Metro logistics distribution intelligent vehicle control method and device, intelligent vehicle, medium

Abstract

The application provides a subway logistics distribution intelligent vehicle control method and device, an intelligent vehicle and a medium, and belongs to the technical field of intelligent vehicle control. The method comprises the following steps: acquiring road surface sensing data in front of the intelligent vehicle, and determining height change information of the road surface in front; when it is determined that an obstacle surmounting action is about to occur, locking the current pose of the intelligent vehicle as a reference pose; determining the motion trajectory of the intelligent vehicle in the obstacle surmounting process based on inertial data of the intelligent vehicle in the obstacle surmounting process, and combining the reference pose to calculate the pose of the intelligent vehicle after the obstacle surmounting ends, to generate a plurality of pose hypotheses and respective corresponding confidence scores; selecting a target position hypothesis from the plurality of pose hypotheses according to the respective confidence scores, and controlling the intelligent vehicle to travel based on the target pose. The subway logistics distribution intelligent vehicle control method and device, the intelligent vehicle and the medium provided by the application can improve the positioning continuity and driving safety of the intelligent vehicle on complex road surfaces.

Description

Technical Field

[0001] This application belongs to the field of intelligent vehicle control technology, and more specifically, relates to a method and device for controlling intelligent vehicles for subway logistics distribution, intelligent vehicles, and media. Background Technology

[0002] Intelligent vehicles used for subway logistics and delivery often transport goods in underground tunnels. Current vehicles primarily rely on external sensors for environmental perception and pose estimation. However, subway tunnels frequently contain obstacles such as track joints and trenches. When vehicles cross these obstacles, the vehicle body experiences severe jolts and abrupt attitude changes, leading to significant distortion or temporary failure of data collected by external sensors. This results in positioning interruptions or large cumulative drift errors, making it impossible for the system to accurately obtain the vehicle's true spatial state after obstacle crossing, severely impacting subsequent motion control. Therefore, a new control scheme is urgently needed to overcome the positioning loss defects caused by obstacle crossing and improve the positioning continuity and driving safety of intelligent vehicles on complex road surfaces. Summary of the Invention

[0003] The purpose of this application is to provide a method and device for controlling intelligent vehicles in subway logistics distribution, as well as an intelligent vehicle and a medium, so as to improve the positioning continuity and driving safety of intelligent vehicles on complex road surfaces.

[0004] A first aspect of this application provides a method for controlling intelligent vehicles used in subway logistics distribution, executed by an intelligent vehicle, comprising: Acquire road perception data in front of the intelligent vehicle. The road perception data includes data collected by LiDAR and millimeter-wave radar. Based on road surface perception data, determine the height change information of the road surface ahead; When it is determined from the height change information that there is a height difference on the road ahead that exceeds the preset height threshold, it is determined that an obstacle crossing action is about to occur, and the current pose of the intelligent vehicle is locked as the reference pose. Acquire inertial data of intelligent vehicles during obstacle crossing, and determine the trajectory of intelligent vehicles during obstacle crossing based on the inertial data; Based on the motion trajectory and combined with the reference pose, the pose of the intelligent vehicle after the obstacle crossing is calculated, and multiple pose hypotheses and their corresponding confidence scores are generated. Based on the confidence scores, the target position hypothesis is selected from multiple pose hypotheses and used as the target pose of the intelligent vehicle after the obstacle crossing is completed. Intelligent vehicle driving is controlled based on target pose.

[0005] A second aspect of this application provides a smart vehicle control device for subway logistics distribution, comprising: The road surface perception module is used to acquire and fuse the road surface perception data ahead to determine the height change information of the road surface ahead; the road surface perception data ahead includes data collected by lidar and data collected by millimeter-wave radar. The obstacle crossing prediction module is used to determine that an obstacle crossing action is about to occur when the height difference on the road ahead exceeds a preset height threshold based on the height change information, and to lock the current pose of the intelligent vehicle as the reference pose. The pose estimation module is used to acquire the inertial data of the intelligent vehicle during obstacle crossing, and combine it with the reference pose to estimate the pose of the intelligent vehicle, thereby obtaining an estimated pose sequence composed of estimated poses at multiple times, and determining the estimated pose at the end of obstacle crossing as the obstacle crossing end pose. The candidate generation module is used to acquire point cloud data collected by LiDAR after obstacle crossing, delineate a local search area centered on the obstacle crossing end pose, and generate multiple candidate poses based on point cloud data within the local search area. The confidence calculation module is used to generate a corresponding pose hypothesis for each candidate pose by combining the inferred pose sequence, and to calculate the confidence score of the pose hypothesis. The control module is used to determine the location recovery result based on the confidence score and control the driving state of the intelligent vehicle based on the location recovery result.

[0006] A third aspect of this application provides an intelligent vehicle, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-described intelligent vehicle control method for subway logistics distribution.

[0007] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described intelligent vehicle control method for subway logistics distribution.

[0008] The beneficial effects of the intelligent vehicle control method and device for subway logistics distribution, the intelligent vehicle, and the medium provided in this application embodiment are as follows: By acquiring road perception data including LiDAR and millimeter-wave radar data, this application embodiment can accurately determine the height change information of the road surface ahead; when the height difference exceeds a preset threshold, the reference pose is locked in advance, avoiding interference from obstacle crossing bumps on the initial positioning; during obstacle crossing, inertial data is used to determine the motion trajectory, compensating for the deficiency of limited external perception during obstacle crossing; after obstacle crossing, multiple pose hypotheses are generated by combining the motion trajectory and the reference pose, and the most reliable target pose is selected based on the confidence score, effectively eliminating accumulated errors and positioning drift. Therefore, this application embodiment achieves precise connection of poses before and after obstacle crossing, improving the positioning continuity and driving safety of intelligent vehicles on complex road surfaces. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a method for controlling intelligent vehicles in subway logistics distribution according to an embodiment of this application; Figure 2 A structural block diagram of a subway logistics distribution intelligent vehicle control device provided in an embodiment of this application; Figure 3 This is a schematic block diagram of an intelligent vehicle provided in an embodiment of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for controlling intelligent vehicles in subway logistics distribution according to an embodiment of this application. This method can be executed by an intelligent vehicle and may include: S101: Acquire road perception data in front of the intelligent vehicle. The road perception data includes data collected by LiDAR and millimeter-wave radar.

[0013] In this embodiment, the intelligent vehicle refers to an autonomous vehicle used within a subway station or subway depot to undertake logistics delivery tasks such as delivering goods, packages, or meals. It possesses autonomous navigation, environmental perception, and motion control capabilities. Road surface perception data refers to the set of environmental information acquired by the intelligent vehicle through real-time detection of the road surface ahead using its multi-source sensors (at least lidar and millimeter-wave radar in this embodiment). LiDAR data primarily refers to three-dimensional point cloud data generated by emitting laser pulses and receiving reflected signals, accurately reflecting the geometric contours of the road surface and obstacles. Millimeter-wave radar data refers to distance, speed, and angle information obtained by emitting millimeter-wave electromagnetic waves and receiving echoes; its advantages include strong penetration through rain and fog and stable ranging.

[0014] For example, an intelligent vehicle operates at a maximum speed of 10 km / h within a subway logistics corridor. Dual LiDAR sensors are installed at the front of the vehicle, positioned on the left and right sides of the front. Their scanning fields overlap directly in front of the vehicle, eliminating near-field blind spots caused by a single LiDAR sensor's low mounting position. The dual LiDAR sensors and millimeter-wave radar achieve microsecond-level time synchronization via a precise time protocol, continuously scanning the road surface within a 5-meter range ahead. The LiDAR outputs a high-density point cloud at a 10Hz frequency, while the millimeter-wave radar outputs a target list at a 20Hz frequency. The 20Hz data from the millimeter-wave radar is linearly interpolated to align with the 10Hz frame rate of the dual LiDAR sensors, ensuring a consistent time reference for data from all sensors during subsequent joint processing.

[0015] S102: Based on road surface perception data, determine the height change information of the road surface ahead.

[0016] In this embodiment, height change information refers to the undulation or abrupt change features of the road surface ahead in the vertical direction extracted based on the fusion of multi-source sensor data.

[0017] In this embodiment, the point cloud data generated by the two LiDARs are first stitched and fused together. The two point cloud frames are uniformly transformed to the vehicle coordinate system, and corresponding point pairs in the overlapping areas are precisely registered using an iterative nearest-point algorithm to merge them into a complete dense point cloud. Subsequently, the merged point cloud is segmented onto the ground, and a height map-based method is used to project the point cloud onto a 5cm×5cm two-dimensional grid map. The minimum elevation value within each grid is calculated to construct a road surface reference plane. When fusing millimeter-wave radar data, the distance and angle information of each target detected by the millimeter-wave radar are extracted and projected onto the same grid map. Candidate obstacle areas detected by both LiDARs are identified as high-confidence obstacles; areas detected only by the LiDARs and whose point cloud reflection intensity is below a preset threshold are identified as false echoes caused by rain, fog, or dust and are removed. Finally, based on the valid point cloud after fusion verification, the height change gradient of each grid on the road surface relative to the reference plane is calculated in real time. This process is executed cyclically on the onboard computing platform at a 50ms cycle to ensure real-time response to road surface undulations.

[0018] S103: When it is determined from the height change information that there is a height difference on the road ahead that exceeds the preset height threshold, it is determined that an obstacle crossing action is about to occur, and the current pose of the intelligent vehicle is locked as the reference pose.

[0019] In this embodiment, the preset height threshold refers to the critical value of the road surface height difference that triggers the obstacle crossing judgment. In this embodiment, it can be set to 20mm. This value can be set according to the vehicle chassis minimum ground clearance of 150mm and the maximum suspension compression travel, while retaining a safety margin. The reference pose refers to the precise spatial state of the intelligent vehicle at the moment when it is determined that an obstacle crossing action is about to occur, including three-dimensional position coordinates and heading angle.

[0020] In this embodiment, when a continuous area with a height difference exceeding 20mm is detected on the road surface ahead, it is determined that an obstacle-crossing action is about to occur. Simultaneously with this determination, the current three-dimensional position and heading angle output by the intelligent vehicle through fusion positioning are immediately locked as the reference pose. The locking timing is chosen approximately 200ms before obstacle crossing. This lead time can be calculated based on the current vehicle speed and signal processing delay, ensuring that the reference pose is not affected by sensor jitter caused by obstacle-crossing bumps, providing a stable and reliable absolute reference starting point for pose estimation during subsequent obstacle-crossing.

[0021] S104: Acquire the inertial data of the intelligent vehicle during obstacle crossing and determine the trajectory of the intelligent vehicle during obstacle crossing based on the inertial data.

[0022] In this embodiment, inertial data refers to the triaxial acceleration and triaxial angular velocity data collected at high frequency by the vehicle-mounted inertial measurement unit during obstacle crossing.

[0023] Upon entering the obstacle-crossing phase, severe distortion of the LiDAR point cloud may occur due to the violent shaking of the vehicle body. Therefore, the reliance on point cloud localization is temporarily reduced, and the raw data from the inertial measurement unit is read at a sampling frequency of 200Hz. During the calculation process, the raw data is first subjected to zero-bias compensation and low-pass filtering to remove high-frequency noise. Then, the angular velocity is integrated to calculate the vehicle's attitude change in real time. The acceleration is first subjected to gravity component elimination before a second integration to calculate displacement change, thus constructing the vehicle's six-degree-of-freedom trajectory during obstacle crossing. To suppress the cumulative drift caused by pure integration, a zero-speed correction strategy is introduced. This involves resetting and correcting the velocity and position during the brief instant when the wheel speed sensor output is detected to be zero. Utilizing the high-frequency self-containment characteristics of the inertial data, the continuity of pose estimation is maintained during the brief window of point cloud failure.

[0024] S105: Based on the motion trajectory and combined with the reference pose, calculate the pose of the intelligent vehicle after the obstacle crossing, and generate multiple pose hypotheses and their corresponding confidence scores.

[0025] In this embodiment, pose assumptions refer to multiple candidate spatial states generated based on the baseline pose before obstacle crossing and the inertial trajectory calculated during obstacle crossing, combined with various environmental matching results that may exist after obstacle crossing. Confidence scores are quantitative indicators of reliability assigned to each pose assumption.

[0026] In this embodiment, approximately 100ms after obstacle crossing, based on the inertial-calculated trajectory and using the reference pose locked before obstacle crossing as the starting anchor point, a set of particles is distributed in a Gaussian distribution near the reference pose using a particle filtering algorithm, thereby generating 3-5 pose hypotheses. For each hypothesis, a scoring factor is calculated from three dimensions: first, displacement continuity deviation, i.e., the Euclidean distance deviation between the hypothesis and the reference pose plus the inertial-calculated displacement; second, trajectory consistency error, i.e., the dynamic time-warped distance between the path corresponding to the hypothesis and the inertial-calculated trajectory; and third, map matching fit, i.e., the normalized correlation coefficient obtained by correlation matching between the current LiDAR point cloud and the pre-stored environmental map. The three scoring factors are weighted and fused to obtain the final confidence score. The weight coefficients can be statistically derived from historical operation data, for example, displacement continuity accounts for 0.3, trajectory consistency accounts for 0.4, and map matching fit accounts for 0.3.

[0027] S106: Based on the confidence scores, select the target position hypothesis from multiple pose hypotheses as the target pose of the intelligent vehicle after the obstacle crossing is completed.

[0028] In this embodiment, the target position hypothesis refers to the candidate pose that is confirmed to be closest to the actual spatial state of the vehicle after being filtered by confidence scores among multiple pose hypotheses.

[0029] In one embodiment of this application, a target position hypothesis is selected from multiple pose hypotheses based on various confidence scores, serving as the target pose of the intelligent vehicle after obstacle crossing; including: Based on the confidence scores, the pose hypothesis with the highest confidence score is selected from multiple pose hypotheses as the target pose of the intelligent vehicle after the obstacle crossing is completed. Among them, controlling the driving of intelligent vehicles based on target pose includes: Based on the target pose, and combined with a preset control strategy, the intelligent vehicle is controlled to drive. The preset control strategies include: During the process of controlling the intelligent vehicle to drive, the driving speed is less than the preset speed threshold and the steering angle is less than the preset steering angle threshold.

[0030] In this embodiment, the preset control strategy refers to a set of conservative motion constraint rules that the intelligent vehicle enforces to ensure driving safety during the positioning recovery phase or under special operating conditions. The preset speed threshold and preset steering angle threshold are specific boundary values ​​used to limit the vehicle's longitudinal speed and lateral steering amplitude, aiming to reduce the motion risks caused by positioning uncertainty.

[0031] In this embodiment, based on the confidence scores of each pose hypothesis, the pose hypothesis with the highest score is directly selected as the target pose of the intelligent vehicle after obstacle crossing. For example, if the confidence scores of the three generated pose hypotheses are 0.85, 0.72, and 0.61, the hypothesis with a score of 0.85 is selected as the target pose.

[0032] During the process of controlling the intelligent vehicle's movement based on this target pose, a preset control strategy is simultaneously activated to strictly limit the vehicle's speed to below a preset speed threshold and the steering angle to within a preset steering angle threshold. Specifically, the preset speed threshold is set to 3 km / h, a value that can be set based on empirical values ​​of safe braking distances in mixed pedestrian and vehicle traffic scenarios within subway logistics channels, ensuring sufficient emergency response time for the vehicle at low speeds; the preset steering angle threshold is set to ±5°, a value that can be calculated based on the vehicle's minimum turning radius of 0.9m and steering response characteristics at low speeds, preventing vehicle instability or deviation from the expected path due to excessive steering.

[0033] Furthermore, if the location cannot be restored within a preset time (e.g., 300ms) after overcoming the obstacle, or if the confidence score remains below a preset confidence threshold (e.g., 0.5) for multiple consecutive cycles, a timeout protection mechanism is triggered, controlling the intelligent vehicle to automatically stop and issue an alarm signal, awaiting manual intervention or restart. This preset confidence threshold can be statistically derived from the lower boundary of the reliable location interval during historical operation.

[0034] As can be seen from the above, this embodiment effectively reduces the driving risk caused by pose estimation errors by implementing a conservative control strategy during the positioning recovery phase. The dual constraints of speed limit and rotation limit ensure the controllability of vehicle movement in uncertain environments, and combined with the overdue automatic stop alarm mechanism, further improve the overall safety level in subway logistics and delivery scenarios.

[0035] S107: Intelligent vehicle driving control based on target pose.

[0036] In this embodiment, controlling the driving of an intelligent vehicle based on a target pose refers to updating the vehicle's navigation state using a determined target pose and restoring environmental perception and path planning functions.

[0037] In this embodiment, the target pose is first written into the navigation state vector to replace the inertial pose estimation during obstacle crossing, thus achieving a smooth switch in the positioning results. Subsequently, the matching and positioning of the LiDAR point cloud and the map is reactivated to verify the long-term consistency of the target pose; and the planned speed is gradually restored within a preset time to allow the vehicle to smoothly transition to normal driving, ensuring the continuity of vehicle behavior after obstacle crossing.

[0038] As can be seen from the above, this embodiment can accurately determine the height change information of the road surface ahead by acquiring road perception data including LiDAR and millimeter-wave radar data; when the height difference exceeds a preset threshold, the reference pose is locked in advance to avoid interference from obstacle crossing bumps on the initial positioning; during obstacle crossing, inertial data is used to determine the motion trajectory, compensating for the deficiency of limited external perception during obstacle crossing; after obstacle crossing, multiple pose hypotheses are generated by combining the motion trajectory and the reference pose, and the most reliable target pose is selected according to the confidence score, effectively eliminating accumulated errors and positioning drift. Therefore, this embodiment achieves precise connection of poses before and after obstacle crossing, improving the positioning continuity and driving safety of intelligent vehicles on complex road surfaces.

[0039] In one embodiment of this application, the method further includes the following steps during the process of controlling the driving of an intelligent vehicle: If the intelligent vehicle is detected to be accelerating, the damping coefficient of the rear suspension shock absorber and / or the rear wheel drive force distribution ratio are adjusted. If the intelligent vehicle is detected to be decelerating, the damping coefficient of the front suspension shock absorber and / or the front wheel braking force distribution ratio are adjusted. The determination of whether an intelligent vehicle is accelerating or decelerating is based on its longitudinal acceleration.

[0040] In this embodiment, acceleration and deceleration refer to two motion modes of the intelligent vehicle during longitudinal travel, categorized by the sign and magnitude of longitudinal acceleration. Longitudinal acceleration refers to the acceleration component of the vehicle in the direction of travel, typically acquired in real-time by the onboard inertial measurement unit and obtained after low-pass filtering. Rear and front suspension shock absorbers are elastic damping elements installed between the rear and front axle wheels and the vehicle frame, respectively, used to absorb road impacts and maintain wheel contact. The damping coefficient is a physical quantity that measures the shock absorber's ability to dissipate vibration energy; a higher damping coefficient results in a stiffer suspension, and vice versa. The rear-wheel drive force distribution ratio and front-wheel braking force distribution ratio refer to the percentage of rear axle drive torque to total drive torque and the percentage of front axle braking torque to total braking torque in the vehicle's four-wheel drive system, used to adjust the vehicle's longitudinal dynamic characteristics.

[0041] In one embodiment of this application, the acceleration state includes a gradual acceleration state, a rapid acceleration state, and a fully loaded rapid acceleration state. If the intelligent vehicle is detected to be accelerating, the damping coefficient of the rear suspension shock absorbers and / or the rear-wheel drive force distribution ratio are adjusted, including: If the intelligent vehicle is detected to be accelerating smoothly and the road surface is flat, the damping coefficient of the rear suspension shock absorber is reduced while maintaining the standard rear-wheel drive force distribution ratio. If the intelligent vehicle is detected to be accelerating rapidly, the rear-wheel drive force distribution ratio is increased while the damping coefficient of the standard rear suspension shock absorber is maintained. If the system detects that the vehicle is under full load and accelerating rapidly, and the road surface is uneven, the damping coefficient of the rear suspension shock absorber will be reduced and the rear wheel drive force distribution ratio will be increased.

[0042] In this embodiment, the terms "gradual acceleration," "rapid acceleration," and "full-load rapid acceleration" are further subdivisions of acceleration states based on the magnitude of longitudinal acceleration and vehicle load. Gradual acceleration refers to an acceleration process with relatively small and gradual longitudinal acceleration; rapid acceleration refers to an acceleration process with large longitudinal acceleration and a strong power demand; and full-load rapid acceleration refers to the extreme condition of rapid acceleration under the vehicle's maximum design load. Road surface smoothness or unevenness is determined qualitatively based on vertical acceleration data from suspension displacement sensors or inertial measurement units.

[0043] In this embodiment, the acceleration state is precisely divided into three sub-states based on the specific value of longitudinal acceleration and load status, and differentiated control strategies are executed accordingly. When the intelligent vehicle is detected to be in a smooth acceleration state, i.e., the longitudinal acceleration does not exceed 0.15g, and the current road surface is smooth (suspension displacement fluctuation is less than 5mm), the damping coefficient of the rear suspension shock absorber is actively reduced by 30% to 50% from the standard value, making the rear suspension softer to absorb the recoil tendency caused by acceleration, while maintaining the rear wheel drive force distribution ratio at the standard 50%. This 0.15g threshold can be calculated based on the vehicle's full load mass and the natural frequency of the suspension system to ensure that the suspension system is in the linear operating range during smooth acceleration. When the vehicle is detected to be in a rapid acceleration state, i.e., the longitudinal acceleration exceeds 0.15g, or there is slight unevenness on the road surface, the rear wheel drive force distribution ratio is temporarily increased, raising the rear axle drive force distribution ratio from the standard 50% to the range of 60% to 70%. The reaction force generated by the increased drive torque is used to counteract the tendency of the front of the vehicle to lift up, while maintaining the damping coefficient of the rear suspension shock absorber at the standard value. When a vehicle is detected to be under full load and accelerating rapidly, and simultaneously encountering uneven road surfaces due to speed bumps or raised obstacles, two measures are simultaneously activated: reducing rear suspension damping and increasing rear-wheel drive force distribution. The rear suspension damping is reduced by 40% of the standard value, and the rear axle drive force distribution ratio is increased to 65%. These 40% and 65% values ​​are determined based on vehicle dynamics simulations and real-vehicle calibration data under full load conditions, aiming to achieve optimal anti-pitch performance and traction balance under extreme conditions.

[0044] As can be seen from the above, this embodiment achieves precise suppression of vehicle pitch motion under different operating conditions through refined division of acceleration states and differentiated control strategies. Comfort is prioritized during smooth acceleration, while power response and attitude stability are emphasized during rapid acceleration. Comprehensive measures are taken under extreme conditions of full-load rapid acceleration and uneven road surfaces, significantly improving the driving stability and cargo safety of intelligent vehicles in all scenarios.

[0045] In one embodiment of this application, the deceleration state includes a smooth deceleration state, a rapid deceleration state, and a fully loaded rapid deceleration state. If the intelligent vehicle is detected to be decelerating, the damping coefficient of the front suspension shock absorbers and / or the front wheel braking force distribution ratio are adjusted, specifically including: If the deceleration is detected to be gradual and the road surface is flat, the damping coefficient of the front suspension shock absorber is reduced while maintaining the standard front wheel braking force distribution ratio. If a rapid deceleration is detected, the front wheel braking force distribution ratio is increased while maintaining the damping coefficient of the standard front suspension shock absorbers. If the deceleration state is detected to be a state of rapid deceleration under full load, and the current road surface is uneven, then the operation of reducing the damping coefficient of the front suspension shock absorber and increasing the front wheel braking force distribution ratio will be executed simultaneously.

[0046] In this embodiment, the terms "gradual deceleration," "rapid deceleration," and "full-load rapid deceleration" are further subdivisions of the deceleration state based on the magnitude of longitudinal deceleration and vehicle load. Gradual deceleration refers to a braking process with relatively small and gradual longitudinal deceleration; rapid deceleration refers to an emergency braking process with large longitudinal deceleration and a strong braking demand; and full-load rapid deceleration refers to the extreme condition of emergency braking under the maximum design load of the vehicle. The front-wheel regenerative braking force refers to the braking torque generated by the reverse-rotation generator of the front axle drive motor, used for energy recovery and auxiliary braking.

[0047] In this embodiment, the deceleration state is precisely divided into three sub-states based on the specific value of the longitudinal deceleration and the load state, and differentiated control strategies are implemented accordingly. When the intelligent vehicle is detected to be in a smooth deceleration state, i.e., the longitudinal deceleration does not exceed 0.15g and the current road surface is smooth, the damping coefficient of the front suspension shock absorber is actively reduced by 30% to 50% from the standard value, making the front suspension softer to absorb the nose-diving tendency caused by deceleration, while maintaining the front wheel braking force distribution ratio at the standard 50%. When the vehicle is detected to be in a rapid deceleration state, i.e., the longitudinal deceleration exceeds 0.15g, or there is slight unevenness on the road surface, the front wheel regenerative braking force distribution ratio is temporarily increased, raising the front axle braking force distribution ratio from the standard 50% to the range of 60% to 70%. The reaction force generated by the increased braking torque is used to counteract the nose-diving tendency, while maintaining the damping coefficient of the front suspension shock absorber at the standard value. This 60% to 70% distribution ratio range can be determined comprehensively based on the braking energy recovery efficiency and the front axle adhesion limit, ensuring that nose-diving is suppressed without triggering front wheel lock-up. When the vehicle is detected to be under full load and undergoing rapid deceleration, and simultaneously encountering uneven road surfaces due to speed bumps or potholes, two measures are activated simultaneously: reducing front suspension damping and increasing regenerative braking force on the front wheels. The front suspension damping is reduced by 40% of the standard value, and the front axle braking force distribution ratio is increased to 65%. This control process is achieved by coordinating the regenerative braking torque of the front axle motor with the hydraulic pressure of the mechanical braking system through the vehicle controller, ensuring a balance between braking smoothness and attitude stability.

[0048] As can be seen from the above, this embodiment achieves precise suppression of vehicle pitch motion under different braking conditions through refined classification of deceleration states and differentiated control strategies. During smooth deceleration, priority is given to ensuring comfort and energy recovery efficiency; during rapid deceleration, emphasis is placed on braking response and attitude stability; and comprehensive measures are taken under extreme conditions of rapid deceleration under full load and on uneven road surfaces, significantly shortening the braking distance and improving driving safety in emergency situations.

[0049] In one embodiment of this application, the method further includes: acquiring the change in the vehicle body pitch angle of the intelligent vehicle; when the vehicle body pitch angle exceeds a preset angle threshold, increasing the damping coefficient adjustment range and / or the power distribution ratio adjustment range until the vehicle body pitch angle returns to a preset safe range.

[0050] In this embodiment, the vehicle pitch angle refers to the angle between the vehicle's longitudinal axis and the horizontal plane, reflecting the degree of tilt of the vehicle body in the longitudinal plane. It can be obtained by fusing data from the gyroscope and accelerometer in the onboard inertial measurement unit. The preset angle threshold refers to the critical value of the vehicle pitch angle that triggers enhanced control measures. When the actual pitch angle exceeds this value, the current anti-pitch measures are considered insufficient to maintain vehicle stability. The preset safety range refers to the safe interval within which the vehicle pitch angle is allowed to fluctuate. It is usually based on the vehicle's horizontal state, with a small angle tolerance set both above and below.

[0051] In this embodiment, the vehicle pitch angle change is acquired and calculated in real time at a frequency of 200Hz using an inertial measurement unit. After the vehicle is accelerating or decelerating and corresponding suspension damping and power distribution adjustments are performed, the real-time pitch angle value is continuously monitored. When the vehicle pitch angle exceeds a preset angle threshold, for example, more than 2°, it is determined that the current adjustment is insufficient to effectively suppress the longitudinal sway of the vehicle body, and the adjustment range of the damping coefficient and / or the power distribution ratio is automatically increased.

[0052] Specifically, if the vehicle is currently accelerating and the rear suspension damping has already decreased by 40%, the reduction will be further increased to 50%~60%; if the rear axle drive force distribution ratio has already increased to 65%, it will be further increased to 70%~75%. This 2° threshold can be calculated based on the vehicle's fully loaded center of gravity height and suspension travel limits, ensuring that the vehicle body remains stable and controllable before the enhancement measures are triggered. After the enhancement adjustment, the pitch angle change continues to be monitored in real time until the pitch angle returns to the preset safe range, for example, to within 1°. At this point, the adjustment range is gradually returned to the initial setting value to avoid over-adjustment leading to suspension system saturation or power distribution imbalance. The response time of the entire closed-loop control process is controlled within 50ms to ensure real-time and precise intervention in the vehicle's attitude.

[0053] As can be seen from the above, this embodiment achieves dynamic evaluation and adaptive enhancement of the anti-pitch control effect by introducing real-time monitoring and closed-loop feedback mechanism of vehicle pitch angle. When the initial adjustment measures are insufficient to suppress vehicle sway, the adjustment range is automatically increased until the attitude returns to a safe range, which significantly improves the attitude stability control capability of intelligent vehicles under complex working conditions and effectively prevents the risk of cargo overturning or tires leaving the ground due to excessive pitch angle.

[0054] In one embodiment of this application, the method further includes: Obtain the load status of the intelligent vehicle, including no-load, full-load, and off-center load status; If the load state of the intelligent vehicle is unloaded, the first suspension setting and the first acceleration limit are used to control the operation of the intelligent vehicle. If the intelligent vehicle is fully loaded, the second suspension setting and the second acceleration limit are used to control the operation of the intelligent vehicle; the stiffness of the second suspension setting is higher than that of the first suspension setting, and the second acceleration limit is lower than that of the first acceleration limit. If the load state of the intelligent vehicle is off-center, the driving force of the left or right wheel or the braking force of the left or right wheel will be adjusted based on the off-center load direction to generate a reverse yaw moment.

[0055] In this embodiment, the load state refers to the mass and distribution of cargo currently carried by the intelligent vehicle, specifically divided into three types: unloaded state, fully loaded state, and off-center load state. An unloaded state refers to a state where the vehicle is not carrying any cargo or has a very small load mass; a fully loaded state refers to a state where the vehicle's load mass reaches or approaches the maximum design load; and an off-center load state refers to a state where the cargo mass is unevenly distributed, concentrated on the left or right side of the vehicle. The first suspension setting and the second suspension setting refer to preset combinations of suspension system stiffness and damping parameters for unloaded and fully loaded conditions, respectively. The second suspension setting has a higher stiffness than the first suspension setting to accommodate the larger sprung mass under full load. The first acceleration limit and the second acceleration limit are the maximum permissible longitudinal acceleration of the vehicle set for different load states. The second acceleration limit is lower than the first acceleration limit to ensure a safe braking distance under full load. The reverse yaw moment refers to the moment around the vehicle's vertical axis generated by adjusting the difference in driving or braking forces between the left and right wheels, used to counteract the lateral deviation tendency caused by off-center load.

[0056] In this embodiment, the current load and center of gravity position of the vehicle are estimated in real time by using displacement sensors installed in the suspension system or pressure sensor arrays at the bottom of the vehicle body, thereby determining the load status. When the vehicle is determined to be unloaded, a first suspension setting is adopted, that is, the suspension damping coefficient is set to a low value (such as 70% of the standard value), making the suspension softer to improve the comfort of unloaded driving. At the same time, the first acceleration limit is set to no more than 0.2g, which is calculated based on the unloaded mass and tire adhesion coefficient to ensure that the drive wheels do not slip during acceleration. When the vehicle is determined to be fully loaded, the system switches to a second suspension setting, increasing the suspension damping coefficient to the range of 120%~130% of the standard value, making the suspension stiffer to support a larger load and suppress vehicle body sway. At the same time, the second acceleration limit is set to no more than 0.3g, which is determined based on the full-load braking distance requirement and the peak torque limit of the motor. When an off-center load is detected, such as when the cargo is concentrated on the left side causing a shift in the left center of gravity exceeding 50mm, the system actively adjusts the distribution of driving or braking force between the left and right wheels. For example, it appropriately increases the driving force ratio of the right wheel during acceleration or increases the braking force ratio of the right wheel during deceleration, thereby generating a rightward yaw moment to counteract the tendency to veer to the left caused by the left-side imbalance. This off-center load compensation is calculated in real-time using a dynamic model based on the center of gravity offset distance and the vehicle's wheelbase.

[0057] As can be seen from the above, this embodiment achieves comprehensive adaptation to different load conditions by sensing the vehicle load status in real time and adaptively adjusting suspension parameters, acceleration limit, and left and right wheel power distribution. When unloaded, it emphasizes comfort and responsiveness; when fully loaded, it emphasizes support and braking safety; and when unevenly loaded, it actively compensates for yaw moment to prevent deviation, thereby improving the driving stability, handling safety, and cargo transportation quality of the intelligent vehicle across the entire load range.

[0058] Corresponding to the intelligent vehicle control method for subway logistics distribution in the above embodiment, Figure 2 This is a structural block diagram of a subway logistics distribution intelligent vehicle control device according to an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 2 The intelligent vehicle control device 20 for subway logistics distribution includes: a road surface perception module 21, a road surface change calculation module 22, a pose estimation module 23, a motion trajectory generation module 24, a confidence calculation module 25, a selection module 26, and a control module 27.

[0059] The road perception module 21 is used to acquire road perception data in front of the intelligent vehicle. The road perception data includes data collected by lidar and data collected by millimeter-wave radar. The road surface change calculation module 22 is used to determine the height change information of the road surface ahead based on the road surface perception data; The pose estimation module 23 is used to determine that an obstacle crossing action is about to occur when the height difference on the road ahead is determined based on the height change information and exceeds the preset height threshold, and to lock the current pose of the intelligent vehicle as the reference pose. The motion trajectory generation module 24 is used to acquire the inertial data of the intelligent vehicle during the obstacle crossing process and determine the motion trajectory of the intelligent vehicle during the obstacle crossing process based on the inertial data. The confidence calculation module 25 is used to calculate the pose of the intelligent vehicle after the obstacle crossing based on the motion trajectory and the reference pose, and generate multiple pose hypotheses and their corresponding confidence scores. Module 26 is used to select a target position hypothesis from multiple pose hypotheses based on each confidence score, as the target pose of the intelligent vehicle after the obstacle crossing is completed. Control module 27 is used to control the driving of intelligent vehicles based on target pose.

[0060] In one embodiment of this application, module 26 is selected and is specifically used for: Based on the confidence scores, the pose hypothesis with the highest confidence score is selected from multiple pose hypotheses as the target pose of the intelligent vehicle after the obstacle crossing is completed. Among them, controlling the driving of intelligent vehicles based on target pose includes: Based on the target pose, and combined with a preset control strategy, the intelligent vehicle is controlled to drive. The preset control strategies include: During the process of controlling the intelligent vehicle to drive, the driving speed is less than the preset speed threshold and the steering angle is less than the preset steering angle threshold.

[0061] In one embodiment of this application, during the control of the intelligent vehicle's operation, the subway logistics delivery intelligent vehicle control device 20 further includes: an adjustment module, used for: If the intelligent vehicle is detected to be accelerating, the damping coefficient of the rear suspension shock absorber and / or the rear wheel drive force distribution ratio are adjusted. If the intelligent vehicle is detected to be decelerating, the damping coefficient of the front suspension shock absorber and / or the front wheel braking force distribution ratio are adjusted. The determination of whether an intelligent vehicle is accelerating or decelerating is based on its longitudinal acceleration.

[0062] In one embodiment of this application, the acceleration state includes a gradual acceleration state, a rapid acceleration state, and a fully loaded rapid acceleration state. The adjustment module is specifically used for: If the intelligent vehicle is detected to be accelerating smoothly and the road surface is flat, the damping coefficient of the rear suspension shock absorber is reduced while maintaining the standard rear-wheel drive force distribution ratio. If the intelligent vehicle is detected to be accelerating rapidly, the rear-wheel drive force distribution ratio is increased while the damping coefficient of the standard rear suspension shock absorber is maintained. If the system detects that the vehicle is under full load and accelerating rapidly, and the road surface is uneven, the damping coefficient of the rear suspension shock absorber will be reduced and the rear wheel drive force distribution ratio will be increased.

[0063] In one embodiment of this application, the deceleration state includes a smooth deceleration state, a rapid deceleration state, and a fully loaded rapid deceleration state. The adjustment module is also specifically used for: If the deceleration is detected to be gradual and the road surface is flat, the damping coefficient of the front suspension shock absorber is reduced while maintaining the standard front wheel braking force distribution ratio. If a rapid deceleration is detected, the front wheel braking force distribution ratio is increased while maintaining the damping coefficient of the standard front suspension shock absorbers. If the deceleration state is detected to be a state of rapid deceleration under full load, and the current road surface is uneven, then the operation of reducing the damping coefficient of the front suspension shock absorber and increasing the front wheel braking force distribution ratio will be executed simultaneously.

[0064] In one embodiment of this application, the intelligent vehicle control device 20 for subway logistics distribution further includes: a vehicle body detection module, used for: Acquire changes in the pitch angle of the intelligent vehicle; When the vehicle pitch angle exceeds the preset angle threshold, the damping coefficient adjustment range and / or the power distribution ratio adjustment range are increased until the vehicle pitch angle returns to the preset safe range.

[0065] In one embodiment of this application, the intelligent vehicle control device 20 for subway logistics distribution further includes: a load detection module, used for: Obtain the load status of the intelligent vehicle, including no-load, full-load, and off-center load status; If the load state of the intelligent vehicle is unloaded, the first suspension setting and the first acceleration limit are used to control the operation of the intelligent vehicle. If the intelligent vehicle is fully loaded, the second suspension setting and the second acceleration limit are used to control the operation of the intelligent vehicle; the stiffness of the second suspension setting is higher than that of the first suspension setting, and the second acceleration limit is lower than that of the first acceleration limit. If the load state of the intelligent vehicle is off-center, the driving force of the left or right wheel or the braking force of the left or right wheel will be adjusted based on the off-center load direction to generate a reverse yaw moment.

[0066] See Figure 3 , Figure 3 This is a schematic block diagram of an intelligent vehicle provided in one embodiment of this application. Figure 3The intelligent vehicle 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 2 The functions of the road surface perception module 21, road surface change calculation module 22, pose estimation module 23, motion trajectory generation module 24, confidence calculation module 25, selection module 26, and control module 27 are shown.

[0067] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0068] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0069] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store information such as road perception data and inertial data.

[0070] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the intelligent vehicle control method for subway logistics distribution provided in the embodiments of this application, or they can execute the implementation method of intelligent vehicles described in the embodiments of this application, which will not be repeated here.

[0071] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0072] The computer-readable storage medium can be an internal storage unit of the intelligent vehicle in any of the foregoing embodiments, such as a hard drive or memory of the intelligent vehicle. The computer-readable storage medium can also be an external storage device of the intelligent vehicle, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped in the intelligent vehicle. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the intelligent vehicle. The computer-readable storage medium is used to store computer programs and other programs and data required by the intelligent vehicle. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0073] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0074] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the intelligent vehicle and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0075] In the several embodiments provided in this application, it should be understood that the disclosed intelligent vehicles and methods can be implemented in other ways. For example, the device 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, modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connections shown or discussed may be indirect coupling or communication connections through some interfaces or units, or they may be electrical, mechanical, or other forms of connection.

[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0077] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The integrated modules described above can be implemented in hardware or as software functional units.

[0078] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered 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 method for controlling intelligent vehicles used in subway logistics distribution, executed by intelligent vehicles, characterized in that, include: The road surface perception data in front of the intelligent vehicle is acquired, including data collected by lidar and data collected by millimeter-wave radar. Based on the road surface perception data, determine the height change information of the road surface ahead; When it is determined from the height change information that there is a height difference on the road ahead that exceeds a preset height threshold, it is determined that an obstacle crossing action is about to occur, and the current pose of the intelligent vehicle is locked as the reference pose. Acquire the inertial data of the intelligent vehicle during obstacle crossing, and determine the trajectory of the intelligent vehicle during obstacle crossing based on the inertial data; Based on the motion trajectory and combined with the reference pose, the pose of the intelligent vehicle after the obstacle crossing is calculated, and multiple pose hypotheses and their corresponding confidence scores are generated. Based on each confidence score, a target position hypothesis is selected from the plurality of pose hypotheses as the target pose of the intelligent vehicle after the obstacle crossing is completed. The intelligent vehicle is controlled to drive based on the target pose.

2. The intelligent vehicle control method for subway logistics distribution as described in claim 1, characterized in that, The step of selecting a target position hypothesis from the plurality of pose hypotheses based on each confidence score, as the target pose of the intelligent vehicle after obstacle crossing, includes: Based on each confidence score, the pose hypothesis with the highest confidence score is selected from the plurality of pose hypotheses as the target pose of the intelligent vehicle after the obstacle crossing is completed. The step of controlling the intelligent vehicle's movement based on the target pose includes: Based on the target pose, and in conjunction with a preset control strategy, the intelligent vehicle is controlled to drive. The preset control strategy includes: During the process of controlling the intelligent vehicle to drive, the driving speed is less than a preset speed threshold and the steering angle is less than a preset steering angle threshold.

3. The intelligent vehicle control method for subway logistics distribution as described in claim 1, characterized in that, In controlling the driving of the intelligent vehicle, the method further includes: If the intelligent vehicle is detected to be accelerating, the damping coefficient of the rear suspension shock absorber and / or the rear wheel drive force distribution ratio are adjusted. If the intelligent vehicle is detected to be decelerating, the damping coefficient of the front suspension shock absorber and / or the front wheel braking force distribution ratio are adjusted. The state of the intelligent vehicle, whether it is accelerating or decelerating, is determined based on its longitudinal acceleration.

4. The intelligent vehicle control method for subway logistics distribution as described in claim 3, characterized in that, The acceleration states include gradual acceleration, rapid acceleration, and full-load rapid acceleration. If the intelligent vehicle is detected to be accelerating, the damping coefficient of the rear suspension shock absorber and / or the rear wheel drive force distribution ratio are adjusted, including: If the intelligent vehicle is detected to be in a state of gentle acceleration and the current road surface is flat, the damping coefficient of the rear suspension shock absorber is reduced, while maintaining the standard rear-wheel drive force distribution ratio. If the intelligent vehicle is detected to be in a state of rapid acceleration, the rear wheel drive force distribution ratio is increased while the damping coefficient of the standard rear suspension shock absorber is maintained. If the acceleration state is detected to be under full load and rapid acceleration, and the current road surface is uneven, the damping coefficient of the rear suspension shock absorber will be reduced and the rear wheel drive force distribution ratio will be increased.

5. The intelligent vehicle control method for subway logistics distribution as described in claim 3, characterized in that, The deceleration states include gentle deceleration, rapid deceleration, and full-load rapid deceleration. If the intelligent vehicle is detected to be decelerating, the damping coefficient of the front suspension shock absorber and / or the front wheel braking force distribution ratio are adjusted, specifically including: If the deceleration state is detected to be a gentle deceleration state and the current road surface is flat, the damping coefficient of the front suspension shock absorber is reduced, while maintaining the standard front wheel braking force distribution ratio. If the deceleration state is detected to be a rapid deceleration state, the front wheel braking force distribution ratio is increased while maintaining the damping coefficient of the standard front suspension shock absorber. If the deceleration state is detected to be a state of rapid deceleration under full load, and the current road surface is uneven, then the operation of reducing the damping coefficient of the front suspension shock absorber and increasing the front wheel braking force distribution ratio will be executed simultaneously.

6. The intelligent vehicle control method for subway logistics distribution as described in claim 4 or 5, characterized in that, Also includes: Obtain the change in the vehicle's pitch angle; When the vehicle pitch angle exceeds a preset angle threshold, the damping coefficient adjustment range and / or the power distribution ratio adjustment range are increased until the vehicle pitch angle returns to the preset safe range.

7. The intelligent vehicle control method for subway logistics distribution as described in claim 1, characterized in that, Also includes: The load status of the intelligent vehicle is obtained, including no-load status, full-load status, and off-center load status. If the load state of the intelligent vehicle is unloaded, the operation of the intelligent vehicle is controlled by the first suspension setting and the first acceleration limit. If the intelligent vehicle is in a fully loaded state, the operation of the intelligent vehicle is controlled by a second suspension setting and a second acceleration limit; the stiffness of the second suspension setting is higher than that of the first suspension setting, and the second acceleration limit is lower than that of the first acceleration limit; If the load state of the intelligent vehicle is off-center, the driving force of the left or right wheel or the braking force of the left or right wheel is adjusted based on the off-center load direction to generate a reverse yaw moment.

8. A smart vehicle control device for subway logistics distribution, characterized in that, include: The road perception module is used to acquire road perception data in front of the intelligent vehicle. The road perception data includes data collected by lidar and data collected by millimeter-wave radar. The road surface change calculation module is used to determine the height change information of the road surface ahead based on the road surface sensing data. The pose estimation module is used to determine that an obstacle crossing action is about to occur when the height difference on the road ahead is determined based on the height change information and exceeds a preset height threshold, and to lock the current pose of the intelligent vehicle as the reference pose. The motion trajectory generation module is used to acquire the inertial data of the intelligent vehicle during obstacle crossing and determine the motion trajectory of the intelligent vehicle during obstacle crossing based on the inertial data. The confidence calculation module is used to calculate the pose of the intelligent vehicle after the obstacle crossing based on the motion trajectory and the reference pose, and generate multiple pose hypotheses and their corresponding confidence scores. The selection module is used to select a target position hypothesis from the plurality of pose hypotheses based on each confidence score, as the target pose of the intelligent vehicle after the obstacle crossing is completed. The control module is used to control the driving of the intelligent vehicle based on the target pose.

9. An intelligent vehicle, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images