes to control the autonomous vehicle to safely pass an obstacle vehicle, it is necessary to immediately train the initial risk training model based on the first perception data matched with the emergency risk, the planned driving data, and the manual intervention driving data matched with the emergency risk. To improve the accuracy of the planned driving data for the later training of the initial risk training model.

[0061] In some embodiments, if it is detected that the risk level information is a non-emergency risk, the risk priority of the risk level is determined according to the first perception data matched with the non-emergency risk. Based on the risk priority, the manual intervention driving data matched with the non-emergency risk is sorted and then added to the training task list. The initial risk training model is trained based on the training task list. In this embodiment, the risk priority includes a first priority and a second priority, and the level of the first priority is higher than the level of the second priority. The vehicle risk model preferentially trains the manual intervention driving data corresponding to the first priority. Specifically, if it is detected that the risk level information is a medium risk, the risk priority of the medium risk is determined to be the first priority. If it is detected that the risk level information is a low risk, the risk priority of the low risk is determined to be the second priority. The vehicle risk model preferentially trains the manual intervention driving data corresponding to the medium risk.

[0062] In some embodiments, the manual intervention driving data matched with the first priority is extracted from the training task list. Based on the first perception data matched with the first priority, the planned driving data, and the manual intervention driving data matched with the first priority, the initial risk training model is trained every first preset time. The manual intervention driving data matched with the second priority is extracted from the training task list. Based on the first perception data matched with the second priority, the planned driving data, and the manual intervention driving data matched with the second priority, the initial risk training model is trained every second preset time, where the duration of the second preset time is greater than the duration of the first preset time.

[0063] For example, in step S300, when the autonomous vehicle is driving at the first turn of the first road, the manual intervention driving data is of medium risk. For ease of understanding, this manual intervention driving data is denoted as medium-risk - manual intervention driving data here, and the risk priority corresponding to the medium-risk - manual intervention driving data is the first priority. When the autonomous vehicle is driving at the pothole of the second road, the manual intervention driving data is of low risk, and this manual intervention driving data is denoted as low-risk - manual intervention driving data, and the risk priority corresponding to the low-risk - manual intervention driving data is the second risk. Then, after arranging the medium-risk - manual intervention driving data and the low-risk - manual intervention driving data, they are added to the training task list.

[0064] In this embodiment, the first preset time is 48 hours, and the second preset time is 7 days. The vehicle risk model training system extracts the medium-risk - manual intervention driving data from the training task list every 48 hours, and trains the initial risk training model based on the medium-risk - manual intervention driving data, the first perception data, and the planned driving data. The vehicle risk model training system extracts the low-risk - manual intervention driving data from the training task list every 7 days, and trains the initial risk training model based on the low-risk - manual intervention driving data, the first perception data, and the planned driving data. In this way, the safety of the planned driving path planned by the initial risk training model is improved, and the occurrence of driving accidents during the driving process is avoided.

[0065] In this embodiment, the risk priority further includes a third priority level. The level of the third priority level is lower than that of the second priority level. For example, in step S300, when the autonomous vehicle is driving at the road traffic signal of the third road, the manual intervention driving data is risk-free. Then this manual intervention driving data is denoted as risk-free - manual intervention driving data, and the risk priority corresponding to the risk-free - manual intervention driving data is the third priority. After arranging the medium-risk - manual intervention driving data, the low-risk - manual intervention driving data, and the risk-free - manual intervention driving data, they are added to the training task list.

[0066] The vehicle risk model training system extracts the risk-free - manual intervention driving data from the training task list every third preset time, and trains the initial risk training model based on the risk-free - manual intervention driving data, the environmental perception data, and the planned driving data. Among them, the third preset time can be one month.

[0067] In other embodiments, the first preset time can be 24 hours or 72 hours, the second preset time can be 5 days or 10 days, and the third preset time can be 15 days or 20 days. The first preset time, the second preset time, and the third preset time can be set according to the actual situation.

[0068] The vehicle risk model training method of the present application obtains manual intervention driving data during the driving process of an autonomous vehicle. And classify the risk level of the manual intervention driving data. When the risk level information of the manual intervention driving data is an emergency risk, immediately train the initial risk training model based on the first perception data, planned driving data, and manual intervention driving data. To improve the driving safety of the autonomous vehicle. When the risk level information of the manual intervention driving data is a non-emergency risk, sort the manual intervention driving data by priority, and add the sorted manual intervention driving data to the training task list. After that, extract the manual intervention driving data with different priorities from the training task list at preset time intervals. Finally, train the initial risk training model based on the first perception data, planned driving data, and manual intervention driving data. When the risk level information corresponding to the manual intervention driving data is not high, train the initial risk training model at preset time intervals to reduce the cost of model training.

[0069] In some embodiments, the present application also discloses a vehicle risk model training device 200. As Figure 2 shown, the vehicle risk model training device 200 includes an acquisition module 210, an extraction module 220, a determination module 230, and a training module 240. The acquisition module 210 is used to acquire the driving data and environmental perception data of the autonomous vehicle, where the driving data includes the actual driving data and the planned driving data generated by the initial risk training model. The extraction module 220 is used to use the data in the actual driving data that is different from the planned driving data as the manual intervention driving data. The determination module 230 is used to determine the first perception data that matches the manual intervention driving data in the environmental perception data, and determine the risk level information that matches the manual intervention driving data according to the first perception data. The training module 240 is used to train the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manual intervention driving data.

[0070] In some embodiments, the present application also discloses an electronic device 100, as Figure 3 shown, the electronic device 100 includes a memory 20, a processor 30, and a computer program 40 stored in the memory 20 and executable on the processor 30. When the processor 30 executes the computer program 40, it implements the steps in the embodiments of the above vehicle risk model training method, such as Figure 1 the steps 100 to 400 shown.

[0071] Exemplarily, the computer program 40 can also be divided into one or more modules / units, which are stored in the memory 20 and executed by the processor 30. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 40 in the electronic device 100.

[0072] Those skilled in the art can understand that the schematic diagram is only an example of the electronic device 100, and does not constitute a limitation on the electronic device 100. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the electronic device 100 may also include input / output devices, network access devices, buses, etc.

[0073] The processor 30 can be a central processing unit (CPU), or can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, a single-chip microcomputer, or the processor 30 can also be any conventional processor, etc.

[0074] The memory 20 can be used to store the computer program 40 and / or modules / units. The processor 30 realizes various functions of the electronic device 100 by running or executing the computer programs and / or modules / units stored in the memory 20, and by invoking the data stored in the memory 20. The memory 20 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the electronic device 100 (such as audio data, etc.). In addition, the memory 20 can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0075] If the modules / units integrated in the electronic device 100 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present application, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0076] In several embodiments provided by the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the above-described electronic device embodiments are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation.

[0077] In addition, in each embodiment of the present application, the various functional units can be integrated in the same processing unit, or each unit can exist physically alone, or two or more units can be integrated in the same unit. The above-mentioned integrated units can be implemented in the form of hardware, or in the form of hardware plus software functional modules.

[0078] For those skilled in the art, it is obvious that the present application is not limited to the details of the above-described exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. In addition, obviously, the word "including" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or electronic devices stated in the electronic device claims can also be implemented by the same unit or electronic device through software or hardware. The words "first", "second", etc. are used to represent names and do not represent any specific order.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A vehicle risk model training method, applied to an autonomous vehicle, characterized in that, The method includes: Obtaining the driving data and environmental perception data of the autonomous vehicle, where the driving data includes the actual driving data and the planned driving data generated by the initial risk training model; Regarding the data in the actual driving data that is different from the planned driving data as the manually intervened driving data; Determining first perception data that matches the manually intervened driving data in the environmental perception data, and determining risk level information that matches the manually intervened driving data according to the first perception data; Training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data.

2. The vehicle risk model training method according to claim 1, wherein The risk level information includes emergency risk and non-emergency risk; before training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data, the method further includes: Detecting whether the risk level information is the emergency risk; The training of the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data includes: When it is detected that the risk level information is the emergency risk, training the initial risk training model based on the first perception data that matches the emergency risk, the planned driving data, and the manually intervened driving data that matches the emergency risk.

3. The vehicle risk model training method according to claim 1, wherein The risk level information includes emergency risk and non-emergency risk; before training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data, the method further includes: Detecting whether the risk level information is the emergency risk; The training of the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data includes: When it is detected that the risk level information is the non-emergency risk, determining the risk priority of the risk level according to the first perception data that matches the non-emergency risk; Based on the risk priority, sorting the manually intervened driving data that matches the non-emergency risk and adding it to the training task list; Training the initial risk training model based on the training task list.

4. The vehicle risk model training method according to claim 3, characterized in that, The non-emergency risk includes medium risk and low risk; the risk priority includes the first priority and the second priority, where the level of the first priority is higher than the level of the second priority, and the vehicle risk model preferentially trains the manually intervened driving data corresponding to the first priority; when it is detected that the risk level information is the non-emergency risk, determining the risk priority of the risk level according to the first perception data that matches the non-emergency risk includes: If it is detected that the risk level information is the medium risk, determining the risk priority of the medium risk as the first priority; If it is detected that the risk level information is the low risk, determining the risk priority of the low risk as the second priority.

5. The vehicle risk model training method according to claim 4, characterized in that Training the initial risk training model based on the training task list includes: Extracting the manually intervened driving data that matches the first priority from the training task list; Training the initial risk training model every first preset time based on the first perception data that matches the first priority, the planned driving data, and the manually intervened driving data that matches the first priority.

6. The vehicle risk model training method according to claim 5, characterized in that It further includes: Extracting the manually intervened driving data that matches the second priority from the training task list; Training the initial risk training model every second preset time based on the first perception data that matches the second priority, the planned driving data, and the manually intervened driving data that matches the second priority, where the duration of the second preset time is greater than the duration of the first preset time.

7. The vehicle risk model training method according to claim 1, wherein Obtaining the driving data and environmental perception data of the autonomous vehicle includes: Obtaining the planned itinerary and environmental perception data of the autonomous vehicle; Based on the environmental perception data, obtaining the actual driving data of the autonomous vehicle; Based on the planned itinerary and the initial risk training model, obtaining the planned driving data generated by the initial risk training model.

8. A vehicle risk model training device, characterized in that, It includes: An acquisition module for acquiring the driving data and environmental perception data of the autonomous vehicle, where the driving data includes the actual driving data and the planned driving data generated by the initial risk training model; An extraction module for using the data in the actual driving data that is different from the planned driving data as the manually intervened driving data; A determination module for determining the first perception data that matches the manually intervened driving data in the environmental perception data, and determining the risk level information that matches the manually intervened driving data according to the first perception data; A training module for training the initial risk training model based on the risk level information, the first perception data, the planned driving data, and the manually intervened driving data.

9. An electronic device, the electronic device comprising a processor and a memory, characterized in that, The memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the vehicle risk model training method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and when the computer instructions run on the electronic device, the electronic device executes the vehicle risk model training method according to any one of claims 1 to 7.

Images