Self-learning method and system of vehicle intelligent driving longitudinal control system

Abstract

The invention discloses a self-learning method and system of a vehicle intelligent driving longitudinal control system. The method comprises the steps that whether a vehicle meets the self-learning condition of the vehicle intelligent driving longitudinal control system or not is judged according to a vehicle state signal; when the conditions are met, the driving state data of the vehicle are obtained, the driving state data of the vehicle are preprocessed, and the current driving scene is recognized; analyzing and training the driving state data of each driving scene to obtain a driving key parameter-collision time graph corresponding to each driving scene; respectively judging whether data in the driving key parameter-collision time graph corresponding to each driving scene meets a preset safety requirement or not; and pushing the driving key parameter-collision time graph meeting the requirements to a vehicle intelligent driving longitudinal control system or pushing the corrected driving key parameter-collision time graph to the vehicle intelligent driving longitudinal control system. The vehicle can be intelligently controlled according to the habits of a driver.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and particularly to a self-learning method and system for an intelligent driving control system. Background Art

[0002] With the development of intelligent driving, more and more vehicles are equipped with intelligent driving technology. By longitudinally controlling the vehicle through intelligent driving, the driving pressure on the driver can be reduced and the driver can be helped to make more accurate judgments to control the vehicle. The above-mentioned longitudinal control is defined as the control in the driving speed direction, that is, the automatic control of the vehicle speed and the distance between the vehicle and the vehicle in front and behind or obstacles.

[0003] Most intelligent driving control systems collect vehicle signals and generate control commands according to the signals and send them to the vehicle computer. For example, in CN116279593A, when using such an intelligent driving control system, the intelligent driving system may not be able to consider more complex situations when controlling the vehicle and may have control methods that do not conform to the driving expectations. Therefore, in the prior art, there are problems that intelligent driving control cannot meet more complex situations and cannot meet driving expectations. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems that intelligent driving control cannot meet more complex situations and cannot meet expectations. To solve the above technical problems, an embodiment of the present invention discloses a self-learning method for a vehicle intelligent driving longitudinal control system, including the following steps:

[0005] S1: Obtain vehicle state signals, and determine whether the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system according to the vehicle state signals;

[0006] If so, execute step S2;

[0007] If not, continue to determine whether the vehicle meets the self-learning conditions of the intelligent driving longitudinal control system;

[0008] S2: Obtain the driving state data of the vehicle, preprocess the driving state data of the vehicle, and identify the current driving scenario of the vehicle; wherein, the driving scenarios include a stable following scenario, a following start scenario, a following braking scenario, a following stop scenario, and a following acceleration scenario; S3: Analyze and train the driving state data of each driving scenario respectively to obtain a driving critical parameter - time-to-collision graph corresponding to each driving scenario; S4: Determine whether the data in the driving critical parameter - time-to-collision graph corresponding to each driving scenario meets the preset safety requirements respectively; S5: Push the driving critical parameter - time-to-collision graph that meets the preset safety requirements in the driving critical parameter - time-to-collision graph corresponding to each driving scenario to the vehicle intelligent driving longitudinal control system; or, according to the preset correction principle, correct the driving critical parameter - time-to-collision graph that does not meet the preset safety requirements in the driving critical parameter - time-to-collision graph corresponding to each driving scenario to obtain an updated updated parameter - time-to-collision graph, and push the updated parameter - time-to-collision graph to the vehicle intelligent driving longitudinal control system.

[0009] Adopting the above technical solution, obtaining vehicle signals to determine whether to activate the above self-learning method, and performing self-learning of intelligent driving control parameters when the vehicle is stable and meets the conditions can effectively ensure driving safety during driving; further, scene recognition can be performed after obtaining vehicle state data to screen out following modes in different scenarios. Different following modes can ensure the comfort and accuracy of intelligent driving control in complex situations, and analyzing and training the above scenarios can ensure that the above following modes are more accurate, safe, and more in line with the driver's personal driving habits, ensuring safe driving of the vehicle during intelligent driving control; furthermore, the above technical solution determines whether the critical parameter - time-to-collision graph after analysis and training meets the safety requirements. If it meets the requirements, the critical parameter - time-to-collision graph can be used to control the vehicle; if not, it is corrected according to the predetermined principle and then used to control the vehicle. Determining whether it meets the safety requirements can ensure the safety of intelligent driving control of the vehicle and correct autonomous driving according to different situations, further ensuring the stability of the vehicle during intelligent control.

[0010] According to another specific embodiment of the present invention, in the self-learning method of a vehicle intelligent driving longitudinal control system disclosed in the embodiment of the present invention, the vehicle state signal in step S1 includes an intelligent driving control state signal and a gear signal; and when the intelligent driving state signal is not activated and the gear signal is in the forward gear, it is determined that the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system; if the intelligent driving state signal is activated and / or the gear signal is not in the forward gear, it is determined that the vehicle does not meet the self-learning conditions of the vehicle intelligent driving longitudinal control system.

[0011] With the above technical solution, the vehicle state signal is judged before entering self-learning. Self-learning is only carried out when the vehicle intelligent driving state signal is not activated and the vehicle gear is in the forward gear, which can prevent the vehicle from entering self-learning when it does not meet the intelligent driving scenario, resulting in unnecessary energy consumption.

[0012] According to another specific embodiment of the present invention, in the self-learning method of a longitudinal control system for vehicle intelligent driving disclosed in the embodiment of the present invention, in step S2, it further includes obtaining weather data for identifying weather scenarios. Wherein, the weather data includes the rain sensor signal of the vehicle, and the weather scenarios include no rain or snow, low amount of rain or snow, medium amount of rain or snow, and high amount of rain or snow; and the driving state data includes adjacent vehicle state data related to the adjacent vehicle in front of the vehicle and self-vehicle state data related to the vehicle itself; wherein, the adjacent vehicle state data at least includes one or more of the following: adjacent vehicle speed, adjacent vehicle acceleration, adjacent vehicle relative distance, adjacent vehicle type signal; the self-vehicle state data at least includes one or more of the following: accelerator pedal position and rate, brake pedal position and rate, steering wheel angle, steering wheel rate, self-vehicle speed, self-vehicle acceleration, brake master cylinder pressure signal, driving torque, gear signal.

[0013] With the above technical solution, different weather scenarios are determined by the rain sensor of the vehicle, and different parameters can be used to control the vehicle in different weather scenarios to ensure that the vehicle can drive safely and comfortably in different weather scenarios. Obtaining one or more driving state data according to requirements can ensure the accuracy of the obtained data and avoid data errors and energy consumption waste caused by simultaneously obtaining multiple useless data.

[0014] According to another specific embodiment of the present invention, in the self-learning method of a longitudinal control system for vehicle intelligent driving disclosed in the embodiment of the present invention, in step S2, preprocessing the driving state data of the vehicle includes data cleaning and data segmentation of the initially obtained vehicle driving data. Wherein, data cleaning includes: removing invalid data from the initially obtained vehicle driving data, and the invalid data includes: data with an adjacent vehicle relative distance greater than a preset distance threshold, data with a steering wheel angle greater than a preset angle threshold, data with a steering wheel rate greater than a preset rate threshold, and fault data indicating a fault of the vehicle itself; data segmentation includes: taking the driving process of the self-vehicle speed from the current 0 km / h to the next 0 km / h as a driving cycle, dividing the vehicle driving data obtained after data cleaning into vehicle driving state data corresponding to multiple driving cycles respectively, and adding an environment label related to the environment scenario of the driving cycle to the vehicle driving state data corresponding to each driving cycle.

[0015] By adopting the above technical solution, the collected status data of adjacent vehicles and the host vehicle are cleaned and eliminated, improving the accuracy of self-learning for intelligent driving. The data after cleaning and elimination can control the vehicle more accurately, bringing a better driving experience. Moreover, by defining a driving cycle from 0 km / h to 0 km / h for the vehicle speed and dividing multiple driving cycles corresponding to driving status data, the data can be processed more precisely.

[0016] According to another specific embodiment of the present invention, a self-learning method for a longitudinal control system of vehicle intelligent driving disclosed in the embodiment of the present invention, identifying the current driving scenario of the vehicle includes: in each driving cycle, determining the relative vehicle speed according to the host vehicle speed and the speed of the adjacent vehicle; and establishing a first curve with time as the abscissa and relative vehicle speed as the ordinate, and establishing a second curve with time as the abscissa and the relative distance of the adjacent vehicle as the ordinate; regarding the scenario corresponding to the data where the change rate of the relative vehicle speed in the first curve falls within the range of the first change rate threshold as a stable following scenario, and regarding the scenario corresponding to the data where the change rate of the relative distance of the adjacent vehicle in the second curve falls within the range of the second change rate threshold as a stable following scenario.

[0017] Regarding the scenario corresponding to the time period starting from the moment when the host vehicle speed starts to accelerate from 0 km / h, the position and rate of the accelerator pedal are not 0%, and the position and rate of the brake pedal are always 0% until the moment when a stable following scenario is reached or the position and rate of the brake pedal are not 0% as the following start scenario; regarding the scenario corresponding to the time period starting from the moment when the host vehicle acceleration is less than 0 km / h to the moment when the host vehicle acceleration is greater than 0 km / h in other scenarios except the stable following scenario as the following braking scenario; or, regarding the scenario corresponding to the time period starting from the end moment of the following start scenario to the moment when the host vehicle acceleration is greater than 0 km / h as the following braking scenario; regarding the scenario corresponding to the time period starting from the end moment of the stable following scenario to the start of the next stable following scenario or the start of the following braking scenario as the following acceleration scenario; and regarding the scenario corresponding to the time period starting from the end of the stable following scenario, the end of the following start scenario, or the end of the following acceleration scenario to the moment when the host vehicle speed is 0 km / h as the following stopping scenario.

[0018] By adopting the above technical solution, regarding the host vehicle speed from 0 km / h to 0 km / h as a driving cycle, dividing the above driving cycle, screening out different following scenarios, establishing the first and second curves to screen the stable following scenario, and screening the following start scenario, following braking scenario, following acceleration scenario, and following stopping scenario according to the start and end of the stable following scenario and the host vehicle status data. By establishing curves and screening the above following scenarios according to the host vehicle status data, the accuracy of determining different following scenarios can be ensured, further improving the safety and comfort of intelligent driving.

[0019] According to another specific embodiment of the present invention, in a self-learning method for a longitudinal control system of vehicle intelligent driving disclosed in the embodiment of the present invention, in step S3, the driving state data of each driving scenario are respectively analyzed and trained to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario, including: respectively performing clustering fitting or deep learning on the driving state data corresponding to each driving scenario to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario.

[0020] Adopting the above technical solution, the data after cleaning and elimination is trained by using the method of clustering fitting or deep learning, which can improve the accuracy of the driving key parameter - time-to-collision graph corresponding to each driving scenario, and can continuously optimize the data model of each driving scenario; further, the method of clustering fitting or deep learning can be selected in different situations, and the specific situation is described in the specific implementation process.

[0021] According to another specific embodiment of the present invention, in a self-learning method for a longitudinal control system of vehicle intelligent driving disclosed in the embodiment of the present invention, in step S4, the preset safety requirements include: determining that the time-to-collision between the vehicle and the adjacent vehicle in front of the vehicle is within a preset range, the driving torque is within a preset torque range, the acceleration of the self-vehicle is within a preset acceleration range, and the relative distance of the adjacent vehicle is within a preset distance range according to the speed of the self-vehicle, the acceleration of the self-vehicle, the speed of the adjacent vehicle, the acceleration of the adjacent vehicle, and the relative distance of the adjacent vehicle; and the preset correction principle includes: when some data in the driving key parameter - time-to-collision graph is greater than the corresponding threshold in the preset first data threshold, replacing the value of the part of the data exceeding the first data threshold with the corresponding threshold in the first data threshold; and when some data in the driving key parameter - time-to-collision graph is less than the corresponding threshold in the preset second data threshold, replacing the value of the part of the data less than the second data threshold with the corresponding threshold in the second data threshold.

[0022] Adopting the above technical solution, presetting the time-to-collision range according to the self-vehicle data and the data such as the speed, acceleration, and distance of the adjacent vehicle, and setting the preset torque, acceleration, and distance ranges can effectively improve the safety of intelligent control of the vehicle; further, if the above driving key parameter - time-to-collision graph does not meet the preset range, it is corrected according to the correction principle to make it meet the safety requirements, which further ensures the safety and comfort of the vehicle when receiving intelligent driving control. At the same time, the correction of the time-to-collision graph can also enable the vehicle to adaptively modify the range in different situations instead of using the same preset range unchanged. Not changing the range may not meet the conditions for intelligent control of the vehicle in complex situations and the safety and comfort of intelligent driving.

[0023] According to another specific embodiment of the present invention, for a self-learning method of a longitudinal control system for vehicle intelligent driving disclosed in the embodiment of the present invention, after obtaining the driving state data of the vehicle in step S2, it further includes: storing the driving state data of the vehicle by using the on-vehicle memory of the vehicle; and determining whether the storage capacity of the memory reaches a preset capacity threshold, and when the capacity threshold is reached, transmitting the driving state data stored in the memory to the cloud storage device.

[0024] By adopting the above technical solution, storing first by using the memory of the vehicle can ensure the integrity of the data, and the collected data can be stored immediately for subsequent operations. Further, transmitting the data in the vehicle memory to the cloud storage device when the preset capacity threshold is reached can ensure the permanent storage of the data and relieve the storage pressure of the memory, preventing the inability to store the newly collected data when the storage space of the on-vehicle memory is insufficient.

[0025] According to another specific embodiment of the present invention, for a self-learning method of a longitudinal control system for vehicle intelligent driving disclosed in the embodiment of the present invention, before step S5, it further includes: determining whether a first external instruction is received, where the first external instruction indicates prohibiting the longitudinal control system for vehicle intelligent driving from receiving the driving key parameter - time-to-collision graph; if so, returning to execute steps S1 - S4; if not, the longitudinal control system for vehicle intelligent driving receives the driving key parameter - time-to-collision graph;

[0026] After step S5, it further includes: determining whether a second external instruction is received; the second external instruction indicates turning off the longitudinal control system for vehicle intelligent driving; if so, stopping the self-learning of the longitudinal control system for vehicle intelligent driving; if not, returning to execute step S1.

[0027] By adopting the above technical solution, setting the first external instruction to control whether the vehicle receives the driving key parameter - time-to-collision graph can further control whether the vehicle updates its intelligent driving style. Setting the first external instruction gives the vehicle more choices. The first external instruction can ensure that the vehicle does not receive the new key parameter - time-to-collision graph. Therefore, the first external instruction can ensure that the vehicle continues to drive under the original key parameters or drive without intelligent driving intervention. Further, the second external instruction can control the vehicle to turn off the self-learning of the intelligent driving control. Setting the second external instruction can avoid the vehicle recording and uploading its own driving data without the driver's knowledge, resulting in privacy leakage and data security risks.

[0028] According to another specific embodiment of the present invention, a self-learning method for a vehicle intelligent driving longitudinal control system disclosed in the embodiment of the present invention. Before step S1, it is judged whether a third external instruction is received; the third external instruction indicates pushing the driving key parameter - collision time calibration map generated by the self-learning method of the intelligent driving longitudinal control system to the vehicle intelligent driving longitudinal control system; if so, the above-mentioned driving key parameter - collision time calibration map is pushed to the vehicle intelligent driving longitudinal control system; if not, step S1 is continued to be executed.

[0029] By adopting the above technical solution, a third external instruction is set, and it can be judged whether to agree to update the intelligent driving control system parameters through the third external instruction. During driving, different drivers may have different driving habits. When a driver finds the calibration parameters suitable for himself, he may not want to change them anymore. This instruction can provide more autonomous choices for the driver.

[0030] According to another specific embodiment of the present invention, a vehicle intelligent driving longitudinal control system disclosed in the embodiment of the present invention is used to implement the self-learning method of the vehicle intelligent driving longitudinal control system. The system includes: a cloud server, including a cloud storage device and a cloud computing unit, and the cloud computing unit includes an AI and a big data platform.

[0031] And a vehicle-mounted terminal, which is communicatively connected to the cloud server via a gateway and includes a controller assembly and a self-learning calibration system; wherein: the vehicle-mounted terminal obtains the vehicle state signal and the driving state data of the vehicle, judges whether the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system according to the vehicle state signal, and transmits the driving state data of the vehicle to the cloud server when the conditions are met; the cloud server preprocesses the driving state data of the vehicle and identifies the current driving scenario of the vehicle; analyzes and trains the driving state data of each driving scenario respectively to obtain the driving key parameter - collision time map corresponding to each driving scenario; and respectively judges whether the data in the driving key parameter - collision time map corresponding to each driving scenario meets the preset safety requirements.

[0032] By adopting the above technical solution, the self-learning method of the vehicle intelligent driving longitudinal control system is realized by using the vehicle driving intelligent driving control system. In the system, the vehicle-mounted terminal judges whether the self-learning conditions are met through the vehicle state signal, obtains the vehicle state data and transmits it to the cloud server when the conditions are met. The vehicle state signal and data can be quickly obtained through direct collection and judgment by the vehicle-mounted terminal; further, the cloud is used to preprocess the data and identify the scenario, and analyze and train the preprocessed data to judge whether it meets the safety requirements. The cloud server has the advantages of high efficiency and convenience in processing data, and the data processed by the cloud server can be permanently stored for later continuous optimization or adoption.

[0033] According to another specific embodiment of the present invention, a longitudinal control system for intelligent vehicle driving disclosed by the embodiment of the present invention, the controller assembly includes: an intelligent driving assistance system controller, a powertrain controller, a chassis controller, a body controller, and a central control entertainment system controller; the intelligent driving assistance system controller, the powertrain controller, the chassis controller, the body controller, and the central control entertainment system controller are respectively communicatively connected to the self-learning calibration system to transmit the acquired driving state data of the vehicle to the self-learning calibration system.

[0034] By adopting the above technical solution, the self-learning calibration system is directly communicatively connected to the controller assembly, and can transmit data to the self-learning calibration system quickly and stably, so that the self-learning calibration system can make a quick judgment and improve the safety of intelligent driving control.

[0035] The embodiment of the present invention also discloses an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the self-learning method of the above-mentioned longitudinal control system for intelligent vehicle driving.

[0036] The embodiment of the present invention also discloses a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the self-learning method of the above-mentioned longitudinal control system for intelligent vehicle driving.

[0037] The beneficial effects of the present invention are: according to different driving environments and driving scenarios, the acquired vehicle state data can be preprocessed and then the scenarios can be recognized and segmented to obtain various following vehicle scenarios encountered in daily driving. Then, algorithms such as clustering fitting or deep learning are used to analyze and train the above-mentioned following vehicle scenarios to obtain a driving key parameter - time-to-collision diagram. The safety of the above-mentioned time-to-collision diagram is judged, and the above-mentioned time-to-collision diagram is judged according to a preset safety range. If the requirements are met, the next step is continued; if not, the data will be corrected according to the correction principle. After the time-to-collision diagram meets the safety requirements, the intelligent driving system will control the vehicle according to the above-mentioned time-to-collision diagram; through the above solution, the present invention can analyze the driver's driving habits and driving environment in real time during vehicle driving, judge the driving mode in complex situations and continuously optimize the driving mode, and can realize safe, stable and comfortable intelligent control of the vehicle during driving, reduce the driver's driving pressure and improve the riding experience of passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flowchart of the present invention;

[0039] Figure 2 is the flowchart of Embodiment 1 of the present invention;

[0040] Figure 3 is the schematic diagram of the system of the present invention;

[0041] Figure 4 is the schematic diagram of signal flow of the present invention;

[0042] Figure 5 is the system structure diagram of the present invention;

[0043] Figure 6 is the schematic diagram of the structure of the electronic device provided by the embodiment of the present invention.

[0044] Description of reference numerals:

[0045] 1, cloud server; 11, cloud memory; 12, cloud computing unit; 2, vehicle-mounted terminal.

[0046] 121, transceiver; 122, processor; 123, memory. Detailed implementation manners

[0047] To make the objectives, technical solutions and advantages of the present invention clearer, the following will further describe the embodiments of the present invention in detail with reference to the accompanying drawings.

[0048] Embodiment 1:

[0049] Based on the problems existing in the prior art that intelligent driving control cannot meet more complex situations and cannot meet driving expectations, the present application proposes a self-learning method for an intelligent driving control system. Refer to Figure 1 , which includes the following steps:

[0050] S1: Obtain the vehicle state signal, and determine whether the vehicle meets the self-learning condition of the longitudinal control system of vehicle intelligent driving according to the vehicle state signal;

[0051] If yes, execute step S2;

[0052] If no, continue to determine whether the vehicle meets the self-learning condition of the longitudinal control system of intelligent driving;

[0053] S2: Obtain the driving state data of the vehicle, preprocess the driving state data of the vehicle and identify the current driving scenario of the vehicle; wherein, the driving scenario includes a stable following scenario, a following start scenario, a following braking scenario, a following stop scenario, and a following acceleration scenario;

[0054] S3: Analyze and train the driving state data of each driving scenario respectively to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario;

[0055] S4: Determine whether the data in the driving key parameter - time - to - collision graph corresponding to each driving scenario meets the preset safety requirements respectively;

[0056] S5: Push the driving key parameter - time - to - collision graph that meets the preset safety requirements in the driving key parameter - time - to - collision graph corresponding to each driving scenario to the vehicle intelligent driving longitudinal control system; or

[0057] According to the preset correction principle, correct the driving key parameter - time - to - collision graph that does not meet the preset safety requirements in the driving key parameter - time - to - collision graph corresponding to each driving scenario to obtain an updated updated parameter - time - to - collision graph, and push the updated parameter - time - to - collision graph to the vehicle intelligent driving longitudinal control system.

[0058] For example, in a specific embodiment, if the vehicle state signal meets the self - learning condition through judgment, then enter the self - learning of intelligent driving. The vehicle can obtain the vehicle state data through components such as the vehicle's own sensors or vehicle controllers that can collect vehicle data. Further, the pre - processing of the collected data can be completed on the in - vehicle unit itself, which has the advantages of fast speed and high efficiency, or can be selected to be pre - processed in the cloud or big data platform according to the actual situation. Suppose when the vehicle is driving on the highway, the time requirements for obtaining and pre - processing vehicle data are not very high. At this time, it can be processed on the cloud server to ensure the accuracy of data processing. However, if the vehicle is driving in the urban area, the collection and pre - processing of data can be completed on the in - vehicle unit itself to ensure the speed of this link. Similarly, different methods mentioned above can also be used for scene recognition, and it can be selected according to the specific situation. If quick judgment is required for intelligent driving self - learning, the above - mentioned processing and recognition can be carried out on the in - vehicle unit. If more accurate data is required and the time requirement is not high, it can be processed in the cloud. After processing the data, it is necessary to calculate and analyze the driving key parameter - time - to - collision graph for the next - step control. It should be noted that in this application, the driving key parameter - time - to - collision graph is a map of the collision time of different key parameters corresponding to different following - vehicle scenarios. For example, in the above - mentioned stable following - vehicle scenario, the speed - time - to - collision map needs to be analyzed and calculated, or in the following - vehicle start scenario: speed - drive torque - time - to - collision map, following - vehicle acceleration scenario and following - vehicle braking scenario: speed - acceleration - time - to - collision map, following - vehicle stop scenario: speed - time - to - collision map. Setting different key parameters in different scenarios can ensure the intelligence and accuracy of the self - learning of the intelligent driving system. Among them, the time to collision is the time required for the host vehicle to collide with the target object in front while maintaining the current motion state, abbreviated as TTC.

[0059] Specifically, after obtaining and processing the above data, it is determined whether the calculated data meets the preset safety requirements. If the data after calculation and training does not meet the preset range, the data will be corrected according to the correction principle. It should be noted that the preset range of the above safety requirements is fixed, but different boundary values can be selected within the preset range according to different scenarios. Suppose when the vehicle is driving on a spacious road with few vehicles, the boundary values of the driving torque and acceleration can be appropriately increased. Conversely, the boundary values can be decreased. As long as the appropriate safety boundary values are selected according to the current situation or the performance of different vehicles. Further, the data passed through the safety judgment, that is, the above-mentioned key parameter of intelligent driving - time-to-collision graph, will be pushed to the intelligent driving control system. The intelligent driving can perform intelligent control on the vehicle through the above time-to-collision graph. For example, if the vehicle is currently in a stable following scenario, after data collection, processing, and safety judgment, it is pushed to the intelligent driving control system. At this time, the intelligent driving system can control the vehicle through the above parameter - time-to-collision graph. After intelligent driving control, the driver can drive the vehicle more easily.

[0060] It has the following steps. By obtaining the vehicle signal to determine whether to activate the above self-learning method, and performing self-learning intelligent driving control when the vehicle is stable and meets the conditions, the driving safety during the driving process can be effectively guaranteed. Further, after obtaining the vehicle state data, scene recognition can screen out the following modes in different scenarios. Different following modes can ensure the comfort and accuracy of intelligent driving control in complex situations. Analyzing and training the above scenarios can ensure that the above following modes are more accurate and safe, and ensure the safe driving of the vehicle during intelligent driving control. Furthermore, the above technical solution determines whether the key parameter - time-to-collision graph after analysis and training meets the safety requirements, which further ensures the safety before the vehicle receives intelligent control. Further, by determining whether the key parameter - time-to-collision graph meets the safety requirements, it can be ensured that the key parameter - time-to-collision graph after training can be applied to vehicle automatic control on the premise of safety. Or in the case of not meeting safety, it can be corrected according to the predetermined principle, which can cope with different vehicle working conditions and road conditions to ensure the safety of autonomous driving. Furthermore, controlling the vehicle can enable the key parameter - time-to-collision graph used to control the vehicle to meet more complex situations and also achieve the driving expectation.

[0061] This embodiment proposes a self-learning method for the longitudinal control system of vehicle intelligent driving. Among them, in step S1, the vehicle state signal includes the intelligent driving control state signal and the gear signal. And when the intelligent driving state signal is not activated and the gear signal is in the forward gear, it is determined that the vehicle meets the self-learning conditions of the longitudinal control system of vehicle intelligent driving. If the intelligent driving state signal is activated and / or the gear signal is not in the forward gear, it is determined that the vehicle does not meet the self-learning conditions of the longitudinal control system of vehicle intelligent driving.

[0062] For example, in a specific embodiment, the vehicle is in a normal driving state. At this time, the vehicle gear signal is in the forward gear, but due to a vehicle-mounted system failure or damage to the intelligent driving control, the state of the intelligent driving control system is activated. At this time, the vehicle will not enter the self-learning of intelligent driving control. If the vehicle intelligent driving system is normal and in an unactivated state, the vehicle enters the self-learning of intelligent driving. Further, if the vehicle intelligent driving system is unactivated at this time, but the vehicle gear is not in the forward gear, the vehicle will not enter self-learning either.

[0063] By judging the vehicle state signal before entering self-learning, self-learning is only carried out when the intelligent driving state signal of the vehicle is not activated and the vehicle gear is in the forward gear, which can avoid the vehicle entering self-learning when it does not meet the intelligent driving scenario, resulting in unnecessary energy consumption.

[0064] This embodiment proposes a self-learning method for the longitudinal control system of vehicle intelligent driving, including: in step S2, it also includes obtaining weather data for identifying weather scenarios. Among them, the weather data includes the rain sensor signal of the vehicle, and the weather scenarios include no rain or snow, light rain or snow, moderate rain or snow, heavy rain or snow. And the driving state data includes the adjacent vehicle state data related to the adjacent vehicle in front of the vehicle and the self-vehicle state data related to the vehicle itself. Among them, the adjacent vehicle state data at least includes one or more of the adjacent vehicle speed, adjacent vehicle acceleration, adjacent vehicle relative distance, adjacent vehicle type signal. The self-vehicle state data at least includes one or more of the accelerator pedal position and rate, brake pedal position and rate, steering wheel angle, steering wheel rate, self-vehicle speed, self-vehicle acceleration, brake master cylinder pressure signal, drive torque, gear signal.

[0065] For example, in a specific embodiment, the vehicle can identify weather scenarios based on a rain sensor. It is assumed that when the rainfall range is 0%-10%, there is no rain or snow; when it is 10%-20%, there is light rain or snow; when it is 20%-30%, there is moderate rain or snow; and when it is 30% or above, there is heavy rain or snow. However, this preset range is not unique because the vehicle driving areas are not uniform. For example, in southern regions, it may rain for a long time, and the preset range can be adjusted adaptively according to the regional environment. Further, the vehicle data obtained by the self-learning method of intelligent driving control includes signals such as the speed, acceleration, relative distance from the ego vehicle, and target type of the first main target directly ahead, as well as signals such as the position and rate of the ego vehicle's accelerator pedal, the position and rate of the ego vehicle's brake pedal, the steering angle and rate of the ego vehicle's steering wheel, the speed of the ego vehicle, the acceleration of the ego vehicle, the brake master cylinder pressure, and the driving torque. The method for obtaining the above data can be obtained by body sensors or from the body controller. The specific acquisition method is not limited, and as long as it is comprehensively considered according to the vehicle's own situation and cost, a suitable method can be selected for data acquisition.

[0066] Further, when obtaining the above vehicle state data, only the data required for subsequent calculation of the following scenario model can be obtained. For example, when calculating a stable following scenario, only the speed of the ego vehicle, the acceleration of the ego vehicle, and the relative distance to the adjacent vehicle need to be obtained. Moreover, when calculating different following scenarios or performing data segmentation for different requirements, the driving state data can be collected according to the specific required data.

[0067] Different weather scenarios are determined through the vehicle's rain sensor, and different parameters can be used to control the vehicle in different weather scenarios to ensure that the vehicle can drive safely and comfortably in different weather scenarios. Obtaining one or more driving state data according to requirements can ensure the accuracy of the acquired data and avoid data errors and energy consumption waste caused by simultaneously obtaining multiple useless data.

[0068] This embodiment proposes a self-learning method for the longitudinal control system of vehicle intelligent driving, including: in step S2, preprocessing the driving state data of the vehicle includes data cleaning and data segmentation of the initially acquired vehicle driving data. Data cleaning includes: removing invalid data from the initially acquired vehicle driving data. The invalid data includes: data with the relative distance to the adjacent vehicle greater than the preset distance threshold, data with the steering angle greater than the preset steering angle threshold, data with the steering rate greater than the preset rate threshold, and fault data indicating that the vehicle itself has a fault.

[0069] Data segmentation includes: taking the driving process from the current vehicle speed of 0 km / h to the next 0 km / h as a driving cycle, dividing the vehicle driving data obtained after data cleaning into vehicle driving state data corresponding to multiple driving cycles respectively, and adding an environment label related to the environment scenario of the driving cycle to the vehicle driving state data corresponding to each driving cycle;

[0070] For example, in a specific embodiment, after obtaining the above data, the data is first cleaned. Data cleaning includes: removing invalid data from the initially obtained vehicle driving data. The invalid data includes: data with the relative distance between adjacent vehicles greater than a preset distance threshold, data with the steering wheel angle greater than a preset angle threshold, data with the steering wheel speed greater than a preset speed threshold, and fault data indicating that the vehicle itself has a fault. For example, in a specific scenario, data with the relative distance between an adjacent vehicle and the host vehicle greater than 150 m, the steering wheel angle greater than 30°, and the steering wheel angle speed greater than 100° / s are removed by cleaning. The preset values of the above data can be adaptively changed according to the situation. For example, when the vehicle is driving at a high speed, the distance threshold can be appropriately increased to ensure safety, and it can be set to 200 m or greater. The steering wheel angle can be adjusted according to the vehicle speed of the host vehicle. When the vehicle speed is slow, the threshold of the steering wheel angle can be increased, and when the vehicle speed is fast, it can be decreased. The same applies to the steering wheel angle speed. Further, data with no first main target, the first main target type being Truck or non-motor vehicle, data loss, incomplete data, vehicle faults, intelligent driving assistance system faults, and data with signal parsing failures are removed. The above specific data types are the fault data. For example, data with no first main target is data with adjacent vehicle data loss or adjacent vehicle detection failure. When the vehicle is driving, it will encounter large vehicles such as trucks or buses. The data of such vehicles is different from that of ordinary vehicles. If the intelligent driving system continues to be used for calculation and control, there may be potential safety hazards, so the driver needs to control the vehicle according to the situation himself. And for non-motor vehicles on the road, if the data of ordinary vehicles is used for calculation, danger will occur. Removing the data of the above vehicles by cleaning can ensure safety. Furthermore, data loss, incomplete data, vehicle faults, intelligent driving assistance system faults, and data with signal parsing failures are all data with vehicle self-faults, and also need to be cleaned to ensure the accuracy of intelligent driving control self-learning.

[0071] By cleaning and removing the collected adjacent vehicle and host vehicle state data, the accuracy of intelligent driving self-learning is improved. The data after cleaning and removing can control the vehicle more accurately, bringing a better driving experience. And by defining a driving cycle from 0 km / h to 0 km / h according to the vehicle speed and dividing the driving state data corresponding to multiple driving cycles, the data can be processed more precisely.

[0072] This embodiment proposes a self - learning method for the longitudinal control system of vehicle intelligent driving, including identifying the current driving scenario of the vehicle: In each driving cycle, determine the relative vehicle speed according to the self - vehicle speed and the speed of the adjacent vehicle; and establish a first curve with time as the abscissa and relative vehicle speed as the ordinate, and establish a second curve with time as the abscissa and the relative distance of the adjacent vehicle as the ordinate; The scenario corresponding to the data where the change rate of the relative vehicle speed in the first curve falls within the first change rate threshold range is regarded as a stable following scenario, and the scenario corresponding to the data where the change rate of the relative distance of the adjacent vehicle in the second curve falls within the second change rate threshold range is regarded as a stable following scenario; The scenario corresponding to the time period from the moment when the self - vehicle speed starts to accelerate from 0 km / h, the position and rate of the accelerator pedal are not 0%, and the position and rate of the brake pedal are always 0% until reaching the stable following scenario or the moment when the position and rate of the brake pedal are not 0% is determined as the following start scenario; The scenario corresponding to the time period from the moment when the self - vehicle acceleration is less than 0 km / h to the moment when the self - vehicle acceleration is greater than 0 km / h in other scenarios outside the stable following scenario is determined as the following braking scenario; Or, the scenario corresponding to the time period from the end of the following start scenario to the moment when the self - vehicle acceleration is greater than 0 km / h is determined as the following braking scenario; The scenario corresponding to the time period from the end of the stable following scenario to the start of the next stable following scenario or the start of the following braking scenario is determined as the following acceleration scenario; And the scenario corresponding to the time period from the end of the stable following scenario, the end of the following start scenario, or the end of the following acceleration scenario to the moment when the self - vehicle speed is 0 km / h is determined as the following stop scenario.

[0073] For example, in a specific embodiment, during the vehicle driving process, one driving cycle is from a speed of 0 km / h to a speed of 0 km / h. In this cycle, the above - mentioned data is segmented. For example, the TTC (time - to - collision) data of each segmented driving cycle can be calculated, and the TTC is integrated into each driving cycle as a signal similar to the self - vehicle speed, relative distance, etc. to participate in the segmentation. According to the self - vehicle speed and the speed of the first main target vehicle in front in each driving cycle, calculate the relative vehicle speed, generate a corresponding curve with time as the abscissa and relative vehicle speed and relative distance as the ordinate, and screen out the stable following scenario. In a specific embodiment, the stable following scenario refers to the data scenario where the change rate of the relative vehicle speed is within the range of ±5%, and the change rate of the relative distance is within the range of ±10%. However, according to different driving environments, the above - mentioned threshold range can be adaptively modified. When driving in the urban area, the range can be reduced to ensure the driving safety of the vehicle and have reaction time to deal with emergencies, such as a vehicle or pedestrian suddenly appearing in the blind spot of vision.

[0074] For each driving cycle, based on the signals of the vehicle's own speed, acceleration, throttle pedal position and rate, and brake pedal position and rate, filter out the following vehicle-following scenarios: vehicle-following start scenario, vehicle-following braking scenario, vehicle-following acceleration scenario, and vehicle-following stop scenario.

[0075] The vehicle-following start scenario refers to the moment when the vehicle's own speed starts to accelerate from 0 kph, the throttle pedal position is not 0%, and the brake pedal position is always 0% until the conditions of the above stable vehicle-following scenario are met or the moment when the brake pedal position is not 0%. The vehicle-following start scenario can be an idling driving scenario of the vehicle, such as moving forward at idle speed in a traffic jam, or a vehicle-following start scenario.

[0076] The vehicle-following braking scenario refers to the moment when the vehicle's own acceleration is always less than 0 m / s 2 from the end moment of the stable vehicle-following scenario until the moment when the vehicle's own acceleration is greater than 0 m / s 2 The data scenario at that moment. The vehicle-following braking scenario can also be the data scenario from the end moment of the above vehicle-following start scenario until the moment when the acceleration is greater than 0 m / s 2 at that moment. The vehicle-following braking scenario is also a vehicle deceleration scenario.

[0077] The vehicle-following acceleration scenario refers to the data scenario from the end moment of the above stable vehicle-following scenario until the start moment of the next stable vehicle-following scenario or the start moment of vehicle-following braking. The vehicle-following acceleration scenario means that after the stable vehicle-following ends, the vehicle starts to speed up to the next scenario of following another vehicle at a stable speed or the start of the next vehicle-following braking scenario.

[0078] The vehicle-following stop scenario refers to the data scenario from the end moment of the above stable vehicle-following scenario, or the end moment of the vehicle-following start scenario, or the end moment of the above vehicle-following acceleration scenario until the moment when the vehicle's own speed is 0 km / h. The vehicle continuously brakes until the speed is 0, which is the vehicle-following stop scenario.

[0079] Taking the vehicle's own speed from 0 km / h to 0 km / h as a driving cycle, dividing the above driving cycle, filtering out different vehicle-following scenarios, establishing the first and second curves to filter the stable vehicle-following scenario, and filtering the vehicle-following start scenario, vehicle-following braking scenario, vehicle-following acceleration scenario, and vehicle-following stop scenario according to the start and end of the stable vehicle-following scenario and the vehicle's own state data. By establishing curves and filtering the above vehicle-following scenarios according to the vehicle's own state data, the accuracy of determining different vehicle-following scenarios can be ensured, and the safety and comfort of intelligent driving can be further improved.

[0080] This embodiment provides a self-learning method for the longitudinal control system of vehicle intelligent driving, including: in step S3, the driving state data of each driving scenario are analyzed and trained respectively to obtain the driving key parameter - time-to-collision graph corresponding to each driving scenario, including: clustering and fitting or deep learning the driving state data corresponding to each driving scenario respectively to obtain the driving key parameter - time-to-collision graph corresponding to each driving scenario.

[0081] For example, in a specific embodiment, if the vehicle is driving in the urban area where there are many road vehicles and the environment is complex, the clustering and fitting algorithm is used to obtain the time-to-collision graph at this time. Using the clustering and fitting algorithm can quickly obtain the calculation result and implement the intelligent control of the vehicle. If the vehicle is driving on the highway section where there are fewer adjacent vehicle targets, the deep learning algorithm can be used to obtain the time-to-collision graph with higher accuracy and more accurate and comfortable control of the vehicle. The above clustering and fitting algorithm is a machine learning algorithm used to divide the samples in the dataset into data groups that do not come from the machine database, making the samples within the same group more similar and the samples between different groups more different. In the clustering algorithm, the centroid fitting curve is a common method, which can be used to represent the center of each cluster. The basic idea of the centroid fitting curve is to calculate the center point of the cluster and then fit a curve that can represent the cluster. The above deep learning algorithm can eliminate part of the data preprocessing process related to machine learning. This algorithm can extract and process unstructured data such as text and images and can also automatically extract features.

[0082] After cleaning and eliminating the data, the method of clustering and fitting or deep learning is used for training, which can improve the accuracy of the driving key parameter - time-to-collision graph corresponding to each driving scenario and can continuously optimize the data model of each driving scenario.

[0083] This embodiment provides a self-learning method for the longitudinal control system of vehicle intelligent driving, including: in step S4, the preset safety requirements include: determining that the time-to-collision between the vehicle and the adjacent vehicle in front of the vehicle is within the preset range, the driving torque is within the preset torque range, the acceleration of the vehicle itself is within the preset acceleration range, and the relative distance of the adjacent vehicle is within the preset distance range according to the speed of the vehicle itself, the acceleration of the vehicle itself, the speed of the adjacent vehicle, the acceleration of the adjacent vehicle, and the relative distance of the adjacent vehicle; and the preset correction principle includes: when some data in the driving key parameter - time-to-collision graph are greater than the corresponding threshold in the preset first data threshold, the values of the part of the data exceeding the first data threshold are replaced with the corresponding threshold in the first data threshold; and when some data in the driving key parameter - time-to-collision graph are less than the corresponding threshold in the preset second data threshold, the values of the part of the data less than the second data threshold are replaced with the corresponding threshold in the second data threshold.

[0084] For example, in a specific embodiment, the preset safety range of vehicle key parameters can be that the time to collision (TTC) ∈ [0.5s, 2.5s], the driving torque ∈ [-30 Nm, Tmax] (Tmax is the maximum driving torque of the vehicle), the acceleration ∈ [-4 m / s 2 , 4 m / s 2 , the relative distance between the host vehicle and the adjacent vehicle ∈ [0.5 m, 5 m]. In different scenarios, the above range can also be adjusted adaptively; further, if the driving key parameter - time to collision graph does not meet the preset range, the preset range can be corrected. For example, if the range of the time to collision graph is [0.5 s, 3.0 s], and the preset range is the time to collision (TTC) ∈ [0.5 s, 2.5 s], then according to the preset correction principle, the preset range can be adjusted to [0.5 s, 3.0 s]. Further, if the range of the acceleration - time to collision graph is [-5 m / s 2 , 3 m / s 2 , then the safety range can be adjusted to [-5 m / s 2 , 4 m / s 2 according to the preset correction principle. Adjusting the preset range according to the calculated and trained time to collision graph can better facilitate the self - learning of the intelligent driving control system. For example, in specific situations, the preset range may not meet the requirements of current vehicle intelligent control. For example, when the vehicle is driving in the urban area, there are many road vehicles and the situation is complex, and the distance range from the adjacent vehicle may be appropriately reduced.

[0085] Presetting the time to collision range based on the host vehicle data and the data of the adjacent vehicle's speed, acceleration, and distance, and setting the preset torque, acceleration, and distance ranges can effectively improve the safety of intelligent vehicle control; further, if the above - mentioned driving key parameter - time to collision graph does not meet the preset range, it is corrected according to the correction principle to meet the safety requirements, which further ensures the safety and comfort of the vehicle when receiving intelligent driving control. At the same time, correcting the time to collision graph can also enable the vehicle to adaptively modify the range in different situations instead of using the same preset range unchanged. Not changing the range may not meet the conditions of intelligent control required by the vehicle in complex situations and the safety and comfort of intelligent driving.

[0086] This embodiment proposes a self - learning method for the longitudinal control system of vehicle intelligent driving. Among them; after obtaining the driving state data of the vehicle in step S2, it further includes: storing the driving state data of the vehicle using the vehicle's memory; and judging whether the storage capacity of the memory reaches a preset capacity threshold, and when the capacity threshold is reached, transmitting the driving state data stored in the memory to the cloud storage device.

[0087] For example, in a specific embodiment, the memory can be a local storage disk on the vehicle side. When the capacity of the local disk on the vehicle side reaches 70% (this preset capacity can be set according to the storage space of the in-vehicle device itself. If a lot of in-vehicle applications have been downloaded on the in-vehicle device and it occupies a large amount of memory on the in-vehicle device, this preset range can be adjusted down to 50% or less. Conversely, if the storage space of the in-vehicle device is large, this preset range can be appropriately adjusted up), the driving state data stored in the local disk will be packaged and uploaded to the cloud storage device through the vehicle network. This cloud storage device can be a cloud server or a big data platform, etc. After the upload is completed, the cloud storage device will permanently save the driving data and feedback a completion signal. After receiving the transmission completion signal, the in-vehicle device will delete the driving data in the local disk to continue saving future driving data.

[0088] Using the local vehicle-side disk of the vehicle for storage first can ensure the integrity of the data and store the collected data immediately for subsequent operations. Further, when the preset capacity threshold is reached, transmitting the data on the vehicle-side disk to the cloud storage device can ensure the permanent storage of the data and relieve the storage pressure of the vehicle-side memory, preventing the inability to store the latest collected data when the storage space of the vehicle-side memory is insufficient.

[0089] This embodiment proposes a self-learning method for the longitudinal control system of vehicle intelligent driving, which further includes: before step S5, it further includes: determining whether a first external instruction is received. The first external instruction indicates prohibiting the longitudinal control system of vehicle intelligent driving from receiving the driving key parameter - time-to-collision diagram; if so, return to execute steps S1 - S4; if not, the longitudinal control system of vehicle intelligent driving receives the driving key parameter - time-to-collision diagram; after step S5, it further includes: determining whether a second external instruction is received; the second external instruction indicates turning off the longitudinal control system of vehicle intelligent driving; if so, stop the self-learning of the longitudinal control system of vehicle intelligent driving; if not, return to execute step S1.

[0090] For example, in a specific embodiment, the vehicle provides a first power-on / off key on the central control entertainment screen. When the user presses the first power-on / off key, it means that the first external instruction is input. The vehicle can use the first external instruction to prohibit the vehicle intelligent driving control system from receiving the driving key parameter - time-to-collision diagram. Suppose the driver presses the first external instruction switch, the vehicle will not update the driving key parameter - time-to-collision diagram and will continue to drive in the current state. If the scene has not changed significantly after the last intelligent control of the vehicle and the parameters trained last time can still meet the safety and comfortable driving of the vehicle, at this time, the driver can press the first external instruction switch to continue the current driving. If it is necessary to update the parameters of intelligent driving, the driver can press the first external instruction switch to make the vehicle continue to normally receive the driving key parameter - time-to-collision diagram.

[0091] The vehicle also provides a second power-on key on the central control entertainment screen. When the user presses the second power-on key, it means that a second external instruction is input. The second external instruction controls the vehicle intelligent driving control system. Further, the driver can turn off the intelligent driving control system by pressing the second external instruction switch, that is, restore the intelligent driving control system to the factory settings. The intelligent driving control system is default turned on when the vehicle leaves the factory. After the driver turns off the intelligent driving control system through the second external instruction switch, the intelligent driving control system enters the sleep mode and stops recording driving data and communicating with the cloud. If the driver clicks the second external instruction switch again, the intelligent driving control system sends a reset request to erase the original parameters and restore to the factory state.

[0092] It should be noted that the first power-on key and the second power-on key can be set as a combination key integrated on the vehicle central control entertainment screen.

[0093] By setting the first external instruction to control whether the vehicle receives the driving key parameter - collision time graph, it is possible to further control whether the vehicle updates the intelligent driving style. Setting the first external instruction gives the vehicle more options. The first external instruction can ensure that the vehicle does not receive the new key parameter - collision time graph. Therefore, the first external instruction can ensure that the vehicle continues to drive under the original key parameters or drive without intelligent driving intervention. Further, the second external instruction can control the vehicle to turn off the self-learning of the intelligent driving control. Setting the second external instruction can prevent the vehicle from recording and uploading its own driving data without the driver's knowledge, causing privacy leakage and data security risks.

[0094] This embodiment proposes a self-learning method for a vehicle intelligent driving longitudinal control system, which further includes: before step S1, determining whether a third external instruction is received; the third external instruction indicates pushing the driving key parameter - collision time calibration graph generated by the self-learning method of the vehicle intelligent driving longitudinal control system to the vehicle intelligent driving longitudinal control system; if so, pushing the above-mentioned driving key parameter - collision time calibration graph to the vehicle intelligent driving longitudinal control system; if not, continuing to execute step S1.

[0095] For example, in a specific embodiment, after data processing is completed, the driving key parameter - collision time graph is packaged to form a calibration parameter package, which is stored in the cloud, and information about new calibration parameters is pushed to the intelligent driving control system through the vehicle network. After receiving the push, the intelligent driving control system notifies the driver that there are new calibration parameters through a pop-up reminder on the central control entertainment screen and asks the driver whether to apply them. If the driver clicks to accept and apply, the intelligent driving control system downloads the calibration parameter package in the cloud to the local through the vehicle network, decompresses it, and writes the calibration parameters into the intelligent driving system through the vehicle Ethernet protocol to control the vehicle. In addition to being reminded through a pop-up window on the central control, the third external instruction can also be prompted on the vehicle liquid crystal instrument panel. The driver can directly click on the steering wheel whether to accept it. As long as it does not affect driving safety, the third external instruction can be displayed in other ways.

[0096] A third external instruction is set, and through the third external instruction, it can be determined whether to agree to update the parameters of the intelligent driving control system. During driving, different drivers may have different driving habits. After a driver finds the calibration parameters that suit them, they may not want to make any more changes. This instruction can provide more autonomous choices for the driver.

[0097] In the specific self-learning method provided by the embodiments of the present invention, refer to Figure 2, during the driving process of the vehicle, first, it is determined whether the status signal of the intelligent driving assistance system collected by the controller is activated. If it is activated, this intelligent driving control ends. If it is not activated, driving data is collected through vehicle body sensors and the controller and stored in the local memory. When the local memory reaches a certain storage capacity (the specific setting of the storage capacity has been specifically explained in Embodiment 1), the data is transmitted to the cloud server (the cloud server includes a cloud memory and a cloud calculator). After the data is uploaded to the cloud, it is cleaned, and scene recognition and segmentation are performed. The weather scene can be classified according to the signal of the rain sensor on the vehicle body; the vehicle speed range from 0 km / h to 0 km / h is defined as a driving cycle. Within this driving cycle, based on the vehicle speed of the host vehicle, the relative distance to the adjacent vehicle and other data, the first and second curves are established to screen out the stable following scene. Then, based on the start and end of the stable following scene, as well as the vehicle acceleration, vehicle speed and other data of the host vehicle, the following acceleration scene, following braking scene, and following stopping scene are screened out; after screening out the above scenes, the cloud calculator of the cloud server uses algorithms such as clustering fitting or deep learning (the specific method of selecting different algorithms according to different situations has been explained in Embodiment 1 above) to analyze and train to obtain the driving key parameter - time-to-collision diagram. The cloud server determines whether the obtained driving key parameter - time-to-collision diagram meets the preset safety requirements. If it does not meet the requirements, it is corrected according to the predetermined correction principle; the driving key parameter - time-to-collision diagram that meets the conditions is used to generate new calibration parameters and stored in the cloud server. The new calibration parameters are pushed to the in-vehicle terminal, and the driver can be asked whether to accept the new calibration parameters through means such as display on the central control screen or instrument panel. After acceptance, the new calibration parameters are used to control the vehicle.

[0098] By judging the driving scenario and environment of the vehicle during the driving process of the vehicle and collecting vehicle status data in real time, and obtaining the optimal control method suitable for the current driving scenario after processing and screening the data, the vehicle can be better controlled; continuously learning the driving habits of the driver can reduce the pressure on the driver during long-term driving of the vehicle and can improve the riding experience of passengers through precise algorithms.

[0099] Embodiment 2:

[0100] This embodiment discloses a vehicle intelligent driving longitudinal control system for implementing the self-learning method of the vehicle intelligent driving longitudinal control system described in the above embodiment. Specifically, referring to Figure 3 , the system includes a cloud server 1 and an in-vehicle terminal 2.

[0101] Among them, the cloud server 1 includes a cloud storage device 11 and a cloud computing unit 12. The cloud computing unit 12 includes an AI and a big data platform.

[0102] The in-vehicle unit 2 is communicatively connected to the cloud server via a gateway, and includes a controller assembly and a self-learning calibration system. Among them, the in-vehicle unit obtains the vehicle status signal and the driving status data of the vehicle, determines whether the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system according to the vehicle status signal, and transmits the driving status data of the vehicle to the cloud server when the conditions are met; the cloud server preprocesses the driving status data of the vehicle and identifies the current driving scenario of the vehicle; analyzes and trains the driving status data of each driving scenario respectively to obtain the driving key parameter - time-to-collision diagram corresponding to each driving scenario; and respectively determines whether the data in the driving key parameter - time-to-collision diagram corresponding to each driving scenario meets the preset safety requirements.

[0103] Specifically, the in-vehicle unit has a gateway and also has a network interface for data interaction with the cloud server; further, the in-vehicle unit has a CAN message receiver, as well as a processor and a memory for receiving and processing data.

[0104] For example, in a specific embodiment, during the driving process of the vehicle, the vehicle status data is obtained through the sensors or controllers of the in-vehicle unit 2. After determining that the vehicle signal meets the intelligent driving conditions, it enters the intelligent driving control system. At this time, the obtained data is transmitted to the cloud server 1. The cloud server 1 will perform preprocessing of data cleaning on the transmitted data, and then perform segmentation and scenario recognition on the processed data. Assuming that after data preprocessing and scenario segmentation, it is determined as a steady following scenario. At this time, the computing unit in the cloud server 1 calculates the data of the steady following scenario to obtain the driving key parameter - time-to-collision diagram, and determines whether the calculated data meets the safety range preset in advance. If it meets, the next operation can be carried out. If it does not meet, the driving key parameter - time-to-collision diagram is corrected, and the corrected data is used for the next operation.

[0105] In addition, refer to Figure 4, an embodiment of the present invention includes an Ethernet interface, a CAN message receiver, a processor, a memory, and a network interface. The cloud structure includes a network interface, a cloud server, and a cloud memory. Among them, the CAN message receiver of the self-learning calibration system included in the in-vehicle unit is used to receive signals such as the status of the intelligent driving assistance system, the speed, acceleration, relative distance from the vehicle, target type, throttle pedal position and rate, brake pedal position and rate, steering wheel angle and rate, vehicle speed, vehicle acceleration, brake master cylinder pressure, etc. output by the in-vehicle unit controller assembly, the rain sensor signal, the driving torque, the gear position, etc., and transmit them to the processor. The in-vehicle unit processor determines whether the current intelligent driving assistance system is activated through the intelligent driving assistance system status signal. If not, it stores signals such as the speed, acceleration, relative distance from the vehicle, target type, throttle pedal position and rate, brake pedal position and rate, steering wheel angle and rate, vehicle speed, vehicle acceleration, brake master cylinder pressure, driving torque, gear position, etc. of the first main target in front of the vehicle in the local memory of the in-vehicle unit.

[0106] A self-learning method for the vehicle intelligent driving longitudinal control system using the vehicle driving intelligent driving control system. In the system, the in-vehicle unit determines whether the self-learning condition is met through the vehicle status signal, obtains the vehicle status data, and transmits it to the cloud server when the condition is met. By directly collecting and judging through the in-vehicle unit, the vehicle status signal and data can be obtained quickly. Further, the cloud is used to preprocess the data and perform scene recognition, and analyze and train the preprocessed data to determine whether the safety requirements are met. The cloud server has the advantages of high efficiency and convenience in processing data, and the data processed by the cloud server can be permanently stored for later continuous optimization or adoption.

[0107] This embodiment also discloses a vehicle intelligent driving longitudinal control system, including; the controller assembly includes: an intelligent driving assistance system controller, a powertrain controller, a chassis controller, a body controller, and a central control entertainment system controller; the intelligent driving assistance system controller, the powertrain controller, the chassis controller, the body controller, and the central control entertainment system controller are respectively communicatively connected to the self-learning calibration system to transmit the obtained vehicle driving state data to the self-learning calibration system.

[0108] Reference Figure 5, an embodiment of the present invention includes a gateway, an intelligent driving assistance system controller, a powertrain controller, a chassis controller, a body controller, a central control entertainment system controller, a self-learning calibration system, a cloud server, etc. Among them, the intelligent driving assistance system controller, the powertrain controller, the chassis controller, the body controller, the central control entertainment system controller, and the self-learning calibration system are respectively connected to the gateway through the CAN / CANFD vehicle bus. The gateway plays a role in signal processing and routing, realizing network interconnection between each controller.

[0109] The self-learning calibration system is also directly connected to the intelligent driving assistance system controller through the Ethernet protocol for the brushing of calibration parameters. Among them, the signals output by the intelligent driving assistance controller are signals such as the state of the intelligent driving assistance system, the speed, acceleration, relative distance from the vehicle itself, and target type of the first main target directly in front. The signals output by the chassis controller include signals such as the position and rate of the accelerator pedal, the position and rate of the brake pedal, the steering wheel angle and rate, the vehicle speed of the vehicle itself, the acceleration of the vehicle itself, and the pressure of the brake master cylinder. The signals output by the body controller include the signal of the rain sensor. The signals output by the powertrain controller are signals such as drive torque and gear position.

[0110] The self-learning calibration system is directly connected to the controller assembly through communication, and can transmit data to the self-learning calibration system quickly and stably, so that the self-learning calibration system can make a quick judgment and improve the safety of intelligent driving control.

[0111] Embodiment 3:

[0112] This embodiment discloses an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the self-learning method of the vehicle intelligent driving longitudinal control system described in the above embodiment.

[0113] Figure 6 It is a schematic structural diagram of the electronic device provided by the embodiment of the present application. As Figure 4 shown, the electronic device may include: a transceiver 121, a processor 122, and a memory 123.

[0114] The processor 122 executes the computer-executable instructions stored in the memory, so that the processor 122 executes the solution in the above embodiment. The processor 122 may be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it may also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field-programmable gate array FPGA, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0115] The memory 123 is connected to the processor 122 via the system bus and completes the communication therebetween. The memory 123 is used to store computer program instructions.

[0116] The transceiver 121 can be used to obtain the task to be run and the configuration information of the task to be run.

[0117] The system bus can be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only a thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to implement the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include a Random Access Memory (RAM), and may also include a non-volatile memory.

[0118] Embodiment 4:

[0119] This embodiment discloses a computer-readable storage medium. Computer-executable instructions are stored in the computer-readable storage medium. When the computer-executable instructions are executed by a processor, they are used to implement the self-learning method of the vehicle intelligent driving longitudinal control system described in the above embodiment.

[0120] Although the present invention has been illustrated and described by referring to some preferred embodiments of the present invention, those of ordinary skill in the art should understand that the above content is a further detailed description of the present invention in combination with specific embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. Those skilled in the art can make various changes in form and details, including making several simple deductions or substitutions, without departing from the spirit and scope of the present invention.

Claims

1. A self-learning method for a longitudinal control system of vehicle intelligent driving, characterized in that, Including the following steps: S1: Obtain vehicle status signals, and determine whether the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system according to the vehicle status signals; If so, execute step S2; If not, continue to determine whether the vehicle meets the self-learning conditions of the intelligent driving longitudinal control system; S2: Obtain the driving status data of the vehicle, preprocess the driving status data of the vehicle and identify the current driving scenario of the vehicle; wherein, the driving scenario includes a stable following scenario, a following start scenario, a following braking scenario, a following stop scenario, and a following acceleration scenario; S3: Analyze and train the driving status data of each driving scenario respectively to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario; S4: Determine whether the data in the driving key parameter - time-to-collision graph corresponding to each driving scenario meets the preset safety requirements respectively; S5: Push the driving key parameter - time-to-collision graph that meets the preset safety requirements in the driving key parameter - time-to-collision graph corresponding to each driving scenario to the vehicle intelligent driving longitudinal control system; or According to the preset correction principle, correct the driving key parameter - time-to-collision graph that does not meet the preset safety requirements in the driving key parameter - time-to-collision graph corresponding to each driving scenario to obtain an updated updated parameter - time-to-collision graph, and push the updated parameter - time-to-collision graph to the vehicle intelligent driving longitudinal control system.

2. The self-learning method of the vehicle intelligent driving longitudinal control system according to claim 1, wherein The vehicle status signals in step S1 include intelligent driving control status signals and gear signals; and When the intelligent driving status signal is not activated and the gear signal is in the forward gear, it is determined that the vehicle meets the self-learning conditions of the intelligent driving longitudinal control system; If the intelligent driving status signal is activated and / or the gear signal is not in the forward gear, it is determined that the vehicle does not meet the self-learning conditions of the intelligent driving longitudinal control system.

3. The self-learning method of the vehicle intelligent driving longitudinal control system according to claim 1, wherein In step S2, it also includes obtaining weather data for weather scenario recognition, wherein the weather data includes the rain sensor signal of the vehicle, and the weather scenario includes no rain or snow, low amount of rain or snow, medium amount of rain or snow, high amount of rain or snow; and The driving status data includes adjacent vehicle status data related to adjacent vehicles in front of the vehicle and self-vehicle status data related to the vehicle itself; wherein, the adjacent vehicle status data at least includes one or more of adjacent vehicle speed, adjacent vehicle acceleration, adjacent vehicle relative distance, adjacent vehicle type signal; the self-vehicle status data at least includes one or more of accelerator pedal position and rate, brake pedal position and rate, steering wheel angle, steering wheel rate, self-vehicle speed, self-vehicle acceleration, brake master cylinder pressure signal, drive torque, gear signal.

4. The self-learning method of the longitudinal control system for vehicle intelligent driving according to claim 3, characterized in that in step S2, the preprocessing of the driving state data of the vehicle includes data cleaning and data segmentation of the initially acquired vehicle driving data; wherein, the data cleaning includes: removing invalid data from the initially acquired vehicle driving data, and the invalid data includes: data with the relative distance to the adjacent vehicle greater than a preset distance threshold, data with the steering wheel angle greater than a preset angle threshold, data with the steering wheel speed greater than a preset speed threshold, and fault data indicating that the vehicle itself has a fault; the data segmentation includes: taking the driving process of the host vehicle speed from the current 0 km / h to the next 0 km / h as a driving cycle, dividing the vehicle driving data obtained after the data cleaning into vehicle driving state data corresponding to multiple driving cycles respectively, and adding an environment label related to the environment scene of the driving cycle to the vehicle driving state data corresponding to each driving cycle.

5. The self-learning method of the longitudinal control system for vehicle intelligent driving according to claim 3, characterized in that The recognition of the current driving scene of the vehicle includes: in each driving cycle, determining the relative vehicle speed according to the host vehicle speed and the speed of the adjacent vehicle; and establishing a first curve with time as the abscissa and the relative vehicle speed as the ordinate, and establishing a second curve with time as the abscissa and the relative distance to the adjacent vehicle as the ordinate; taking the scene corresponding to the data where the change rate of the relative vehicle speed in the first curve falls within the first change rate threshold range as the stable following scene, and taking the scene corresponding to the data where the change rate of the relative distance to the adjacent vehicle in the second curve falls within the second change rate threshold range as the stable following scene; determining the scene corresponding to the time period from the moment when the host vehicle speed starts to accelerate from 0 km / h, the position and rate of the accelerator pedal are not 0%, and the position and rate of the brake pedal are always 0% until the stable following scene or the position and rate of the brake pedal are not 0% as the following start scene; determining the scene corresponding to the time period from the moment when the host vehicle acceleration is less than 0 km / h to the moment when the host vehicle acceleration is greater than 0 km / h in other scenes outside the stable following scene as the following braking scene; or, determining the scene corresponding to the time period from the end moment of the following start scene to the moment when the host vehicle acceleration is greater than 0 km / h as the following braking scene; determining the scene corresponding to the time period from the end moment of the stable following scene to the start of the next stable following scene or the start of the following braking scene as the following acceleration scene; and determining the scene corresponding to the time period from the end of the stable following scene, the end of the following start scene, or the end of the following acceleration scene to the moment when the host vehicle speed is 0 km / h as the following stop scene.

6. The self-learning method of the longitudinal control system for vehicle intelligent driving according to claim 1, characterized in that In step S3, the driving state data of each driving scenario are respectively analyzed and trained to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario, including: The driving state data corresponding to each driving scenario are respectively subjected to clustering fitting or deep learning to obtain the driving key parameter - time-to-collision graph corresponding to each driving scenario.

7. The self-learning method of the vehicle intelligent driving longitudinal control system according to claim 3, wherein: In step S4, the preset safety requirements include: Determine that the time to collision between the vehicle and the adjacent vehicle in front of the vehicle is within a preset range, the driving torque is within a preset torque range, the acceleration of the host vehicle is within a preset acceleration range, and the relative distance of the adjacent vehicle is within a preset distance range according to the speed of the host vehicle, the acceleration of the host vehicle, the speed of the adjacent vehicle, the acceleration of the adjacent vehicle, and the relative distance of the adjacent vehicle; and The preset correction principle includes: If some data in the driving key parameter - time-to-collision graph are greater than the corresponding threshold in the preset first data threshold, replace the values of the part of the data exceeding the first data threshold with the corresponding threshold in the first data threshold; and If some data in the driving key parameter - time-to-collision graph are less than the corresponding threshold in the preset second data threshold, replace the values of the part of the data less than the second data threshold with the corresponding threshold in the second data threshold.

8. The self-learning method of the vehicle intelligent driving longitudinal control system according to any one of claims 1-7, wherein: After obtaining the driving state data of the vehicle in step S2, it further includes: Storing the driving state data of the vehicle by using the vehicle's memory; and Judging whether the storage capacity of the memory reaches a preset capacity threshold, and when reaching the capacity threshold, transmitting the driving state data stored in the memory to the cloud storage device.

9. The self-learning method of the longitudinal control system for vehicle intelligent driving according to claim 8, wherein It further includes: Before step S5, it further includes: Judging whether a first external instruction is received, where the first external instruction indicates prohibiting the vehicle intelligent driving longitudinal control system from receiving the driving key parameter - time-to-collision graph; If so, return to execute steps S1-S4; If not, the vehicle intelligent driving longitudinal control system receives the driving key parameter - time-to-collision graph; After step S5, it further includes: Judging whether a second external instruction is received; the second external instruction indicates shutting down the vehicle intelligent driving longitudinal control system; If so, stop the self-learning of the vehicle intelligent driving longitudinal control system; If not, return to execute step S1.

10. The self-learning method of the longitudinal control system for vehicle intelligent driving according to claim 9, characterized in that, It further includes: Before step S1, judging whether a third external instruction is received; the third external instruction indicates pushing the driving key parameter - time-to-collision calibration graph generated by the self-learning method of the vehicle intelligent driving longitudinal control system to the vehicle intelligent driving longitudinal control system; If so, push the driving key parameter - time-to-collision calibration graph to the vehicle intelligent driving longitudinal control system; If not, continue to execute step S1.

11. A vehicle intelligent driving longitudinal control system for implementing the self-learning method of the vehicle intelligent driving longitudinal control system according to any one of claims 1-10, characterized in that, The system includes: A cloud server, including a cloud storage device and a cloud computing unit, where the cloud computing unit includes an AI and a big data platform; A vehicle head unit, which is communicatively connected to the cloud server via a gateway and includes a controller assembly and a self-learning calibration system; wherein: The vehicle head unit obtains the vehicle status signal and the driving status data of the vehicle, determines whether the vehicle meets the self-learning conditions of the vehicle intelligent driving longitudinal control system according to the vehicle status signal, and transmits the driving status data of the vehicle to the cloud server when the conditions are met; The cloud server preprocesses the driving status data of the vehicle and identifies the current driving scenario of the vehicle; analyzes and trains the driving status data of each driving scenario to obtain a driving key parameter - time-to-collision graph corresponding to each driving scenario; and respectively determines whether the data in the driving key parameter - time-to-collision graph corresponding to each driving scenario meets the preset safety requirements.

12. The vehicle intelligent driving longitudinal control system according to claim 11, wherein The controller assembly includes: An intelligent driving assistance system controller, a powertrain controller, a chassis controller, a body controller, and a central control entertainment system controller; The intelligent driving assistance system controller, the powertrain controller, the chassis controller, the body controller, and the central control entertainment system controller are respectively communicatively connected to the self-learning calibration system to transmit the obtained driving status data of the vehicle to the self-learning calibration system.

13. An electronic device, characterized in that, including: A processor and a memory communicatively connected to the processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to implement the self-learning method of the vehicle intelligent driving longitudinal control system according to any one of claims 1-10.

14. A computer-readable storage medium, wherein The computer-readable storage medium stores computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the self-learning method of the vehicle intelligent driving longitudinal control system according to any one of claims 1-10.

Images