Vehicle control methods and electronic equipment

Abstract

This application provides a vehicle control method and electronic device, relating to the field of intelligent driving assistance technology. The method includes: acquiring an image sequence of at least one surrounding vehicle during its journey; the surrounding vehicles are vehicles in the current lane and adjacent lanes of the vehicle, located in front of the vehicle; detecting the vehicle attitude characteristics of each surrounding vehicle based on the image sequence; determining, based on the vehicle attitude characteristics, whether each surrounding vehicle has experienced an attitude change event due to passing through the road surface ahead; if an attitude change event is determined to have occurred in any surrounding vehicle, controlling the vehicle to pass through the road surface ahead based on driving control parameters associated with the attitude change event. Thus, by performing advance passage control on the vehicle based on the driving control parameters associated with the attitude change event, the vehicle avoids passing through abnormal road surfaces at high speeds, preventing issues such as chassis bottoming out and body collisions, and eliminating driving safety hazards caused by the failure to perceive road bumps during intelligent driving.

Description

Technical Field

[0001] This application relates to the field of intelligent driving assistance technology, and more specifically, to a vehicle control method and electronic device. Background Technology

[0002] With the rapid development of automotive intelligence, Advanced Driver Assistance Systems (ADAS) and advanced autonomous driving systems have become one of the core configurations of vehicles. The recognition and active control of road obstacles during vehicle cruising are key factors in ensuring intelligent driving safety and driving comfort.

[0003] In related technologies, advanced driver assistance systems and high-level autonomous driving systems mainly use vision cameras as sensing devices to identify static obstacles on the road (such as speed bumps, road protrusions, steps, etc.) in order to achieve the detection and passage control of road obstacles.

[0004] However, in practical applications, such solutions all rely on the direct identification of the visual features of road obstacles. However, obstacles such as raised sections of the road ahead generally lack visual features or are not obvious. For example, pure cement speed bumps have no obvious texture, and speed bumps with worn paint are similar in color to the road surface, making it impossible for the visual model to directly segment and identify them. Because the system cannot directly perceive road abnormalities, it is difficult to determine whether there are raised obstacles on the road ahead from a safe distance. As a result, it cannot perform reasonable advance passage control. When cruising, vehicles are prone to pass through such road surfaces at high speeds, causing chassis bottoming out and collisions. This not only damages vehicle parts and reduces driving comfort, but also poses driving safety hazards. Summary of the Invention

[0005] This application provides a vehicle control method and electronic device. The method performs advance passage control on the vehicle based on driving control parameters associated with attitude change events, avoiding the vehicle from passing through abnormal road surfaces at high speeds, which could lead to chassis bottoming out and body collisions. This not only protects vehicle components from damage but also significantly improves driving comfort and effectively eliminates driving safety hazards caused by the failure to perceive road bumps and obstacles during intelligent driving.

[0006] In a first aspect, a vehicle control method is provided, the method comprising: acquiring an image sequence of at least one surrounding vehicle in motion; the surrounding vehicles are vehicles located in front of the vehicle in the current lane and adjacent lanes of the vehicle; detecting vehicle attitude features of each surrounding vehicle based on the image sequence; the vehicle attitude features are used to characterize the relative vertical displacement of the wheels and the vehicle body of the surrounding vehicle, and / or the pitch attitude of the vehicle body; and determining, based on the vehicle attitude features, whether each surrounding vehicle has triggered an attitude change event due to passing through the road surface in front. If a sudden attitude change event is detected in any surrounding vehicle, the vehicle is controlled to pass through the road ahead based on the driving control parameters associated with the attitude change event.

[0007] In the above technical solution, by acquiring image sequences of at least one surrounding vehicle in front of the vehicle in the current lane and adjacent lanes during its travel, the traditional solution abandons the direct reliance on the visual features of road surface protrusions and obstacles. Then, based on this image sequence, the vehicle attitude features of each surrounding vehicle are detected. These vehicle attitude features are specifically used to characterize the relative vertical displacement of the wheels and body of the surrounding vehicles, and / or the pitch attitude of the vehicle body. This can describe the inherent attitude response of surrounding vehicles when passing through the road surface from dimensions such as vehicle suspension deformation and body sway, providing standardized and quantifiable feature support for the indirect perception of road surface anomalies. Based on this, the extracted surrounding vehicle attitude features are used to determine whether each surrounding vehicle has experienced a sudden attitude change event caused by passing through the road surface ahead. This utilizes the relative vertical displacement of the wheels and body and the pitch attitude of the surrounding vehicles when they pass through the abnormal road surface area. The system utilizes a realistic physical response to sudden changes in road surface conditions to indirectly perceive hidden protruding obstacles. This effectively addresses the technical challenge of visually missing or similarly colored obstacles, such as pure cement speed bumps and faded speed bumps, which cannot be segmented and identified by visual models. Furthermore, it can predict the presence of protruding anomalies on the road surface ahead from a safe distance based on sudden changes in the posture of neighboring vehicles. This overcomes the limitations of traditional systems that cannot directly perceive road anomalies and predict risks in advance. Subsequently, upon determining that any surrounding vehicle has experienced a sudden change in posture, the system immediately matches the driving control parameters associated with that event to implement advance passage control for the vehicle. This prevents the vehicle from passing over abnormal road surfaces at high speeds, avoiding issues like chassis bottoming out or body collisions. It protects vehicle components from damage, improves driving comfort, and eliminates the driving safety hazards caused by the failure to perceive road surface protrusions during intelligent driving.

[0008] Secondly, a vehicle control device is provided, comprising: an acquisition module for acquiring an image sequence of at least one surrounding vehicle during its driving process; the surrounding vehicles are vehicles located in front of the vehicle in its current lane and adjacent lanes; a detection module for detecting vehicle attitude features of each surrounding vehicle based on the image sequence; the vehicle attitude features are used to characterize the relative vertical displacement of the wheels and the vehicle body of the surrounding vehicle, and / or the pitch attitude of the vehicle body; a determination module for determining, based on the vehicle attitude features, whether each surrounding vehicle has triggered an attitude change event due to passing through the road surface ahead; and a control module for controlling the vehicle to pass through the road surface ahead based on driving control parameters associated with the attitude change event if it is determined that any surrounding vehicle has triggered an attitude change event.

[0009] Thirdly, an electronic device is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, causing the electronic device to perform the methods of the first aspect or any possible implementation thereof.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of this application more easily understood, specific embodiments of this application are given below. Attached Figure Description

[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of a vehicle control system provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 3 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 1 ; Figure 4 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 2 ; Figure 5 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 3 ; Figure 6 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 4 ; Figure 7 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 5 ; Figure 8 This is a schematic diagram of the structure of a vehicle control device provided in an embodiment of this application. Detailed Implementation

[0012] The technical solutions of this application will now be described clearly and in detail with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " signifies "or," for example, A / B can mean A or B. "And / or" in the text merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more.

[0013] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0014] With the continuous popularization of intelligent driving technology, ADAS and advanced autonomous driving systems have been widely used in vehicle cruise control scenarios. These systems mostly rely on visual cameras to directly identify static obstacles such as road bumps and speed bumps to complete road risk detection and vehicle passage control. However, this direct perception solution is highly dependent on the visual characteristics of the obstacles themselves. For bumps with weak visual characteristics or those highly integrated with the road surface, such as concrete speed bumps or worn-out speed bumps, the visual model struggles to effectively segment and identify them. This makes it difficult for the system to determine road anomalies within a safe distance, hindering the implementation of advance and reasonable passage control. Vehicles are prone to chassis collisions and component damage when passing through such hidden bumps at high speeds, reducing driving comfort and posing safety hazards. Therefore, this application proposes a technical concept that eliminates reliance on direct visual recognition of road obstacles and instead indirectly determines road anomalies by monitoring sudden changes in the driving posture of surrounding vehicles. This addresses the perception failure and control lag problems of existing technologies. The following detailed description, in conjunction with the accompanying drawings, provides a series of embodiments of the vehicle control method and electronic equipment of this application.

[0015] Before elaborating on the technical solution, the technical terms used in this application will be explained to facilitate subsequent understanding.

[0016] "Side vehicles" or "reference vehicles" refer to vehicles traveling normally in all lanes surrounding the vehicle, including vehicles traveling in the same direction, vehicles traveling in the opposite direction, and vehicles in adjacent lanes. The dynamic responses of these vehicles when passing unknown road obstacles can serve as an important reference for road anomaly perception. Among them, "surrounding vehicles" specifically refers to vehicles in front of the vehicle within the current lane and adjacent lanes of the vehicle.

[0017] Vehicle attitude characteristics are the vertical motion response generated when a vehicle drives over a raised obstacle on the road surface. Specifically, they include, but are not limited to, the vertical displacement of the wheel center relative to the wheel arch (i.e., suspension compression), the pitch angular velocity corresponding to the shark fin antenna on the roof or the edge of the roof, and the overall vertical acceleration of the vehicle.

[0018] A sudden attitude change event refers to a sudden change in the vehicle attitude characteristics of a vehicle passing over a road obstacle, which exceeds a preset attitude threshold. This sudden change is determined to be a valid road anomaly observation.

[0019] The road surface anomaly map is a dynamic map built on a spatiotemporal grid. It is used to record the number of times attitude change events are observed at each geographical location, the vehicle type involved, the corresponding confidence level, and distance information.

[0020] The safe passing speed is the maximum permissible speed that can be calculated to avoid the vehicle bottoming out or experiencing severe bumps, taking into account the vehicle's minimum ground clearance (such as the vertical height of the lowest part of the vehicle body to the ground when the vehicle is parked on flat ground), the vehicle's wheelbase, and the estimated height of obstacles.

[0021] Figure 1 This is a schematic diagram of a vehicle control system provided in an embodiment of this application. Figure 1 As shown, the vehicle control system 100 includes at least: an environmental perception module 110, a vehicle dynamic analysis module 120, a multi-source fusion and map building module 130, and a decision and control module 140.

[0022] The environmental perception module 110, as the front-end perception unit of this system, is integrated on the vehicle body. Its main function is to collect and output visual image sequences of surrounding vehicles and the vehicle's own driving status information in real time, providing multi-source basic perception data for subsequent vehicle dynamic analysis and road anomaly perception. Surrounding vehicles specifically refer to vehicles traveling in the current driving lane and the adjacent lanes to the left and right of the vehicle, and whose spatial orientation is in front of the vehicle. These vehicles serve as reference carriers for indirectly perceiving road protrusions in this application. The attitude changes generated when they pass through road protrusions can be used as a key basis for judging road anomalies. The environmental perception module 110 achieves comprehensive perception of the surrounding vehicles and the vehicle's status through the collaborative work of multiple sensing devices, ensuring the accuracy of subsequent analysis data.

[0023] The environmental perception module 110 includes at least: a front-view high-definition camera, a side-view or surround-view camera, a millimeter-wave radar or lidar, and a vehicle status sensor.

[0024] Among them, the forward-looking high-definition camera is used to detect vehicles and track the attitude of adjacent vehicles; the side-looking or surround-view cameras are used to expand the observation range and capture vehicles in adjacent lanes; millimeter-wave radar or lidar is used to assist in the direct detection of road surface protrusions, complementing vision; the vehicle status sensor is used to compensate for the vehicle's motion, which includes at least: IMU (Inertial Measurement Unit) and wheel speed sensor.

[0025] The vehicle dynamic analysis module 120 is communicatively connected to the environment perception module 110 and is used to process image sequences to determine whether each surrounding vehicle has experienced a sudden attitude change event caused by passing the road ahead. The vehicle dynamic analysis module 120 achieves the above functions through its internal vehicle detection and tracking submodule, key point detection submodule, and attitude temporal analysis submodule. Specifically, the vehicle detection and tracking submodule detects and identifies target vehicles in the image sequence and tracks their continuous trajectories, locking onto surrounding vehicles that meet the required range. The key point detection submodule detects preset attitude key points on surrounding vehicles, such as wheel contact points, wheel arches, roof shark fins, and suspension towers, and obtains their corresponding pixel coordinates. The attitude temporal analysis submodule calculates vehicle attitude features such as suspension compression and vehicle pitch rate based on the key point coordinates, constructs a feature sequence through temporal analysis, calculates a stable attitude baseline value, and compares the feature changes in the current frame to ultimately determine the attitude change event. Among them, the vehicle detection and tracking submodule is the first execution unit, which forms a front input connection with the key point detection submodule; the key point detection submodule is the intermediate execution unit, which forms a data receiving connection with both the front and rear submodules; the attitude timing analysis submodule is the end execution unit, which forms a final calculation connection with the key point detection submodule; the three are connected in series and the data is passed step by step in a linear execution connection relationship.

[0026] The multi-source fusion and map building module 130 is communicatively connected to the adjacent vehicle dynamic analysis module 120. It receives the judgment results of surrounding vehicle attitude change events output by the adjacent vehicle dynamic analysis module 120, and performs coordinate transformation, spatial alignment, and confidence assessment on the relevant data corresponding to the vehicle attitude change events. Simultaneously, it completes the construction and updating of a local dynamic map. Specifically, coordinate transformation converts the two-dimensional pixel coordinates of the image into the actual three-dimensional spatial coordinates of the road. Spatial alignment combines the vehicle's state information to eliminate position offset errors caused by the vehicle's movement. Confidence assessment combines multi-source perception data to perform weighted verification of the reliability of attitude change events, eliminating false judgments. Finally, the spatial area corresponding to the verified road surface protrusions and obstacles is written into and stored in the local dynamic map, providing spatial location basis for subsequent vehicle decision-making and control. The multi-source fusion and map building module 130 includes at least: a coordinate transformation and spatial alignment submodule, a confidence assessment and fusion submodule, and a local dynamic map storage submodule. Among them, the coordinate transformation and spatial alignment submodule is the first-level execution unit, which forms a front-end data supply connection with the confidence assessment and fusion submodule; the confidence assessment and fusion submodule is the intermediate-level execution unit, which forms a data receiving connection with both the preceding and following submodules; the local dynamic map storage submodule is the final-level execution unit, which forms a final landing connection with the confidence assessment and fusion submodule. Thus, the three together form a linearly connected execution and data transmission relationship.

[0027] The decision-making and control module 140 is communicatively connected to the multi-source fusion and map building module 130. It is used to access road anomaly spatial information from a local dynamic map, generate appropriate driving control parameters and corresponding vehicle control commands based on this information, and regulate the vehicle's driving state by executing these commands to safely and smoothly pass through areas with protruding obstacles. The decision-making and control module 140 internally includes a safe speed calculation submodule, a deceleration strategy and trajectory planning submodule, and a human-machine interaction prompt submodule. Specifically, the safe speed calculation submodule calculates the safe passing speed based on obstacle spatial information; the deceleration strategy and trajectory planning submodule formulates corresponding deceleration control strategies and driving trajectory optimization schemes; and the human-machine interaction prompt submodule simultaneously outputs road anomaly prompt information to the driver. These submodules work collaboratively to form a complete decision-making and control logic, ensuring both vehicle traffic safety and improved driving comfort. The system comprises three main components: a safe speed calculation submodule, a front-end core calculation unit that directly inputs data to the deceleration strategy and trajectory planning submodule; a deceleration strategy and trajectory planning submodule, an intermediate core decision-making unit that connects to both the front and rear modules for data reception and triggering; and a human-machine interaction prompt submodule, a parallel auxiliary unit that triggers a linkage with the deceleration strategy and trajectory planning submodule. These three components together form a collaborative connection between serial core control and parallel human-machine prompts.

[0028] The vehicle control system provided in this application includes an environmental perception module for acquiring image sequences containing surrounding vehicles and the vehicle's status information. The surrounding vehicles are those in front of the vehicle in its current lane and adjacent lanes, providing comprehensive and reliable multi-source raw perception data for subsequent indirect perception of road anomalies. A vehicle dynamic analysis module communicates with the environmental perception module and, through targeted processing of the image sequences, can determine whether each surrounding vehicle has experienced a sudden attitude change event caused by passing a protruding obstacle on the road ahead. This allows for indirect perception of hidden protruding obstacles on the road surface based on the driving responses of the surrounding vehicles, effectively overcoming the shortcomings of traditional direct visual recognition. Multi-source fusion and mapping are also included. The construction module communicates with the adjacent vehicle dynamic analysis module, and can sequentially perform coordinate transformation, spatial alignment and confidence assessment on the identified attitude change events, and build a local dynamic map accordingly, thereby completing the spatial calibration and reliability verification of attitude change events, and providing a standardized and unified environmental spatial basis for vehicle decision-making. The decision and control module communicates with the multi-source fusion and map construction module, and can reasonably generate appropriate driving control parameters and corresponding control commands based on the updated local dynamic map, so as to smoothly control the vehicle to safely pass through the road surface with protruding obstacles, thereby avoiding chassis collisions, ride bumps and other problems caused by high-speed vehicle passage, and improving the driving safety and ride comfort of intelligent driving.

[0029] Optionally, Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 2 As shown, the electronic device 200 may include a processor 210 and a memory 220.

[0030] The memory 220 stores machine-executable instructions that can be executed by the processor 210. When the electronic device 200 is running, these machine-executable instructions are executed. The processor 210 and the memory 220 communicate via a bus. The processor 210 can execute these machine-executable instructions to implement a vehicle control method.

[0031] The memory 220, processor 210, and buses are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected via one or more communication buses or signal lines. The memory 220 includes at least one software functional module, which is stored or embedded in the operating system (OS) of the electronic device in the form of software or firmware. This software functional module includes at least one executable module. The processor 210 is used to execute the executable modules stored in the memory 220, such as the software functional modules and computer programs included in the vehicle control method.

[0032] The memory 220 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0033] The electronic device 200 can be selected according to the actual situation; for example, it can be selected as a vehicle controller. Furthermore, the electronic device 200 has software capable of executing vehicle control methods or is equipped with a vehicle control system.

[0034] The vehicle control method provided in this application embodiment can be executed by a processor in the electronic device 200. The vehicle control method provided in this application embodiment will be explained further below. Figure 3 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 1 .like Figure 3 As shown, the method may include: S310. Acquire an image sequence of at least one surrounding vehicle during its movement.

[0035] Among them, surrounding vehicles refer to vehicles in front of the vehicle in its current lane and adjacent lanes. That is, vehicles that simultaneously meet the dual constraints of lane affiliation and spatial orientation. In terms of lane affiliation, these vehicles are limited to the vehicle's current lane and the adjacent left and right lanes, excluding irrelevant vehicles in oncoming lanes or non-adjacent lanes. In terms of spatial orientation, these vehicles are further limited to those in front of the vehicle in its direction of travel, excluding vehicles behind or to the side that are not relevant references. This limitation allows for the selection of effective target vehicles that can serve as a reference for perceiving road surface protrusions. Because these vehicles are in front of the vehicle and in adjacent or same lanes, the sudden changes in their posture when passing a road surface protrusion directly reflect the abnormal conditions of the road surface the vehicle is about to pass through. This avoids redundant analysis of irrelevant vehicles, improves the computational efficiency and detection accuracy of subsequent dynamic analysis, and ensures that the acquired vehicle posture response information is highly correlated with the road surface risk in front of the vehicle, providing a reasonable and effective reference for subsequently indirectly determining road surface protrusions through vehicle posture changes.

[0036] In one possible implementation, an environmental perception module mounted on the vehicle is used to acquire and obtain image sequences. This environmental perception module uses a front-view high-definition camera as the visual acquisition device, supplemented by side-view or surround-view cameras to expand the field of view. At the same time, it works with vehicle status sensors such as IMU and wheel speed sensors for motion compensation. The front-view high-definition camera is placed inside the windshield or at a preset position at the front of the vehicle to acquire the main field of view image in the direction of the vehicle's travel. The side-view or surround-view cameras are distributed on the sides of the vehicle body, rearview mirrors, etc., to cover the visual area of ​​adjacent lanes. The two types of cameras work together to continuously acquire multiple frames of time-series images of the road scene at a preset frame rate, forming an image sequence that includes the driving posture of surrounding vehicles and road environment information. During the acquisition process, the IMU collects the vehicle's angular velocity and acceleration data in real time, while the wheel speed meter collects the vehicle's wheel speed information. Together, they compensate for image offsets and distortions caused by the vehicle's movement, steering, acceleration, and deceleration in real time, ensuring that the acquired image sequence is clear, stable, and free of motion distortion. This provides real raw visual data for subsequent vehicle detection, key point recognition, and attitude temporal analysis. The image sequence is a set of visual data composed of continuous temporal frames, which can record the dynamic changes of surrounding vehicles during their movement.

[0037] S320: Based on the image sequence, detect the vehicle posture features of each surrounding vehicle.

[0038] Vehicle attitude features are used to characterize the relative vertical displacement between the wheels and the body of surrounding vehicles, and / or the pitch attitude of the vehicle body. For example, when surrounding vehicles are traveling on a smooth, unobstructed road surface, the relative vertical displacement between the wheels and the body exhibits only small, stable, and regular fluctuations, while the pitch attitude of the vehicle body remains stable without significant angle changes. When the vehicle suspension undergoes rapid extension and contraction deformation, it directly causes a sudden change in the relative vertical displacement between the wheels and the body, resulting in noticeable forward and backward pitching swaying of the vehicle body. These two types of changes can be quantified using the corresponding vehicle attitude features. Therefore, vehicle attitude features can objectively and realistically characterize the wheel-body relative vertical motion characteristics and the vehicle body pitch motion characteristics of surrounding vehicles. This allows for the differentiation between normal, smooth vehicle driving and abnormal attitude changes when passing over road bumps, and provides crucial information for subsequent determination of attitude change events and indirect identification of road bump obstacles based on attitude features. Among them, the relative vertical displacement between the wheels and the body of surrounding vehicles refers to the change in the relative pixel distance divided by the physical distance between the wheel center or wheel contact point of the same wheel and the corresponding wheel arch point above the vehicle body in the vertical height direction, with the vehicle's own structure as the reference. This relative vertical displacement is generated by the compression and tension deformation of the vehicle's suspension, which can intuitively reflect the degree of suspension deformation when the vehicle travels through road bumps and uneven road sections. It is the core quantitative feature of the vehicle's vertical attitude change, which is different from the overall translation of the vehicle body and only represents the difference in the vertical relative motion of the wheels relative to the body body. The pitch attitude of a vehicle body refers to the tilting and swaying state of the front and rear of the vehicle body relative to each other in the vertical direction, with the vehicle body's lateral axis as the rotation reference. Specifically, it is reflected in the temporal relative change of the vertical pixel position of the front and rear reference points of the roof in the image. The pitch indicator can be formed by the vertical pixel difference between the front and rear reference points, and the pitch angular velocity can be further solved. It can characterize the attitude change law of the vehicle body tilting forward and backward when the vehicle travels through road bumps and bumpy sections, and is a vehicle body attitude feature that is different from the local vertical displacement of the wheels.

[0039] In one possible implementation, since the image sequence acquired by the environmental perception module possesses continuous temporal and spatial imaging characteristics, it can retain information on the vehicle body structure, local key points, and attitude changes of surrounding vehicles at different times. Therefore, when processing this image sequence, the vehicle detection and tracking submodule in the adjacent vehicle dynamic analysis module first performs target detection, lane area filtering, and multi-frame trajectory correlation tracking on the continuous image sequence, locking each surrounding vehicle in the vehicle's lane and adjacent lanes, and eliminating interfering objects such as stationary targets on the roadside and vehicles in non-adjacent lanes. Subsequently, the key point detection submodule detects preset vehicle body attitude key points frame by frame in the image sequence for the tracked surrounding vehicles, identifying and acquiring wheel contact points, wheel arches, and roof sharks. The two-dimensional pixel coordinates of key feature points such as fins and suspension towers in each temporal image frame are obtained. Finally, the attitude temporal analysis submodule performs quantization calculations based on the above temporal pixel coordinates. On the one hand, it calculates the longitudinal relative pixel difference between the wheel center or wheel contact point and the wheel arch point of the same wheel to quantize the attitude features of the relative vertical displacement between the wheel and the vehicle body. On the other hand, it uses the temporal changes of the pixel coordinates of the front and rear reference points of the roof to calculate the attitude features of the vehicle body pitch attitude and pitch angular velocity. This enables the detection of vehicle attitude features of each surrounding vehicle based on the image sequence, which respectively represent at least one of the relative vertical displacement between the wheel and the vehicle body and the vehicle body pitch attitude, providing reliable feature inputs for subsequent attitude change determination, road obstacle recognition and autonomous vehicle driving control.

[0040] S330. Based on the vehicle's attitude characteristics, determine whether each surrounding vehicle has experienced a sudden attitude change event.

[0041] Among them, attitude change events are abnormal changes in vehicle body posture that occur when surrounding vehicles drive over or pass over protruding obstacles on the road surface (such as worn speed bumps, concrete protrusions, road steps, etc.). These attitude change events are specifically manifested as sudden vertical motion responses of the vehicle, including at least: sudden changes in the vertical displacement of the wheel center relative to the wheel arch, sudden changes in suspension compression, sudden changes in the pitch angular velocity of the roof fin or roof edge, and sudden changes in the overall vertical acceleration of the vehicle. Therefore, by defining these abnormal changes induced by road conditions as attitude change events, a unique association can be established between the abnormal posture of each surrounding vehicle and the protruding road obstacle, providing a basis for subsequent identification of hidden road obstacles through indirect sensing.

[0042] In one possible implementation, after the attitude temporal analysis submodule obtains the surrounding vehicle attitude features output by the key point detection submodule, it first constructs a temporal feature sequence for the vehicle attitude features corresponding to multiple consecutive frames of images. The vehicle attitude features include at least the following features: the vertical displacement of the wheel center relative to the wheel arch (i.e., the suspension compression feature), the pitch angular velocity of the roof shark fin or the edge of the roof (i.e., the vehicle pitch angular velocity), and the overall vertical acceleration of the vehicle, which can reflect the vertical motion and pitch changes of the vehicle body. Subsequently, attitude feature data of the vehicle during the stable driving phase are extracted from the temporal feature sequence, and a stable attitude baseline value is calculated. This stable attitude baseline value is used to characterize the normal attitude level of surrounding vehicles under normal driving conditions without road anomalies. Then, the difference between the current vehicle attitude feature and the stable attitude baseline value is calculated to obtain the attitude feature change. The calculated attitude feature change is then compared with a preset attitude change threshold. When the attitude feature change is greater than the preset attitude change threshold, it is determined that the vehicle body attitude of the surrounding vehicle has undergone an abnormal change beyond the normal driving range, that is, the surrounding vehicle has experienced an attitude change event caused by passing through a protruding obstacle on the road ahead. When the attitude feature change is less than or equal to the preset attitude change threshold, it is determined that the surrounding vehicle has not experienced an attitude change event. This completes the determination of whether each surrounding vehicle has experienced an attitude change event due to passing through the road ahead based on the vehicle attitude features.

[0043] It should be noted that, since the image sequence is not a single static image, but rather a series of continuous temporal images collected by the environmental perception module, it can reconstruct the dynamic information of surrounding vehicles throughout their journey on the road, providing the necessary data foundation for capturing instantaneous changes in vehicle attitude. Secondly, when surrounding vehicles are driving normally and smoothly on a road surface without abnormalities, their vehicle attitude remains relatively stable. This is reflected in the image sequence as small, gentle, and regular fluctuations in the pixel coordinates of preset key points such as wheel contact points, wheel arches, roof shark fins, and suspension towers. However, when surrounding vehicles pass over protruding obstacles on the road ahead, sudden vertical jumps and vehicle pitch occur due to abrupt changes in road height, leading to a sudden change in suspension compression and vehicle attitude. Sudden changes in pitch angular velocity at the top and vertical displacement of the vehicle body are abnormal attitudes. These abnormalities are directly reflected in the sudden and significant shifts in the pixel coordinates of the aforementioned attitude key points in the image sequence, creating visual feature differences that are completely different from the stable driving state. Finally, the vehicle dynamic analysis module uses the vehicle detection and tracking submodule to lock the vehicles around the target, the key point detection submodule to locate the attitude key points, and the attitude temporal analysis submodule to compare the temporal feature changes with preset thresholds. This can distinguish between the normal stable attitude of the vehicle and the attitude change caused by the road surface bump. Based on the temporal visual feature processing results of the image sequence, it can reliably determine whether each surrounding vehicle has experienced an attitude change event caused by passing through the road surface bump ahead.

[0044] S340. If it is determined that any of the surrounding vehicles has experienced a sudden attitude change event, then based on the driving control parameters associated with the attitude change event, control the vehicle to pass through the road ahead.

[0045] In one possible implementation, the vehicle dynamic analysis module detects and tracks vehicles in the image sequence, extracts key points, and performs time-series attitude analysis. Once it determines that a vehicle has experienced a sudden attitude change due to a road surface bump, i.e., an attitude change event, the module transmits the attitude change event data to the multi-source fusion and map building module. The coordinate transformation and spatial alignment submodule converts the pixel coordinates corresponding to the attitude change event to three-dimensional spatial coordinates and performs spatial alignment. Then, the confidence assessment and fusion submodule performs confidence-weighted verification by combining visual attitude data, radar or lidar perception data, and the vehicle's IMU and wheel speed sensor status information to eliminate false judgments caused by non-road surface bumps, lane changes, and other factors. When the multi-source fusion verification confirms that any vehicle has actually experienced an attitude change event, the subsequent vehicle decision-making and control process is immediately triggered, providing an accurate triggering basis for the vehicle to intervene in road traffic control in advance.

[0046] When a sudden attitude change event is confirmed for any surrounding vehicle, the decision-making and control module retrieves information such as the spatial location and distribution range of road surface protrusions and obstacles marked in the local dynamic map storage submodule. Combined with the current vehicle speed and driving posture information obtained by the environmental perception module, the internal safe speed calculation submodule calculates the optimal safe speed threshold for the vehicle to pass through the obstacle section based on preset safe passage rules. Then, the deceleration strategy and trajectory planning submodule formulates a graded deceleration strategy, a smooth braking control curve, and an adapted driving trajectory planning scheme based on the safe speed threshold. Finally, these are integrated to form driving control parameters directly related to the attitude change event. These driving control parameters include at least higher-level control quantities such as vehicle speed control parameters and driving trajectory constraint parameters, providing standardized and executable parameter basis for subsequent vehicle control actions. Finally, the decision-making and control module generates corresponding vehicle control commands based on the generated driving control parameters and sends these commands to the vehicle's underlying actuators. Simultaneously, through the internal human-machine interaction prompt submodule, it outputs interactive information such as road bump warnings and vehicle deceleration prompts to the driver. The vehicle actuators adjust the power output, braking force, and driving trajectory in real time according to the vehicle control commands, enabling the vehicle to smoothly pass through the road surface with protruding obstacles along the planned safe trajectory and at the target speed. This avoids problems such as chassis bottoming out, body collisions, and bumpy rides caused by high-speed passage. It not only makes up for the technical defects of traditional visual direct perception in being unable to identify hidden road bumps, but also ensures the driving safety and ride comfort of intelligent driving through proactive control in advance. It realizes indirect road anomaly perception and adaptive vehicle passage control based on sudden changes in the posture of surrounding vehicles.

[0047] It should be noted that the attitude change event is not a normal vehicle attitude fluctuation, but rather an abnormal attitude event caused by the sudden change in road surface height, such as bumps or speed bumps, leading to vertical bouncing and pitch changes in the vehicle body. This attitude change event is directly equivalent to the presence of a raised obstacle on the road ahead that is difficult to identify with traditional vision. This compensates for the failure of direct perception solutions to identify obstacles with weak visual features, providing an accurate trigger signal for the vehicle to perceive road risks in advance. Furthermore, this attitude change event has undergone coordinate transformation, spatial alignment, and confidence assessment through multi-source fusion and map building modules, and is consistent with the actual situation of the raised obstacle on the road. The system deeply binds spatial and temporal information such as the vehicle's location and distance from the vehicle itself. The driving control parameters generated based on this binding relationship are targeted control parameters obtained by combining the spatial characteristics of obstacles and the vehicle's current driving state, through safe speed calculation and deceleration strategy planning, rather than general control parameters. They have a high degree of adaptability to road risks. Therefore, the system can directly implement control operations such as speed adjustment and trajectory optimization based on these driving control parameters that are strongly correlated with attitude change events. This allows the vehicle to pass through the road surface with protruding obstacles in advance in a safe and smooth state, effectively avoiding problems such as chassis collisions and ride bumps, and realizing intelligent traffic control based on indirect perception.

[0048] The vehicle control method provided in this application acquires image sequences of at least one surrounding vehicle in front of the vehicle in the current lane and adjacent lanes during its movement. This eliminates the direct reliance on the visual features of road surface protrusions and obstacles found in traditional methods. Based on these image sequences, the method detects the vehicle attitude features of each surrounding vehicle. These vehicle attitude features are specifically designed to characterize the relative vertical displacement of the wheels and body of the surrounding vehicles, and / or the pitch attitude of the vehicle body. This allows for the description of the inherent attitude responses of surrounding vehicles when passing through the road surface from dimensions such as vehicle suspension deformation and body sway, providing standardized and quantifiable feature support for the indirect perception of road anomalies. Furthermore, based on the extracted surrounding vehicle attitude features, the method determines whether each surrounding vehicle has experienced a sudden attitude change event caused by passing through the road surface ahead. This is achieved by utilizing the relative vertical displacement of the wheels and body of the surrounding vehicles when they pass through an abnormal road surface area, and the pitch attitude of the vehicle body. The system utilizes realistic physical responses to sudden changes in vehicle posture to indirectly perceive hidden protruding obstacles. This effectively addresses the technical challenge of visually identifying and segmenting obstacles with missing visual features, such as pure concrete speed bumps and faded speed bumps, or those similar in color to the road surface. Furthermore, it can predict the presence of protruding anomalies on the road surface ahead from a safe distance based on sudden changes in the posture of neighboring vehicles. This overcomes the limitations of traditional systems that cannot directly perceive road anomalies or predict risks in advance. Subsequently, upon detecting a sudden change in the posture of any surrounding vehicle, the system immediately matches the driving control parameters associated with that event to implement advance passage control for the vehicle. This prevents the vehicle from passing over abnormal road surfaces at high speeds, avoiding issues like chassis bottoming out or body collisions. It protects vehicle components from damage, improves driving comfort, and eliminates the safety hazards caused by the failure to perceive road protruding obstacles during intelligent driving.

[0049] Figure 4 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 2 .like Figure 4 As shown, the above method detects the vehicle pose features of each surrounding vehicle based on the image sequence, including: S410. Perform vehicle detection on the image sequence to identify at least one surrounding vehicle.

[0050] In one possible implementation, vehicle detection can identify vehicle targets from complex road image scenes. Combined with preset range rules to lock onto surrounding vehicles in the current lane and adjacent lanes, it effectively eliminates interference from non-target objects such as oncoming vehicles, vehicles behind, and stationary vehicles on the roadside. This significantly reduces the data processing load for subsequent attitude reference point detection and attitude feature calculation, improving overall analysis efficiency and real-time performance. Therefore, the vehicle detection and tracking submodule first receives a continuous temporal image sequence collected and transmitted by the environmental perception module. Based on a visual target detection algorithm, it performs global vehicle target recognition on each frame of the image sequence, extracting the appearance, contour, and position features of vehicle targets in the image to complete the initial localization of all vehicle targets within the frame. On this basis, the vehicle detection and tracking submodule filters targets according to preset range rules for surrounding vehicles, retaining only those vehicles traveling within the current lane and adjacent lanes and whose spatial position is within the vehicle's driving range. The system identifies at least one surrounding vehicle from the image sequence as a reference for road anomaly perception, eliminating invalid targets such as oncoming vehicles, vehicles behind the vehicle, and other unrelated vehicles. This ensures that at least one surrounding vehicle provides a reliable road driving reference. The actual driving response of this vehicle on the road ahead replaces the traditional method of direct visual recognition of road obstacles, fundamentally solving the technical challenge of directly detecting obstacles with weak visual features, such as wear-resistant speed bumps and concrete protrusions. Furthermore, the clearly defined surrounding vehicle targets delineate the analysis area for subsequent preset attitude reference point detection, pixel coordinate extraction, and vehicle attitude feature calculation, avoiding computational redundancy and feature mis-extraction caused by targetless full-domain analysis. This ensures the accuracy and stability of subsequent vehicle attitude change judgment. Simultaneously, the detection results of multiple surrounding vehicles can form cross-validation, further improving the confidence of road anomaly perception and providing a solid and reliable preliminary basis for subsequent safe passage control of the vehicle.

[0051] S420: Detects the pixel coordinates of multiple preset attitude reference points on each surrounding vehicle.

[0052] The preset attitude reference points are used to measure the relative vertical displacement between the wheels and the vehicle body of surrounding vehicles. These primarily select points on the vehicle body that directly reflect suspension extension and contraction, such as the wheel center, wheel arch, and suspension tower top. Changes in the pixel coordinates of these reference points quantify the vertical displacement of the wheels relative to the vehicle body. The preset attitude reference points are used to measure the vehicle's pitch attitude. These primarily select points that visually represent pitch changes, such as the roof fins and roof edges. The relative offset of the pixel coordinates of these reference points reflects changes in the vehicle's pitch angle. Therefore, these two types of reference points can be used individually or in combination, comprehensively covering the vertical and pitch motion characteristics of a vehicle passing over road bumps. This provides a standardized and effective detection benchmark for calculating vehicle attitude characteristics and reliably determining attitude change events.

[0053] In one possible implementation, the difference in vehicle posture between smooth driving and traversing road bumps cannot be directly represented by a single feature point. It requires the relative positional changes of feature points across different parts of the vehicle body for objective characterization. Multiple preset posture reference points correspond to two key measurement objects: the relative vertical displacement between the wheels and the vehicle body, and the vehicle's pitch posture. This covers the core motion patterns such as vertical bounce and pitch generated when the vehicle passes over bumps. Therefore, after the vehicle detection and tracking submodule identifies and locates surrounding vehicles, the key point detection submodule uses the target area corresponding to each surrounding vehicle in the image as its processing range. Based on a visual key point detection algorithm, it identifies and locates feature points in this local area, sequentially extracting multiple preset posture reference points on each surrounding vehicle. Using the image's built-in pixel coordinate system, it calculates and determines the horizontal and vertical pixel positions of each posture reference point in the corresponding image frame. Finally, it outputs the pixel coordinate set of all preset posture reference points on each surrounding vehicle. This transforms the vehicle's actual physical posture changes into calculable and comparable quantitative data within the image domain, providing a unique and direct data foundation for subsequent calculations of vehicle posture features such as suspension compression and pitch angular velocity. Meanwhile, by using multiple preset posture reference points for collaborative detection, the posture judgment error caused by the occlusion of a single feature point, detection noise, or image distortion can be effectively avoided, thereby improving the stability and reliability of vehicle posture feature extraction. This provides solid data support for determining whether surrounding vehicles have experienced a sudden posture change event caused by road surface protrusion, ensuring the accuracy of the indirect perception of hidden road obstacles in this application.

[0054] S430: Calculate the vehicle posture features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of multiple preset posture reference points.

[0055] In one possible approach, since the change in vehicle posture is essentially a relative movement between different structural points of the vehicle body, rather than an absolute change in the position of a single point, and using relative positional relationships can effectively eliminate external interference and truly reflect the physical posture changes of the vehicle caused by road conditions. On the one hand, the suspension compression and pitch angle of a vehicle when it passes over a road bump are manifested as relative displacements or angular changes between key points such as the wheels and wheel arches, and the front and rear of the roof, rather than the absolute pixel coordinates of these points in the image. Therefore, by calculating the relative distances and offsets between them, the vehicle's actual vertical bounce and pitch motion can be directly correlated and quantified. On the other hand, interference factors such as vehicle movement, camera viewpoint shifts, and changes in vehicle distance during image acquisition can cause meaningless fluctuations in the absolute pixel coordinates of reference points. However, relative positional relationships are not affected by such global displacements, exhibiting stronger anti-interference capabilities and stability. Furthermore, by combining multiple sets of relative positional relationships between reference points used to measure the vertical displacement of the wheels and the vehicle body, and reference points used to measure the vehicle's pitch attitude, standardized vehicle attitude features such as suspension compression and pitch angular velocity can be calculated. This provides an objective and reliable quantitative basis for subsequently determining whether the vehicle has experienced a sudden change in attitude due to a road bump. Therefore, after obtaining the pixel coordinates of multiple preset attitude reference points corresponding to each surrounding vehicle from the key point detection submodule, the attitude temporal analysis submodule uses the relative positional relationship of each preset attitude reference point in the image coordinate system as the basis for calculation to perform quantitative calculation of vehicle attitude features. During the calculation process, the attitude temporal analysis submodule first groups the preset attitude reference points of different functional types. For reference points used to measure the relative vertical displacement between the wheel and the vehicle body, it calculates the longitudinal pixel relative distance between the wheel-related reference points and vehicle body reference points such as the wheel arch and suspension tower. For reference points used to measure the vehicle body pitch attitude, it calculates the relative offset and positional change trend of pixel coordinates between points such as the roof shark fin and the roof edge. Then, through image space mapping, it converts the above pixel-level relative positional changes into physical motion features, ultimately calculating vehicle attitude features that objectively reflect the vehicle's driving state. These vehicle attitude features specifically include: the vertical displacement of the wheel center relative to the wheel arch (i.e., suspension compression), the vehicle body pitch angular velocity, and other feature quantities characterizing the vehicle's vertical and pitch motion. Therefore, by using relative positional relationships instead of absolute pixel coordinates for calculation, irrelevant interference such as vehicle movement and camera angle changes can be effectively eliminated, so that the vehicle's posture characteristics can truly reflect the posture changes of surrounding vehicles caused by road conditions, providing a stable and reliable quantitative data foundation for subsequent determination of whether a vehicle has experienced a sudden posture change event.

[0056] The vehicle control method provided in this application performs vehicle detection processing on the image sequence acquired by the environmental perception module, which can identify at least one surrounding vehicle in front of the vehicle's current lane and adjacent lanes, thereby effectively eliminating non-target interference and locking in the effective analysis object. On this basis, it further detects the pixel coordinates of multiple preset attitude reference points on each surrounding vehicle. The preset attitude reference points include reference points for measuring the relative vertical displacement between the wheels and the vehicle body of the surrounding vehicles, as well as reference points for measuring the vehicle body's pitch attitude, which can comprehensively cover the key feature dimensions characterizing the vertical motion and pitch change of the vehicle. Subsequently, based on the relative positional relationship between the pixel coordinates of the multiple preset attitude reference points, geometric calculations are performed to obtain the vehicle attitude features of each surrounding vehicle, realizing the transformation of the vehicle body's dynamic attitude change into quantifiable feature data, providing stable data support for subsequent reliable determination of whether the vehicle has experienced a sudden attitude change caused by road surface protrusions or obstacles.

[0057] Optionally, the above method performs vehicle detection on the image sequence to determine at least one surrounding vehicle, including: Vehicle detection is performed on the image sequence to identify at least one tracked vehicle.

[0058] In one possible implementation, since the image sequence is a continuous multi-frame temporal visual data, it can record the dynamic driving state and continuous position changes of vehicles in the road environment. By performing vehicle detection processing on the image sequence, the vehicle target in each frame can be identified based on the visual features such as the vehicle's contour and texture. Then, combined with the target tracking algorithm, feature matching, trajectory association and unique identifier binding are performed on the same vehicle in the continuous frames. At the same time, the effective vehicles in front of the current lane and adjacent lanes of the vehicle are filtered out according to preset rules. In actual road driving scenarios, there are usually detectable and identifiable vehicles in front. Therefore, by performing vehicle detection and tracking processing on the image sequence, at least one tracked vehicle with a continuous tracking trajectory can be stably obtained, providing a reliable analysis carrier for subsequent vehicle posture feature extraction and posture change determination. Based on this, the vehicle detection and tracking submodule first receives continuous temporal image sequences captured and transmitted by the forward-looking high-definition camera, side-looking or surround-view camera in the environmental perception module. Based on the visual target detection algorithm, it performs global vehicle target recognition on each frame of the image sequence. By extracting the contour features, texture features and structural features of the vehicles, it locates the position areas of all vehicle targets in the image. After completing the initial vehicle detection, the vehicle detection and tracking submodule filters according to preset target selection conditions, retaining only valid vehicles located in the current lane and adjacent lanes and in front of the vehicle, and removing oncoming vehicles, vehicles behind the vehicle, stationary vehicles on the roadside and other unrelated vehicles. Subsequently, the target tracking algorithm performs trajectory association, position tracking and unique identification binding on the selected valid vehicles in continuous multi-frame images to achieve continuous and stable tracking of the target vehicles. Finally, at least one tracked vehicle with a continuous tracking trajectory is obtained from the image sequence processing results, providing a fixed, continuous and effective analysis object for subsequent attitude reference point detection and attitude feature calculation, ensuring the coherence and reliability of subsequent attitude temporal analysis and attitude change event judgment.

[0059] Lane line recognition is performed on the image sequence to determine the areas of the vehicle's lane and adjacent lanes in the image.

[0060] In one possible implementation, the image sequence is acquired in real time by an onboard camera, recording visual information about the road surface, lane lines, and surrounding environment. Lane lines typically possess visual features such as high contrast, continuous linearity, and regular geometric shapes, making them stable for extraction and recognition through image processing algorithms. Secondly, after lane line recognition, the position, direction, and boundary contours of each lane line can be fitted based on the image coordinate system. Combining this with the imaging pattern that the vehicle is usually located in the lower center of the image, the image area enclosed by the two nearest lane lines on the left and right sides is directly defined as the image area corresponding to the vehicle's lane, using the vehicle's position as a reference. Simultaneously, the image area adjacent to the outer side of the vehicle's lane, separated by adjacent lane lines, is the corresponding area of ​​the adjacent lane in the image. Thus, this method maps the lane spatial relationships of the real road onto the image plane, providing clear spatial constraints for subsequent selection of the vehicle's lane and vehicles in adjacent lanes, while effectively eliminating interference from non-target areas such as oncoming lanes and non-motorized vehicle lanes, ensuring the effectiveness of subsequent vehicle tracking and attitude analysis. Based on this, the vehicle detection and tracking submodule in the adjacent vehicle dynamic analysis module works in conjunction with the environmental perception module to achieve this. After acquiring the road image sequence collected by the environmental perception module, the vehicle detection and tracking submodule extracts lane line features from each frame of the image based on a visual lane line detection algorithm. By identifying the edge features, grayscale gradient features, and linear structure features of the lane lines in the image, it locates the pixel positions of the lane lines within the image and performs curve fitting on the discrete lane line points to form a continuous lane line contour. Based on the lane line recognition, the vehicle detection and tracking submodule uses the vehicle's driving position in the image as a reference and divides the image region according to the fitted lane line boundaries. The region between the nearest lane lines on both sides of the vehicle is determined as the corresponding region of the vehicle's lane in the image. The outer lane region adjacent to the vehicle's lane and located in the same driving direction is determined as the corresponding region of the adjacent lane in the image. This delineates the image range boundaries between the vehicle's lane and adjacent lanes, providing a clear spatial basis for subsequent screening of vehicles in the effective lane and eliminating interference from vehicles in non-target lanes, ensuring the targeting and effectiveness of subsequent vehicle tracking and attitude analysis.

[0061] If it is determined that the position of at least one tracked vehicle in the image sequence falls within the area of ​​the vehicle's lane and adjacent lanes in the image, then at least one tracked vehicle is identified as a surrounding vehicle.

[0062] In one possible implementation, since the area represents the current and future path of the vehicle, changes in vehicle attitude within this path range can accurately and directly reflect the actual road conditions ahead of the vehicle, serving as a valid reference for determining whether there are any protruding obstacles on the road. Vehicles outside this area are unrelated to the vehicle's path, and their attitude fluctuations cannot characterize the state of the road surface the vehicle is about to pass through. By filtering out surrounding vehicles through this position determination rule, valid analysis targets can be identified, while interference from vehicles in unrelated lanes can be eliminated, ensuring the accuracy and effectiveness of subsequent vehicle attitude analysis and attitude change event determination. Based on this, after completing the localization and tracking of the tracked vehicle in the image sequence and the division of the image regions of the vehicle's lane and adjacent lanes, the vehicle detection and tracking submodule first retrieves the pixel coordinate range of the target detection box corresponding to at least one tracked vehicle in each frame of the image to represent the real-time position of the tracked vehicle in the image sequence. Then, it performs point-by-point comparison and spatial matching of the pixel coordinate range of the vehicle with the region boundaries of the vehicle's lane and adjacent lanes in the image previously determined by lane line recognition to determine whether the overall position or detection area of ​​the tracked vehicle falls within the image region of the aforementioned effective lane. When the determination result is yes, the vehicle detection and tracking submodule formally identifies this tracked vehicle that meets the lane position condition as the surrounding vehicle referred to in this application, thereby completing the final screening and confirmation of the effective analysis target. Only vehicles that are highly related to the vehicle's driving path and can truly reflect the road conditions ahead are included in the subsequent attitude analysis process, effectively eliminating the interference caused by vehicles in irrelevant lanes and ensuring the relevance and reliability of subsequent attitude feature calculation and attitude change event determination.

[0063] The vehicle control method provided in this application, based on a continuous temporal image sequence transmitted by an environmental perception module, simultaneously performs two preprocessing operations: vehicle detection and tracking, and lane line recognition. This provides dual data support for selecting effective analysis targets: First, vehicle detection is performed on the image sequence. A visual target detection algorithm extracts features such as the appearance and contour of vehicles in the image to achieve initial global target localization. Then, a trajectory tracking algorithm continuously tracks and uniquely identifies the selected potential targets, ultimately obtaining at least one tracked vehicle. This operation can stably lock continuously analyzable vehicle targets in dynamic road scenes, defining fixed analysis objects for subsequent region matching and ensuring the continuity and coherence of the analysis process. Second, lane line recognition is performed on the image sequence. Based on feature extraction and curve fitting of lane lines, such as edge gradients and linear structures, the position and direction of lane lines in the image are located. Then, the vehicle's lane and adjacent lanes are divided based on the vehicle's imaging position. In the corresponding area of ​​the image, this step can clearly define the effective spatial range for analysis, eliminating interference from non-target areas such as oncoming lanes and non-driving lanes at the image level, providing clear spatial constraints for target selection. On this basis, the side vehicle dynamic analysis module spatially matches the pixel coordinates of at least one tracked vehicle in the image with the image area boundaries of the vehicle's lane and adjacent lanes. When it is confirmed that the overall position or detection area of ​​the tracked vehicle falls into the above-mentioned effective area, the tracked vehicle is determined to be the surrounding vehicle referred to in this application. Through the progressive processing logic of target locking, range definition and matching, surrounding vehicles that are highly related to the vehicle's driving path and can truly reflect the road conditions ahead are selected, effectively eliminating interference from unrelated vehicles, and greatly improving the pertinence and accuracy of subsequent attitude reference point detection, attitude feature calculation and attitude change event judgment, laying the target foundation for indirect perception of road protrusion obstacles based on the dynamic response of surrounding vehicles.

[0064] Optionally, vehicle attitude characteristics include: suspension compression characteristics.

[0065] Among them, the suspension compression feature is a quantity calculated based on the image pixel coordinates, which is the change in the vertical distance between the wheels of surrounding vehicles and the wheel arches of the vehicle body, to reflect the compression and rebound amplitude of the suspension when the vehicle is in motion.

[0066] The above method calculates the vehicle pose features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of multiple preset pose reference points, including: For each surrounding vehicle, obtain the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel from multiple preset attitude reference points, and the pixel coordinates of the wheel arch point corresponding to the same wheel.

[0067] In one possible implementation, since the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel among multiple preset attitude reference points, and the pixel coordinates of the wheel arch point corresponding to the same wheel are both paired feature points that directly reflect the changes in suspension extension and contraction, the relative difference between their pixel coordinates can quantify the vertical displacement between the wheel and the vehicle body. This intuitively reflects the suspension compression or extension state when the vehicle passes over a road bump, thus objectively characterizing the vehicle's vertical jumping attitude. At the same time, by limiting the reference points to the same wheel, calculation deviations caused by cross-wheel and cross-vehicle points can be avoided, ensuring the accuracy of subsequent vehicle attitude feature calculations. This provides a core and reliable quantitative basis for determining whether the vehicle has experienced a sudden attitude change event caused by a road bump. Based on this, the key point detection submodule in the vehicle dynamic analysis module is implemented independently for each screened and confirmed surrounding vehicle. After completing the detection and pixel coordinate extraction of all preset attitude reference points for a single surrounding vehicle, the key point detection submodule classifies and matches all preset attitude reference points according to the wheel dimension based on the preset vehicle body structure association logic, filters out the wheel contact point or wheel center point corresponding to the same wheel, and locks the wheel arch point that corresponds one-to-one with the wheel in physical structure, while eliminating other wheel-related reference points and non-wheel attitude reference points. On this basis, the key point detection submodule extracts and stores the pixel coordinates of the matched wheel contact point or wheel center point and the pixel coordinates of the wheel arch point corresponding to the same wheel from the acquired global attitude reference point pixel coordinate data, forming two sets of coordinate data pairs with direct correlation under the same wheel. This provides the corresponding original coordinate basis for subsequent calculation of the vertical displacement of the wheel relative to the vehicle body and quantification of the suspension compression, a core vehicle attitude feature, ensuring the pertinence and physical rationality of the attitude feature calculation.

[0068] The initial compressed pixel value is obtained by calculating the difference between the ordinate of the pixel coordinates of the wheel contact point or wheel center point and the ordinate of the pixel coordinates of the wheel arch point.

[0069] In one possible implementation, in the image pixel coordinate system, the vertical coordinate corresponds to the vertical height direction of the road scene. The difference in the vertical coordinate between the wheel contact point or wheel center point and the corresponding wheel arch point of the same wheel can directly reflect the vertical displacement change of the wheel relative to the vehicle body, that is, the degree of compression or extension of the vehicle suspension. By calculating this vertical coordinate difference, the initial compressed pixel value can be obtained, and the suspension deformation of the vehicle in physical space can be converted into a quantifiable and comparable pixel value in the image domain. At the same time, calculating only the vertical difference can effectively eliminate irrelevant interference from the lateral pixel displacement, allowing the value to purely represent the vertical jumping posture of the vehicle, providing a key basic quantitative basis for subsequent judgment on whether the surrounding vehicles have undergone a sudden change in posture due to running over the road surface protrusion. Based on this, the attitude timing analysis submodule in the adjacent vehicle dynamic analysis module, after obtaining the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel of each surrounding vehicle, and the pixel coordinates of the wheel arch point corresponding to the same wheel, extracts the vertical pixel coordinates (ordinate) from the wheel contact point or wheel center point pixel coordinates and the wheel arch point pixel coordinates according to the coordinate rules of the image pixel coordinate system. By performing a numerical difference operation on the two sets of ordinates, the pixel offset difference between the two in the vertical dimension of the image is obtained, and the calculated pixel-level difference result is determined as the initial compressed pixel value. In this way, the relative vertical displacement between the wheel and the wheel arch is converted into a quantifiable pixel feature value, providing basic quantitative data for further analysis of the vehicle suspension compression degree and judgment of whether the vehicle has experienced a sudden attitude change caused by road bumps.

[0070] Based on the preset size of each surrounding vehicle in the image sequence, the initial compressed pixel values ​​are normalized to obtain the suspension compression feature of each surrounding vehicle.

[0071] In one possible implementation, the varying distances between surrounding vehicles and the vehicle's camera lead to significant differences in their image size. Directly using initial compressed pixel values ​​cannot objectively and uniformly reflect the true degree of suspension compression. However, a preset size can characterize the vehicle's imaging proportion in the image. Normalization can eliminate computational biases caused by differences in vehicle distance and imaging scale, uniformly converting pixel-level compression amounts under different operating conditions into standardized relative features. This results in stable, comparable, and accurate suspension compression feature quantities that truly reflect the degree of vehicle suspension deformation, ensuring that subsequent judgments of attitude change events are not affected by the vehicle's imaging size, thus improving the accuracy and versatility of recognition. Based on this, the attitude temporal analysis submodule first determines the preset size of each surrounding vehicle in the current image frame based on visual information such as the height of the outer detection box and the pixel span of the wheel area obtained by vehicle detection in the image sequence. This preset size is used to characterize the actual imaging size of the vehicle in the image, thereby eliminating pixel value deviations caused by different distances and imaging scales between surrounding vehicles and the vehicle-mounted camera. Subsequently, the initial compressed pixel value calculated in the previous step is normalized with the preset size of the surrounding vehicle, that is, the initial compressed pixel value is divided by the corresponding preset size, completing the standardization calibration of the suspension compression pixel differences under different imaging scales. The dimensionless feature value obtained after the above scale normalization process is the suspension compression feature of each surrounding vehicle. This feature can uniformly and objectively reflect the degree of suspension compression of different vehicles under the same evaluation criteria, providing a stable and comparable quantitative basis for accurately determining whether the vehicle has experienced a sudden change in attitude caused by road bumps or obstacles.

[0072] For example, when performing attitude analysis on vehicles tracking the road, a temporal queue of 10-30 frames is first established for each vehicle to continuously record the pixel coordinates of key points. Then, based on a pre-defined lightweight deep learning key point detection model (such as HRNet or ViT), the pixel coordinates of the front wheel center point Pw=(uw,vw) and the front wheel arch point Pf=(uf,vf) are detected and output for each frame of the tracked vehicle. The front suspension is the spring and shock absorber connection structure between the two front wheels and the vehicle body. Its compression state can be reflected by the positional changes of the wheel arch and wheel center. The compression amount C of the front suspension in a single frame is then calculated as the difference between the y-coordinate of the wheel arch point and the y-coordinate of the wheel center point (i.e., C=vf). vw can also be used. (vf represents the upward compression displacement of the wheel). To eliminate scale differences caused by the distance between different vehicles and the camera, the image height or width corresponding to the vehicle detection box is selected as a normalization factor. The vertical pixel distance from the wheel arch to the center of the wheel under unloaded, stationary, and normal posture conditions is used as a reference static value to calculate the dimensionless normalized front suspension compression ratio Rc. This compression ratio is the ratio of the difference between the y-coordinate of the wheel arch point and the y-coordinate of the wheel center point to the reference static value. If the ratio is ≈1, it means that the suspension is in a normal uncompressed state. If the ratio is <1, it means that the wheel arch has moved down and the suspension is compressed. The smaller the ratio, the more severe the compression. A sudden increase in the compression ratio indicates that the front wheel of the vehicle has run over a protruding obstacle on the road.

[0073] The vehicle control method provided in this application includes at least a suspension compression feature quantity as the vehicle attitude characteristics. The suspension compression feature quantity is a quantitative index calculated based on image pixel coordinates and used to characterize the change in the vertical distance between the wheels of surrounding vehicles and the wheel arches of the vehicle body. It can intuitively reflect the real-time extension and contraction deformation state of the suspension when the vehicle drives over a road bump. For each surrounding vehicle, the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel and the pixel coordinates of the corresponding wheel arch point are extracted from multiple preset attitude reference points. The initial compressed pixel value is obtained by calculating the difference between the vertical coordinates of the two in the pixel coordinates, thereby converting the change in the physical vertical distance between the wheel and the wheel arch into a quantifiable value in the image domain. Then, the initial compressed pixel value is normalized based on the preset size of each surrounding vehicle in the image sequence, effectively eliminating the calculation deviation caused by the different distances between the vehicle and the camera and the difference in imaging size. Finally, a stable suspension compression feature quantity for each surrounding vehicle is obtained, providing an objective and unified quantitative data basis for subsequent reliable determination of whether the vehicle has experienced a sudden attitude change caused by a road bump.

[0074] Optionally, vehicle attitude characteristics include: vehicle pitch rate.

[0075] The vehicle pitch rate is the magnitude of the rotational angular velocity of each surrounding vehicle about its lateral axis.

[0076] The above method calculates the vehicle pose features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of multiple preset pose reference points, including: For each surrounding vehicle, obtain the pixel coordinates of the first reference point located in the front area of ​​the roof and the pixel coordinates of the second reference point located in the rear area of ​​the roof from multiple preset attitude reference points.

[0077] In one possible implementation, since a vehicle experiences significant pitch motion when passing over road bumps, the reference points at the front and rear of the roof are paired feature points that can intuitively represent the vehicle's pitch attitude. The relative changes in their pixel coordinates can directly quantify the vehicle's pitch amplitude and angular velocity, thus supplementing the key attitude dimension beyond the vertical displacement of the wheels. This provides a more comprehensive and realistic reflection of the sudden attitude changes caused by road anomalies. At the same time, the feature points in the roof area are prominent and not easily obscured, resulting in higher detection stability. By obtaining this set of paired coordinates, a basis can be provided for subsequent calculation of pitch attitude features, complementing the feature data of wheel-related reference points. This effectively improves the accuracy and reliability of attitude change event judgment and avoids misjudgment problems caused by single attitude dimension detection. Based on this, after the key point detection submodule completes the detection and pixel coordinate extraction of all preset attitude reference points of a single surrounding vehicle, it locates and extracts the first reference point located in the front area of ​​the vehicle roof and the second reference point located in the rear area of ​​the vehicle roof from multiple preset attitude reference points according to the preset vehicle body area division rules, and removes attitude reference points in other non-roof areas such as wheels and vehicle sides. Subsequently, the key point detection submodule acquires and records the pixel coordinate information corresponding to the first reference point and the second reference point in the image pixel coordinate system, forming two sets of paired coordinate data to represent the vehicle pitch attitude, providing the corresponding original coordinate basis for subsequent calculation of the vehicle pitch angle by the relative change of the coordinates of the two points and for determining whether the vehicle has a pitch attitude change due to road surface protrusion.

[0078] The pitch indicator within a frame of an image is calculated based on the ordinate of the pixel coordinates of the first and second reference points.

[0079] Among them, the pitch indicator is used to characterize the pixel change in the pitch state of surrounding vehicles.

[0080] In one possible implementation, since the vertical coordinate of the image corresponds to the vertical direction in real space, when the vehicle pitches, the front and rear of the roof will produce opposite or different vertical displacements, which is directly reflected in the regular change of the relative difference between the two points. By calculating the relative difference or proportional relationship of the vertical pixel coordinates, the pitch angle change of the vehicle body can be converted into a quantitative value. Moreover, using only the vertical coordinate can eliminate lateral position interference and focus on the vertical attitude shift, thereby obtaining a pitch indicator that can intuitively reflect the pitch degree and direction of the vehicle in a single frame image, providing a reliable quantitative basis for determining whether the vehicle has a pitch-like attitude change due to road bumps. Based on this, after obtaining the pixel coordinates of the first reference point in the front area of ​​the roof and the second reference point in the rear area of ​​the roof of the same surrounding vehicle, the attitude timing analysis submodule extracts the corresponding vertical pixel coordinates (i.e., ordinates) of the two reference points in the image pixel coordinate system. By calculating the numerical difference between the two ordinates, the relative pixel offset of the front and rear parts of the roof in the vertical dimension of the image is converted into a quantized value. This quantized value is determined as the pitch indicator in the current single frame image, so that it can objectively reflect the change of the vehicle's pitch attitude at that frame moment. This provides a directly comparable quantitative feature basis for subsequent judgment of whether the vehicle has a sudden pitch attitude change due to road bumps.

[0081] Based on the difference in pitch indication between preset consecutive image frames and the time interval between preset consecutive image frames, the pitch angular velocity of each surrounding vehicle is calculated.

[0082] In one possible implementation, since the physical definition of angular velocity is the change in angle per unit time, the vehicle pitch angular velocity is used to measure the rate of change in vehicle pitch attitude. The pitch indicator is a characteristic value that can quantify the degree of vehicle pitch. The difference in pitch indicators between consecutive image frames directly corresponds to the actual change in vehicle pitch angle within adjacent frames, reflecting the magnitude of pitch attitude change. The time interval between consecutive image frames is fixed by the camera's frame rate, representing the length of time the pitch change has taken. By calculating the two using the physical formula that angular velocity is the change in angle divided by time, the vehicle pitch angular velocity can be objectively obtained. This quantifies the rate of vehicle pitch change, thus distinguishing between slow attitude changes during smooth driving and sudden, violent pitch jumps caused by passing road bumps, providing crucial dynamic attitude information for determining attitude change events. Based on this, the attitude timing analysis submodule first obtains the pitch indication corresponding to each preset continuous image frame based on the pixel coordinates of the first reference point at the front of the roof and the second reference point at the rear of the roof of each surrounding vehicle. Then, it calculates the numerical difference between the pitch indications of adjacent preset continuous image frames to characterize the change range of the vehicle body pitch attitude between frames. Subsequently, the attitude timing analysis submodule determines the fixed time interval between preset continuous image frames based on the actual acquisition frame rate of the vehicle camera. Finally, the difference in pitch indication and the time interval are substituted into the following formula (1) to calculate the vehicle body pitch angular velocity of each surrounding vehicle.

[0083] Vehicle pitch rate = Difference in pitch indication between preset consecutive image frames ÷ Time interval between preset consecutive image frames Formula (1) Based on the above formula (1), the dynamic changes in vehicle pitch are transformed into quantifiable rate characteristics, providing a dynamic characteristic basis for judging whether the vehicle experiences sudden pitch-like attitude changes due to rolling over road bumps.

[0084] For example, in the process of vehicle pitch attitude analysis, key point detection is first carried out. The midpoints of the front and rear door handles are selected as reference points, and the front point of the roof is detected and obtained respectively. Prf =( urf , vrf (i.e., the midpoint of the front door handle) and the rear point of the roof. Prr =( urr , vrr The pixel coordinates of the center point of the rear door handle are determined; then, a single-frame pitch angle estimation is performed, and the vertical pixel difference between the front and rear reference points (i.e., Δv=vrf) is calculated. In scenarios where the distance between the vehicle and the vehicle changes slowly within a short, continuous frame, the vertical pixel difference is directly defined as the pitch indicator (i.e., P=Δv), which serves as a surrogate variable for the vehicle's pitch state. Finally, the pitch angular velocity is calculated, and the pitch indicator P is temporally differentiald to obtain the inter-frame rate of change (i.e., P(t)-P(t-1)), which is then divided by the frame time interval Δv. t (seconds) This gives the vehicle body pitch rate in pixels per second. ωp To convert to the actual angular velocity in radians per second, multiply by the calibration coefficient obtained from the actual vehicle experiment fitting. α When simplifying applications, the normalized pixel change rate is used directly and an empirical threshold is configured to complete the determination.

[0085] The vehicle control method provided in this application includes at least the vehicle pitch angular velocity as a vehicle attitude feature, where the vehicle pitch angular velocity is the magnitude of the rotational angular velocity of each surrounding vehicle around its lateral axis. For each surrounding vehicle, the pixel coordinates of a first reference point located in the front region of the roof and a second reference point located in the rear region of the roof are first obtained from multiple preset attitude reference points. Then, based on the ordinate of the pixel coordinates of the first and second reference points, a pitch indication quantity is calculated within a single frame image to characterize the pixel change in the vehicle pitch state of the surrounding vehicles. Finally, based on the difference in pitch indication quantities between preset consecutive image frames and the inter-frame time interval, the vehicle pitch angular velocity of each surrounding vehicle is calculated, quantifying the dynamic rate of change of vehicle pitch. This method sensitively captures sudden pitch attitude changes when a vehicle passes over a road bump, providing a reliable dynamic quantitative basis for accurately determining whether a sudden attitude change event has occurred.

[0086] Figure 5 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 3 .like Figure 5 As shown, the above method, based on vehicle attitude characteristics, determines whether each surrounding vehicle has experienced a sudden attitude change event due to passing through the road surface ahead, including: S510. Perform time-series analysis on the vehicle attitude characteristics of each surrounding vehicle to obtain a time-series feature sequence.

[0087] In one possible approach, since single-frame attitude features can only reflect the instantaneous state of a vehicle at a certain moment, they cannot distinguish between slight attitude fluctuations during normal driving and sudden attitude changes caused by road bumps. They are also susceptible to misjudgments due to image noise and minor road bumps. However, by organizing the attitude features of multiple consecutive frames into a temporal feature sequence through temporal analysis, the dynamic change process and magnitude of the vehicle's attitude can be reconstructed. This allows for the capture of sudden and drastic changes in features such as suspension compression and vehicle pitch rate, effectively filtering out gradual attitude changes not caused by road bumps. Furthermore, it provides a continuous, comparable, and more interference-resistant quantitative basis for subsequent reliable determination of whether the vehicle has experienced an attitude change event caused by driving over road bumps. Based on this, the attitude temporal analysis submodule, for each screened and confirmed surrounding vehicle, first retrieves the various vehicle attitude feature data calculated frame by frame in consecutive multi-frame images, including at least: initial compressed pixel values, vehicle pitch angular velocity, and other quantitative results that can reflect vertical sway and vehicle pitch changes. Then, according to the chronological order of the images captured by the camera, the attitude feature values ​​corresponding to each frame are sequentially sorted, correlated between frames, and normalized. At the same time, discrete abnormal feature points caused by image noise, local occlusion, and other factors are removed, while continuous and effective attitude change data are retained. Finally, the ordered feature set containing multi-dimensional attitude change information, arranged in a time sequence, is used as the temporal feature sequence of the surrounding vehicle, thereby reconstructing the dynamic process of vehicle attitude change over time and providing temporal data support for subsequent identification of whether the vehicle has experienced sudden or abnormal attitude changes due to road surface protrusions.

[0088] S520. Based on the time-series feature sequence, determine whether the vehicle attitude features have undergone a sudden change that exceeds the preset attitude threshold.

[0089] The preset attitude threshold is a critical value used to distinguish between minor attitude fluctuations during normal, stable vehicle driving and abnormal attitude changes caused by road surface bumps. The preset attitude threshold can be selected based on actual conditions; a value of 0.05 is a suitable option.

[0090] In one possible implementation, since vehicle attitude changes are a continuous and dynamic physical process, the attitude feature data of a single frame image is subject to randomness and interference, and cannot objectively and accurately reflect the true amplitude and trend of attitude changes. However, a temporal feature sequence can capture the dynamic change patterns of attitude features and enable judgment. That is, vehicle attitude features (such as initial compressed pixel values, vehicle pitch angular velocity, etc.) are calculated from multiple consecutive frames of images. These attitude features of the same type and dimension are arranged according to the image acquisition sequence to construct a temporal feature sequence. This sequence can record the entire process of attitude changes of surrounding vehicles from normal driving to passing over road bumps, effectively eliminating false detections or fluctuations in attitude features caused by non-road factors such as camera noise, lighting fluctuations, and instantaneous vehicle shaking in a single frame image. Simultaneously, the preset attitude threshold is statistically calibrated through extensive road tests to distinguish between small attitude fluctuations during normal vehicle driving and large attitude changes caused by road bumps. The critical judgment value cannot be determined by the feature value of a single frame alone to determine whether the change range exceeds the preset attitude threshold. However, based on the time-series feature sequence, the actual change amount, change rate and cumulative change trend of the attitude feature between consecutive frames can be calculated. By comparing the actual change amount with the preset attitude threshold, it is possible to objectively and reliably determine whether the vehicle attitude feature has undergone a sudden change exceeding the preset attitude threshold. This fundamentally improves the accuracy of identifying sudden changes in vehicle attitude caused by road surface protrusions, avoids missed detection or false triggering of road risks due to misjudgment of single frame data, and provides accurate triggering signals for subsequent traffic control of the vehicle. Based on this, the attitude temporal analysis submodule first calculates the vehicle attitude features of the same surrounding vehicles from multiple consecutive frames of images, including initial compressed pixel values ​​and vehicle pitch angular velocity, and sorts them according to the acquisition sequence of the image frames to construct a temporal feature sequence, thereby recording the dynamic change process of vehicle attitude features over time. Then, the attitude feature values ​​in adjacent frames or within a set consecutive frame interval in the temporal feature sequence are differentially calculated to obtain the actual change amplitude of vehicle attitude features in the temporal dimension. Next, the actual change amplitude is compared with the preset attitude threshold pre-calibrated by the system. Finally, a judgment is made based on the comparison result. If the temporal change amplitude of the attitude feature is greater than the preset attitude threshold, it is determined that the vehicle attitude feature of the surrounding vehicle has undergone a sudden change exceeding the threshold. Otherwise, it is determined to be a normal driving attitude fluctuation without a sudden change in attitude, thereby identifying the abnormal bouncing state of the vehicle caused by the road surface protrusion in front.

[0091] S530. If the vehicle's posture characteristics undergo a sudden change that exceeds a preset posture threshold, then it is determined that each surrounding vehicle caused the posture change event by passing through the road surface ahead.

[0092] In one possible implementation, when a vehicle is driving normally on a smooth road, its attitude characteristics, such as suspension compression and vehicle pitch rate, only fluctuate slightly and gradually, always remaining within a preset attitude threshold range. However, when the vehicle passes over obstacles such as road bumps or speed bumps, it will experience sudden and violent vertical swaying and pitch changes, directly causing significant and abrupt changes in attitude characteristics such as initial compressed pixel values ​​and vehicle pitch rate. This preset attitude threshold serves as a critical standard for distinguishing between normal driving attitude fluctuations and abnormal attitude changes caused by road obstacles. Therefore, abrupt changes in attitude characteristics exceeding this preset attitude threshold can directly correspond to the vehicle's swaying behavior caused by uneven road surfaces. Based on this, it is possible to accurately determine the attitude change event caused by each surrounding vehicle passing through the road ahead. Based on this, after the attitude timing analysis submodule completes the numerical comparison of the temporal change amplitude of vehicle attitude features with the preset attitude threshold, if it determines that the temporal change amplitude of vehicle attitude features such as the initial compressed pixel value and the vehicle pitch angular velocity exceeds the preset attitude threshold that distinguishes between normal driving attitude fluctuations and abnormal attitude changes, it is determined that the surrounding vehicles have experienced sudden and violent attitude changes caused by road surface protrusions, thus meeting the core triggering conditions of the attitude change event.

[0093] After the above-mentioned vehicle attitude characteristics change abruptly beyond the preset attitude threshold, the attitude timing analysis submodule of the adjacent vehicle dynamic analysis module will formally confirm the surrounding vehicle that meets the conditions according to the preset event calibration rules, directly determine and confirm that the surrounding vehicle has experienced an attitude change event. This determination result will be used as a key signal that there is a hidden protruding obstacle on the road ahead, and will be transmitted to the multi-source fusion and map building module for subsequent spatial association and confidence verification. At the same time, it will provide a clear trigger basis for the decision and control module to start the autonomous vehicle safe passage control process.

[0094] For example, if the vehicle's attitude characteristics exceed the preset attitude threshold abrupt change judgment condition, and the duration meets the condition (e.g., exceeding the threshold for 2 consecutive frames), then it is determined that the vehicle's front axle is experiencing a compression event, i.e., an attitude abrupt change event has occurred. In addition, the absolute value of the pitch angular velocity is monitored simultaneously. When it exceeds the preset attitude threshold calibrated by the actual vehicle (e.g., corresponding to a 20% increase in the compression ratio threshold and a pitch angular velocity threshold set to 0.05 rad / s), and this pitch attitude change is temporally correlated with the suspension compression event (e.g., the vehicle pitches up approximately 0.2 seconds after the front wheels hit a speed bump), then it is finally determined that the vehicle has experienced a valid bounce, i.e., an attitude abrupt change event has occurred.

[0095] The vehicle control method provided in this application performs time-series analysis on the vehicle posture characteristics of each surrounding vehicle to obtain a time-series feature sequence. This captures the dynamic changes in vehicle posture characteristics with continuous image frames, effectively avoiding posture interference caused by noise in single-frame images and slight road bumps, thus improving the stability and reliability of vehicle posture feature analysis. Based on further comparative analysis of this time-series feature sequence, it can determine whether the vehicle posture characteristics have undergone abrupt changes exceeding a preset posture threshold. This clearly distinguishes between small posture fluctuations during normal and stable vehicle driving and drastic abnormal posture changes caused by driving over road bumps. If it is determined that the vehicle posture characteristics have undergone such abrupt changes exceeding the preset posture threshold, it can be determined that the surrounding vehicle's posture change event was caused by passing through the road ahead. This can accurately reflect the actual state of the road ahead having bumps or obstacles, providing a reliable triggering basis for the vehicle to detect road abnormalities in advance and implement active traffic control.

[0096] Optionally, determining whether the vehicle attitude characteristics undergo a sudden change exceeding a preset attitude threshold in the above method includes: Calculate the change in vehicle pose features in the current frame relative to the baseline value.

[0097] The baseline value is used to characterize the vehicle attitude characteristics of each surrounding vehicle before it has passed the road surface ahead. Specifically, this baseline value is obtained by the system collecting multiple consecutive frames of normal attitude features and calculating the mean or median when surrounding vehicles are in a stable driving state and have not yet entered a road surface area where there may be protruding obstacles ahead. It represents the standard feature value of a vehicle on a flat road surface without abnormal attitude fluctuations, serving as a benchmark for subsequent comparison and judgment. This effectively eliminates interference from non-road protrusion factors such as normal vehicle driving vibrations and changes in viewing angle, ensuring the accuracy of attitude change judgment.

[0098] In one possible implementation, after the attitude timing analysis submodule obtains the initial compressed pixel value and vehicle pitch angular velocity of the surrounding vehicles corresponding to the current image frame, it retrieves the baseline value of the vehicle that has been predetermined, and calculates the offset value of the current attitude feature relative to the reference state through the following formula (2), thereby quantifying the real-time change amplitude of the vehicle attitude and providing a calculation basis for subsequent judgment on whether a threshold change has occurred.

[0099] Change = Vehicle attitude feature value in the current frame Baseline value formula (2) If the change is greater than the preset attitude threshold, it is determined that the vehicle attitude characteristics have undergone a sudden change that exceeds the preset attitude threshold.

[0100] In one possible implementation, the attitude temporal analysis submodule first retrieves the temporal variation quantities obtained from previous calculations, which reflect the dynamic changes in the attitude characteristics of surrounding vehicles over time. These variations specifically include the inter-frame difference of the initial compressed pixel values ​​of the same wheel, the temporal fluctuation value of the vehicle's pitch angular velocity, etc., which are quantitative indicators characterizing the actual change in vehicle attitude. Subsequently, the attitude temporal analysis submodule retrieves a preset attitude threshold, which is a critical judgment value obtained through extensive real-vehicle road tests and calibration under different road conditions. This threshold is used to strictly distinguish small attitude fluctuations caused by slight road bumps and vehicle acceleration / deceleration during normal vehicle driving. To ensure the objectivity and accuracy of the judgment criteria, the system detects significant attitude changes caused by obstacles such as road bumps and speed bumps. Finally, the attitude temporal analysis submodule directly compares the retrieved temporal change values ​​with preset attitude thresholds. When the comparison verifies that the value of the temporal change value is greater than the preset attitude threshold, it can be clearly determined that the vehicle attitude characteristics of the surrounding vehicles have undergone a sudden change exceeding the preset attitude threshold. This judgment result will serve as an effective indirect sensing signal of the presence of hidden bump-like obstacles on the road ahead, providing a reliable decision basis for triggering subsequent autonomous vehicle traffic control strategies, and realizing the road anomaly identification logic based on vehicle attitude changes.

[0101] The vehicle control method provided in this application calculates the change in the vehicle's attitude features relative to a baseline value in the current frame. The baseline value is used to characterize the stable attitude feature reference value of each surrounding vehicle before it passes the road surface ahead. This effectively filters out normal attitude fluctuations caused by normal vehicle driving and extracts the true attitude offset caused by road surface anomalies. When the change exceeds a preset attitude threshold, it can be determined that the vehicle's attitude features have undergone a sudden change exceeding the preset attitude threshold. This reliably identifies the sudden abnormal jumps caused by the vehicle rolling over road surface protrusions, providing an accurate and stable basis for subsequent triggering of autonomous vehicle passage control.

[0102] Figure 6 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 4 .like Figure 6 As shown, in the above method, if a sudden attitude change event is determined to occur in any surrounding vehicle, the method further includes controlling the vehicle to pass the road surface ahead based on the driving control parameters associated with the attitude change event. S610. In response to determining that any surrounding vehicle has experienced a sudden attitude change event, extract the vehicle attitude features of the surrounding vehicles corresponding to the attitude change event and the image coordinates of the preset attitude reference point.

[0103] Among them, vehicle posture features are used to characterize the changes in motion state of surrounding vehicles when they pass through the road ahead.

[0104] In one possible implementation, to quantify the amplitude and severity of the vehicle's bouncing through its posture features, thereby determining the size and risk level of road bumps ahead, and providing a quantitative basis for the vehicle to formulate safe passage strategies such as speed adjustment and suspension adaptation; and to bind posture change events with image spatial locations using the image coordinates of preset posture reference points, combining the lane area and imaging mapping relationship to locate the actual location of the road anomaly ahead, while verifying the authenticity of the posture change event through temporal changes in the reference point coordinates, eliminating algorithm detection noise or misjudgment, and ensuring that the vehicle can subsequently avoid the driving risks caused by road bumps based on reliable position and posture data. Based on this, when the attitude timing analysis submodule determines, through timing feature sequence comparison, that the vehicle attitude features of any surrounding vehicle have undergone a sudden change exceeding a preset attitude threshold, and thus confirms that the vehicle has experienced an attitude change event, the system immediately triggers a feature extraction command. The attitude timing analysis submodule retrieves the specific data corresponding to the surrounding vehicle that experienced the attitude change event from the pre-stored timing feature database, including the calculated initial compressed pixel value, vehicle pitch angular velocity, and other vehicle attitude features, as well as the image pixel coordinates of all preset attitude reference points such as wheel contact points or wheel center points, wheel arch points, and front and rear roof reference points, and completes targeted data extraction operations to provide raw data for subsequently associating vehicle attitude features with abnormal road surface spatial locations and generating driving control parameters adapted to the vehicle.

[0105] S620: Determine the safe passing speed of the vehicle based on its attitude characteristics.

[0106] In one possible implementation, since vehicle posture characteristics can directly quantify the severity of road bumps ahead, the greater the change in posture, suspension compression, and body pitch rate of surrounding vehicles, the higher the road bump and the worse the road conditions. Conversely, the road conditions are relatively smooth. The system can reverse-match the corresponding driving speed threshold based on the degree of road abnormality represented by the posture characteristics, and thus determine a safe passing speed that can avoid severe bumps and suspension impacts while maintaining driving stability. Based on the actual road response of the vehicles ahead, the system can plan the vehicle's passage strategy and effectively avoid the driving risks caused by road bumps. Based on this, the vehicle driving decision module first receives and analyzes the extracted vehicle attitude characteristics of surrounding vehicles that have experienced attitude change events. By measuring the magnitude of these vehicle attitude characteristics, it determines the height and severity of the bumps caused by road surface protrusions ahead. Subsequently, it compares and matches these vehicle attitude characteristic values ​​with the pre-defined attitude characteristic and safe speed correlation mapping relationship within the system. The larger the magnitude of the attitude characteristic change, the more significant the impact of the road surface anomaly on the vehicle's driving, and the lower the corresponding safe passing speed limit. After completing the basic speed matching, the vehicle driving decision module adaptively calibrates and corrects its own driving parameters, such as the vehicle's current real-time driving speed, vehicle weight, and suspension characteristics. Finally, it comprehensively calculates and determines a safe passing speed that ensures smooth passage for the vehicle while effectively avoiding severe bumps or loss of control caused by road surface protrusions, providing a reliable speed decision basis for subsequent vehicle speed adjustment and driving control.

[0107] S630: Based on the image coordinates of a preset attitude reference point, determine the distance of the road ahead relative to the vehicle.

[0108] The driving control parameters include: safe passing speed and the distance of the road ahead relative to the vehicle.

[0109] In one possible implementation, the attitude timing analysis submodule first retrieves the image coordinates of preset attitude reference points corresponding to the surrounding vehicles that experienced attitude change events. Specifically, this includes the coordinate data of wheel contact points, wheel center points, wheel arch points, and front and rear roof reference points in the image pixel coordinate system. Then, combining the pre-calibrated intrinsic parameters (such as focal length and principal point coordinates) and extrinsic parameters (such as installation position and viewing angle) of the vehicle camera, the reference point image coordinates in the pixel coordinate system are converted into three-dimensional spatial coordinates in the road world coordinate system through a pinhole camera projection model. The wheel contact point, as the point where the vehicle directly contacts the road surface, has world coordinates that directly correspond to the spatial location of the road anomaly ahead. Finally, using the vehicle's reference position in the world coordinate system as a reference, the straight-line distance between the road anomaly location and the vehicle is calculated using a spatial distance calculation formula. This ultimately determines the actual distance of the road surface ahead relative to the vehicle, providing a spatial distance basis for the vehicle to predict the location of road obstacles.

[0110] The vehicle control method provided in this application, in response to determining that any surrounding vehicle has experienced a sudden attitude change event, extracts the vehicle attitude features of the surrounding vehicles corresponding to the attitude change event and the image coordinates of a preset attitude reference point. The vehicle attitude features are used to characterize the changes in motion state of the surrounding vehicles when passing through the road surface ahead. Based on the extracted data, the decision and control module first combines the impact intensity and risk level of the road surface protrusion reflected by the vehicle attitude features, and derives the safe passing speed of the vehicle for the risky road section through the internal safe speed calculation submodule. This avoids structural damage such as chassis bottoming out and body collision when the vehicle travels at high speed, while ensuring the smoothness and comfort of the driving process. At the same time, based on the preset attitude... The reference point image coordinates are transformed from image space to physical space and distance is calculated to determine the real-time distance of the protruding road surface ahead relative to the vehicle. This provides a clear spatial and temporal basis for the vehicle to plan a smooth deceleration trajectory and intervene in traffic control in advance. The final integrated driving control parameters clearly include two types of parameters: safe passage speed and the distance of the road surface ahead relative to the vehicle. This enables the quantification of road risk based on indirect responses from surrounding vehicles and the customized adaptation of vehicle traffic control. It effectively makes up for the technical shortcomings of traditional direct visual perception in recognizing low-contrast road obstacles such as worn speed bumps and hidden cement protrusions, and improves the response speed of the intelligent driving system to road anomalies and the safety of driving in all scenarios.

[0111] Optionally, determining the distance of the road surface ahead relative to the vehicle based on the image coordinates of a preset attitude reference point in the above method includes: Based on the image coordinates of a preset attitude reference point and the mapping relationship between the preset image coordinates and the actual distance, the distance of the road ahead relative to the vehicle is determined.

[0112] In one possible implementation, the vehicle-side dynamic analysis module first retrieves the image coordinates of preset attitude reference points corresponding to the extracted surrounding vehicles that have experienced attitude change events. It focuses on selecting the image coordinates of reference points that are in direct contact with the road surface, such as wheel contact points and wheel center points, as the core calculation basis. Then, it calls the preset mapping relationship between image coordinates and actual road distances, which is obtained by the system through pre-calibration of the vehicle-mounted camera. This mapping relationship is used to convert the position of the point in the pixel coordinate system into the physical distance in the real road space. Then, it substitutes the image coordinates of the preset attitude reference points into the mapping relationship for conversion processing, converting the pixel position of the reference point in the image into the corresponding actual spatial position. Finally, it uses this actual spatial position to represent the point where the road surface anomaly occurs ahead, and combines it with the current position of the vehicle to finally determine the actual distance of the abnormal position on the road surface ahead relative to the vehicle, providing spatial distance data for the vehicle to predict road obstacles in advance and plan driving strategies.

[0113] The vehicle control method provided in this application, based on the image coordinates of preset attitude reference points of surrounding vehicles that have experienced a sudden attitude change event, and combined with the mapping relationship between the image coordinates and the actual distance pre-calibrated by the system, can convert the reference point position information in the pixel domain into physical scale information in the real road space, calculate and determine the actual distance of the abnormal position on the road ahead relative to the vehicle. This method not only realizes the spatial positioning of the abnormal point on the road by relying on the attitude reference points, but also effectively eliminates the scale deviation between the image pixel coordinates and the actual physical distance through the preset mapping relationship, which greatly improves the accuracy and real-time performance of distance measurement, and provides a reliable spatial distance basis for the vehicle to predict the position of road obstacles in advance and adjust its driving state.

[0114] Optionally, the method described above determines the safe passing speed of the vehicle based on its attitude characteristics, including: The height of obstacles on the road ahead is estimated based on the amplitude of the vehicle's attitude characteristics.

[0115] In one possible implementation, since the height of road obstacles is directly positively correlated with the magnitude of attitude change when a vehicle passes over them, the higher the obstacle, the more pronounced the lifting effect on the vehicle wheels, the greater the suspension compression, the greater the change in the initial compression pixel value, and the more severe the vehicle pitch, the greater the amplitude of the vehicle pitch angular velocity. The amplitude of the vehicle attitude features is a quantitative representation of the degree of attitude changes such as suspension compression and vehicle pitch. Therefore, based on the amplitude of the vehicle attitude features, the height of the road obstacle that causes the vehicle to bounce can be estimated, thus providing a basis for the vehicle to judge the size of the road obstacle and the risk of passage. Based on this, the vehicle dynamic analysis module first acquires the vehicle attitude characteristics of the surrounding vehicles that experienced the attitude change event. Then, it compares the actual amplitude of these vehicle attitude characteristics with the normal attitude values ​​of the vehicles under stable driving conditions to obtain the amplitude of the attitude change. Subsequently, the vehicle dynamic analysis module uses a pre-established physical mapping relationship between the amplitude of vehicle attitude characteristics and the height of road obstacles. This physical mapping relationship is calibrated from experimental data on the height of road bumps corresponding to suspension compression and vehicle pitch. The amplitude of the vehicle attitude characteristics is substituted into this physical mapping relationship for conversion, and finally the height of the obstacle on the road ahead that caused the vehicle to bounce is estimated. Thus, the height information of the road obstacle is deduced from the change in vehicle attitude, providing a basis for the vehicle to judge the size of the obstacle and the risk of passage.

[0116] The safe passage speed is determined based on the relationship between the obstacle height and the vehicle's preset minimum ground clearance.

[0117] The preset minimum ground clearance is the minimum vertical distance between the vehicle chassis and the horizontal road surface.

[0118] In one possible implementation, since the obstacle height is the actual vertical dimension of the external obstacle, directly representing the minimum distance the vehicle needs to lift when passing through, and the vehicle's preset minimum ground clearance is the critical physical threshold for rigid contact between the vehicle's chassis and the obstacle, it is the inherent safety boundary for the vehicle to avoid collision with the obstacle during driving, directly determining the maximum vertical lift space allowed by the chassis when passing through the obstacle. Together, these two constitute the physical constraints for whether the vehicle can pass safely. If only the obstacle height is used as a basis, it is impossible to determine its matching relationship with the vehicle's chassis clearance, which may easily lead to the risk of "the obstacle height is lower than the ground clearance but the vehicle speed is too high, resulting in vertical impact exceeding the suspension's tolerance range and chassis contact with the ground"; if only the minimum ground clearance is used as a basis, it cannot adapt to the passage requirements of obstacles of different heights, which may lead to problems such as "passing through at high speed when the obstacle height is close to the ground clearance, greatly increasing the probability of collision" or "excessive deceleration when the obstacle height is much lower than the ground clearance, reducing passage efficiency". Therefore, only by combining the two and constructing a mapping relationship between the two and the safe passage speed can the vehicle speed range be determined in order to avoid collision accidents and vehicle damage caused by improper vehicle speed and ensure the safe passage of the vehicle in complex road environments. Based on this, the driving decision calculation unit first reads and calls two key parameters. One is the height of the road obstacle ahead, which was previously estimated based on the vehicle's attitude characteristic amplitude. The other is the preset minimum ground clearance of the vehicle, which is the limit height index of the vehicle chassis without scraping and safe passage. Then, the obstacle height is compared with the vehicle's preset minimum ground clearance to clarify the size relationship between the obstacle height and the vehicle's passage limit, thereby judging the risk level of passing the obstacle. Then, according to the height and speed correlation rules calibrated by the actual vehicle test in advance, the following formula (3) is substituted into the calculation.

[0119] Safe passing speed = Preset reference speed × (Vehicle preset minimum ground clearance ÷ Obstacle height) × Working condition correction coefficient Formula (3) Among them, the working condition correction coefficient is used to adapt to actual scenarios such as road surface adhesion and driving conditions; the final calculation yields a safe passing speed that can smoothly cross obstacles without causing chassis collisions, providing a direct and quantitative control basis for adjusting the vehicle's driving speed.

[0120] Specifically, when the obstacle height is close to the vehicle's preset minimum ground clearance, the safe passing speed needs to be reduced to minimize vertical impact and avoid excessive suspension compression or chassis contact with the ground. When the obstacle height is significantly lower than the preset minimum ground clearance, the vehicle speed can be appropriately increased to ensure traffic efficiency. Ultimately, by balancing collision risk and traffic efficiency, the system provides compliant safe passing speed parameters for the vehicle, providing a basis for active driving control. For example, if the obstacle height is less than the preset minimum ground clearance, it indicates that the obstacle will not pose a risk of scraping or impacting the vehicle's chassis, meeting the conditions for safe direct passage. Based on this determination, the system sets the safe passing speed applicable to this type of low-height obstacle to a predefined first speed value according to preset speed configuration rules. This first speed value is a conventional safe driving speed that balances driving efficiency and ride comfort, thereby completing the target passing speed setting for the corresponding road conditions and providing standardized speed parameters for subsequent output of driving control commands to the vehicle's power and braking systems.

[0121] If the height of the obstacle is equal to the preset minimum ground clearance, it indicates that the height of the obstacle has reached the critical height for safe passage of the vehicle chassis. In order to avoid scraping the vehicle chassis and ensure smooth passage through the road obstacle, according to the preset height and speed matching strategy, the safe passage speed for this critical height obstacle is directly set to the second speed value preset by the system. This determines the target safe driving speed for the vehicle to pass through the road obstacle and provides speed parameter basis for the subsequent execution control of the vehicle's power and braking system.

[0122] If the obstacle's height exceeds the preset minimum ground clearance, it indicates that the obstacle has exceeded the vehicle's normal safe passage height range. Driving at normal speeds carries a high risk of chassis scraping and collisions. Based on this risk assessment, and according to pre-set graded speed control rules, the safe passage speed for such high-risk obstacles is directly set as the third speed value. This third speed value is a low-speed limit adapted for passing high obstacles, allowing the vehicle to cautiously pass through the obstacle area at a low speed, minimizing the risk of chassis impact and loss of control. The first speed value is greater than the second speed value, and the second speed value is greater than the third speed value. Therefore, this stepped speed setting method dynamically matches the optimal driving speed according to the actual height of the obstacle, maintaining efficient passage in low-risk scenarios while effectively avoiding chassis damage and severe vehicle vibration in high-risk scenarios, thus balancing vehicle safety, smoothness, and passage efficiency.

[0123] For example, based on the estimated obstacle height, the preset minimum ground clearance, and a safety factor α (e.g., 0.8) that can be calibrated on the actual vehicle, a corresponding speed strategy is formulated: When the obstacle height is less than the preset minimum ground clearance × α / 2, the obstacle is considered low, and the cruising speed can be maintained or only slightly reduced. The safe passing speed can be set to 5-8 km / h (i.e., the first speed value); when the obstacle height is close to the preset minimum ground clearance or close to the preset minimum ground clearance × α, the speed needs to be significantly reduced to a crawling speed. The safe passing speed can be set to 5 km / h (i.e., the second speed value); when the obstacle height is greater than the preset minimum ground clearance, the vehicle cannot theoretically pass the obstacle, and the speed needs to be reduced to 0 (i.e., the third speed value). The entire deceleration process uses a comfortable deceleration curve to achieve a smooth deceleration in advance, and issues a "uneven road surface ahead, slow down or cannot pass" warning to the driver through the vehicle's instrument panel or central control screen.

[0124] The vehicle control method provided in this application can estimate the height of obstacles on the road ahead based on the vehicle attitude characteristic amplitude of surrounding vehicles that have experienced a sudden attitude change event, providing core dimensional basis for the quantitative assessment of road obstacles; and by combining the relationship between the obstacle height and the vehicle's preset minimum ground clearance, the safe passing speed of the vehicle can be derived, which can effectively avoid the risk of chassis scraping against obstacles, and maintain reasonable driving efficiency while ensuring traffic safety, thus achieving the adaptation of road obstacle perception and vehicle traffic control.

[0125] Figure 7 A flowchart illustrating a vehicle control method provided in this application embodiment. Figure 5 .like Figure 7 As shown, the above method controls the vehicle to pass through the road surface ahead based on driving control parameters associated with attitude change events, including: S710: Based on the distance and safe passing speed associated with attitude change events, plan the control process for the vehicle to decelerate to a safe passing speed before reaching the road ahead.

[0126] In one possible implementation, the distance information of obstacles associated with sudden changes in road surface attitude is first retrieved, along with the safe passage speed value obtained by matching the obstacle height. Using the vehicle's current real-time speed as the initial parameter, and combining it with the actual distance between the vehicle and the road surface ahead, a deceleration control process is planned. During the planning process, the deceleration start point and deceleration duration are determined based on the distance parameter. Following the control logic of smooth deceleration, a reasonable deceleration rate is calculated, and the speed change curve of the vehicle gradually decreasing from the current speed to the safe passage speed is determined. This ensures that the vehicle's speed has smoothly decreased to the safe passage speed before reaching the road surface with obstacles ahead, while avoiding driving jerks caused by sudden deceleration. Ultimately, a continuous and smooth deceleration control process is formed, providing a standardized control basis for subsequent execution of vehicle braking and speed adjustment commands.

[0127] S720, based on the control process, controls the vehicle to pass through the road ahead.

[0128] In one possible implementation, after acquiring key road information such as the distance of the road surface ahead relative to the vehicle and the estimated height of obstacles, a corresponding smooth passage control strategy is formulated based on a preset safe driving control process, combined with the vehicle's current real-time speed, driving posture, and road environment conditions. This strategy includes adjusting the vehicle speed, controlling braking force or drive torque, and fine-tuning the steering angle. Subsequently, this control strategy is converted into electrical control commands and sent to the vehicle's power, braking, and steering actuators. Through the coordinated action of these actuators, the vehicle passes through the road surface with obstacles ahead in a safe and smooth manner, avoiding bumps or passage risks caused by road surface protrusions, and ensuring the stability and safety of the vehicle's driving.

[0129] The vehicle control method provided in this application plans a refined control process for the vehicle to smoothly decelerate to the safe passage speed before approaching the road surface, based on the relative distance to the road surface ahead associated with a sudden attitude change event and the appropriate safe passage speed. Then, the vehicle is controlled in real time according to the control process. This method can match the spatial position and height risk of road obstacles, and can effectively avoid the risk of the vehicle shaking and losing control due to high speed passing over raised road surfaces, thereby improving the safety and ride comfort of the vehicle.

[0130] Based on the same inventive concept, this application also provides a vehicle control device. Since the principle of the device in this application is similar to the vehicle control method described above in this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0131] Figure 8 This is a schematic diagram of the structure of a vehicle control device provided in an embodiment of this application. Figure 8 As shown, the vehicle control device 800 may include: The acquisition module 801 is used to acquire an image sequence of at least one surrounding vehicle in motion; the surrounding vehicles are vehicles in front of the vehicle in the current lane and adjacent lanes of the vehicle. The processing module 802 is used to detect the vehicle attitude features of each surrounding vehicle based on the image sequence; the vehicle attitude features are used to characterize the relative vertical displacement of the wheels and the body of the surrounding vehicles, and / or the pitch attitude of the body. The determination module 803 is used to determine, based on the vehicle's attitude characteristics, whether each surrounding vehicle has caused a sudden attitude change event due to passing through the road surface ahead; The control module 804 is used to control the vehicle to pass through the road ahead based on driving control parameters associated with any attitude change event if it is determined that any surrounding vehicle has experienced an attitude change event.

[0132] In one optional implementation, the processing module 802 is specifically configured to: perform vehicle detection on the image sequence to identify at least one surrounding vehicle; detect the pixel coordinates of multiple preset attitude reference points on each surrounding vehicle; the preset attitude reference points are reference points used to measure the relative vertical displacement between the wheels and the vehicle body of the surrounding vehicle, and / or reference points used to measure the pitch attitude of the vehicle body; and calculate the vehicle attitude features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of the multiple preset attitude reference points.

[0133] In one optional implementation, the processing module 802 is specifically configured to: perform vehicle detection on the image sequence to obtain at least one tracked vehicle; perform lane line recognition on the image sequence to determine the area of ​​the vehicle lane and adjacent lanes in the image; if it is determined that the position of at least one tracked vehicle in the image sequence falls within the area of ​​the vehicle lane and adjacent lanes in the image, then at least one tracked vehicle is determined to be a surrounding vehicle.

[0134] In one optional implementation, the vehicle attitude features include: a suspension compression feature quantity, which is the amount of change in the vertical distance between the wheels of surrounding vehicles and the wheel arches of the vehicle body, calculated based on image pixel coordinates; the processing module 802 is specifically used for: for each surrounding vehicle, obtaining the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel from multiple preset attitude reference points, and the pixel coordinates of the wheel arch point corresponding to the same wheel; calculating the difference between the ordinate of the pixel coordinates of the wheel contact point or wheel center point and the ordinate of the pixel coordinates of the wheel arch point to obtain an initial compressed pixel value; and normalizing the initial compressed pixel value based on the preset size of each surrounding vehicle in the image sequence to obtain the suspension compression feature quantity of each surrounding vehicle.

[0135] In one optional implementation, the vehicle attitude features include: vehicle pitch angular velocity, which is the magnitude of the rotational angular velocity of each surrounding vehicle about a lateral axis; the processing module 802 is specifically used to: for each surrounding vehicle, obtain the pixel coordinates of a first reference point located in the front region of the roof and the pixel coordinates of a second reference point located in the rear region of the roof among multiple preset attitude reference points; calculate the pitch indication within a frame of image based on the ordinate of the pixel coordinates of the first and second reference points; the pitch indication is used to characterize the pixel change in the vehicle pitch state of the surrounding vehicles; and calculate the vehicle pitch angular velocity of each surrounding vehicle based on the difference in pitch indications between preset consecutive image frames and the time interval between preset consecutive image frames.

[0136] In one optional implementation, the determining module 803 is specifically used to: perform time-series analysis on the vehicle posture features of each surrounding vehicle to obtain a time-series feature sequence; based on the time-series feature sequence, determine whether the vehicle posture features have undergone a sudden change exceeding a preset posture threshold; if the vehicle posture features have undergone a sudden change exceeding the preset posture threshold, then determine that each surrounding vehicle has caused a posture change event due to passing through the road surface in front.

[0137] In one optional implementation, the determining module 803 is specifically used to: calculate the amount of change of the vehicle posture features of the current frame relative to the baseline value; the baseline value is used to characterize the reference value of the vehicle posture features of each surrounding vehicle when it has not passed the road surface ahead; if the amount of change is greater than a preset posture threshold, then it is determined that the vehicle posture features have undergone a sudden change exceeding the preset posture threshold.

[0138] In an optional implementation, the control module 804 is further configured to: in response to determining that any surrounding vehicle has experienced a sudden attitude change event, extract the vehicle attitude features of the surrounding vehicle corresponding to the attitude change event and the image coordinates of a preset attitude reference point, wherein the vehicle attitude features are used to characterize the changes in motion state of the surrounding vehicle when it passes the road ahead; determine the safe passing speed of the vehicle based on the vehicle attitude features; and determine the distance of the road ahead relative to the vehicle based on the image coordinates of the preset attitude reference point; the driving control parameters include the safe passing speed and the distance of the road ahead relative to the vehicle.

[0139] In one optional implementation, the control module 804 is specifically used to: estimate the height of obstacles on the road ahead based on the amplitude of the vehicle's posture characteristics; calculate the safe passing speed based on the obstacle height and the vehicle's preset minimum ground clearance; the preset minimum ground clearance is the minimum vertical distance between the vehicle chassis and the horizontal road surface.

[0140] It should be noted that for details not disclosed in the vehicle control device of this application embodiment, please refer to the details disclosed in the vehicle control method of this application embodiment, which will not be repeated here.

[0141] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more microprocessors, or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).

[0142] Optionally, embodiments of this application also provide a computer-readable storage medium storing a computer program. When the computer program is run by a processor, the processor executes the steps of the vehicle control method for the mobile storage medium described in the above embodiments. The specific implementation and technical effects are similar and will not be repeated here.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the functional units in the various embodiments of this application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The integrated unit described above can be implemented in hardware or in the form of hardware plus software functional units.

[0144] Optionally, this embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to implement a vehicle control method provided in the above embodiment.

[0145] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0146] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0147] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A vehicle control method, characterized in that, include: Acquire an image sequence of at least one surrounding vehicle during its movement; The surrounding vehicles refer to vehicles in front of the vehicle in its current lane and adjacent lanes. Based on the image sequence, the vehicle pose features of each of the surrounding vehicles are detected; The vehicle attitude features are used to characterize the relative vertical displacement of the wheels and body of surrounding vehicles, and / or the pitch attitude of the body. Based on the vehicle attitude characteristics, determine whether each surrounding vehicle has experienced a sudden attitude change event due to passing through the road surface ahead; If it is determined that any of the surrounding vehicles has experienced the attitude change event, the vehicle is controlled to pass through the road surface ahead based on the driving control parameters associated with the attitude change event.

2. The method according to claim 1, characterized in that, The step of detecting the vehicle pose features of each of the surrounding vehicles based on the image sequence includes: Vehicle detection is performed on the image sequence to identify the at least one surrounding vehicle; Detect the pixel coordinates of multiple preset attitude reference points on each of the surrounding vehicles; the preset attitude reference points are reference points used to measure the relative vertical displacement between the wheels and the body of the surrounding vehicles, and / or reference points used to measure the pitch attitude of the body. Based on the relative positional relationship between the pixel coordinates of the multiple preset posture reference points, the vehicle posture features of each surrounding vehicle are calculated.

3. The method according to claim 2, characterized in that, The process of performing vehicle detection on the image sequence to determine the at least one surrounding vehicle includes: Vehicle detection is performed on the image sequence to obtain at least one tracked vehicle; Lane line recognition is performed on the image sequence to determine the areas of the vehicle lane and adjacent lanes in the image; If it is determined that the position of the at least one tracked vehicle in the image sequence falls within the area of ​​the vehicle lane and adjacent lanes in the image, then the at least one tracked vehicle is determined to be one of the surrounding vehicles.

4. The method according to claim 2, characterized in that, The vehicle attitude features include: suspension compression features, which are the amount of change in the vertical distance between the wheels of the surrounding vehicles and the wheel arches of the vehicle body, calculated based on image pixel coordinates. The calculation of the vehicle pose features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of the plurality of preset pose reference points includes: For each surrounding vehicle, obtain the pixel coordinates of the wheel contact point or wheel center point corresponding to the same wheel among the multiple preset posture reference points, and the pixel coordinates of the wheel arch point corresponding to the same wheel; The difference between the ordinate of the pixel coordinates of the wheel contact point or wheel center point and the ordinate of the pixel coordinates of the wheel arch point is calculated to obtain the initial compressed pixel value. Based on the preset size of each surrounding vehicle in the image sequence, the initial compressed pixel value is normalized to obtain the suspension compression feature of each surrounding vehicle.

5. The method according to claim 2, characterized in that, The vehicle attitude characteristics include: vehicle pitch rate, which is the magnitude of the rotational angular velocity of each surrounding vehicle about the lateral axis; The calculation of the vehicle pose features of each surrounding vehicle based on the relative positional relationship between the pixel coordinates of the plurality of preset pose reference points includes: For each surrounding vehicle, obtain the pixel coordinates of the first reference point located in the front area of ​​the roof and the pixel coordinates of the second reference point located in the rear area of ​​the roof among the plurality of preset attitude reference points; Based on the ordinate of the pixel coordinates of the first reference point and the second reference point, the pitch indicator within a frame of image is calculated; the pitch indicator is used to characterize the pixel change in the body pitch state of the surrounding vehicles. Based on the difference in pitch indication between preset consecutive image frames and the time interval between the preset consecutive image frames, the vehicle body pitch angular velocity of each surrounding vehicle is calculated.

6. The method according to claim 1, characterized in that, The step of determining whether each surrounding vehicle has experienced a sudden attitude change event due to passing through the road surface ahead, based on the vehicle's attitude characteristics, includes: A temporal analysis is performed on the vehicle attitude features of each surrounding vehicle to obtain a temporal feature sequence; Based on the time-series feature sequence, determine whether the vehicle attitude features have undergone a sudden change exceeding a preset attitude threshold; If the vehicle's posture characteristics undergo a sudden change exceeding the preset posture threshold, then it is determined that each surrounding vehicle triggered the posture change event by passing through the road surface ahead.

7. The method according to claim 6, characterized in that, Determining whether the vehicle attitude characteristics have undergone a sudden change exceeding a preset attitude threshold includes: Calculate the change in vehicle attitude features in the current frame relative to a baseline value; the baseline value is used to characterize the vehicle attitude feature reference value of each surrounding vehicle before it passes the road surface ahead; If the change is greater than a preset attitude threshold, then it is determined that the vehicle attitude feature has undergone a sudden change exceeding the preset attitude threshold.

8. The method according to claim 1, characterized in that, If it is determined that any of the surrounding vehicles has experienced the attitude change event, the method further includes controlling the vehicle to pass the road surface ahead based on driving control parameters associated with the attitude change event. In response to determining that any of the surrounding vehicles has experienced the attitude change event, the vehicle attitude features of the surrounding vehicles corresponding to the attitude change event and the image coordinates of a preset attitude reference point are extracted. The vehicle attitude features are used to characterize the changes in motion state of the surrounding vehicles when they pass the road surface in front. Based on the vehicle's attitude characteristics, determine the safe passing speed of the vehicle; Based on the image coordinates of a preset attitude reference point, the distance of the road surface ahead relative to the vehicle is determined; The driving control parameters include: the safe passing speed and the distance of the road surface ahead relative to the vehicle.

9. The method according to claim 8, characterized in that, Determining the safe passing speed of the vehicle based on the vehicle attitude characteristics includes: Based on the amplitude of the vehicle's attitude characteristics, the height of the obstacle on the road ahead is estimated; The safe passing speed is determined based on the relationship between the height of the obstacle and the preset minimum ground clearance of the vehicle; the preset minimum ground clearance is the minimum vertical distance between the vehicle chassis and the horizontal road surface.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable program code; A processor for calling and running the executable program code from the memory, causing the electronic device to perform the method as described in any one of claims 1 to 9.

Images