Intelligent suspension control method and device, electronic equipment and intelligent vehicle

Abstract

The invention relates to the technical field of intelligent suspensions, in particular to an intelligent suspension control method and device, electronic equipment and an intelligent vehicle. Acquiring operation state data corresponding to the current operation state of the target vehicle; obtaining target person information corresponding to a target person in the target vehicle; inputting the running state data and the target personnel information into a preset reinforcement learning model, and outputting target intelligent suspension control information corresponding to the target vehicle; and based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled, so that the intelligent suspension responds in advance, and then accurate adjustment and intelligence can be realized. Besides, the method not only considers the running state data corresponding to the target vehicle, but also considers the information of the target personnel, so that the safety of the target vehicle and the comfort of the target personnel are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent suspension, and in particular to an intelligent suspension control method, device, electronic equipment and an intelligent vehicle. Background Art

[0002] As an important component of the vehicle chassis, intelligent suspension is related to the comfort and safety of the vehicle.

[0003] At present, traditional air intelligent suspension systems have some inherent limitations in stiffness and damping adjustment, which makes the current air suspension generally have response lag and cannot be accurately adjusted.

[0004] As a result, in many cases, the adjustment of the intelligent suspension cannot take into account both the safety and comfort of the vehicle. Summary of the invention

[0005] In view of this, the present invention provides an intelligent suspension control method, device, electronic device and intelligent vehicle to solve the problem that the adjustment of the intelligent suspension cannot take into account both the safety and comfort of the vehicle.

[0006] In a first aspect, the present invention provides an intelligent suspension control method, the method comprising: Acquire operating status data corresponding to the current operating status of the target vehicle; the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data; Obtaining target person information corresponding to a target person in a target vehicle; the target person information includes at least one of age information, gender information, health status information, and driving style information; Input the running status data and target personnel information into the preset reinforcement learning model, and output the target intelligent suspension control information corresponding to the target vehicle; Based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled.

[0007] The intelligent suspension control method provided in the embodiment of the present application obtains the operating status data corresponding to the current operating status of the target vehicle; obtains the target personnel information corresponding to the target personnel in the target vehicle; inputs the operating status data and the target personnel information into the preset reinforcement learning model, and outputs the target intelligent suspension control information corresponding to the target vehicle, thereby ensuring the accuracy of the output target intelligent suspension control information. Then, based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled, and the intelligent suspension responds in advance, thereby accurately adjusting the intelligence. In addition, the above method takes into account both the operating status data corresponding to the target vehicle and the target personnel information, thereby ensuring the safety of the target vehicle and the comfort of the target personnel.

[0008] In an optional implementation, the preset reinforcement learning model includes a first target feature extraction network, a second target feature extraction network, and a feature fusion network; the operation status data and the target personnel information are input into the preset reinforcement learning model, and the target intelligent suspension control information corresponding to the target vehicle is output, including: The running state data is input into the first target feature extraction network in the preset reinforcement learning model, and the target environment risk feature corresponding to the current running state of the target vehicle is output; the target environment risk feature is used to characterize the danger level of the current driving environment of the target vehicle; The target person information is input into the second target feature extraction network in the preset reinforcement learning model, and the target person comfort feature corresponding to the current operating state of the target vehicle is output; the target person comfort feature is used to characterize the target person's requirement for comfort; The target environment risk characteristics and target personnel comfort characteristics are fused to output the target intelligent suspension control information corresponding to the target vehicle.

[0009] The intelligent suspension control method provided in the embodiment of the present application inputs the operating status data into the first target feature extraction network in the preset reinforcement learning model, and outputs the target environment risk characteristics corresponding to the current operating status of the target vehicle; the accuracy of the output target environment risk characteristics is ensured, so that the feature extraction of the driving environment risk can be realized. The target personnel information is input into the second target feature extraction network in the preset reinforcement learning model, and the target personnel comfort characteristics corresponding to the current operating status of the target vehicle are output, thereby ensuring the accuracy of the output target personnel comfort characteristics. The target environment risk characteristics and the target personnel comfort characteristics are feature fused, and the target intelligent suspension control information corresponding to the target vehicle is output, thereby ensuring the accuracy of the output target intelligent suspension control information corresponding to the target vehicle.

[0010] In an optional implementation, the feature fusion network includes a target actor network and a target critic network; the target environment risk feature and the target personnel comfort feature are feature fused, and the target intelligent suspension control information corresponding to the target vehicle is output, including: The target environment risk characteristics and the target personnel comfort characteristics are input into the target actor network, and the target actor network extracts the target environment risk characteristics and the target personnel comfort characteristics, and outputs the candidate intelligent suspension control information corresponding to the target vehicle; The target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information are input into the target critic network, the target critic network extracts the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information, outputs the candidate score corresponding to the candidate intelligent suspension control information, and transmits the candidate score to the target actor network; the candidate score is used to characterize the safety and comfort corresponding to the control of the target vehicle based on the candidate intelligent suspension control information; If the candidate score is greater than or equal to the preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information and outputs the target intelligent suspension control information; If the candidate score is less than a preset score threshold, the candidate intelligent suspension control information is modified according to the candidate score, and the target intelligent suspension control information is output.

[0011] The intelligent suspension control method provided in the embodiment of the present application inputs the target environment risk characteristics and the target personnel comfort characteristics into the target actor network, and the target actor network extracts the target environment risk characteristics and the target personnel comfort characteristics, and outputs the candidate intelligent suspension control information corresponding to the target vehicle, thereby ensuring the accuracy of the output candidate intelligent suspension control information corresponding to the target vehicle. The target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information are input into the target critic network, and the target critic network extracts the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information, outputs the candidate score corresponding to the candidate intelligent suspension control information, and transmits the candidate score to the target actor network; thereby ensuring the accuracy of the output candidate score. If the candidate score is greater than or equal to the preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information, and outputs the target intelligent suspension control information, thereby ensuring that the score corresponding to the output target intelligent suspension control information is high. If the candidate score is less than the preset score threshold, the candidate intelligent suspension control information is corrected according to the candidate score, and the target intelligent suspension control information is output, thereby ensuring the accuracy of the output target intelligent suspension control information and avoiding inaccurate output target intelligent suspension control information, which may lead to dangerous driving of the target vehicle and / or discomfort of the target person.

[0012] In an optional implementation, the candidate intelligent suspension control information is modified according to the candidate scores using the following formula to output the target intelligent suspension control information, including: , in, is the candidate score corresponding to the candidate intelligent suspension control information, s is the target environmental risk characteristics and target personnel comfort characteristics corresponding to the current moment, is the candidate intelligent suspension control information corresponding to the target vehicle, The target environmental risk characteristics and target personnel comfort characteristics corresponding to the next moment; The target intelligent suspension control information corresponding to the next moment; is a preset scoring threshold; is the learning rate; Rewards for the environment; is the discount factor; The weight coefficient is used to control the influence of the candidate score on the value function update.

[0013] The intelligent suspension control method provided in the embodiment of the present application ensures the accuracy of the output target intelligent suspension control information.

[0014] In an optional implementation, the target person information includes age information, gender information, and health status information, and obtaining the target person information corresponding to the target person in the target vehicle includes: Acquire a person image corresponding to a target person in a target vehicle; Recognize the person image and determine a target region of interest from the person image, where the target region of interest includes the target person; Identify the target region of interest and determine the target person's facial image and target person's body image; Recognize the target person’s facial image and determine the target person’s age and gender information; The target person's facial image and body image are identified to detect whether the target person has any injuries, and the target person's health information is determined based on the detection results.

[0015] The intelligent suspension control method provided in the embodiment of the present application obtains a person image corresponding to a target person in a target vehicle; identifies the person image, and determines a target region of interest from the person image, wherein the target region of interest includes the target person, thereby ensuring the accuracy of the determined target region of interest. The target region of interest is identified, and the facial image and the body image of the target person are determined, thereby ensuring the accuracy of the determined facial image and the body image of the target person. The facial image of the target person is identified, and the age information and the gender information of the target person are determined, thereby ensuring the accuracy of the determined age information and the gender information. The facial image and the body image of the target person are identified, and it is detected whether the target person has any injured position. Based on the detection results, the health information of the target person is determined, thereby ensuring the accuracy of the determined health information of the target person.

[0016] In an optional implementation, the training process of the preset reinforcement learning model includes: Obtaining a training data set, the training data set including training vehicle operation data, training personnel information, training intelligent suspension control information, and training scores corresponding to the training intelligent suspension control information; wherein the training intelligent suspension control information and the training scores are label information; The training vehicle operation data, the training personnel information, the training intelligent suspension control information and the training scores corresponding to the training intelligent suspension control information are input into the initial reinforcement learning network, the initial reinforcement learning network is trained, and a preset reinforcement learning model is generated.

[0017] The intelligent suspension control method provided in the embodiment of the present application obtains a training data set, which includes training vehicle operation data, training personnel information, training intelligent suspension control information and training scores corresponding to the training intelligent suspension control information; wherein the training intelligent suspension control information and the training scores are label information; the training vehicle operation data, training personnel information, training intelligent suspension control information and the training scores corresponding to the training intelligent suspension control information are input into an initial reinforcement learning network, the initial reinforcement learning network is trained, and a preset reinforcement learning model is generated, thereby ensuring the accuracy of the generated preset reinforcement learning model.

[0018] In an optional implementation, the initial reinforcement learning network includes a first initial feature extraction network, a second initial feature extraction network, an initial actor network, and an initial critic network, and inputs the training vehicle operation data, the training personnel information, the training intelligent suspension control information, and the training score corresponding to the training intelligent suspension control information into the initial reinforcement learning network, trains the initial reinforcement learning network, and generates a preset reinforcement learning model, including: Inputting the training vehicle operation data into a first initial feature extraction network and outputting the training environment risk features; Inputting the trainee information into the second initial feature extraction network, and outputting the trainee comfort feature; The risk characteristics of the training environment and the comfort characteristics of the trainees are input into the initial actor network, and the initial actor network outputs the virtual intelligent suspension control information; Calculating a first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information, updating a first initial feature extraction network, a second initial feature extraction network, and an initial actor network based on the first loss function, and generating a first candidate feature extraction network, a second candidate feature extraction network, and a candidate actor network; Inputting the training vehicle operation data into a first candidate feature extraction network and outputting candidate environmental risk features; Input the training personnel information into the second candidate feature extraction network, and output the candidate personnel comfort features; Inputting the candidate environmental risk characteristics and the candidate occupant comfort characteristics into the candidate actor network, the candidate actor network outputs the candidate intelligent suspension control information; Input candidate environmental risk features, candidate occupant comfort features, and candidate intelligent suspension control information into the initial critic network, and output virtual scores corresponding to the candidate intelligent suspension control information; calculating a second loss function between the virtual score and the training score; Based on the second loss function, the initial critic network is updated to generate a candidate critic network; Inputting the training vehicle operation data into the first candidate feature extraction network and outputting the candidate environmental risk features; Input the training personnel information into the second candidate feature extraction network, and output the comfort features of the candidate personnel; Inputting candidate environmental risk characteristics and candidate personnel comfort characteristics into the candidate actor network, and the candidate actor network outputs candidate intelligent suspension control information; Calculating a first loss function between the candidate intelligent suspension control information and the training intelligent suspension control information, updating a first candidate feature extraction network, a second candidate feature extraction network, and a candidate actor network based on the first loss function, and generating a first target feature extraction network, a second target feature extraction network, and a target actor network; Inputting candidate environmental risk characteristics, candidate personnel comfort characteristics, and candidate intelligent suspension control information into the candidate critic network, and outputting candidate scores corresponding to the candidate intelligent suspension control information; calculating a second loss function between the candidate scores and the training scores; Based on the second loss function, the candidate critic network is updated to generate the target critic network.

[0019] The intelligent suspension control method provided in the embodiment of the present application inputs the training vehicle operation data into the first initial feature extraction network, outputs the training environment risk feature, inputs the training personnel information into the second initial feature extraction network, outputs the training personnel comfort feature; inputs the training environment risk feature and the training personnel comfort feature into the initial actor network, and the initial actor network outputs the virtual intelligent suspension control information, thereby ensuring that the output virtual intelligent suspension control information comprehensively considers the training environment risk feature and the training personnel comfort feature. Calculate the first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information, and based on the first loss function, update the first initial feature extraction network, the second initial feature extraction network, and the initial actor network, generate the first candidate feature extraction network, the second candidate feature extraction network, and the candidate actor network, thereby ensuring the accuracy of the generated first candidate feature extraction network, the second candidate feature extraction network, and the candidate actor network.

[0020] Input the training vehicle operation data into the first candidate feature extraction network, and output the candidate environmental risk features; input the training personnel information into the second candidate feature extraction network, and output the candidate personnel comfort features; input the candidate environmental risk features and the candidate personnel comfort features into the candidate actor network, and the candidate actor network outputs the candidate intelligent suspension control information, thereby ensuring the accuracy of the output candidate intelligent suspension control information. Input the candidate environmental risk features, the candidate personnel comfort features, and the candidate intelligent suspension control information into the initial critic network, and output the virtual scores corresponding to the candidate intelligent suspension control information, thereby ensuring the accuracy of the output virtual scores. Calculate the second loss function between the virtual scores and the training scores; based on the second loss function, update the initial critic network and generate the candidate critic network; thereby ensuring the accuracy of the generated candidate critic network. Input the training vehicle operation data into the first candidate feature extraction network, and output the alternative environmental risk features; input the training personnel information into the second candidate feature extraction network, and output the alternative personnel comfort features; input the alternative environmental risk features and the alternative personnel comfort features into the candidate actor network, and the candidate actor network outputs the alternative intelligent suspension control information; calculate the first loss function between the alternative intelligent suspension control information and the training intelligent suspension control information, update the first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network based on the first loss function, generate the first target feature extraction network, the second target feature extraction network and the target actor network, and ensure the accuracy of the generated first target feature extraction network, the second target feature extraction network and the target actor network. Input the alternative environmental risk features, the alternative personnel comfort features and the alternative intelligent suspension control information into the candidate critic network, and output the alternative scores corresponding to the alternative intelligent suspension control information; calculate the second loss function between the alternative scores and the training scores. Based on the second loss function, update the candidate critic network, generate the target critic network, and ensure the accuracy of the generated target critic network.

[0021] In a second aspect, the present invention provides an intelligent suspension control device, the device comprising: A first acquisition module is used to acquire operating status data corresponding to the current operating status of the target vehicle; the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data; The second acquisition module is used to acquire target person information corresponding to the target person in the target vehicle; the target person information includes at least one of age information, gender information, health status information, and driving style information; An output module, used to input the operation status data and target personnel information into a preset reinforcement learning model, and output the target intelligent suspension control information corresponding to the target vehicle; A control module for target-based intelligent suspension control information.

[0022] The intelligent suspension control device provided in the embodiment of the present application obtains the operating status data corresponding to the current operating status of the target vehicle; obtains the target personnel information corresponding to the target personnel in the target vehicle; inputs the operating status data and the target personnel information into the preset reinforcement learning model, and outputs the target intelligent suspension control information corresponding to the target vehicle, thereby ensuring the accuracy of the output target intelligent suspension control information. Then, based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled, and the intelligent suspension responds in advance, thereby accurately adjusting the intelligence. In addition, the above-mentioned device takes into account both the operating status data corresponding to the target vehicle and the target personnel information, thereby ensuring the safety of the target vehicle and the comfort of the target personnel.

[0023] In a third aspect, the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the intelligent suspension control method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0024] In a fourth aspect, the present invention provides an intelligent vehicle, which includes a vehicle body, an intelligent suspension, and an electronic device, wherein the electronic device is used to execute the intelligent suspension control method of the first aspect or any corresponding embodiment thereof.

[0025] In a fifth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the intelligent suspension control method of the first aspect or any corresponding embodiment thereof.

[0026] In a sixth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the intelligent suspension control method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 is a flow chart of an intelligent suspension control method according to an embodiment of the present invention; Figure 2is a flow chart of another intelligent suspension control method according to an embodiment of the present invention; Figure 3 is a flow chart of another intelligent suspension control method according to an embodiment of the present invention; Figure 4 is a flow chart of another intelligent suspension control method according to an embodiment of the present invention; Figure 5 is a structural block diagram of an intelligent suspension control device according to an embodiment of the present invention; Figure 6 It is a schematic diagram of the hardware structure of the electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0030] It should be noted that the execution subject of the intelligent suspension control method provided in the embodiment of the present application may be an intelligent suspension control device, and the intelligent suspension control device may be implemented as part or all of an electronic device through software, hardware, or a combination of software and hardware, wherein the electronic device may be an intelligent suspension control device in a target vehicle. In the following method embodiments, the execution subject is an electronic device as an example for explanation.

[0031] According to an embodiment of the present invention, an embodiment of an intelligent suspension control method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0032] In this embodiment, an intelligent suspension control method is provided, which can be used in the above-mentioned electronic device. Figure 1 is a flow chart of an intelligent suspension control method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps: Step S101, obtaining operating status data corresponding to the current operating status of the target vehicle.

[0033] Among them, the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data.

[0034] Specifically, the electronic device may receive operating status data corresponding to the current operating status of the target vehicle collected by the sensor device.

[0035] Specifically, the target vehicle can be equipped with various sensors, such as speed sensors, acceleration sensors, gyroscopes, GPS positioning systems, etc., to collect target vehicle operation data, such as speed, acceleration, position, etc. In addition, the target vehicle can also be equipped with cameras, lidar sensors, millimeter wave radar sensors and other equipment to monitor the location and behavior of surrounding vehicles. Electronic devices can also collect road environment images through cameras, identify road environment images, and determine current road status data, such as road congestion, road surface conditions, etc. Electronic devices can also identify road environment images and determine traffic flow data.

[0036] In addition, electronic devices can also obtain current weather information, including temperature, humidity, rainfall, snowfall, etc., through weather data providers, or electronic devices can recognize weather images taken by cameras to determine current weather information, such as cloudy, rainy, snowy, foggy, and sunny.

[0037] In addition, electronic devices can also use driver behavior monitoring equipment, such as cameras, sensors, etc., to monitor driver behavior, such as distraction and fatigue driving.

[0038] Step S102, obtaining target person information corresponding to the target person in the target vehicle.

[0039] The target person information includes at least one of age information, gender information, health status information, and driving style information. The target person may include a driver and a passenger.

[0040] Specifically, the electronic device can receive target person information corresponding to a target person in a target vehicle input by a user, and can also collect an image of the target person, recognize the image of the target person, determine the age information, gender information, and health status information corresponding to the target person, and receive driving style information corresponding to the target person input by the user.

[0041] The embodiment of the present application does not specifically limit the manner in which the electronic device obtains the target person information corresponding to the target person in the target vehicle.

[0042] This step will be described in detail below.

[0043] Step S103, inputting the operating status data and target personnel information into a preset reinforcement learning model, and outputting target intelligent suspension control information corresponding to the target vehicle.

[0044] Specifically, the electronic device can input the operation status data and the target person information into a preset reinforcement learning model. Then, the preset reinforcement learning model can perform feature extraction on the operation status data and the target person information, thereby extracting the target environment risk characteristics and the target person comfort characteristics corresponding to the target vehicle. Next, based on the target environment risk characteristics and the target person comfort characteristics, the target intelligent suspension control information corresponding to the target vehicle is output.

[0045] This step will be described in detail below.

[0046] Step S104: Controlling the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

[0047] Specifically, the electronic device can control the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

[0048] Exemplarily, the target intelligent suspension control information may include intelligent suspension stiffness and intelligent suspension damping, and the electronic device may adjust the damping and stiffness of the intelligent suspension of the target vehicle according to the intelligent suspension stiffness and intelligent suspension damping included in the target intelligent suspension control information.

[0049] The intelligent suspension control method provided in the embodiment of the present application obtains the operating status data corresponding to the current operating status of the target vehicle; obtains the target personnel information corresponding to the target personnel in the target vehicle; inputs the operating status data and the target personnel information into the preset reinforcement learning model, and outputs the target intelligent suspension control information corresponding to the target vehicle, thereby ensuring the accuracy of the output target intelligent suspension control information. Then, based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled, and the intelligent suspension responds in advance, thereby accurately adjusting the intelligence. In addition, the above method takes into account both the operating status data corresponding to the target vehicle and the target personnel information, thereby ensuring the safety of the target vehicle and the comfort of the target personnel.

[0050] In this embodiment, an intelligent suspension control method is provided, which can be used in the above-mentioned electronic device. Figure 2 is a flow chart of an intelligent suspension control method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps: Step S201, obtaining the running status data corresponding to the current running status of the target vehicle.

[0051] Among them, the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data.

[0052] For details about this step, please refer to the above description of step S101, which will not be elaborated here.

[0053] Step S202, obtaining target person information corresponding to the target person in the target vehicle.

[0054] The target person information includes at least one of age information, gender information, health status information, and driving style information.

[0055] For details about this step, please refer to the above description of step S102, which will not be elaborated here.

[0056] Step S203, inputting the operating status data and target personnel information into a preset reinforcement learning model, and outputting target intelligent suspension control information corresponding to the target vehicle.

[0057] Specifically, the preset reinforcement learning model includes a first target feature extraction network, a second target feature extraction network, and a feature fusion network; the above step S203 may include the following steps: Step S2031, input the operating status data into the first target feature extraction network in the preset reinforcement learning model, and output the target environment risk feature corresponding to the current operating status of the target vehicle.

[0058] Among them, the target environment risk feature is used to characterize the danger level of the current driving environment of the target vehicle.

[0059] Among them, the first target feature extraction network can be composed of a recurrent neural network (RNN) plus an attention mechanism.

[0060] Specifically, the electronic device can pre-process the vehicle's operating status data, including data cleaning, feature engineering and other operations, for subsequent model training and prediction. The recurrent neural network can extract the time series features of the operating status data. In addition, the attention mechanism is introduced into the recurrent neural network. The attention mechanism can assign different weights according to the importance of the input operating status data, so as to better extract the key information in the vehicle's operating status data.

[0061] Then, the RNN network and attention mechanism in the first target feature extraction network are used to extract features from the vehicle operation status data to obtain a feature vector corresponding to the current operation status of the target vehicle. The extracted feature vector is mapped to the target environment risk feature space to obtain the target environment risk feature corresponding to the current operation status of the target vehicle.

[0062] Step S2032: input the target person information into the second target feature extraction network in the preset reinforcement learning model, and output the target person comfort feature corresponding to the current operating state of the target vehicle.

[0063] Among them, the comfort characteristics of the target personnel are used to characterize the comfort requirements of the target personnel.

[0064] Specifically, the input layer in the second target feature extraction network can take the target person information as input. The target person information can include at least one of age information, gender information, health status information, and driving style information. The convolutional neural network (CNN) in the second target feature extraction network extracts features from the input target person information. Among them, the convolutional layer can automatically learn the feature pattern related to the comfort of the target person. Then, the pooling layer in the second target feature extraction network performs a pooling operation on the output of the convolutional layer to reduce the feature dimension and extract the main features. The recurrent neural network (RNN) layer in the second target feature extraction network processes time series data because the current operating state of the target vehicle changes over time. RNN can capture the temporal dynamics of the comfort features of the target person. In addition, in the RNN layer, the long short-term memory network (LSTM) or the gated recurrent unit (GRU) in the second target feature extraction network better handles long-term dependencies, thereby more accurately extracting the comfort features of the target person. In addition, an attention mechanism is added to the second target feature extraction network to focus on the part of the input information that is most relevant to the comfort of the target person. The attention mechanism can assign weights according to the importance of the input, thereby improving the accuracy of feature extraction. Finally, the fully connected layer in the second target feature extraction network connects the output of the RNN layer to the fully connected layer for final feature extraction and classification. The output layer in the second target feature extraction network outputs the target occupant comfort features corresponding to the current operating state of the target vehicle.

[0065] Step S2033, perform feature fusion on the target environment risk characteristics and the target personnel comfort characteristics, and output the target intelligent suspension control information corresponding to the target vehicle.

[0066] Specifically, the feature fusion network includes a target actor network and a target critic network; the above step S2033 may include the following steps: Step a1: input the target environment risk characteristics and the target personnel comfort characteristics into the target actor network, the target actor network extracts the target environment risk characteristics and the target personnel comfort characteristics, and outputs the candidate intelligent suspension control information corresponding to the target vehicle.

[0067] Specifically, the electronic device can input the target environmental risk characteristics and the target personnel comfort characteristics into the target actor network, and the target actor network can use a series of convolutional layers, pooling layers, and fully connected layers to extract and transform the input target environmental risk characteristics and target personnel comfort characteristics. These layers can automatically learn the representation of the features to better understand the information of environmental risks and personnel comfort. Then, based on the extracted features, the target actor network generates candidate intelligent suspension control information. Among them, the candidate intelligent suspension control information may include the stiffness, damping or other related parameters of the intelligent suspension.

[0068] In step a2, the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information are input into the target critic network, the target critic network extracts the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information, outputs the candidate scores corresponding to the candidate intelligent suspension control information, and transmits the candidate scores to the target actor network.

[0069] Among them, the candidate score is used to characterize the safety and comfort of controlling the target vehicle based on the candidate intelligent suspension control information.

[0070] In an optional implementation manner of the present invention, the candidate intelligent suspension control information is modified according to the candidate score using the following formula, and the target intelligent suspension control information is output, including: , in, is the candidate score corresponding to the candidate intelligent suspension control information, s is the target environmental risk characteristics and target personnel comfort characteristics corresponding to the current moment, is the candidate intelligent suspension control information corresponding to the target vehicle, The target environmental risk characteristics and target personnel comfort characteristics corresponding to the next moment; The target intelligent suspension control information corresponding to the next moment; is a preset scoring threshold; is the learning rate; Rewards for the environment; is the discount factor; The weight coefficient is used to control the influence of the candidate score on the value function update.

[0071] Step a3: If the candidate score is greater than or equal to the preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information and outputs the target intelligent suspension control information.

[0072] Specifically, the target actor network may receive a preset score threshold set by a user. Then, the candidate score transmitted by the target critic network is compared with the preset score threshold. If the candidate score is greater than or equal to the preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information and outputs the target intelligent suspension control information.

[0073] Step a4: if the candidate score is less than a preset score threshold, the candidate intelligent suspension control information is modified according to the candidate score, and the target intelligent suspension control information is output.

[0074] Specifically, if the candidate score is less than the preset score threshold, the target actor network determines that the candidate intelligent suspension control information currently output is unqualified. Therefore, the target actor network extracts the target environmental risk characteristics and the target personnel comfort characteristics again until the target intelligent suspension control information is output, that is, the target score corresponding to the target intelligent suspension control information is greater than the preset score threshold.

[0075] Step S204: Control the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

[0076] For details about this step, please refer to the above description of step S104, which will not be elaborated here.

[0077] The intelligent suspension control method provided by the embodiment of the present application inputs the operating state data into the first target feature extraction network in the preset reinforcement learning model, and outputs the target environment risk feature corresponding to the current operating state of the target vehicle; the accuracy of the output target environment risk feature is ensured, so that the feature extraction of the driving environment risk can be realized. The target personnel information is input into the second target feature extraction network in the preset reinforcement learning model, and the target personnel comfort feature corresponding to the current operating state of the target vehicle is output, which ensures the accuracy of the output target personnel comfort feature. The target environment risk feature and the target personnel comfort feature are input into the target actor network, and the target actor network extracts the target environment risk feature and the target personnel comfort feature, and outputs the candidate intelligent suspension control information corresponding to the target vehicle, which ensures the accuracy of the output candidate intelligent suspension control information corresponding to the target vehicle. The target environment risk feature, the target personnel comfort feature and the candidate intelligent suspension control information are input into the target critic network, and the target critic network extracts the target environment risk feature, the target personnel comfort feature and the candidate intelligent suspension control information, outputs the candidate score corresponding to the candidate intelligent suspension control information, and transmits the candidate score to the target actor network; the accuracy of the output candidate score is ensured. If the candidate score is greater than or equal to the preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information and outputs the target intelligent suspension control information, thereby ensuring that the score corresponding to the output target intelligent suspension control information is high. If the candidate score is less than the preset score threshold, the candidate intelligent suspension control information is corrected according to the candidate score and the target intelligent suspension control information is output, thereby ensuring the accuracy of the output target intelligent suspension control information and avoiding the inaccurate output target intelligent suspension control information, which may cause dangerous driving of the target vehicle and / or discomfort to the target person.

[0078] In this embodiment, an intelligent suspension control method is provided, which can be used in the above-mentioned electronic device. Figure 3 is a flow chart of an intelligent suspension control method according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps: Step S301, obtaining the running status data corresponding to the current running status of the target vehicle.

[0079] Among them, the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data.

[0080] For details about this step, please refer to the above description of step S201, which will not be elaborated here.

[0081] Step S302, obtaining target person information corresponding to the target person in the target vehicle.

[0082] The target person information includes at least one of age information, gender information, health status information, and driving style information.

[0083] Specifically, the target person information includes age information, gender information, and health status information. The above step S302 may include the following steps: Step S3021, obtaining a person image corresponding to a target person in a target vehicle.

[0084] Specifically, the electronic device can receive a person image corresponding to a target person in a target vehicle captured by a camera.

[0085] Step S3022: recognize the person image and determine the target region of interest from the person image.

[0086] Among them, the target area of ​​interest includes the target person.

[0087] Specifically, the electronic device may utilize a target detection algorithm to identify a person image and determine a target region of interest from the person image.

[0088] Among them, the target detection algorithm can be a model based on manual features, such as DPM (Deformable Parts Model), and the target detection algorithm can also be a model based on a convolutional neural network, such as YOLO (You Only Look Once) detector, R-CNN, (Region-based Convolutional Neural Networks, region-based convolutional neural network) model, SSD (Single Shot MultiBox, single shot multi-box) detector and Mask R-CNN (Mask Region-based Convolutional Neural Networks, masked region-based convolutional neural network) model, etc. The embodiment of the present application does not specifically limit the target detection algorithm.

[0089] Step S3023, identifying the target region of interest, and determining the target person's facial image and the target person's body image.

[0090] Specifically, the electronic device may use a facial recognition algorithm to identify the target region of interest, and the facial recognition algorithm may extract features from the target region of interest to obtain facial features, and determine the facial image of the target person based on the facial features.

[0091] Specifically, the electronic device may use a body recognition algorithm to identify the target region of interest, and the body recognition algorithm may extract features from the target region of interest to obtain body features, and determine the body image of the target person based on the body features.

[0092] Step S3024, identifying the target person's facial image and determining the target person's age and gender information.

[0093] Specifically, the electronic device inputs the target person's facial image into the age recognition model and the gender recognition model. The age recognition model extracts features from the target person's facial image, extracts features such as facial contour features, facial wrinkle features, facial texture features, and facial skin color features, and then outputs the target person's age information based on the extracted features such as facial contour features, facial wrinkle features, facial texture features, and facial skin color features.

[0094] The gender recognition model extracts features from the target person's facial image and outputs the target person's gender information based on the extracted features.

[0095] Among them, the age recognition model and the gender recognition model can be models based on manual features, such as DPM (Deformable Parts Model), and the age recognition model and the gender recognition model can also be models based on convolutional neural networks, such as YOLO (You Only Look Once) detector, R-CNN, (Region-based Convolutional Neural Networks, region-based convolutional neural network) model, SSD (Single Shot MultiBox, single shot multi-box) detector and Mask R-CNN (Mask Region-based Convolutional Neural Networks, masked region-based convolutional neural network) model, etc. The embodiment of the present application does not specifically limit the age recognition model and the gender recognition model.

[0096] Step S3025, identifying the target person's facial image and body image, detecting whether the target person has any injured parts, and determining the target person's health information based on the detection results.

[0097] Specifically, the electronic device can identify the facial image and body image of the target person, detect whether the target person has any injuries, and determine the health information of the target person based on the detection results.

[0098] Step S303: input the operating status data and target personnel information into a preset reinforcement learning model, and output target intelligent suspension control information corresponding to the target vehicle.

[0099] For details about this step, please refer to the above description of step S203, which will not be elaborated here.

[0100] Step S304: Control the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

[0101] For details about this step, please refer to the above description of step S204, which will not be elaborated here.

[0102] The intelligent suspension control method provided in the embodiment of the present application obtains a person image corresponding to a target person in a target vehicle; identifies the person image, and determines a target region of interest from the person image, wherein the target region of interest includes the target person, thereby ensuring the accuracy of the determined target region of interest. The target region of interest is identified, and the facial image and the body image of the target person are determined, thereby ensuring the accuracy of the determined facial image and the body image of the target person. The facial image of the target person is identified, and the age information and the gender information of the target person are determined, thereby ensuring the accuracy of the determined age information and the gender information. The facial image and the body image of the target person are identified, and it is detected whether the target person has any injured position. Based on the detection results, the health information of the target person is determined, thereby ensuring the accuracy of the determined health information of the target person.

[0103] In an optional embodiment of the present application, Figure 4 As shown, the training process of the preset reinforcement learning model may include the following steps: Step S401, obtaining a training data set.

[0104] Among them, the training data set includes training vehicle operation data, training personnel information, training intelligent suspension control information and training scores corresponding to the training intelligent suspension control information; among them, the training intelligent suspension control information and the training scores are label information.

[0105] Specifically, the electronic device may receive a training data set input by a user, or may receive a training data set sent by other devices, and the electronic device may also search for a training data set from a storage space.

[0106] The embodiment of the present application does not specifically limit the manner in which the electronic device obtains the training data set.

[0107] Step S402, inputting the training vehicle operation data, the training personnel information, the training intelligent suspension control information and the training score corresponding to the training intelligent suspension control information into the initial reinforcement learning network, training the initial reinforcement learning network, and generating a preset reinforcement learning model.

[0108] Specifically, the above step S402 may include the following steps: Step S4021, input the training vehicle operation data into the first initial feature extraction network, and output the training environment risk features.

[0109] Specifically, the first initial feature extraction network can be composed of a recurrent neural network (RNN) plus an attention mechanism.

[0110] Specifically, the electronic device can pre-process the training vehicle operation data, including data cleaning, feature engineering and other operations, for subsequent model training and prediction. The recurrent neural network can extract the time series features of the training vehicle operation data. In addition, the attention mechanism is introduced into the recurrent neural network. The attention mechanism can assign different weights according to the importance of the input training vehicle operation data, so as to better extract the key information in the vehicle operation status data.

[0111] Then, the first initial feature extraction network extracts features from the training vehicle operation data to obtain feature vectors corresponding to the training vehicle operation data. The extracted feature vectors are mapped to the initial environmental risk feature space to obtain training environmental risk features corresponding to the training vehicle operation data.

[0112] Step S4022: input the trainee information into the second initial feature extraction network, and output the trainee comfort features.

[0113] Specifically, the input layer in the second initial feature extraction network can take the trainee information as input. The trainee information can include at least one of age information, gender information, health status information, and driving style information. The convolutional neural network (CNN) in the second initial feature extraction network extracts features from the input trainee information. Among them, the convolutional layer can automatically learn the feature pattern related to the comfort of the target person. Then, the pooling layer in the second initial feature extraction network performs a pooling operation on the output of the convolutional layer to reduce the feature dimension and extract the main features. The recurrent neural network (RNN) layer in the second initial feature extraction network processes the time series data. Specifically, the RNN layer can capture the temporal dynamics of the trainee comfort feature. In addition, in the RNN layer, the long short-term memory network (LSTM) or the gated recurrent unit (GRU) in the second initial feature extraction network better handles long-term dependencies, thereby more accurately extracting the trainee comfort feature. In addition, an attention mechanism is added to the second initial feature extraction network to focus on the part of the input information that is most relevant to the trainee comfort. The attention mechanism can assign weights according to the importance of the input, thereby improving the accuracy of feature extraction. Finally, the fully connected layer in the second initial feature extraction network connects the output of the RNN layer to the fully connected layer for final feature extraction and classification. The output layer in the second initial feature extraction network outputs the trainee comfort features corresponding to the trainee information.

[0114] Step S4023, inputting the risk characteristics of the training environment and the comfort characteristics of the trainees into the initial actor network, and the initial actor network outputs the virtual intelligent suspension control information.

[0115] Specifically, the electronic device can input the risk characteristics of the training environment and the comfort characteristics of the trainees into the initial actor network, and the initial actor network can extract the risk characteristics of the training environment and the comfort characteristics of the trainees, and output the virtual intelligent suspension control information.

[0116] Step S4024, calculate the first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information, and based on the first loss function, update the first initial feature extraction network, the second initial feature extraction network, and the initial actor network to generate the first candidate feature extraction network, the second candidate feature extraction network, and the candidate actor network.

[0117] Specifically, the electronic device calculates a first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information.

[0118] Among them, the first loss function can be a mean squared error (MSE) loss function, a cross entropy loss function, a binary cross entropy loss function, a categorical cross entropy loss function, a Hinge loss function, an absolute value loss function, a logarithmic loss function, etc.

[0119] Step S4025, input the training vehicle operation data into the first candidate feature extraction network, and output the candidate environmental risk features.

[0120] Specifically, the electronic device inputs the training vehicle operation data into the first candidate feature extraction network and outputs the candidate environmental risk features.

[0121] For details about this step, please refer to the above description of step S4021, which will not be repeated here.

[0122] Step S4026, input the training personnel information into the second candidate feature extraction network, and output the candidate personnel comfort features.

[0123] Specifically, the electronic device inputs the training personnel information into the second candidate feature extraction network and outputs the candidate personnel comfort features.

[0124] For details about this step, please refer to the above description of step S4022, which will not be repeated here.

[0125] Step S4027, inputting the candidate environmental risk characteristics and the candidate actor comfort characteristics into the candidate actor network, and the candidate actor network outputs the candidate intelligent suspension control information.

[0126] For details about this step, please refer to the above description of step S4023, which will not be repeated here.

[0127] Step S4028, input the candidate environmental risk characteristics, candidate occupant comfort characteristics and candidate intelligent suspension control information into the initial critic network, and output the virtual score corresponding to the candidate intelligent suspension control information.

[0128] Specifically, the electronic device can input candidate environmental risk characteristics, candidate occupant comfort characteristics, and candidate intelligent suspension control information into the initial critic network, and output a virtual score corresponding to the candidate intelligent suspension control information.

[0129] Step S4029, calculating a second loss function between the virtual score and the training score.

[0130] Specifically, the electronic device may calculate a second loss function between the virtual score and the training score.

[0131] Among them, the second loss function can be a mean squared error (MSE) loss function, a cross entropy loss function, a binary cross entropy loss function, a categorical cross entropy loss function, a Hinge loss function, an absolute value loss function, a logarithmic loss function, etc.

[0132] Step S40210: Based on the second loss function, update the initial critic network and generate a candidate critic network.

[0133] Specifically, the electronic device can calculate the gradient of each parameter in the initial critic network through the back propagation algorithm according to the value of the second loss function. Then, the parameters of the initial critic network are updated according to the calculated gradient using an optimization algorithm such as gradient descent. The above steps are iterated multiple times until the value of the second loss function converges or reaches a preset number of iterations, and the obtained initial critic network is the candidate critic network.

[0134] Step S40211, input the training vehicle operation data into the first candidate feature extraction network, and output the candidate environmental risk features.

[0135] For details about this step, please refer to the above description of step S4021, which will not be repeated here.

[0136] Step S40212: input the training personnel information into the second candidate feature extraction network, and output the comfort features of the candidate personnel.

[0137] For details about this step, please refer to the above description of step S4022, which will not be repeated here.

[0138] Step S40213, inputting the alternative environmental risk characteristics and the alternative personnel comfort characteristics into the candidate actor network, and the candidate actor network outputs the alternative intelligent suspension control information.

[0139] For details about this step, please refer to the above description of step S4023, which will not be repeated here.

[0140] Step S40214, calculate the first loss function between the alternative intelligent suspension control information and the training intelligent suspension control information, based on the first loss function, update the first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network, and generate the first target feature extraction network, the second target feature extraction network and the target actor network.

[0141] Among them, the first loss function is used to measure the difference between the alternative intelligent suspension control information and the training intelligent suspension control information. Specifically, the electronic device can use the back propagation algorithm to update the parameters of the first candidate feature extraction network, the second candidate feature extraction network, and the candidate actor network based on the first loss function. This process adjusts the weights of the network so that the network can better fit the training data, thereby improving the accuracy of the strategy. Then, after multiple iterative updates, when the loss function converges or reaches a certain performance indicator, the updated network is saved as the first target feature extraction network, the second target feature extraction network, and the target actor network. These target networks will be used in actual intelligent suspension control applications to generate the optimal intelligent suspension control strategy based on the input feature information.

[0142] Step S40215, inputting the alternative environmental risk characteristics, the alternative personnel comfort characteristics and the alternative intelligent suspension control information into the candidate critic network, and outputting the alternative scores corresponding to the alternative intelligent suspension control information.

[0143] Specifically, the electronic device inputs alternative environmental risk characteristics, alternative personnel comfort characteristics, and alternative intelligent suspension control information into the candidate critic network, and the candidate critic network outputs alternative scores corresponding to the alternative intelligent suspension control information.

[0144] Step S40216, calculating a second loss function between the candidate score and the training score.

[0145] Specifically, the electronic device may calculate a second loss function between the candidate scores and the training scores.

[0146] Among them, the second loss function can be a mean squared error (MSE) loss function, a cross entropy loss function, a binary cross entropy loss function, a categorical cross entropy loss function, a Hinge loss function, an absolute value loss function, a logarithmic loss function, etc.

[0147] Step S40217: based on the second loss function, update the candidate critic network to generate a target critic network.

[0148] Specifically, the electronic device can calculate the gradient of each parameter in the candidate critic network through the back propagation algorithm according to the value of the second loss function. Then, the parameters of the candidate critic network are updated according to the calculated gradient using an optimization algorithm such as gradient descent. The above steps are iterated multiple times until the value of the second loss function converges or reaches a preset number of iterations, and the obtained candidate critic network is the target critic network.

[0149] The intelligent suspension control method provided in the embodiment of the present application obtains a training data set, which includes training vehicle operation data, training personnel information, training intelligent suspension control information, and training scores corresponding to the training intelligent suspension control information; wherein the training intelligent suspension control information and the training scores are label information.

[0150] The training vehicle operation data is input into the first initial feature extraction network to output the training environment risk characteristics, and the training personnel information is input into the second initial feature extraction network to output the training personnel comfort characteristics; the training environment risk characteristics and the training personnel comfort characteristics are input into the initial actor network, and the initial actor network outputs the virtual intelligent suspension control information, which ensures that the output virtual intelligent suspension control information comprehensively considers the training environment risk characteristics and the training personnel comfort characteristics. The first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information is calculated, and based on the first loss function, the first initial feature extraction network, the second initial feature extraction network and the initial actor network are updated to generate the first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network, which ensures the accuracy of the generated first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network.

[0151] Input the training vehicle operation data into the first candidate feature extraction network, and output the candidate environmental risk features; input the training personnel information into the second candidate feature extraction network, and output the candidate personnel comfort features; input the candidate environmental risk features and the candidate personnel comfort features into the candidate actor network, and the candidate actor network outputs the candidate intelligent suspension control information, thereby ensuring the accuracy of the output candidate intelligent suspension control information. Input the candidate environmental risk features, the candidate personnel comfort features, and the candidate intelligent suspension control information into the initial critic network, and output the virtual scores corresponding to the candidate intelligent suspension control information, thereby ensuring the accuracy of the output virtual scores. Calculate the second loss function between the virtual scores and the training scores; based on the second loss function, update the initial critic network and generate the candidate critic network; thereby ensuring the accuracy of the generated candidate critic network. Input the training vehicle operation data into the first candidate feature extraction network, and output the alternative environmental risk features; input the training personnel information into the second candidate feature extraction network, and output the alternative personnel comfort features; input the alternative environmental risk features and the alternative personnel comfort features into the candidate actor network, and the candidate actor network outputs the alternative intelligent suspension control information; calculate the first loss function between the alternative intelligent suspension control information and the training intelligent suspension control information, update the first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network based on the first loss function, generate the first target feature extraction network, the second target feature extraction network and the target actor network, and ensure the accuracy of the generated first target feature extraction network, the second target feature extraction network and the target actor network. Input the alternative environmental risk features, the alternative personnel comfort features and the alternative intelligent suspension control information into the candidate critic network, and output the alternative scores corresponding to the alternative intelligent suspension control information; calculate the second loss function between the alternative scores and the training scores. Based on the second loss function, update the candidate critic network, generate the target critic network, and ensure the accuracy of the generated target critic network.

[0152] In this embodiment, an intelligent suspension control device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0153] This embodiment provides an intelligent suspension control device, such as Figure 5 As shown, including: The first acquisition module 501 is used to acquire the operating state data corresponding to the current operating state of the target vehicle; the operating state data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road state data, weather information, traffic flow data, and driver behavior data; The second acquisition module 502 is used to acquire target person information corresponding to the target person in the target vehicle; the target person information includes at least one of age information, gender information, health status information, and driving style information; The output module 503 is used to input the operation status data and the target personnel information into the preset reinforcement learning model, and output the target intelligent suspension control information corresponding to the target vehicle; The control module 504 is used to control the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

[0154] In some optional embodiments, the preset reinforcement learning model includes a first target feature extraction network, a second target feature extraction network and a feature fusion network; the output module 503 is specifically used to input the operating status data into the first target feature extraction network in the preset reinforcement learning model, and output the target environment risk characteristics corresponding to the current operating status of the target vehicle; the target environment risk characteristics are used to characterize the degree of danger of the current driving environment of the target vehicle; the target personnel information is input into the second target feature extraction network in the preset reinforcement learning model, and the target personnel comfort characteristics corresponding to the current operating status of the target vehicle are output; the target personnel comfort characteristics are used to characterize the degree of comfort requirements of the target personnel; the target environment risk characteristics and the target personnel comfort characteristics are feature fused, and the target intelligent suspension control information corresponding to the target vehicle is output.

[0155] In some optional embodiments, the feature fusion network includes a target actor network and a target critic network; the output module 503 is specifically used to input the target environment risk characteristics and the target personnel comfort characteristics into the target actor network, the target actor network extracts the target environment risk characteristics and the target personnel comfort characteristics, and outputs the candidate intelligent suspension control information corresponding to the target vehicle; the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information are input into the target critic network, the target critic network extracts the target environment risk characteristics, the target personnel comfort characteristics and the candidate intelligent suspension control information, outputs the candidate score corresponding to the candidate intelligent suspension control information, and transmits the candidate score to the target actor network; the candidate score is used to characterize the safety and comfort corresponding to the control of the target vehicle based on the candidate intelligent suspension control information; if the candidate score is greater than or equal to a preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information, and outputs the target intelligent suspension control information; if the candidate score is less than the preset score threshold, the candidate intelligent suspension control information is corrected according to the candidate score, and the target intelligent suspension control information is output.

[0156] In some optional implementations, the output module 503 is specifically configured to use the formula , output target intelligent suspension control information.

[0157] In some optional embodiments, the target person information includes age information, gender information, and health status information. The second acquisition module 502 is specifically used to obtain a person image corresponding to the target person in the target vehicle; identify the person image, and determine a target area of ​​interest from the person image, where the target area of ​​interest includes the target person; identify the target area of ​​interest to determine a facial image and a body image of the target person; identify the facial image of the target person to determine the age information and gender information of the target person; identify the facial image and the body image of the target person to detect whether the target person has any injured position, and determine the health information of the target person based on the detection result.

[0158] In some optional embodiments, the output module 503 is also used to obtain a training data set, which includes training vehicle operation data, training personnel information, training intelligent suspension control information, and training scores corresponding to the training intelligent suspension control information; wherein the training intelligent suspension control information and the training scores are label information; the training vehicle operation data, training personnel information, training intelligent suspension control information, and the training scores corresponding to the training intelligent suspension control information are input into the initial reinforcement learning network, the initial reinforcement learning network is trained, and a preset reinforcement learning model is generated.

[0159] In some optional embodiments, the initial reinforcement learning network includes a first initial feature extraction network, a second initial feature extraction network, an initial actor network and an initial critic network, and the output module 503 is specifically used to input the training vehicle operation data into the first initial feature extraction network, and output the training environment risk characteristics; input the training personnel information into the second initial feature extraction network, and output the training personnel comfort characteristics; input the training environment risk characteristics and the training personnel comfort characteristics into the initial actor network, and the initial actor network outputs the virtual intelligent suspension control information; calculate the first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information, and based on the first loss function, update the first initial feature extraction network, the second initial feature extraction network and the initial actor network to generate a first candidate feature extraction network, a second candidate feature extraction network and a candidate actor network; input the training vehicle operation data into the first candidate feature extraction network, and output the candidate environment risk characteristics; input the training personnel information into the second candidate feature extraction network, and output the candidate personnel comfort characteristics; input the candidate environment risk characteristics and the candidate personnel comfort characteristics into the candidate actor network, and the candidate actor network outputs the candidate intelligent suspension control information; The features and candidate intelligent suspension control information are input into the initial critic network, and the virtual score corresponding to the candidate intelligent suspension control information is output; the second loss function between the virtual score and the training score is calculated; based on the second loss function, the initial critic network is updated to generate a candidate critic network; the training vehicle operation data is input into the first candidate feature extraction network, and the alternative environmental risk features are output; the training personnel information is input into the second candidate feature extraction network, and the alternative personnel comfort features are output; the alternative environmental risk features and the alternative personnel comfort features are input into the candidate actor network, and the candidate actor network outputs the alternative intelligent suspension control information; the candidate A first loss function is calculated between the selected intelligent suspension control information and the trained intelligent suspension control information; based on the first loss function, the first candidate feature extraction network, the second candidate feature extraction network and the candidate actor network are updated to generate a first target feature extraction network, a second target feature extraction network and a target actor network; the alternative environmental risk features, the alternative personnel comfort features and the alternative intelligent suspension control information are input into the candidate critic network, and the alternative scores corresponding to the alternative intelligent suspension control information are output; a second loss function is calculated between the alternative scores and the training scores; based on the second loss function, the candidate critic network is updated to generate a target critic network.

[0160] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0161] The intelligent suspension control device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0162] The embodiment of the present invention also provides an electronic device having the above Figure 5 Intelligent Suspension Control shown.

[0163] See also Figure 6 , Figure 6 is a schematic diagram of the structure of an electronic device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0164] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0165] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0166] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0167] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0168] The electronic device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0169] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the electronic device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator rod, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0170] An embodiment of the present invention further provides an intelligent vehicle, which includes a vehicle body, an intelligent suspension, and an electronic device, and an intelligent suspension control method according to any one of the above-mentioned implementation modes of the electronic device.

[0171] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0172] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0173] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. An intelligent suspension control method, characterized in that: The method comprises: Acquire operating status data corresponding to the current operating status of the target vehicle; the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data; Acquire target person information corresponding to the target person in the target vehicle; the target person information includes at least one of age information, gender information, health status information, and driving style information; Inputting the operating status data and the target person information into a preset reinforcement learning model, and outputting target intelligent suspension control information corresponding to the target vehicle; Based on the target intelligent suspension control information, the intelligent suspension of the target vehicle is controlled.

2. The method according to claim 1, characterized in that The preset reinforcement learning model includes a first target feature extraction network, a second target feature extraction network, and a feature fusion network; the operation status data and the target personnel information are input into the preset reinforcement learning model, and the target intelligent suspension control information corresponding to the target vehicle is output, including: The operating state data is input into the first target feature extraction network in the preset reinforcement learning model, and the target environment risk feature corresponding to the current operating state of the target vehicle is output; the target environment risk feature is used to characterize the danger level of the current driving environment of the target vehicle; Inputting the target person information into the second target feature extraction network in the preset reinforcement learning model, and outputting the target person comfort feature corresponding to the current running state of the target vehicle; the target person comfort feature is used to characterize the target person's requirement for comfort; The target environmental risk characteristics and the target personnel comfort characteristics are fused, and target intelligent suspension control information corresponding to the target vehicle is output.

3. The method according to claim 2, characterized in that The feature fusion network includes a target actor network and a target critic network; the feature fusion of the target environmental risk feature and the target personnel comfort feature, and outputting the target intelligent suspension control information corresponding to the target vehicle, includes: Inputting the target environment risk feature and the target person comfort feature into the target actor network, the target actor network extracts the target environment risk feature and the target person comfort feature, and outputs candidate intelligent suspension control information corresponding to the target vehicle; The target environment risk feature, the target person comfort feature and the candidate intelligent suspension control information are input into the target critic network, the target critic network performs feature extraction on the target environment risk feature, the target person comfort feature and the candidate intelligent suspension control information, outputs a candidate score corresponding to the candidate intelligent suspension control information, and transmits the candidate score to the target actor network; the candidate score is used to characterize the safety and comfort corresponding to the control of the target vehicle based on the candidate intelligent suspension control information; If the candidate score is greater than or equal to a preset score threshold, the target actor network determines the candidate intelligent suspension control information as the target intelligent suspension control information, and outputs the target intelligent suspension control information; If the candidate score is less than the preset score threshold, the candidate intelligent suspension control information is modified according to the candidate score, and the target intelligent suspension control information is output.

4. The method according to claim 3, characterized in that The candidate intelligent suspension control information is modified according to the candidate score using the following formula, and the target intelligent suspension control information is output, including: ， in, is the candidate score corresponding to the candidate intelligent suspension control information, s is the target environmental risk feature and the target personnel comfort feature corresponding to the current moment, is the candidate intelligent suspension control information corresponding to the target vehicle, The target environment risk characteristics and the target personnel comfort characteristics corresponding to the next moment; The target intelligent suspension control information corresponding to the next moment; is a preset scoring threshold; is the learning rate; Rewards for the environment; is the discount factor; The weight coefficient is used to control the influence of the candidate score on the value function update.

5. The method according to claim 1, characterized in that The target person information includes age information, gender information, and health status information. The step of obtaining the target person information corresponding to the target person in the target vehicle includes: Acquire a person image corresponding to a target person in the target vehicle; Recognize the person image, and determine a target region of interest from the person image, wherein the target region of interest includes the target person; Identify the target region of interest to determine the target person's facial image and the target person's body image; Recognize the facial image of the target person and determine the age information and gender information of the target person; The facial image and body image of the target person are identified to detect whether the target person has any injured parts, and the health information of the target person is determined based on the detection results.

6. The method according to claim 1, characterized in that The training process of the preset reinforcement learning model includes: Acquire a training data set, wherein the training data set includes training vehicle operation data, training personnel information, training intelligent suspension control information, and training scores corresponding to the training intelligent suspension control information; wherein the training intelligent suspension control information and the training scores are label information; The training vehicle operation data, the training personnel information, the training intelligent suspension control information and the training score corresponding to the training intelligent suspension control information are input into the initial reinforcement learning network, the initial reinforcement learning network is trained, and the preset reinforcement learning model is generated.

7. The method according to claim 6, characterized in that The initial reinforcement learning network includes a first initial feature extraction network, a second initial feature extraction network, an initial actor network, and an initial critic network. The training vehicle operation data, the training personnel information, the training intelligent suspension control information, and the training score corresponding to the training intelligent suspension control information are input into the initial reinforcement learning network, the initial reinforcement learning network is trained, and the preset reinforcement learning model is generated, including: Inputting the training vehicle operation data into the first initial feature extraction network, and outputting the training environment risk features; Inputting the trainee information into the second initial feature extraction network, and outputting the trainee comfort feature; Inputting the risk characteristics of the training environment and the comfort characteristics of the trainees into the initial actor network, and the initial actor network outputs virtual intelligent suspension control information; Calculating a first loss function between the virtual intelligent suspension control information and the training intelligent suspension control information, and based on the first loss function, updating the first initial feature extraction network, the second initial feature extraction network, and the initial actor network to generate a first candidate feature extraction network, a second candidate feature extraction network, and a candidate actor network; Inputting the training vehicle operation data into the first candidate feature extraction network, and outputting candidate environmental risk features; Inputting the training personnel information into the second candidate feature extraction network, and outputting the candidate personnel comfort features; Inputting the candidate environmental risk characteristics and the candidate actor comfort characteristics into the candidate actor network, and the candidate actor network outputs candidate intelligent suspension control information; Inputting the candidate environmental risk characteristics, the candidate occupant comfort characteristics and the candidate intelligent suspension control information into the initial critic network, and outputting a virtual score corresponding to the candidate intelligent suspension control information; calculating a second loss function between the virtual score and the training score; Based on the second loss function, updating the initial critic network to generate a candidate critic network; Inputting the training vehicle operation data into the first candidate feature extraction network to output candidate environmental risk features; Inputting the training personnel information into the second candidate feature extraction network, and outputting the comfort features of the candidate personnel; Inputting the candidate environmental risk characteristics and the candidate personnel comfort characteristics into the candidate actor network, and the candidate actor network outputs the candidate intelligent suspension control information; Calculating a first loss function between the candidate intelligent suspension control information and the training intelligent suspension control information, and based on the first loss function, updating the first candidate feature extraction network, the second candidate feature extraction network, and the candidate actor network to generate a first target feature extraction network, a second target feature extraction network, and a target actor network; Inputting the candidate environmental risk characteristics, the candidate personnel comfort characteristics and the candidate intelligent suspension control information into the candidate critic network, and outputting the candidate scores corresponding to the candidate intelligent suspension control information; Calculating a second loss function between the candidate score and the training score; Based on the second loss function, the candidate critic network is updated to generate a target critic network.

8. An intelligent suspension control device, characterized in that: The device comprises: A first acquisition module is used to acquire operating status data corresponding to the current operating status of the target vehicle; the operating status data includes at least one of the target vehicle operating data, surrounding vehicle operating data, current road status data, weather information, traffic flow data, and driver behavior data; A second acquisition module is used to acquire target person information corresponding to the target person in the target vehicle; the target person information includes at least one of age information, gender information, health status information, and driving style information; An output module, used to input the operating status data and the target personnel information into a preset reinforcement learning model, and output target intelligent suspension control information corresponding to the target vehicle; A control module is used to control the intelligent suspension of the target vehicle based on the target intelligent suspension control information.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the intelligent suspension control method according to any one of claims 1 to 7 by executing the computer instructions.

10. An intelligent vehicle, characterized in that: The intelligent vehicle comprises a vehicle body, an intelligent suspension and electronic equipment, and the electronic equipment is used to execute the intelligent suspension control method according to any one of claims 1 to 7.

Images