Multi-sensor fusion cooperative automatic driving decision-making method for complex road conditions

Abstract

The invention provides a complex road condition-oriented multi-sensor fusion cooperative automatic driving decision-making method, and relates to the technical field of image processing, and the multi-sensor fusion cooperative automatic driving decision-making method specifically comprises the steps of collecting a road scene image through a camera, recognizing a front region target, and judging that a temporary construction region is entered; dividing a passable area based on the road scene image; constructing a virtual lane line based on the passable area and the detected vehicle position information; and constructing an automatic driving decision according to the virtual lane line in combination with the passable region and the region target, and through the above steps, collecting the environmental information around the vehicle in real time, and identifying the key information of the unstructured road by using the image visual processing algorithm, thereby improving the accuracy and robustness of environmental perception.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and specifically to a multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions. It is particularly suitable for autonomous driving systems in unstructured road environments and aims to solve the perception and decision-making problems of existing autonomous driving technologies in complex road conditions. Background Art

[0002] With the support of high-precision maps, current autonomous driving systems are already able to achieve many complex driving tasks, such as automatic lane change, path navigation, and traffic signal recognition. These technologies are particularly mature for structured roads, such as urban main roads and highways. In these scenarios, high-precision maps can provide detailed environmental information, including the location of lane lines, road signs, traffic rules, and the precise location of surrounding buildings, thereby providing reliable prior information for the environmental perception and path planning of autonomous vehicles. With the help of high-precision maps, the autonomous driving system can accurately determine the location of the vehicle and the obstacle information around it, and can even perceive future road conditions in advance, so as to make efficient and safe driving decisions. Therefore, with the dual support of high-precision maps and structured roads, the application of autonomous driving technology has been promoted worldwide, especially in some limited areas where commercialization has achieved phased results, such as unmanned freight vehicles on highways and unmanned shuttle buses in closed parks.

[0003] However, despite the excellent performance of the above technologies under specific conditions, the current scope of application of autonomous driving technology is still greatly limited. The core problem is that the existing technology is highly dependent on high-precision maps and structured roads, and the construction and maintenance costs of these conditions are extremely high. First of all, the generation process of high-precision maps is very complicated, requiring the collection of massive environmental data through professional equipment and accurate modeling with the help of efficient algorithms. This process not only requires a lot of manpower and material investment, but also requires continuous updates to ensure the timeliness of the data. For example, the construction of urban roads, the opening of new roads, and the adjustment of traffic signs will affect the accuracy of map data, and these dynamic changes are often difficult to reflect in the map in a timely manner. In addition, the construction and maintenance of structured roads also require long-term investment in transportation infrastructure, such as clear road markings, a complete traffic signal system, and standardized road design. However, the road environment in reality is complex and diverse, especially in some rural, suburban or underdeveloped areas, where the roads are not fully structured, and there may even be roads without markings, narrow roads or roads with complex obstacles. The lack of high-precision map support in these unstructured road environments makes it difficult for existing autonomous driving systems to adapt.

[0004] At the same time, the complexity of unstructured roads is also reflected in the dynamic nature of traffic changes. On the one hand, urban and rural traffic construction is changing every day. Factors such as temporary road closures, adjustments to construction areas, and newly added forks make the update and maintenance of high-precision maps face huge challenges. On the other hand, unstructured roads such as rural roads, dirt roads, and gravel roads often do not have clear lane markings and standardized traffic signs, which makes it difficult for the autonomous driving system to obtain sufficient environmental perception information to complete safe driving. In addition, temporary obstacles place higher requirements on the real-time perception and decision-making of the autonomous driving system. In this case, the defects and summary of the existing technology are: on the one hand, the autonomous driving system lacking high-precision map support is prone to perception blind spots; on the other hand, in unstructured roads and complex dynamic road conditions, the decision reliability of the autonomous driving system is difficult to guarantee. These problems seriously limit the implementation of autonomous driving technology in a wider range of scenarios.

[0005] In response to the above problems, the present invention proposes a multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions. The present invention collects environmental information around the vehicle in real time and uses image vision processing algorithms to identify key information of unstructured roads, thereby improving the accuracy and robustness of environmental perception. Based on this environmental model, a decision-making algorithm is used to generate a driving strategy suitable for complex road conditions to ensure safe driving of the vehicle in an unstructured road environment. Summary of the invention

[0006] The present invention provides a multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions, which specifically includes the following steps: S1: Collect road scene images through cameras and identify construction markers to determine whether to enter the temporary construction area; S2: Divide the drivable area based on the road scene image; S3: Construct virtual lane lines based on the passable area and extracted vehicle driving trajectory feature points; S4: Build autonomous driving decisions based on virtual lane lines and infrared light images.

[0007] The present invention provides a multi-sensor fusion collaborative automatic driving decision-making system for new energy vehicles facing complex road conditions, the system comprising: Image acquisition equipment: collects road scene images through cameras and identifies targets in the front area to determine whether to enter the temporary construction area; Semantic segmentation module: the semantic segmentation module divides the drivable area based on the road scene image; Virtual lane line generation module: The virtual lane line generation module constructs virtual lane lines based on the passable area and the detected vehicle position information; Autonomous driving decision module: The autonomous driving decision module builds autonomous driving decisions based on virtual lane lines combined with drivable areas and regional targets.

[0008] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-sensor fusion collaborative autonomous driving decision method for complex road conditions is implemented.

[0009] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions.

[0010] Compared with the prior art, the present invention aims to solve the driving decision-making problem on unstructured roads in the prior art. The present invention designs a multi-branch feature extraction network for temporary construction scenarios of unstructured roads, and extracts features of multiple targets in combination with the target feature characteristics existing in the scene. At the same time, the feature enhancement part introduces enhancement and suppression processing of foreground and background features to improve the detection accuracy of targets. In addition, the present invention obtains a passable area based on the target guided semantic segmentation model obtained by detection, and constructs virtual lanes and virtual lane lines based on vehicle occlusion conditions. The detection of virtual lane lines makes full use of the positional relationship in the current road environment, and solves the problem of automatic driving decision-making in temporary traffic environments on unstructured roads. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in 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 paying creative work.

[0012] Figure 1 This is the multi-sensor fusion collaborative autonomous driving decision-making for new energy vehicles under complex road conditions in this application. DETAILED DESCRIPTION

[0013] The embodiments of the present application are described in detail with reference to the accompanying drawings.

[0014] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0015] The embodiments of this specification propose a multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions, such as Figure 1 As shown, the method specifically comprises the following steps: S1: The camera collects road scene images and identifies targets in the front area to determine whether to enter the temporary construction area; S2: Divide the drivable area based on the road scene image; S3: Construct virtual lane lines based on the passable area and detected vehicle position information S4: Build autonomous driving decisions based on virtual lane lines combined with drivable areas and regional targets.

[0016] During the implementation process, the system adopts the collaborative design of the front camera and the side camera to ensure comprehensive coverage of complex road conditions, especially in unstructured road environments, and can efficiently capture key road features. In order to ensure complete road scene coverage, the front camera is installed at the center of the upper edge of the front windshield of the new energy vehicle, and the side cameras are installed near the rearview mirrors on the left and right sides of the vehicle. The combined camera ensures that the front camera effectively covers the road in front of the vehicle, and at the same time expands the vehicle's lateral perception ability through the side camera, especially in unstructured scenes such as narrow roads, it can capture potential obstacles and special road conditions on both sides of the vehicle. The front camera adopts a horizontal viewing angle of 120° and a vertical viewing angle of 70°, while the horizontal viewing angle of the side camera is designed to be 90° and the vertical viewing angle is 60°, so as to achieve seamless connection of the perception range of multiple cameras and avoid the appearance of perception blind spots. In order to ensure the quality of the collected images, the resolution of the cameras is set to 1920×1080 pixels and the frame rate is 30 frames / second, which can continuously capture clear road scene images when the vehicle is driving at high speed.

[0017] During the data collection process, the image data from the front camera and the side camera are transmitted to the on-board computing platform via a high-speed bus. To ensure the synchronization of multi-camera data, the system uses timestamp synchronization technology to match each frame of the image with the vehicle's position information and motion status, ensuring accurate alignment when multi-sensor data is fused. In addition, in order to reduce transmission delays, the system introduces the H.265 video compression algorithm to losslessly compress image data, saving bandwidth while maintaining the high quality of the original image.

[0018] The collected video images are processed in real time through the on-board computing platform, mainly including denoising and enhancement steps. The system denoises the video images to reduce the impact of environmental noise on image quality, such as vehicle vibration, sensor interference or image noise under low light conditions; then, the image contrast is enhanced through histogram equalization technology to make road boundaries, lane lines and markers clearer; in addition, in order to further improve the practicality of the image, the system also adjusts the local brightness of the abnormal lighting areas in the video image to ensure that the key features of the road scene can be fully preserved. Based on the video image processing, the system extracts key frames from the video stream as the final road scene image. The key frame extraction process is achieved through the inter-frame change detection algorithm, that is, analyzing the changes in scene features in consecutive video frames and extracting frames containing significant environmental changes or key road information.

[0019] The front area target is realized by using a target detection network, which includes a backbone network, a feature enhancement module, a multi-scale feature fusion module and a target detection head; The backbone network consists of four sequentially connected feature extraction modules to obtain four different depth feature maps of the road scene image. , the feature extraction module is defined as: ; ; ; ; ; in, For convolution calculation, the lower subscript is the convolution kernel size, and the upper subscript is the number of consecutive executions. are the intermediate feature maps of the feature extraction module, and are the input and output of the feature extraction module respectively. is pixel-by-pixel multiplication.

[0020] In the temporary maintenance scenario of unstructured road sections, the detection of targets such as fences, roadblocks, construction workers and vehicles is relatively complicated. The reason is that there are significant differences in the shapes, sizes, texture features and distribution patterns of these targets, and the scene background may also be relatively chaotic. In order to solve this problem, the present invention adopts a parallel feature extraction structure to fully explore the features of each target. First, the present invention can quickly extract the basic edge features and low-level semantic information of the target through a simple convolution operation. This simple structure can efficiently capture the target features with regular shapes or clear boundaries such as fences and roadblocks, and lay the foundation for subsequent multi-scale feature extraction.

[0021] At the same time, in order to solve the problem that the targets such as enclosures, roadblocks, construction personnel and vehicles have different scales and sizes, the present invention also designs a multi-scale extraction method to obtain , using a single convolution path to extract local features of the target, which is suitable for capturing detailed features of small targets such as construction workers and distant vehicles; the double convolution path uses a deeper receptive field to enhance the perception of medium-sized targets, such as fences or medium-sized roadblocks; the triple convolution path extracts deeper semantic features, which helps to better understand complex targets, such as large roadblocks or multiple targets overlapping in the background. Multi-size target extraction can effectively deal with the problem of large changes in target scale in unstructured sections, while retaining enough detailed information.

[0022] Furthermore, the fences in temporary maintenance scenes are usually distributed in long strips, and the distribution of vehicles also has a certain directionality. In order to improve the directional sensitivity and spatial perception capabilities of feature extraction, the present invention also extracts The feature map is supplemented by 1*3 and 3*1 directional convolution calculations, which can flexibly extract the horizontal, vertical and oblique features of the target, thereby enhancing the recognition ability of different target shapes.

[0023] The feature map They are input into the feature enhancement module to enhance the foreground features, and the ; ; ; in, , are the input and output of the feature enhancement module, respectively. is the weight feature map, is pixel-by-pixel multiplication, , , FC represent normalization, activation function and full connection respectively; In the temporary maintenance road scene of unstructured road sections, background information such as the sky and asphalt road surface usually occupies most of the image area, but this background information is irrelevant to the target detection task and may even interfere with the accurate recognition of the foreground target. Therefore, the present invention generates a weight map for the extracted foreground features, guides the attention of the foreground and background to suppress the background information. Through this method, the interference of the background area on feature extraction can be effectively reduced, thereby improving the subsequent network's feature enhancement ability for the foreground target features. In addition, this attention guidance mechanism can not only focus the network's computing resources on the key area where the foreground target is located, but also help the network extract valuable context clues from the background information. For example, in a temporary maintenance scenario, the relative position relationship between the asphalt road surface and the foreground target is usually relatively fixed, such as the relationship between a roadblock or fence and the road surface. By suppressing irrelevant road surface texture information while retaining its associated features with the foreground target, the target detection task can be better assisted.

[0024] The multi-scale feature fusion module transforms the feature map and After fusion, the feature map is obtained , and then Fusion is performed to obtain the fusion output feature map ; ; in, , They are the two input feature maps of the fusion module. and When fusion , , when the feature map and When fusion, , ; The target detection head is based on Get the detection result map containing the target category and corresponding location information ; Detected The target set is , each target Contains: Bounding box center coordinates , target category .

[0025] Divide the drivable area based on the road scene image; The present invention uses the detection result map to guide the road production image to divide the passable area. First, each pixel Construct a composite weight graph:

[0026] in: is the target distance, , is the global background Gaussian radius; is the category radius parameter; In an exemplary embodiment, the class-dependent Gaussian parameters may be specifically defined as and : Suppression amplitude : Fence = -1.2, Construction Workers = 0.8, Roadblock = -1.0, Promotion Amplitude : Driving vehicle = 0.6, effective radius : Fence = 20px, construction workers = 30px, moving vehicles = 10px, roadblocks = 40px; The encoder is used to extract features from the road scene image and obtain three levels of feature coding maps: , downsample the composite weight map to guide the feature map to segment the key areas; ; ; ; in, To downsample the composite weight map W to Consistent size, and They are the m-1th and nth attention-guided feature maps respectively; and Calculate the channel mean and standard deviation respectively; The semantic segmentation map of the drivable area in the road scene image is obtained by inputting it into the decoding network which is symmetrical with the encoder structure; In the process of dividing the passable area, the extracted targets are used to generate a Gaussian weight map to guide the semantic segmentation of the passable area. In the temporary maintenance scenario of unstructured road sections, the boundary of the passable area is usually disturbed by complex background and dynamic targets. By generating a Gaussian weight map, the network's attention can be effectively focused on the key areas associated with the target, thereby enhancing the segmentation's perception of the passable area and filtering out the interference of background areas or irrelevant information on the segmentation results. The generation of the Gaussian weight map is based on the target information extracted previously, and can dynamically adjust the weight distribution according to the spatial distribution of the target. Specifically, higher weights are given to areas close to construction personnel, fences or roadblocks, while lower weights are given to background areas far away from these targets, thereby guiding semantic segmentation to pay more attention to potential passable areas.

[0027] In addition, in temporary maintenance scenarios, the shape of the passable area may change due to the complexity of the road conditions or the diversity of obstacle distribution. For example, some areas may be occupied by roadblocks or fences, making it difficult for traditional segmentation methods to accurately divide the pass path. Guided by the Gaussian weight map, the segmentation process can automatically adapt to these complex changes and dynamically adjust the degree of attention to different areas, so as to more accurately identify the actual passable road areas. Especially when construction workers or vehicles are close to the boundaries of the passable area, the Gaussian weight map can effectively highlight the significance of these areas and help the network more accurately segment safe passage paths.

[0028] Construct virtual lane lines based on the passable area and detected vehicle position information; The set of vehicles detected by the target detection network is , where N is the number of vehicles, and Respectively represent The coordinates of each vehicle and the target width and height; Based on vehicle collection Query the unobstructed vehicle set , and Generate an initial lane for each vehicle: ; ; in, is the lateral position sensitivity coefficient, is the occlusion determination threshold, Representing a collection The t-th vehicle in the set, T is The number of vehicles in is the tth lane; For any vehicle to be assigned , assigned to lanes The following conditions must be met: ; in, , is the standard lane width, is the perspective effect compensation coefficient For each lane , sorted by vertical coordinate: ; represents the pth vehicle belonging to the tth lane The vertical coordinate of Lane Based The vehicle position is used to generate the tth virtual lane line: ; is the lane curvature compensation coefficient, Y is the maximum visible distance, and y is the distance from the current position to the road ahead; In construction areas, due to the setting of fences, roadblocks or other temporary facilities, the original lane lines may be partially blocked or completely disappear. In some cases, the lane lines may no longer reflect the actual lane conditions. If the autonomous driving system still relies on these physical lane lines for path planning, it may cause misjudgment, causing the vehicle to enter a dangerous area or deviate from the safe passage path unnecessarily. However, by analyzing the occlusion relationship between vehicles, the actual lane structure can be inferred from the distribution of vehicles. For example, if there is occlusion between two vehicles and their lateral position deviation is small, it can be judged with a high degree of confidence that they are in the same lane. This judgment based on the position relationship of the target can effectively replace the invalid static lane line information.

[0029] Generating virtual lane lines based on the position information of vehicles on the same lane also improves the adaptability and robustness of the system. Virtual lane lines no longer rely on physical lane lines in the construction area, but reflect the actual travel path of the road by analyzing the actual distribution of vehicles. For example, in the construction area, the original lane may be occupied by fences, and vehicles need to travel along the edge of the construction area or other temporarily open channels. At this time, the virtual lane lines generated by analyzing the position information of vehicles in the same lane can accurately describe the current travel trajectory and avoid path planning errors caused by misleading original lane lines.

[0030] The autonomous driving decision is constructed based on the virtual lane lines combined with the drivable area and regional targets.

[0031] The regional target recognition results show that a circular high-risk area with a radius of 50 cm is generated at the location of the enclosure and roadblock, and its risk value decays exponentially with the distance from the center point; a dynamic elliptical risk field is generated in the construction personnel area, and the long axis direction is perpendicular to the direction of vehicle movement; For high-risk target areas, the original virtual lane line is shifted to the opposite side to form an avoidance area, and the width of adjacent lanes is compressed to ensure a safe distance; For the construction worker area, a dynamic right of way switching point is set 3 meters in front of the construction worker's target position to trigger a temporary bifurcation of the lane line, interrupting the current lane and activating the detour guide line.

[0032] For virtual lane lines, the coordinates of the virtual lane lines are mapped to the passable semantic segmentation map. If the overlap rate with the impassable area exceeds 15%, the path reconstruction is started. During the path reconstruction, an attempt is made to offset the current lane line in parallel, and the offset step increases by 0.2 meters. The maximum offset does not exceed the standard lane width. If the conditions are still not met, the road section is marked as a restricted access area and the degraded control mode is triggered.

[0033] The present invention provides a multi-sensor fusion collaborative automatic driving decision-making system for new energy vehicles facing complex road conditions, the system comprising: Image acquisition equipment: collects road scene images through cameras and identifies targets in the front area to determine whether to enter the temporary construction area; Semantic segmentation module: the semantic segmentation module divides the drivable area based on the road scene image; Virtual lane line generation module: The virtual lane line generation module constructs virtual lane lines based on the passable area and the detected vehicle position information; Autonomous driving decision module: The autonomous driving decision module builds autonomous driving decisions based on virtual lane lines combined with drivable areas and regional targets.

[0034] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-sensor fusion collaborative autonomous driving decision method for complex road conditions is implemented.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions.

[0036] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0037] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0038] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions, characterized by: The method comprises: S1: The camera collects road scene images and identifies targets in the front area to determine whether to enter the temporary construction area; S2: Divide the drivable area based on the road scene image; S3: construct virtual lane lines based on the passable area and the detected vehicle position information; S4: Build autonomous driving decisions based on virtual lane lines combined with drivable areas and regional targets.

2. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 1 is characterized in that: The targets in the front area are realized by using a target detection network, which includes a backbone network, a feature enhancement module, a multi-scale feature fusion module and a target detection head. The targets in the front area include: fences, construction workers, moving vehicles and roadblocks.

3. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 2, characterized in that: The backbone network consists of four sequentially connected feature extraction modules to obtain four different depth feature maps of the road scene image. , the feature extraction module is defined as: ; ; ; ; ; For convolution calculation, the lower subscript is the convolution kernel size, and the upper subscript is the number of consecutive executions. are the intermediate feature maps of the feature extraction module, and are the input and output of the feature extraction module respectively. It is pixel-by-pixel addition.

4. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 3, characterized in that: The feature map They are input into the feature enhancement module to enhance the foreground features, and the ; ; ; , are the input and output of the feature enhancement module, respectively. is the weight feature map, is the pixel-by-pixel multiplication, , , FC are normalization, activation function and full connection respectively.

5. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 4, characterized in that: The encoder can be used to extract features from the road scene image to obtain three levels of feature coding map: , downsample the composite weight map to guide the feature map to segment the key areas; ； ; ; To downsample the composite weight map W to Consistent size, and They are the m-1th and nth attention-guided feature maps respectively; and Calculate the channel mean and standard deviation respectively; The semantic segmentation map of the drivable area in the road scene image is obtained by inputting it into a decoding network that is symmetrical to the encoder structure.

6. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 5, characterized in that: Constructing a virtual lane includes generating an initialization lane; The set of vehicles detected by the target detection network is , where N is the number of vehicles, , and Respectively represent The coordinates of each vehicle and the target width and height; based on the vehicle set Query the unobstructed vehicle set , and Generate an initial lane for each vehicle: ； ; in, is the lateral position sensitivity coefficient, is the occlusion determination threshold, Representing a collection The t-th vehicle in the set, T is The number of vehicles in is the tth lane.

7. The multi-sensor fusion collaborative automatic driving decision-making method for complex road conditions according to claim 6, characterized in that: Constructing a virtual lane also includes: , assigned to lanes The following conditions should be met: ； in, is the standard lane width, is the perspective effect compensation coefficient, For each lane , sorted by vertical coordinate: ; represents the pth vehicle belonging to the tth lane The vertical coordinate of Lane Based The vehicle position is used to generate the tth virtual lane line: ; is the lane curvature compensation coefficient, Y is the maximum visible distance, and y is the distance from the current position to the road ahead.

8. A multi-sensor fusion collaborative automatic driving decision-making system for new energy vehicles in complex road conditions, characterized by The system includes: Image acquisition equipment: collects road scene images through cameras and identifies targets in the front area to determine whether to enter the temporary construction area; Semantic segmentation module: the semantic segmentation module divides the drivable area based on the road scene image; Virtual lane line generation module: The virtual lane line generation module constructs virtual lane lines based on the passable area and the detected vehicle position information; Autonomous driving decision module: The autonomous driving decision module builds autonomous driving decisions based on virtual lane lines combined with drivable areas and regional targets.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for making decisions based on a multi-sensor fusion collaborative automatic driving for complex road conditions as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the multi-sensor fusion collaborative autonomous driving decision-making method for complex road conditions as described in any one of claims 1 to 7.

Images