Automatic driving method, device and equipment for dragged vehicle and medium

Abstract

The invention provides a towed vehicle automatic driving method, device and equipment and a medium, the method comprises the following steps that a trailer drags a vehicle through a traction rope, and the method comprises the following steps that a front camera is controlled to continuously collect image data in front of the vehicle; inputting the acquired front image data into a preset trailer trend identification model to generate trend information of the trailer and tightness change information of the traction rope, the trend information including inclination angle data and speed data of the trailer; the trend information and the tightness change information are input into a preset driving processing model, driving information of the vehicle is generated, and the driving information comprises steering data and braking data; and controlling the steering and speed of the vehicle based on the driving information. Based on the method, various data of the trailer and the traction rope can be acquired based on the acquired image data in front of the vehicle, the driving information is generated, and the vehicle can be automatically driven based on the driving information, so that the driving difficulty of the vehicle is reduced.

Description

Technical Field

[0001] The present application belongs to the field of autonomous driving technology, and specifically relates to an autonomous driving method, device, equipment and medium for a towed vehicle. Background Art

[0002] In the field of vehicle repair and rescue, when a vehicle breaks down or is damaged and cannot drive autonomously, the traditional practice is to use a towing service to tow the damaged vehicle to a repair yard for repair. In this process, the tow truck tows the damaged vehicle with a towing rope, and uses the traction of the towing rope to drive the tires of the damaged vehicle to rotate, thereby moving the vehicle. However, the current towing method has a significant safety hazard: it requires a driver to be in the damaged vehicle, who must remain highly alert throughout the towing process, pay close attention to the driving status of the trailer, and brake and steer the damaged vehicle in a timely manner to ensure the safety of the towing process and avoid traffic accidents such as rear-end collisions, rollovers, or deviations from the road. However, the traditional practice of relying on a towing rope to connect the trailer and the damaged vehicle, and the driver in the damaged vehicle to control the brakes and steering, has exposed many deficiencies in practice.

[0003] On the one hand, from a safety perspective, the driver in the damaged vehicle plays a vital role in the towing process, but the pressure he faces cannot be ignored. Vehicle damage itself may be accompanied by complex mechanical failures, electrical system failures and other problems, which may interfere with the driver's judgment and operation. At the same time, changes in road conditions, mismatched trailer speeds, and sudden changes in towing rope tension that may occur during the towing process all require the driver to have a high degree of adaptability and concentration. In actual operations, the driver may fail to take correct countermeasures in time due to fatigue, tension, negligence or misjudgment of the vehicle's condition, which may lead to traffic accidents.

[0004] On the other hand, from the perspective of efficiency and cost, traditional towing methods also have obvious limitations. For example, the practice of requiring personnel to accompany the damaged vehicle increases the manpower and time costs of the rescue operation. In emergency rescue scenarios, this additional staffing may delay precious rescue time and lead to greater losses. If the driver causes vehicle damage or traffic accidents due to improper operation or negligence during the towing process, additional repair costs and legal liabilities will also be incurred, further increasing the rescue costs. Summary of the invention

[0005] The purpose of the present application is to provide a method, device, equipment and medium for automatic driving of a towed vehicle.

[0006] According to one aspect of the present application, a method for automatically driving a towed vehicle is provided, wherein the trailer tows the vehicle through a towing rope, comprising the following steps:

[0007] Control the front camera to continuously collect image data in front of the vehicle;

[0008] Inputting the acquired front image data into a preset trailer trend recognition model to generate the trailer trend information and the tension change information of the towing rope, wherein the trend information includes the trailer inclination data and speed data;

[0009] Inputting the trend information and the looseness / tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data;

[0010] Based on the driving information, the steering and speed of the vehicle are controlled.

[0011] In one embodiment, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle:

[0012] Get the inclination data of the vehicle;

[0013] The difference between the inclination data of the vehicle and the inclination data of the trailer is determined, and the difference, the trend information and the tightness change information are input into a preset driving processing model to generate the driving information.

[0014] In one embodiment, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle:

[0015] acquiring the turn signal information of the trailer from the trend information;

[0016] The turn signal information, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

[0017] In one embodiment, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle:

[0018] obtaining brake light information of the trailer from the trend information;

[0019] The brake light information, the trend information and the tightness change information are input into a preset driving processing model to generate the brake data.

[0020] In one embodiment, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle:

[0021] Control the vehicle-mounted radar to send signals to the outside, and determine the distance data between the vehicle and the obstacle based on the corresponding reflected signals from the external obstacles;

[0022] The distance data, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

[0023] In one embodiment, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle:

[0024] Get the speed data of the vehicle;

[0025] The speed data of the vehicle, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

[0026] In one embodiment, before the acquired front image data is input into a preset trailer trend recognition model, the following steps are included:

[0027] Obtaining a preset training set, the training set comprising a plurality of training samples and supervisory labels thereof, the training samples being a plurality of the front image data collected in the trailer scene, the supervisory labels representing the trailer trend information and the tension change information of the towing rope corresponding to the training samples;

[0028] The training set is called to train the trailer trend recognition model to a convergence state, so that it can learn the ability to recognize the trailer trend information and the tension change information of the towing rope according to the input front image data.

[0029] According to one aspect of the present application, there is provided an automatic driving device for a towed vehicle, wherein the trailer tows the vehicle through a towing rope, comprising:

[0030] Image acquisition module, controlling the front camera to continuously collect image data in front of the vehicle;

[0031] A trend recognition module, inputting the acquired front image data into a preset trailer trend recognition model to generate trend information of the trailer and tightness change information of the towing rope, wherein the trend information includes inclination data and speed data of the trailer;

[0032] A driving processing module, inputting the trend information and the looseness / tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data;

[0033] The driving control module controls the steering and speed of the vehicle based on the driving information.

[0034] According to one aspect of the present application, a computer device is provided, comprising a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the method described in any one of the preceding purposes.

[0035] According to one aspect of the present application, a computer-readable storage medium is provided, which stores a computer program implemented by the method described in any one of the preceding purposes in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

[0036] Compared with the prior art, this application has many advantages, including but not limited to:

[0037] (1) The method of the present application can continuously collect image data in front of the vehicle through the front camera, and analyze the trend information of the trailer (such as inclination data and speed data) and the change information of the tightness of the towing rope in real time. The method can quickly respond to any subtle changes in the trailer state. This instant feedback mechanism enables the towed vehicle to predict in advance and take corresponding driving actions, such as adjusting the steering or braking in time, thereby effectively avoiding the risk of loss of control or collision caused by sudden changes in the trailer state (such as sharp turns, acceleration or deceleration), and significantly improving the safety during driving.

[0038] (2) The method of the present application inputs the identified trailer trend information and the tension change information of the towing rope into a preset driving processing model, and the driving processing model can generate the optimal driving information (including steering data and braking data) based on these precise data. This data-based intelligent decision-making method ensures that the towed vehicle can closely follow the movement trajectory of the trailer and maintain a stable driving state even in complex and changeable road conditions, reducing shaking or deviation from the trajectory caused by improper operation.

[0039] (3) In the traditional trailer driving process, the driver needs to concentrate highly, constantly observe the status of the trailer and the towing rope, and manually adjust the driving operation based on experience. However, this method greatly reduces the burden on the driver by automating this process, allowing him to focus more on monitoring the overall road conditions and emergency response, thereby improving driving comfort and efficiency.

[0040] (4) Since the method of the present application can respond quickly and accurately to changes in trailer status, it reduces stagnation or delay time caused by human misjudgment or delayed response. This is particularly important for scenarios that require efficient completion of trailer tasks (such as logistics transportation, accident rescue, etc.), and can significantly improve operational efficiency and reduce operating costs.

[0041] (5) The introduction of the automatic driving method for towed vehicles is not only an innovation in the traditional trailer driving mode, but also an important attempt to apply intelligent driving technology in specific fields. As the technology continues to mature and become more popular, it is expected to promote the intelligent and automated development of the entire transportation industry and promote industrial upgrading and transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 This is a schematic diagram of the structure of the trailer system provided for this application.

[0044] Figure 2 This is a circuit diagram of the vehicle provided for this application.

[0045] Figure 3 Schematic diagram of the flow of the automatic driving method for a towed vehicle according to a typical embodiment of the present application.

[0046] Figure 4 This is a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in one of the embodiments of the present application.

[0047] Figure 5 This is a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in another embodiment of the present application.

[0048] Figure 6 This is a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in another embodiment of the present application.

[0049] Figure 7 This is a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in another embodiment of the present application.

[0050] Figure 8 The present application also provides a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in another embodiment.

[0051] Fig. 9This is a flowchart diagram of the steps before the acquired front image data is input into a preset trailer trend recognition model in one embodiment of the present application.

[0052] Fig.10 The present application also provides a flowchart of the steps before the trend information and the tightness change information are input into a preset driving processing model to generate driving information of the vehicle in another embodiment.

[0053] Fig.11 It is a schematic diagram of the structure of the automatic driving device of the towed vehicle in the embodiment of the present application.

[0054] Fig.12 It is a schematic diagram of the structure of the computer device in the embodiment of the present application. DETAILED DESCRIPTION

[0055] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and cannot be interpreted as limiting the present application.

[0056] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0057] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0058] The present application provides an automatic driving method for a towed vehicle. Based on the method, the inclination data and speed data of the trailer and the tightness change data of the towing rope can be obtained based on the image data in front of the vehicle, and driving information can be generated. The vehicle (towed vehicle) can automatically drive based on the driving information to reduce the driving difficulty of the vehicle and improve the user experience.

[0059] In this embodiment, the automatic driving method of the towed vehicle is implemented based on the trailer system. Figure 1 and Figure 2 The trailer system includes a trailer 100, a towed vehicle 200 (hereinafter referred to as the vehicle 200) and a towing rope 300. The power system of the vehicle 200 fails and cannot run, but the brake assembly 240 and the steering assembly 250 of the vehicle 200 can operate normally. The trailer 100 is connected to the vehicle 200 through the towing rope 300, so that the trailer 100 can apply traction to the vehicle 200 through the towing rope 300, so that the vehicle 200 can run under the traction of the trailer 100.

[0060] The vehicle 200 includes a control component 210, a front camera 220, a vehicle-mounted radar 230, a brake component 240 and a steering component 250; the control component 210 includes a first controller 211 and a second controller 212, and the first controller 211 and the second controller 212 are electrically connected; the front camera 220 is arranged at the front of the vehicle 200, so that the front camera 220 can collect image data in front of the vehicle 200 (the image data is called front image data), and the front camera 220 is electrically connected to the first controller 211. The vehicle-mounted radar 230 detects external obstacle information by emitting signals to the outside, and the vehicle-mounted radar 230 is electrically connected to the second controller 212; the brake assembly 240 is used to control the braking of the vehicle 200 to reduce the speed of the vehicle 200, and the brake assembly 240 is electrically connected to the second controller 212, for example, the brake assembly 240 is a brake pedal; the steering assembly 250 is used to control the steering of the vehicle 200, and the steering assembly 250 is electrically connected to the second controller 212, for example, the steering assembly 250 is a steering wheel.

[0061] In another embodiment, the towed vehicle can also be applied to a trailer system, which includes a tractor head (equivalent to the trailer) and a trailer (equivalent to the vehicle), and the towed vehicle automatic driving method is used to achieve safe driving of the trailer system. The specific implementation of the towed vehicle automatic driving method is described below. After knowing the towed vehicle automatic driving method based on the trailer system, those skilled in the art can apply the towed vehicle automatic driving method to the trailer system without creative work. For example, the trailer system includes a tractor head and one or more trailers, such as a road train.

[0062] The automatic driving method of the towed vehicle of the present application is implemented based on the trailer system described above. In this embodiment, the automatic driving method of the towed vehicle is described by taking the control component of the vehicle as an example. Please refer to Figure 3 , the specific steps of the automatic driving method of the towed vehicle are as follows:

[0063] Step S1100, controlling the front camera to continuously collect image data in front of the vehicle:

[0064] When the trailer is towing the vehicle, the trailer will be located in front of the vehicle, and the trailer will tow the vehicle through the towing rope. When the first controller of the control component controls the front camera of the vehicle to work, the front camera will capture the image in front of the vehicle, which is called the front image data. It can be understood that the front camera will capture the image of the trailer and the towing rope, that is, the front image data includes the image data of the trailer and the towing rope.

[0065] Step S1200: input the acquired front image data into a preset trailer trend recognition model to generate the trailer trend information and the tension change information of the towing rope, wherein the trend information includes the trailer inclination data and speed data:

[0066] The first controller is preset with a trailer trend recognition model, which includes an image feature extraction network. The first controller inputs the acquired front image data into the image feature extraction network, and the image feature extraction network can extract image features such as the shape, outline, color, inclination, brake light status, shape, outline, tightness of the towing rope, etc. of the trailer from the front image data. After the image feature extraction network acquires the multiple image features, it will combine the classifier to output the trailer trend information and the towing rope tightness change information, and the trailer trend information includes the trailer inclination data and speed data.

[0067] Specifically, the image feature extraction network is followed by a plurality of classifiers. One of the plurality of classifiers is a first classifier, and the first classifier is used to classify the inclination state of the trailer, where the inclination is the inclination of the trailer relative to the central axis of the road. The following description of the present application takes the inclination of the trailer relative to the central axis of the road as an example generally between 0-90°, but this should not be construed as a limitation on the present application. The first classifier is used to classify at least 9 categories based on the output structure of the image feature extraction network, and the 9 categories equally divide the inclination range of 0-90°. For example, the first category represents the inclination range of 0-10°, the second category represents the inclination range of 10-20°, the third category represents the inclination range of 20-30°, and so on. To save space, it will not be described in detail here.

[0068] The first classifier extracts a category with the largest classification probability from the multiple categories as the classification result. For example, the classification probability of the first category of the first classifier is 0.80, and the classification probabilities of the other multiple categories are all less than 0.80, then the first category is taken as the classification result, that is, the classification result is that the inclination data of the trailer is between 0-10°.

[0069] Another classifier among the multiple classifiers is a second classifier, and the second classifier is used to classify the speed of the trailer. Generally speaking, the speed of the trailer is below 100KM / h. The following describes the present application by taking the speed of the trailer below 100KM / h as an example, but it should not be understood as a limitation on the present application. The second classifier is used to classify at least 10 categories based on the output results of the image feature extraction network. The 10 categories are evenly divided into a speed range of 0-100KM / h. For example, the first category represents a speed range of 0-10KM / h, the second category represents a speed range of 10-20KM / h, the third category represents a speed range of 20-30KM / h, and so on. In order to save space, it is not repeated here.

[0070] The second classifier extracts a category with the largest classification probability from the multiple categories as the classification result. For example, the classification probability of the second category of the second classifier is 0.85, and the classification probabilities of the other multiple categories are all less than 0.85, then the second category is taken as the classification result, which means that the speed data of the trailer is between 10-20KM / h.

[0071] In another embodiment, the image feature extraction network may be connected to a regression model, and based on the features of the trailer obtained by the image feature extraction network, a continuous value, namely the speed of the trailer, is predicted through a regression output layer.

[0072] In a typical embodiment of the present application, another classifier among the multiple classifiers is a third classifier, and the third classifier is used to classify the tension change information of the traction rope. The tension change information of the traction rope is divided into multiple states. The tension change information of the traction rope is divided into three states, and the three states are tight, semi-tight and loose as an example to describe the present application, but it should not be understood as a limitation of the present application. The third classifier is used to classify three states based on the output result of the image feature extraction network. For example, the first category is the tension change information representing tightness, the second category is the tension change information representing semi-tightness, and the third category is the tension change information representing looseness, and so on. In order to save space, it will not be repeated here.

[0073] The third classifier extracts a category with the largest classification probability from the multiple categories as the classification result. For example, the classification probability of the third category of the third classifier is 0.70, and the classification probabilities of the other two categories are both less than 0.70, then the third category is taken as the classification result, that is to say, the tightness change information of the traction rope represents the relaxation state.

[0074] In this embodiment, the image feature extraction network can be CNN, or other neural network models suitable for extracting image features, such as U-Net, Resnet, Vit, etc. The classifier is recommended to use MLP. After understanding the technical solution of the present application, those skilled in the art can also flexibly change the selection of the classifier. To save space, it will not be described here. The front image data is imported into the image feature extraction network, and the trailer inclination data, trailer speed information and towing rope tightness change information of the trailer are obtained through the first classifier, the second classifier and the third classifier.

[0075] Step S1300: input the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data:

[0076] After acquiring the trend information and the looseness and tightness change information, the first controller inputs the trend information and the looseness and tightness change information into the second controller. The second controller is preset with a driving processing model, and the driving processing model can generate driving information of the vehicle based on the trend information and the looseness and tightness change information, and the driving information includes steering data and braking data of the vehicle.

[0077] It can be understood that this vehicle has lost power and needs to be towed by a trailer to move. When the speed of the trailer decreases, the power will be transmitted to the vehicle after a period of time (the interval is called the first interval). In other words, the speed of this vehicle remains unchanged, and the speed of this vehicle may be greater than the speed of the trailer, which reduces the distance between this vehicle and the trailer. When the distance between this vehicle and the trailer is reduced to zero, this vehicle will collide with the trailer, causing damage to both the vehicle and the trailer.

[0078] In this embodiment, the driving processing model can generate the braking data of the vehicle based on the trailer speed data, the inclination data and the tension change information of the towing rope, so as to control the braking assembly of the vehicle to brake, reduce the speed of the vehicle, and prevent the vehicle from hitting the trailer. For example, when the trailer speed data indicates that the speed of the trailer is decreasing, and the tension change information of the towing rope indicates that the tension of the towing rope is 40%-50%, the driving processing model determines that the speed of the trailer is decreasing, and the speed of the vehicle is greater than the speed of the trailer, and the driving processing model generates the braking data to control the braking of the vehicle and reduce the speed of the vehicle.

[0079] In one embodiment, the driving processing model includes a braking calculation model, and the braking calculation model is Braking=k1·θ+k2·v+k3·T, wherein Braking represents the braking data, θ represents the inclination data of the trailer, v represents the speed data of the trailer, T represents the looseness change information, k1, K2 and K3 are all coefficients, and the coefficients represent the degree of influence of various factors on braking, and when the coefficients such as k1, K2 and K3 are multiplied by the relative parameters, the units of the corresponding parameters will be eliminated to convert the braking data into data of the same dimension. Because it is necessary to process the output to obtain fixed k1, K2 and K3, the model is required to output θ, v and T of the same dimension according to each scene, for example, all are converted and output as θ, v and T of the torque dimension. In this embodiment, the second controller digitizes the obtained trailer speed data, inclination data and the looseness change information to obtain specific values, so as to obtain the braking data through the braking calculation model.

[0080] In this embodiment, the coefficients k1, K2 and K3 are obtained by calculating through a machine learning model, for example, through a deep learning model or simulation calculation. For specific contents of obtaining the coefficients through the machine learning model, please see below.

[0081] In a further embodiment, the driving processing model is provided with a plurality of braking calculation models, each of which has a different weight. The weight is directly multiplied with the braking data to obtain new braking data, so that the driving processing model can adapt to different driving states. For example, the driving state is divided into the first gear, the second gear, and so on and so forth to the Nth gear. The weight corresponding to each gear is different, thereby affecting the output result of the final braking calculation model. The weight of the first gear is less than the weight of the second gear, the weight of the second gear is less than the weight of the third gear, and so on and so forth. The weight of the N-1th gear is less than the weight of the Nth gear. The user can freely set the driving state according to the external environment, so that under the control of the user, the vehicle can be more adapted to the driving of the external environment, so that the vehicle can follow the trailer well.

[0082] In a typical embodiment of the present application, it can be understood that the vehicle is moved by being towed by the trailer. When the trailer turns, the steering torque of the trailer will be transmitted to the vehicle only after a period of time (the interval is referred to as the second interval) because the trailer pulls the vehicle through the towing rope. During the second interval, the vehicle may run off the road and cause an accident, such as falling into a ditch. Even when the road is in a mountainous area, the vehicle may run off a cliff, dragging the trailer off the cliff, causing a serious traffic accident.

[0083] In this embodiment, the driving processing model can generate steering data of the vehicle based on the inclination data, speed data and looseness change information of the trailer, so as to facilitate the subsequent steering control of the vehicle, so that the vehicle can follow the steering of the trailer well, and improve the safety of the vehicle during the towing process. For example, when the inclination data of the trailer is between 0-10°, steering data representing a 10° steering is generated.

[0084] In one embodiment, the driving processing model includes a steering calculation model, and the steering calculation model is Steering=k4·θ+k5·v+k6·T, where Steering represents the steering data. In this embodiment, the second controller digitizes the obtained trailer speed data, inclination data and the looseness and tightness change information to obtain specific values, so as to obtain the steering data through the steering calculation model. When coefficients such as k4, K5 and K6 are multiplied by relative parameters, the units of the corresponding parameters will be eliminated to convert the steering data into data of the same dimension. Because it is necessary to process the output to obtain fixed k4, K5 and K6, the model is required to output θ, v and T of the same dimension according to each scenario, for example, all are converted and output as θ, v and T of the torque dimension. In this embodiment, the coefficients such as k4, K5 and K6 will be obtained by calculation through a machine learning model, such as through a deep learning model or simulation calculation. For the specific content of obtaining the coefficients through the machine learning model, please refer to the following.

[0085] In a further embodiment, the driving processing model is also provided with a plurality of steering calculation models, each of which has a different weight, and the weight is directly multiplied with the steering data to obtain new steering data, so that the driving processing model can adapt to different driving states, and the driving state can be divided into different gears. The specific gear and weight settings can refer to the gear and weight settings above, and will not be repeated here to save space. The user can freely set the driving state according to the external environment, so that under the control of the user, the vehicle can be more adapted to the driving of the external environment, so that the vehicle can follow the trailer well.

[0086] Step S1400, controlling the steering and speed of the vehicle based on the driving information.

[0087] After the second controller generates the braking data and steering data of the vehicle based on the driving processing model, the second controller controls the braking component of the vehicle to brake based on the braking data to control the speed of the vehicle so that the vehicle and the trailer are kept at a safe distance to avoid a collision between the vehicle and the trailer; the second controller controls the steering component of the vehicle to steer based on the steering data to control the steering of the vehicle so that the vehicle can synchronously follow the steering of the trailer to avoid the vehicle running off the road and causing a traffic accident.

[0088] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Figure 4 , before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0089] Step S1330, obtaining the inclination data of the vehicle;

[0090] The vehicle is provided with an inclination sensor, through which the inclination data of the vehicle can be detected and obtained. The inclination sensor outputs the detected inclination data to the second controller. In this embodiment, it is recommended that the inclination sensor is an accelerometer, but this should not be understood as a limitation to the present application.

[0091] Step S1340, determining the difference between the inclination data of the vehicle and the inclination data of the trailer, and inputting the difference, the trend information and the tightness change information into a preset driving processing model to generate the driving information:

[0092] Because the trailer is connected to the vehicle through a towing rope, the trailer and the vehicle are prone to inclination difference during driving, that is, the inclination data of the trailer is different from the inclination data of the vehicle, especially when the road is curved, the difference between the inclination data of the trailer and the inclination data of the vehicle (the difference is called the inclination difference) is rapidly enlarged. When the inclination data of the vehicle is different from the inclination data of the trailer, the second controller inputs the inclination data of the vehicle and the inclination data of the trailer into the driving processing model to generate the driving data of the vehicle, so as to control the steering and braking of the vehicle, so that the vehicle can follow the trailer well.

[0093] In this embodiment, the size of the inclination angle difference is proportional to the steering data. It can be understood that when the inclination angle difference is larger, the driving processing model will increase the steering data and the braking data accordingly, so that the vehicle can turn and brake quickly to avoid the vehicle running off the road; when the inclination angle difference is smaller, the driving processing model will reduce the steering data and the braking data accordingly, so that the vehicle can turn slowly and follow the trailer well.

[0094] In one embodiment, in the driving processing model, the inclination data of the vehicle is digitized and added as a dimension to the braking calculation model and the steering calculation model of the driving processing model to improve the accuracy of the driving information output by the driving processing model, so that the vehicle can further accurately follow the trailer and the trailer can tow the vehicle well.

[0095] Among them, after adding the inclination data of the vehicle as a dimension, the steering calculation model is corrected to: Steering = k7·Δθ+k8·v+k9·T+k 10 ·θ 2 car +k 11 ·sin(θ car ), where θ car Represents the inclination angle of the vehicle, Δθ=θ-θ car , Δθ represents the difference between the trailer's inclination angle and the vehicle's inclination angle. Δθ will affect the stability of the steering. The larger the difference, the greater the steering angle may be required to maintain balance. 2 car is the square of the vehicle's inclination angle, indicating the nonlinear effect of the vehicle's inclination angle on steering. The greater the vehicle's inclination angle, the more significant the effect on steering. car ) is the sine value of the vehicle's inclination angle, to further consider the influence of the vehicle's inclination angle on steering, which may be more obvious at a large inclination angle, so as to generate more accurate steering data. 11 When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the steering data into data of the same dimension.11 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0096] Correspondingly, after adding the inclination angle data of the vehicle as a dimension, the braking calculation model is modified to: Braking = k 12 ·v 2 +k 13 ·Δθ+k 14 ·T+k 15 ·v·sin(θ car ), where v 2 is the square of the trailer speed. The greater the speed, the greater the braking force required to ensure a safe distance between the vehicle and the trailer. car ) is the product of the speed and the sine value of the vehicle's inclination angle, which indicates the complexity of braking when the vehicle is in an inclined state, so as to improve the accuracy of the braking data. K 12 To K 15 When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the brake data into data of the same dimension. 12 To K 15 So the model is required to output v of the same dimension according to each scenario. 2 ,Δθ,v,T,v·sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0097] In this embodiment, k7 to K 15 The coefficients will be obtained through machine learning model calculation, such as through a deep learning model or simulation calculation. For specific content about obtaining the coefficients through the machine learning model, please see below.

[0098] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Figure 5 , before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0099] Step S1350, obtaining the turn signal information of the trailer from the trend information:

[0100] The multiple classifiers also include a fourth classifier, which is used to classify the state of the turn signal of the trailer. The first classifier is used to classify into three categories based on the output result of the image feature extraction network, the first category representing that the turn signal of the trailer indicates no turn, the second category representing that the turn signal of the trailer indicates turning left, and the third category representing that the turn signal of the trailer indicates turning right.

[0101] The fourth classifier extracts a category with the largest classification probability from the multiple categories as the classification result. For example, the classification probability of the first category of the fourth classifier is 0.75, and the classification probabilities of the other two categories are both less than 0.75, then the first category is taken as the classification result, that is, the turn signal information represents no turn.

[0102] After the second controller obtains the trend information output by the trend recognition model, the turn signal information of the trailer can be obtained from the trend information.

[0103] Step S1360: input the turn signal information, the trend information and the tightness change information into a preset driving processing model to generate the driving information:

[0104] After the second controller obtains the turn signal information, the turn signal information and the trailer's tilt angle data are input into the driving processing model. The driving processing model modifies the vehicle's steering data based on the turn signal information, narrowing or widening the steering angle, that is, making the vehicle's steering data more accurate, so that the vehicle can turn synchronously with the trailer, so that the vehicle can follow the trailer well. At the same time, the vehicle's brake data is also modified based on the turn signal information, so that the vehicle can brake in time.

[0105] Specifically, when the turn signal information indicates a left turn or a right turn, the driving processing model may correspondingly increase the steering angle of the vehicle to turn left or right, so that the vehicle can turn in advance, thereby allowing the vehicle to follow the trailer well.

[0106] For example, the inclination difference between the vehicle and the trailer is zero. It is understandable that the vehicle and the trailer are traveling in the same direction. At this time, the trailer does not generate inclination data, and the steering data of the vehicle is zero. When the trailer turns on the left turn signal, the vehicle obtains the turn signal information representing the left turn. The driving processing model can generate the steering data representing the left turn based on the turn signal information representing the left turn, control the vehicle to turn left in advance, and expand the left steering angle so that the vehicle can follow the trailer well.

[0107] In one embodiment, in the driving processing model, the turn signal information of the vehicle is digitized and added as a dimension to the braking calculation model and the steering calculation model of the driving processing model to improve the accuracy of the driving information output by the driving processing model, so that the vehicle can further accurately follow the trailer and the trailer can tow the vehicle well.

[0108] Among them, after adding the turn signal information of the vehicle as a dimension, the steering calculation model is corrected to: Steering = Turn Signal · Turn Direction | (k 16 ·Δθ+k 17 ·v+k 18 ·T+k 19 ·θ 2 car +k 20 ·sin(θ car ))|, where Turn Signal·Turn Direction is the turn signal information (left turn / right turn / no turn), which represents the influence of the trailer turn signal on the steering of the vehicle. For the convenience of calculation, when the turn signal information represents a right turn, it is +1, when the turn signal information represents a left turn, it is -1, and when the turn signal information represents no turn, it is also +1. Thus, ±1 indicates the steering of the vehicle, so that the vehicle can well follow the trailer to turn. K 14 To K 20 When the coefficients are multiplied by the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the steering data into data of the same dimension. 14 To K 20 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0109] Correspondingly, after adding the inclination angle data of the vehicle as a dimension, the braking calculation model is modified to: Braking = k 21 ·v 2 +k 22 ·Δθ+k 23 ·T+k 24 ·v·sin(θ car )+k 25 ·|Turn Signal·TurnDirection|, thereby adding the turn signal information to correct the vehicle's braking data so that the vehicle can brake in time and maintain a safe distance from the trailer. K 21 To K 25When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the brake data into data of the same dimension. 21 To K 25 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0110] In this embodiment, k 16 To K 25 The coefficients will be obtained through machine learning model calculation, such as through a deep learning model or simulation calculation. For specific content about obtaining the coefficients through the machine learning model, please see below.

[0111] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Figure 6 , before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0112] Step S1370, obtaining the trailer's brake light information from the trend information:

[0113] The multiple classifiers include a fifth classifier, which is used to classify the state of the brake lights of the trailer. The fifth classifier is used to classify into two categories based on the output results of the image feature extraction network, the first category represents that the brake lights of the trailer are in the on state, and the second category represents that the brake lights of the trailer are in the off state.

[0114] The fifth classifier extracts a category with the largest classification probability from the multiple categories as the classification result. For example, the classification probability of the second category of the fifth classifier is 0.90, and the classification probability of the first category is less than 0.90, then the second category is taken as the classification result, which means that the brake light information represents the off state.

[0115] After the second controller obtains the trend information output by the trend recognition model, the brake light information of the trailer can be obtained from the trend information.

[0116] Step S1380, inputting the brake light information, the trend information and the tightness change information into a preset driving processing model to generate the brake data:

[0117] The second controller inputs the brake light information into the driving processing model of the vehicle, and the driving processing model corrects the braking data of the vehicle based on the brake light information, so that the second controller can control the braking of the vehicle in advance, control the speed of the vehicle, avoid the distance between the vehicle and the trailer being too close, and ensure that the vehicle and the trailer are always within a safe distance.

[0118] In one embodiment, in the driving processing model, the brake light information of the vehicle is digitized and added as a dimension to the brake calculation model of the driving processing model to improve the accuracy of the driving information output by the driving processing model, so that the vehicle can further accurately follow the trailer and the trailer can tow the vehicle well.

[0119] After adding the brake light information of the vehicle as a dimension, the braking calculation model is modified to: Braking = k 26 ·v 2 +k 27 ·Δθ+k 28 ·T+k 29 ·v·sin(θ car )+k 30 ·|Turn Signal·TurnDirection|+k 31 ·BrakeLight, where BrakeLight represents the brake light information. For the convenience of calculation, when the brake light is on red, the value is 1, and when the brake light is off, the value is 0. The brake light can be used to correct the brake data of the vehicle so that the vehicle can brake in time and keep a safe distance between the vehicle and the trailer. K 26 To K 31 When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the brake data into data of the same dimension. 26 To K 31 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0120] In this embodiment, k 25 To K 31 The coefficients will be obtained through machine learning model calculation, such as through a deep learning model or simulation calculation. For specific content about obtaining the coefficients through the machine learning model, please see below.

[0121] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Figure 7, before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0122] Step S1510, control the vehicle-mounted radar to send a signal to the outside, and determine the distance data between the vehicle and the obstacle according to the model of the corresponding reflection of the external obstacle:

[0123] The vehicle is provided with an on-board radar, which acquires obstacle information around the vehicle by transmitting signals to the outside. In this embodiment, the second controller is electrically connected to the on-board radar, and the on-board radar outputs the detected obstacle information to the second controller. Because the trailer is always located in front of the vehicle, the on-board radar sends a signal to the outside to receive the obstacle signal, and the obstacle signal includes the obstacle information. The obstacle information includes the trailer information. After receiving the obstacle information, the second controller will remove the trailer information from the obstacle information, so that the obstacle information can well assist the vehicle in generating the driving information.

[0124] The second controller can calculate the distance between the vehicle and the obstacle based on the time between the vehicle-mounted radar sending the signal and receiving the obstacle signal, and obtain the distance data.

[0125] Step S1520: input the distance data, the trend information and the tightness change information into a preset driving processing model to generate the driving information.

[0126] The second controller inputs the distance data into the driving processing model to correct the driving information of the vehicle, so that the second controller controls the braking and steering of the vehicle in advance to avoid the distance between the vehicle and the trailer being too close, so that the vehicle and the trailer are always within a safe distance.

[0127] In one embodiment, in the driving processing model, the distance data of the vehicle is added as a dimension to the braking calculation model and the steering calculation model of the driving processing model to improve the accuracy of the driving information output by the driving processing model, so that the vehicle can further accurately follow the trailer and the trailer can tow the vehicle well.

[0128] Among them, after adding the turn signal information of the vehicle as a dimension, the steering calculation model is corrected to: Steering = Turn Signal · Turn Direction | (k 32 ·Δθ+k 33 ·v+k 34 ·T+k 35 ·θ 2 car +k 36·sin(θ car )+k 37 Obstacle Distance)|, where Obstacle Distance is the distance data. The steering calculation model is corrected by the distance data so that the vehicle can well follow the steering of the trailer. 32 To K 37 When the coefficients are multiplied by the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the steering data into data of the same dimension. 32 To K 37 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0129] Correspondingly, after adding the brake light information of the vehicle as a dimension, the braking calculation model is modified to: Braking = k 38 ·v 2 +k 39 ·Δθ+k 40 ·T+k 41 ·v·sin(θ car )+k 42 ·|Turn Signal·TurnDirection|+k 43 ·Brake Lightk 44 Obstacle Distance: Use the distance data to correct the vehicle’s braking data so that the vehicle can brake in time and maintain a safe distance between the vehicle and the trailer. 38 To K 44 When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the brake data into data of the same dimension. 38 To K 44 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0130] In this embodiment, k 32 To K 44 The coefficients will be obtained through machine learning model calculation, such as through a deep learning model or simulation calculation. For specific content about obtaining the coefficients through the machine learning model, please see below.

[0131] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Figure 8 , before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0132] Step S1530, obtaining the speed data of the vehicle;

[0133] The second controller may obtain the speed data of the vehicle based on the speed sensor of the vehicle.

[0134] Step S1540, inputting the speed data of the vehicle, the trend information and the tightness change information into a preset driving processing model to generate the driving information;

[0135] The second controller inputs the speed data of the vehicle into the driving processing model to correct the driving information of the vehicle, so that the second controller controls the braking and steering of the vehicle in advance to avoid the distance between the vehicle and the trailer being too close, so that the vehicle and the trailer are always within a safe distance.

[0136] In one embodiment, in the driving processing model, the speed data of the vehicle is added as a dimension to the braking calculation model and the steering calculation model of the driving processing model to improve the accuracy of the driving information output by the driving processing model, so that the vehicle can further accurately follow the trailer and the trailer can tow the vehicle well.

[0137] Among them, after adding the speed data of the vehicle as a dimension, the steering calculation model is corrected to:

[0138] Where Δv=vv car ,v car Characterizes the speed data of the vehicle. When |Δv| is larger, it indicates that the speed difference between the vehicle and the trailer is larger, and a stronger steering adjustment is needed to make the corresponding steering of the vehicle follow the trailer well. α is a constant, and α·v is the influence of the trailer speed on the overall formula, which is used to adjust the overall influence of the trailer speed on the steering. The higher the trailer speed, the greater the steering adjustment of the vehicle should be. K 45 To K 51 When the coefficients are multiplied by the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the steering data into data of the same dimension. 45 To K 51 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θcar ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0139] Correspondingly, after adding the vehicle's speed data as a dimension, the braking calculation model is modified to: Braking = k 52 ·v 2 +k 53 ·Δθ+k 54 ·T+k 55 ·v·sin(θ car )+k 56 ·|Turn Signal·TurnDirection|+k 57 ·Brake Light 57 ·Obstacle Distance+k 58 ·v car , the vehicle's braking data is corrected by the vehicle's speed data, so that the vehicle can brake in time and keep a safe distance between the vehicle and the trailer. K 52 To K 58 When the coefficients are multiplied with the corresponding parameters, the units of the corresponding parameters will be eliminated to convert the brake data into data of the same dimension. 52 To K 58 So the model is required to output Δθ, v, T, θ of the same dimension according to each scene. 2 car， sin(θ car ), for example, they are all converted to output in the torque dimension of θ, v and T.

[0140] In this embodiment, k 45 To K 58 The coefficients will be obtained through machine learning model calculation, such as through a deep learning model or simulation calculation. For specific content about obtaining the coefficients through the machine learning model, please see below.

[0141] Based on any embodiment of the towed vehicle automatic driving method of the present application, the coefficients K1 to K 58 The coefficients can be obtained based on the machine learning model. In this embodiment, it is recommended to use a deep learning algorithm to obtain these coefficients, such as using MLP to obtain these coefficients, wherein all data except the coefficients on the right side of the equation in the steering calculation model are used as the input data in the input layer of the MLP, and the Steering on the left side of the equation in the steering calculation model is used as the output result of the output layer, and the various coefficients of the steering calculation model are used as the various weight parameters that need to be learned in the hidden layer. The MLP under the architecture is used as a training set, and the MLP is trained to a convergence state, and then the values ​​of the various coefficients in the hidden layer are obtained.

[0142] Specifically, in the experimental scenario, test data other than the coefficients required by the steering calculation model can be collected, and the test data includes the steering data of the vehicle, the speed data of the vehicle, the speed data of the trailer, the tightness change information, the inclination data of the vehicle, the inclination data of the trailer, the turn signal information of the trailer, the brake light information of the trailer, and the distance data of the obstacle. The obtained sets of test data are imported into the MLP to train and obtain the various coefficients required by the steering calculation model.

[0143] In this embodiment, the method for obtaining each coefficient in the braking calculation model is the same as the method for obtaining the coefficient of the steering calculation model. The method for obtaining the coefficient of the steering calculation model can be referred to to obtain each coefficient of the braking calculation model. To save space, it will not be repeated here.

[0144] Furthermore, in the experimental scenario, test data other than the coefficients required by the brake calculation model can be collected, and the test data includes the brake data of the vehicle, the speed data of the vehicle, the speed data of the trailer, the tightness change information, the inclination data of the vehicle, the inclination data of the trailer, the turn signal information of the trailer, the brake light information of the trailer, and the distance data of the obstacle. The obtained sets of test data are imported into the MLP to train and obtain the various coefficients required by the brake calculation model.

[0145] In another embodiment, the coefficients in the steering calculation model and the coefficients in the braking calculation model may be obtained through simulation or experimental testing.

[0146] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Fig. 9 Before inputting the acquired front image data into a preset trailer trend recognition model, the method further includes the following specific steps to pre-train the trailer trend recognition model.

[0147] Step S1210, obtaining a preset training set, wherein the training set includes a plurality of training samples and supervisory labels thereof, wherein the training samples are a plurality of the front image data collected in the trailer scene, and the supervisory labels represent the trailer trend information and the tension change information of the towing rope corresponding to the training samples:

[0148] The front image data collected in the trailer scene is used as a training set, and each front image is used as a single training sample. The supervision label of each training sample is manually annotated according to the training sample. The supervision label is the speed data, inclination data, brake light information, turn signal information and the tightness change information of the towing rope of the trailer.

[0149] Step S1220, calling the training set to train the trailer trend recognition model to a convergence state, so that it can learn the ability to recognize the trailer trend information and the tension change information of the towing rope according to the input front image data:

[0150] A single training sample and its supervision label in the training set are called, and the trailer trend recognition model takes the training sample as input. The image features such as speed data, inclination data, brake light information, turn signal information, and tightness change information of the trailer in the training sample are extracted through the image feature extraction network in the trailer trend recognition model, and the image features are given to the multiple classifiers to obtain the classification results of each of the multiple classifiers as the prediction results.

[0151] The cross entropy loss function is used to calculate the loss value corresponding to the cross entropy loss of the prediction results of the multiple classifiers based on the supervisory labels of the training samples. When the loss values ​​are all lower than or equal to the corresponding preset thresholds, it indicates that the trailer trend recognition model has been trained to a convergence state, so that the training of the trailer trend recognition model can be terminated; when the loss value exceeds the preset threshold, it indicates that the trailer trend recognition model has not converged, so the trailer trend recognition model is gradient updated according to the loss value, usually back-propagating to correct the weight parameters of each link in the trailer trend recognition model so that the trailer trend recognition model is closer to convergence, and then continue to call other training samples in the training set and their supervisory labels to perform iterative training on the trailer trend recognition model until the trailer trend recognition model is trained to a convergence state. The various preset thresholds can be set as needed by those skilled in the art according to the relevant disclosures herein.

[0152] Based on any embodiment of the towed vehicle automatic driving method of the present application, please refer to Fig.10 , before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, the following specific steps are also included:

[0153] Step S1550, controlling the vehicle-mounted radar to send signals to the outside to obtain external obstacle information;

[0154] The content of step S1550 can refer to the content of step S1510, and will not be repeated here to save space.

[0155] Step S1560, call out the pre-stored map information to obtain the current road information:

[0156] The vehicle can communicate data with external terminal devices through the vehicle-mounted communication device to obtain electronic map information. The vehicle can determine the specific location information of the vehicle through positioning devices, such as GPS, Beidou satellite positioning system, etc., and mark the location information in the electronic map information, so as to assist the driving of the vehicle through the electronic map information. In this embodiment, the electronic map information can be pre-stored in the second controller.

[0157] Step S1570, based on the obstacle information and the current road information, correct the steering data and braking data of the vehicle:

[0158] The second controller imports the obstacle information and the current road information into the driving processing model to correct the steering data and braking data of the vehicle so that the vehicle can follow the trailer well.

[0159] Wherein, based on the current road information, the second controller can obtain the extended path of the road in advance, and generate turning information or non-turning information in advance to correct the steering data of the vehicle so that the vehicle can follow the trailer well.

[0160] Furthermore, the obstacle information is used to obtain in advance whether there are obstacles in front of the road, on the left, and on the right, and the steering data and braking data of the vehicle are corrected. For example, when the obstacle information indicates that there is an obstacle in front of the vehicle, the braking data of the vehicle is increased in advance to control the braking depth of the vehicle to avoid the vehicle from hitting the obstacle in front. Alternatively, when the obstacle information indicates that there is an obstacle on the left or right side of the vehicle, the steering data is corrected to increase or decrease the turning angle of the vehicle in advance to avoid the vehicle from hitting the obstacle on the left or right side.

[0161] Another embodiment of the present application also provides an automatic driving device for a towed vehicle. Fig.11 , which includes an image acquisition module 5100, a trend recognition module 5200, a driving processing module 5300, and a driving control module 5400; wherein the image acquisition module 5100 controls the front camera to continuously acquire image data in front of the vehicle; the trend recognition module 5200 inputs the acquired front image data into a preset trailer trend recognition model to generate trend information of the trailer and tightness change information of the towing rope, wherein the trend information includes inclination data and speed data of the trailer; the driving processing module 5300 inputs the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data; the driving control module 5400 controls the steering and speed of the vehicle based on the driving information.

[0162] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: an inclination calculation unit, which obtains the inclination data of the vehicle; a first driving calculation unit, which determines the difference between the inclination data of the vehicle and the inclination data of the trailer, and inputs the difference, the trend information and the tightness change information into a preset driving processing model to generate the driving information.

[0163] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: a turn signal recognition unit, which obtains the turn signal information of the trailer from the trend information; and a second driving calculation unit, which inputs the turn signal information, the trend information and the tension change information into a preset driving processing model to generate the driving information.

[0164] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: a brake light recognition unit, which obtains the brake light information of the trailer from the trend information; and inputs the brake light information, the trend information and the tightness change information into a preset driving processing model to generate the brake data.

[0165] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: a distance acquisition unit, which controls the vehicle-mounted radar to send signals to the outside, and determines the distance data between the vehicle and the obstacle according to the corresponding reflected signal of the external obstacle; the distance data, the trend information and the tightness change information are input into a preset driving processing model to generate the driving information.

[0166] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: a vehicle speed acquisition unit, which acquires the speed data of the vehicle; and a fourth driving calculation unit, which inputs the speed data of the vehicle, the trend information and the tightness change information into a preset driving processing model to generate the driving information.

[0167] Based on any embodiment of the device of the present application, the automatic driving device of the towed vehicle also includes: an obstacle acquisition unit, which controls the vehicle-mounted radar to send signals to the outside to obtain external obstacle information; a road acquisition unit, which calls out pre-stored map information to obtain current road information; and a driving correction unit, which corrects the steering data and braking data of the vehicle based on the obstacle information and the current road information.

[0168] Based on any embodiment of the device of the present application, the automatic driving device for the towed vehicle also includes: a training unit, which obtains a preset training set, wherein the training set includes multiple training samples and their supervisory labels, wherein the training samples are several of the front image data collected in the trailer scene, and the supervisory labels represent the trend information of the trailer and the tightness change information of the towing rope corresponding to the training samples; a convergence unit, which calls the training set to train the trailer trend recognition model to a convergence state, so that it learns the ability to recognize the trend information of the trailer and the tightness change information of the towing rope based on the input front image data.

[0169] Based on any embodiment of this application, please refer to Fig.12 Another embodiment of the present application further provides a computer device, which can be used as a controller of the vehicle, such as Fig.12 As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and a computer program encapsulating computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor may implement a method for automatic driving of a towed vehicle. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor may execute the method for automatic driving of a towed vehicle of the present application. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art will understand that Fig.12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0170] In this embodiment, the processor is used to execute Fig.11 The memory stores the program code and various data required to execute the above modules or submodules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the automatic driving device of the towed vehicle of this application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0171] The present application also provides a non-volatile computer-readable storage medium storing computer-readable instructions, which, when executed by one or more processors, enables the one or more processors to execute the steps of the automatic driving method for a towed vehicle described in any embodiment of the present application.

[0172] The present application also provides a computer program product, including a computer program / instruction, which, when executed by one or more processors, implements the steps of the automatic driving method for a towed vehicle described in any embodiment of the present application.

[0173] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0174] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

[0175] To summarize, the automatic driving method for the towed vehicle of the present application collects image data in front, identifies the tightness of the towing rope and the speed and inclination data of the trailer, calculates and obtains the braking and steering data of the vehicle to control the timely braking and steering of the vehicle, so that the vehicle and the trailer are always at a safe distance, so that the vehicle can follow the trailer well and no traffic accidents will occur.

Claims

1. A method for automatically driving a towed vehicle, wherein the tow vehicle is towed by a towing rope, characterized in that: The steps include: Control the front camera to continuously collect image data in front of the vehicle; Inputting the acquired front image data into a preset trailer trend recognition model to generate the trailer trend information and the tension change information of the towing rope, wherein the trend information includes the trailer inclination data and speed data; Inputting the trend information and the looseness / tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data; Based on the driving information, the steering and speed of the vehicle are controlled.

2. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle: Get the inclination data of the vehicle; The difference between the inclination data of the vehicle and the inclination data of the trailer is determined, and the difference, the trend information and the tightness change information are input into a preset driving processing model to generate the driving information.

3. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle: acquiring the turn signal information of the trailer from the trend information; The turn signal information, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

4. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle: obtaining brake light information of the trailer from the trend information; The brake light information, the trend information and the tightness change information are input into a preset driving processing model to generate the brake data.

5. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle: Control the vehicle-mounted radar to send signals to the outside, and determine the distance data between the vehicle and the obstacle based on the corresponding reflected signals from the external obstacles; The distance data, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

6. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before the step of inputting the trend information and the tightness change information into a preset driving processing model to generate driving information of the vehicle: Get the speed data of the vehicle; The speed data of the vehicle, the trend information and the tightness change information are inputted into a preset driving processing model to generate the driving information.

7. The automatic driving method for a towed vehicle according to claim 1, characterized in that: Before inputting the acquired front image data into a preset trailer trend recognition model, the following steps are included: Obtaining a preset training set, the training set comprising a plurality of training samples and supervisory labels thereof, the training samples being a plurality of the front image data collected in the trailer scene, the supervisory labels representing the trailer trend information and the tension change information of the towing rope corresponding to the training samples; The training set is called to train the trailer trend recognition model to a convergence state, so that it can learn the ability to recognize the trailer trend information and the tension change information of the towing rope according to the input front image data.

8. An automatic driving device for a towed vehicle, wherein the towed vehicle is towed by a towing rope, characterized in that: include: Image acquisition module, controlling the front camera to continuously collect image data in front of the vehicle; A trend recognition module, inputting the acquired front image data into a preset trailer trend recognition model to generate trend information of the trailer and tightness change information of the towing rope, wherein the trend information includes inclination data and speed data of the trailer; A driving processing module, inputting the trend information and the looseness / tightness change information into a preset driving processing model to generate driving information of the vehicle, wherein the driving information includes steering data and braking data; The driving control module controls the steering and speed of the vehicle based on the driving information.

9. A computer device comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images