Safety device for reducing back-up collision risk in automatic driving test process

Abstract

The invention discloses a safety device for reducing the back-up collision risk in the automatic driving test process, and belongs to the technical field of automatic driving test safety. Aiming at the collision risk problem caused by rapid movement of dynamic obstacles and communication delay when a vehicle backs up in a closed test site, the device obtains millimeter-level position coordinates and course angles through positioning units deployed on a test vehicle and a target object; a C-V2X protocol is utilized to construct a real-time data link to share vehicle position, speed, backing track prediction data and target object outer contour feature point coordinates, potential conflicts are identified based on future minimum spacing calculation and dynamic threshold, and a cooperative collision avoidance path is generated in combination with a distributed optimization algorithm. And the high-risk vehicle preferentially preempts a communication channel to issue deceleration, course adjustment and braking instructions to surrounding vehicles. The device can effectively improve the reversing safety in a multi-vehicle cooperative test scene, and is suitable for collision risk real-time early warning and active intervention of an automatic driving vehicle closed test site.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving test safety technology, and in particular to a safety device for reducing the risk of reverse collision during autonomous driving testing. Background Art

[0002] In closed-field autonomous driving tests, collision risk control in reversing scenarios is one of the key technical challenges. In the existing technology, there are still several deficiencies in the anti-collision system for multi-vehicle cooperative reversing tests. First, the positioning accuracy is difficult to meet the requirements of high-dynamic scenarios. Traditional positioning solutions mostly rely on a single sensor (such as GPS or low-cost lidar), but in closed venues, GPS signals are easily blocked or reflected by buildings, resulting in positioning errors increasing to the decimeter level, and it is impossible to accurately capture the real-time relative position of the vehicle and the obstacle. For example, during the reversing process, the deviation in the distance measurement between the vehicle and the adjacent dynamic obstacle may cause misjudgment, especially under the millimeter-level avoidance requirements, the existing system is difficult to support accurate trajectory prediction. Secondly, communication delay and channel contention are prominent. In the existing vehicle-to-everything (V2X) communication, in intensive test scenarios, multiple vehicles simultaneously transmitting data can easily lead to channel congestion, and the conventional priority allocation mechanism cannot dynamically respond to sudden risks. For example, when the reversing trajectories of multiple test vehicles intersect, a communication delay of more than 30ms may cause the coordinated instructions to fail to execute synchronously, exacerbating the risk of collision. In addition, the existing collision prediction model is not adaptable enough to dynamic obstacles. Traditional methods use fixed safety thresholds and do not consider the dynamic changes of the target size (such as length and width parameters W, L), resulting in overly conservative or overly loose threshold settings. For example, when the size of obstacles varies greatly, the fixed threshold may not accurately reflect the actual safety boundary, resulting in missed reports or false reports. Furthermore, the global optimization capability of multi-vehicle collaborative collision avoidance strategies is limited. Existing systems are mostly based on single-vehicle decisions and lack a distributed coordination mechanism, which may cause new conflicts due to local avoidance paths. For example, after vehicle A adjusts its path to avoid an obstacle, it may infringe on the preset safety area of ​​vehicle B, and existing technologies make it difficult to coordinate such chain reactions in real time. Finally, the rapid movement of dynamic obstacles places higher requirements on the real-time performance of the system. Traditional trajectory prediction algorithms have high computational complexity, and it is difficult to complete multi-target trajectory calculations and risk assessments within milliseconds in dense target scenarios, resulting in delayed responses. The root causes of the above problems are: the data fusion algorithm for high-precision multi-source positioning is not robust enough in complex environments (such as changing lighting and dense targets); the communication resource allocation lacks a risk-adaptive dynamic scheduling mechanism; the collision risk model does not deeply integrate the physical parameters and motion state of the target; and the real-time performance and global convergence of the distributed collaborative optimization algorithm are difficult to balance. These defects make the existing system face risks such as missed collision warnings and asynchronous collision avoidance instructions in multi-vehicle reversing tests in closed venues, which restricts the improvement of test safety and efficiency. Summary of the invention

[0003] One purpose of the present invention is to solve the risk of reverse collision caused by insufficient positioning accuracy, communication delay and poor adaptability to dynamic obstacles in closed-field multi-vehicle testing. Existing systems cannot achieve millimeter-level positioning, adaptive adjustment of dynamic thresholds and real-time response of multi-vehicle collaborative collision avoidance.

[0004] Another object of the present invention is to ensure the synchronous execution of coordinated deceleration, heading adjustment and emergency braking instructions in multi-vehicle scenarios, and to solve the problem that traditional control instructions lack a hierarchical response mechanism, resulting in delayed collision avoidance actions or intensified conflicts.

[0005] Another object of the present invention is to define the constraints of the collaborative collision avoidance path to balance safety and vehicle kinematic limitations, thereby solving the problem that the existing path planning does not combine the minimum turning radius of the vehicle and the dynamic safety threshold, and is prone to generating an infeasible path.

[0006] Another object of the present invention is to improve the positioning accuracy in complex environments through multi-source fusion positioning, and to solve the problem that a single sensor (such as GPS or lidar) is prone to failure when the signal is blocked or the environment is interfered with, resulting in positioning deviation.

[0007] Another object of the present invention is to solve the problem of maintaining positioning stability when the GPS signal is lost or the sensor fails, and the existing positioning system lacks a dynamic anti-interference mechanism and cannot quickly switch to a backup positioning mode.

[0008] Another purpose of the present invention is to optimize the robustness of the visual SLAM algorithm under lighting changes and target dense scenes, and to solve the problem that traditional visual SLAM relies on fixed feature weights and is prone to tracking failure or mapping errors in complex environments.

[0009] Another purpose of the present invention is to achieve rapid recovery of positioning when visual SLAM continuous tracking fails, and to solve the problem that the existing system relies on single sensor data and lacks a multi-source collaborative repair mechanism, resulting in the risk of positioning interruption.

[0010] Another object of the present invention is to ensure communication reliability and real-time performance in high-density test scenarios, and to solve the problem that the traditional C-V2X protocol cannot dynamically allocate resources when the channel is congested, resulting in data transmission delays or losses.

[0011] Another object of the present invention is to realize a mechanism for dynamically adjusting communication priority and instruction synchronization based on collision risk, so as to solve the problem that the existing algorithm does not quantify the risk value as a basis for channel preemption and is difficult to achieve millisecond-level strong synchronous response.

[0012] Another object of the present invention is to ensure strict timing of instruction execution through hardware-level interrupts and preloaded parameters, thereby solving the problem that traditional control signals rely on software layer processing and are difficult to meet the synchronization error tolerance of 20-25ms.

[0013] In order to achieve the above object, the present invention adopts the following scheme: Safety devices used to reduce the risk of reverse collision during autonomous driving testing, including: Positioning units are deployed on all test vehicles and risk targets. The positioning units obtain the millimeter-level position coordinates and heading angles of each vehicle in real time. The communication unit based on the C-V2X protocol builds a real-time data link between vehicles and between vehicles and objects based on the position coordinates and heading angles obtained by the positioning unit to share the following information: the real-time position, speed, and reverse trajectory prediction data of each vehicle, as well as the coordinates of the outer contour feature points and safety boundary parameters of the risky object; The communication unit receives the reverse trajectory prediction data uploaded by all vehicles, the coordinates of the outer contour feature points of the risk target and the safety boundary parameters, and performs the following operations: Based on the predicted reversing trajectory data, the future minimum distance between vehicles and between vehicles and risk targets is calculated. If the future minimum distance is less than the dynamic threshold D, it is marked as a potential conflict. The dynamic threshold D ranges from 1.3 to 1.5 times the larger value of the width and length of the risk target. Through a distributed optimization algorithm, a collaborative collision avoidance path is dynamically generated based on the position and speed data of each vehicle. At the same time, a dynamic priority algorithm is used to calculate the collision risk value of the vehicle in real time based on the dynamically generated collaborative collision avoidance path and potential conflict data, and communication channel resources are preferentially allocated to vehicles with the highest collision risk value. Synchronous adjustment instructions are sent to their adjacent vehicles. The synchronous adjustment instructions include deceleration, adjustment of heading angle and braking, and force adjacent vehicles to respond to and execute the synchronous adjustment instructions within 20-25ms.

[0014] As a preferred embodiment, the specific response method for issuing the synchronization adjustment instruction is: When a deceleration control command is issued, a coordinated deceleration command is sent to all associated vehicles through V2X communication, so that the driving speed of each vehicle is synchronously reduced to a preset safety value; When issuing a control command to adjust the heading angle, the heading angle of each vehicle is dynamically adjusted based on the optimized path to ensure that its trajectory deviates from the safety boundary of the risk target; When issuing a braking control command, if the path conflict cannot be eliminated, the cascade braking protocol is triggered, and emergency braking is performed in sequence from the nearest to the conflict point.

[0015] Preferably, the constraints used when dynamically generating a collaborative collision avoidance path for each vehicle include: The minimum distance between the reversing paths of each vehicle ≥ the dynamic threshold D; and The curvature radius of the collision avoidance path ≥ the minimum turning radius of the vehicle.

[0016] Preferably, the positioning unit includes a multi-source fusion positioning submodule, which integrates GPS, laser radar point cloud matching and visual SLAM algorithm, and performs millimeter-level positioning through the following steps: The initial position coordinates are obtained through GPS, and the feature point cloud of the surrounding environment is scanned by lidar, which is matched with the high-precision preset site map in real time to correct the position deviation; based on the image data collected by the on-board camera, the visual SLAM algorithm is used to generate a local dense map to calibrate the heading angle.

[0017] Preferably, the positioning unit further includes a dynamic anti-interference submodule, which performs the following operations when the signal is interfered with: When it detects that the GPS signal is blocked, it starts the collaborative positioning mode based on vehicle-to-vehicle V2X communication, receives the positioning data of adjacent vehicles, and calculates its own relative position through the least squares method; If the LiDAR point cloud fails to match for at least three consecutive frames, it switches to pure visual inertial odometry (VIO) mode, fusing IMU data with visual feature tracking results to maintain positioning accuracy. Real-time self-calibration is performed and multi-source data consistency check is executed every 30 seconds. When the positioning deviation between GPS and lidar exceeds 2 cm, the global repositioning process is triggered to correct the positioning deviation.

[0018] Preferably, the method using the visual SLAM algorithm comprises the following steps: Set up a dynamic feature weight allocation submodule, which dynamically adjusts the extraction weight of visual feature points according to the ambient light intensity and the density of risk targets. The specific operation mode is as follows: If the ambient light intensity is strong or low, high-contrast edge feature points are extracted first, and the extraction ratio and weight of texture features are reduced to reduce texture dependence; when the density of risk targets reaches the high-density threshold, the semantic segmentation network is enabled to identify the passable area, and only the ground plane feature points are retained for SLAM mapping; Set up a multi-sensor tight coupling calibration submodule, which does the following: The local dense map generated by visual SLAM is aligned with the lidar point cloud frame by frame, and the heading angle deviation is calculated using the ICP algorithm. When the heading angle deviation exceeds 0.5°, the extended Kalman filter EKF is constructed by integrating the IMU angular velocity data for coupling calibration, and the calibrated heading angle is output.

[0019] Preferably, the operation logic of the visual SLAM algorithm also includes an abnormal state self-repair mechanism, which operates as follows: When visual SLAM tracking fails for 5 consecutive frames, it switches to the collaborative positioning mode based on V2X communication, receives the heading angle data of adjacent vehicles and performs weighted averaging, reversely infers its own posture through the outer contour feature points of the risk target, and resets the SLAM initialization parameters.

[0020] Preferably, the communication unit comprises: The dynamic channel optimization submodule performs the following operations: monitor the load status and interference intensity of the communication channel in real time, dynamically allocate time-frequency resource blocks based on the reinforcement learning algorithm, and prioritize the data transmission of vehicles with collision risk values ​​above the risk threshold according to the collision risk values ​​of the vehicles calculated in real time; start the multi-hop relay transmission mode in the scenario where there are ≥20 test vehicles and / or ≥5 risk targets within a radius of 100 meters: set the central decision node, and transmit the vehicle data that is more than 200 meters away from the central decision node in segments through the relay nodes of adjacent vehicles, with a maximum delay of ≤10ms per hop; the relay node selection criteria are signal strength ≥-90dBm and remaining power >30%; The data priority encryption submodule encrypts the transmitted data in a hierarchical manner: quantum key distribution QKD is used to encrypt collision warning data, and the lightweight national encryption algorithm SM4 is used to encrypt position and speed status data; Among them, the central decision node dynamically adjusts the number of iterations of the path optimization algorithm according to the communication delay and packet loss rate; if the packet loss rate is greater than 5%, the local collision avoidance mode is triggered so that each vehicle only relies on local sensor data to perform emergency braking.

[0021] Preferably, the method of calculating the collision risk value of the vehicle and allocating communication channel resources is: Calculate the collision risk value Risk based on dynamic weighting: Among them, θ is the vehicle heading angle deviation, R is the path curvature radius, α, β, γ are the environment adaptation weight coefficients, and the values ​​are: α=8, β=2, γ=1, TTC is the collision time, v is the vehicle speed, and D is the dynamic threshold; Communication channel resources are allocated through a channel preemption hierarchical strategy, specifically: Vehicles with a collision risk value of Risk ≥ 8 occupy a dedicated emergency channel, the bandwidth of which is set to 10 MHz, and the instruction transmission delay is ≤ 15 ms; vehicles with a collision risk value of 5 ≤ Risk < 8 share a high priority channel, the bandwidth of which is set to 5 MHz, and the delay is ≤ 25 ms; The command strong synchronization mechanism is used when forcing adjacent vehicles to execute synchronous adjustment commands. The operation mode of the mechanism is: after the adjacent vehicle receives the command, the control signal is triggered through the hardware-level interrupt, the coordination parameters are preloaded in the reversing motor controller and the steering system, the trajectory correction or braking response time is set within 20-25ms, and the error tolerance is ≤0.2 meters.

[0022] Preferably, the specific operation mode of the instruction strong synchronization mechanism includes: The hardware-level interrupt trigger module is directly connected to the underlying hardware interface between the reversing motor controller and the steering system, and a priority interrupt signal is generated through a programmable logic device to forcibly interrupt the current control instruction and load the coordination parameters; The coordination parameters are preloaded in the following manner: the coordination parameters generated by the dynamic priority algorithm are written into the controller and the steering system by the coordination parameter preloading module before the command is issued, wherein the coordination parameters include the target deceleration gradient, the heading angle correction value and the braking pressure threshold; Set up a timing constraint module, which aligns the start time of command execution based on the inter-vehicle clock synchronization protocol, and monitors the entire delay from command reception to execution completion through a hardware timer, strictly controlling the delay within the range of 20-25ms; An error tolerance verification module is set up, which feeds back the actual displacement in real time through the positioning unit after trajectory correction or braking is completed. If the deviation from the target displacement exceeds 0.2 meters, the emergency braking locking protocol is triggered.

[0023] The present invention has at least the following beneficial effects: (1) Through multi-source fusion positioning (GPS, LiDAR, visual SLAM) and dynamic anti-interference mechanism (cooperative positioning, VIO mode switching), millimeter-level positioning accuracy is achieved in complex environments (such as GPS signal obstruction, illumination change), and the heading angle deviation is calibrated in real time, which significantly improves the accuracy of the relative position measurement between the vehicle and the obstacle, and provides a reliable data basis for collision prediction and collision avoidance path planning; (2) Based on the dynamic channel optimization and multi-hop relay transmission mode of the C-V2X protocol, combined with the reinforcement learning algorithm, the communication resources of high-risk vehicles are preferentially allocated to ensure that the command transmission delay is ≤25ms in high-density test scenarios (≥20 vehicles / 100 meters). At the same time, hierarchical encryption (QKD, SM4) is used to ensure data transmission security, effectively solving the channel congestion and data loss problems of traditional V2X systems; (3) The distributed optimization algorithm is used to dynamically generate multi-vehicle collaborative collision avoidance paths, and the dynamic priority algorithm is combined to quantify the collision risk value (Risk), prioritize the communication channel to issue commands, and pass Through layered control strategies such as cascade braking and heading angle correction, global synchronization of multi-vehicle actions (20-25ms response) is achieved to avoid chain conflicts caused by single-vehicle collision avoidance; (4) The safety threshold is dynamically adjusted based on the target size (W, L), combined with multi-parameter risk value calculation models such as collision time (TTC) and heading angle deviation, to accurately identify potential conflicts, reduce misjudgments caused by fixed thresholds, and improve the adaptability and reliability of collision warnings; (5) The underlying controller is directly controlled through the hardware interrupt module, and the coordination parameters (deceleration gradient, braking pressure threshold) are preloaded. Combined with the clock synchronization protocol and hardware timer monitoring, the delay of the entire command execution process is strictly controlled within 20-25ms, and the error tolerance is ≤0.2 meters, which greatly improves the real-time control and execution consistency in emergency scenarios; when the visual SLAM fails or the sensor is abnormal, it quickly switches to the V2X collaborative positioning mode, and reversely infers the position based on the outer contour features of the risk target to achieve positioning self-recovery, ensure the continuity of the test process and the fault tolerance of the system, and reduce the need for manual intervention.

[0024] Together, these effects solve core challenges such as positioning deviation, communication delay, path conflict and environmental mutation in multi-vehicle reversing tests in closed venues, significantly improving the safety and efficiency of autonomous driving tests. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A principle block diagram of a safety device provided by the present invention for reducing the risk of reversing collision during autonomous driving testing. DETAILED DESCRIPTION

[0026] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0027] Safety devices used to reduce the risk of reverse collision during autonomous driving testing, including: Positioning units are deployed on all test vehicles and risk targets. The positioning units obtain the millimeter-level position coordinates and heading angles of each vehicle in real time. The communication unit based on the C-V2X protocol builds a real-time data link between vehicles and between vehicles and objects based on the position coordinates and heading angles obtained by the positioning unit to share the following information: the real-time position, speed, and reverse trajectory prediction data of each vehicle, as well as the coordinates of the outer contour feature points and safety boundary parameters of the risky object; The communication unit receives the reverse trajectory prediction data uploaded by all vehicles, the coordinates of the outer contour feature points of the risk target and the safety boundary parameters, and performs the following operations: Based on the predicted reversing trajectory data, the future minimum distance between vehicles and between vehicles and risk targets is calculated. If the future minimum distance is less than the dynamic threshold D, it is marked as a potential conflict. The dynamic threshold D ranges from 1.3 to 1.5 times the larger value of the width and length of the risk target. Through a distributed optimization algorithm, a collaborative collision avoidance path is dynamically generated based on the position and speed data of each vehicle. At the same time, a dynamic priority algorithm is used to calculate the collision risk value of the vehicle in real time based on the dynamically generated collaborative collision avoidance path and potential conflict data, and communication channel resources are preferentially allocated to vehicles with the highest collision risk value. Synchronous adjustment instructions are sent to their adjacent vehicles. The synchronous adjustment instructions include deceleration, adjustment of heading angle and braking, and force adjacent vehicles to respond to and execute the synchronous adjustment instructions within 20-25ms.

[0028] Through millimeter-level positioning and C-V2X real-time communication, a vehicle-target full-link data sharing mechanism is built, combined with dynamic thresholds (adaptively adjusted based on the target size) to accurately identify potential conflicts, and through distributed optimization and dynamic priority algorithms, multi-vehicle collaborative collision avoidance is achieved. Its advantages are: 1) The dynamic threshold D=1.3~1.5·max(W,L) comprehensively considers the size of the target to avoid missed reports or false reports caused by fixed thresholds; 2) The distributed optimization algorithm generates collision avoidance paths on a global scale to avoid chain conflicts caused by single vehicle decisions; 3) The dynamic priority algorithm seizes channel resources based on risk values ​​to ensure that high-risk commands are transmitted first, and the command response time is compressed to 20-25ms, significantly reducing the impact of communication delays on safety.

[0029] In another technical solution, the specific response method for issuing the synchronization adjustment instruction is: When a deceleration control command is issued, a coordinated deceleration command is sent to all associated vehicles through V2X communication, so that the driving speed of each vehicle is synchronously reduced to a preset safety value; When issuing a control command to adjust the heading angle, the heading angle of each vehicle is dynamically adjusted based on the optimized path to ensure that its trajectory deviates from the safety boundary of the risk target; When issuing a braking control command, if the path conflict cannot be eliminated, the cascade braking protocol is triggered, and emergency braking is performed in sequence from the nearest to the conflict point.

[0030] The refined execution of multi-vehicle coordinated actions is achieved through hierarchical control instructions (deceleration, heading adjustment, braking). The coordinated deceleration instruction ensures that the vehicle speed is synchronously reduced to a safe range to avoid secondary collisions caused by speed differences; the heading angle is dynamically adjusted to correct the trajectory in real time based on the optimized path to ensure that the vehicle deviates from the safety boundary of the target object; the cascade braking protocol triggers braking according to the priority of the distance from the conflict point to reduce the interference of emergency braking on the overall test scenario. This solution solves the problem of the lack of hierarchical response of traditional instructions. Especially when the path conflict cannot be eliminated, cascade braking can maximize the reduction of collision energy and improve the reliability of collision avoidance in complex scenarios.

[0031] The constraints used when dynamically generating a collaborative collision avoidance path for each vehicle include: the minimum spacing between the reversing paths of each vehicle ≥ the dynamic threshold D; and the collision avoidance path curvature radius ≥ the vehicle's minimum turning radius.

[0032] By introducing path constraints (minimum spacing and curvature radius), the generated collision avoidance path is ensured to meet both the dynamic safety threshold and the vehicle kinematic constraints. The minimum spacing constraint avoids the close-range risk caused by the intersection of multiple vehicle paths; the curvature radius constraint ensures the feasibility of the path and prevents control failure caused by insufficient turning radius. This solution makes up for the defect of the separation of safety and feasibility in traditional path planning, and is particularly suitable for high-precision reversing scenarios. It can reduce the frequency of path replanning and improve the overall efficiency of the system.

[0033] In another technical solution, the positioning unit includes a multi-source fusion positioning submodule, which integrates GPS, laser radar point cloud matching and visual SLAM algorithm, and performs millimeter-level positioning through the following steps: The initial position coordinates are obtained through GPS, and the feature point cloud of the surrounding environment is scanned by lidar, which is matched with the high-precision preset site map in real time to correct the position deviation; based on the image data collected by the on-board camera, the visual SLAM algorithm is used to generate a local dense map to calibrate the heading angle.

[0034] Millimeter-level accuracy in complex environments is achieved through multi-source fusion positioning (GPS, LiDAR, and visual SLAM). GPS provides initial coordinates, LiDAR point cloud matching corrects map deviations, and visual SLAM further calibrates the heading angle. The three complement each other to overcome the limitations of a single sensor. For example, when the GPS signal is blocked, LiDAR and visual SLAM can maintain positioning continuity; in low-light environments, LiDAR point cloud matching prioritizes accuracy. This solution significantly improves the robustness of the positioning system, and is particularly suitable for scenarios with dense dynamic obstacles in closed venues.

[0035] The positioning unit also includes a dynamic anti-interference submodule that performs the following operations when the signal is interfered with: When it detects that the GPS signal is blocked, it starts the collaborative positioning mode based on vehicle-to-vehicle V2X communication, receives the positioning data of adjacent vehicles, and calculates its own relative position through the least squares method; If the LiDAR point cloud fails to match for at least three consecutive frames, it switches to pure visual inertial odometry (VIO) mode, fusing IMU data with visual feature tracking results to maintain positioning accuracy. Real-time self-calibration is performed and multi-source data consistency check is executed every 30 seconds. When the positioning deviation between GPS and lidar exceeds 2 cm, the global repositioning process is triggered to correct the positioning deviation.

[0036] The dynamic anti-interference mechanism ensures the stability of the positioning system when the sensor fails. The collaborative positioning mode uses V2X communication to obtain neighboring vehicle data and calculates the relative position through the least squares method to make up for the loss of GPS signals. The VIO mode integrates IMU and visual data when the lidar fails to maintain positioning accuracy. The real-time self-calibration logic regularly verifies the consistency of multi-source data and triggers repositioning to eliminate accumulated errors. This solution solves the risk of downtime of traditional positioning systems when signals are interrupted, and ensures continuous and reliable operation in high-dynamic scenarios.

[0037] The way to use the visual SLAM algorithm includes the following steps: Set up a dynamic feature weight allocation submodule, which dynamically adjusts the extraction weight of visual feature points according to the ambient light intensity and the density of risk targets. The specific operation mode is as follows: If the ambient light intensity is strong or low, high-contrast edge feature points are extracted first, and the extraction ratio and weight of texture features are reduced to reduce texture dependence; when the density of risk targets reaches the high-density threshold, the semantic segmentation network is enabled to identify the passable area, and only the ground plane feature points are retained for SLAM mapping; Set up a multi-sensor tight coupling calibration submodule, which does the following: The local dense map generated by visual SLAM is aligned with the lidar point cloud frame by frame, and the heading angle deviation is calculated using the ICP algorithm. When the heading angle deviation exceeds 0.5°, the extended Kalman filter EKF is constructed by integrating the IMU angular velocity data for coupling calibration, and the calibrated heading angle is output.

[0038] The performance of visual SLAM in complex environments is optimized through dynamic feature weight allocation and multi-sensor tightly coupled calibration. The dynamic feature weight module adaptively adjusts the feature extraction strategy according to the illumination and target density. For example, it relies on high-contrast edge features in strong light to reduce texture noise interference; the semantic segmentation network filters redundant features when the targets are dense, improving the efficiency of mapping. The tightly coupled calibration module uses the ICP algorithm and EKF to fuse multi-sensor data, and corrects the heading angle deviation in real time (with an accuracy of 0.5°) to avoid SLAM mapping drift. This solution significantly improves the robustness and accuracy of visual SLAM in scenes with variable illumination and dense targets.

[0039] The operation logic of the visual SLAM algorithm also includes an abnormal state self-repair mechanism, which operates as follows: When visual SLAM tracking fails for 5 consecutive frames, it switches to the collaborative positioning mode based on V2X communication, receives the heading angle data of adjacent vehicles and performs weighted averaging, reversely infers its own posture through the outer contour feature points of the risk target, and resets the SLAM initialization parameters.

[0040] The positioning interruption problem caused by the failure of visual SLAM continuous tracking is solved through the abnormal self-repair mechanism. When SLAM fails, the system switches to V2X collaborative positioning mode and uses the weighted average of the heading angle data of neighboring vehicles to infer its own posture; at the same time, the posture is reversely calculated based on the outer contour feature points of the risk target, and the SLAM parameters are reset to quickly resume tracking. This solution avoids the defect of traditional SLAM systems relying on manual intervention under abnormal conditions, shortens the positioning recovery time (<100ms in typical scenarios), and ensures the continuity of the test process.

[0041] In another technical solution, the communication unit includes: The dynamic channel optimization submodule performs the following operations: monitor the load status and interference intensity of the communication channel in real time, dynamically allocate time-frequency resource blocks based on the reinforcement learning algorithm, and prioritize the data transmission of vehicles with collision risk values ​​above the risk threshold according to the collision risk values ​​of the vehicles calculated in real time; start the multi-hop relay transmission mode in the scenario where there are ≥20 test vehicles and / or ≥5 risk targets within a radius of 100 meters: set the central decision node, and transmit the vehicle data that is more than 200 meters away from the central decision node in segments through the relay nodes of adjacent vehicles, with a maximum delay of ≤10ms per hop; the relay node selection criteria are signal strength ≥-90dBm and remaining power >30%; The data priority encryption submodule encrypts the transmitted data in a hierarchical manner: quantum key distribution QKD is used to encrypt collision warning data, and the lightweight national encryption algorithm SM4 is used to encrypt position and speed status data; Among them, the central decision node dynamically adjusts the number of iterations of the path optimization algorithm according to the communication delay and packet loss rate; if the packet loss rate is greater than 5%, the local collision avoidance mode is triggered so that each vehicle only relies on local sensor data to perform emergency braking.

[0042] Through dynamic channel optimization and multi-hop relay transmission, communication congestion and delay problems in high-density test scenarios are solved. The reinforcement learning algorithm dynamically allocates resource blocks to prioritize data transmission of high-risk vehicles; the multi-hop relay mode expands the communication coverage (delay per hop ≤ 10ms) to avoid data loss of long-distance vehicles; QKD and SM4 hierarchical encryption take into account both security and transmission efficiency. In addition, the local collision avoidance mode is triggered when the packet loss rate exceeds the threshold to ensure basic security under communication interruption. In scenarios with vehicle density ≥ 20 vehicles / 100 meters, this solution can still control the average communication delay within 15ms, which is significantly better than the traditional V2X protocol.

[0043] In another technical solution, the method of calculating the collision risk value of the vehicle and allocating communication channel resources is as follows: Calculate the collision risk value Risk based on dynamic weighting: Among them, θ is the vehicle heading angle deviation, R is the path curvature radius, α, β, γ are the environment adaptation weight coefficients, and the values ​​are: α=8, β=2, γ=1, TTC is the collision time, v is the vehicle speed, and D is the dynamic threshold; note that the values ​​of α, β, and γ can be fine-tuned and corrected according to actual conditions.

[0044] Communication channel resources are allocated through a channel preemption hierarchical strategy, specifically: Vehicles with a collision risk value of Risk ≥ 8 occupy a dedicated emergency channel, the bandwidth of which is set to 10 MHz, and the instruction transmission delay is ≤ 15 ms; vehicles with a collision risk value of 5 ≤ Risk < 8 share a high priority channel, the bandwidth of which is set to 5 MHz, and the delay is ≤ 25 ms; The command strong synchronization mechanism is used when forcing adjacent vehicles to execute synchronous adjustment commands. The operation mode of the mechanism is: after the adjacent vehicle receives the command, the control signal is triggered through the hardware-level interrupt, the coordination parameters are preloaded in the reversing motor controller and the steering system, the trajectory correction or braking response time is set within 20-25ms, and the error tolerance is ≤0.2 meters.

[0045] Through the quantitative risk value, precise control of channel preemption and command synchronization is achieved. The risk value formula integrates multi-dimensional parameters such as TTC, speed, and heading angle deviation to dynamically reflect the urgency of the collision; the hierarchical channel strategy (dedicated / shared channel) allocates bandwidth and latency resources according to the risk level to ensure that high-risk commands are transmitted first; the hardware-level interrupt and parameter preloading mechanism compresses the command response time to 20-25ms, and the error tolerance is ≤0.2 meters. In sudden risk scenarios (such as vehicles suddenly cutting into the reversing path), this solution can reduce the probability of collision to less than 1 / 5 of the traditional system.

[0046] The specific operation modes of the instruction strong synchronization mechanism include: The hardware-level interrupt trigger module is directly connected to the underlying hardware interface between the reversing motor controller and the steering system, and a priority interrupt signal is generated through a programmable logic device to forcibly interrupt the current control instruction and load the coordination parameters; The coordination parameters are preloaded in the following manner: the coordination parameters generated by the dynamic priority algorithm are written into the controller and the steering system by the coordination parameter preloading module before the command is issued, wherein the coordination parameters include the target deceleration gradient, the heading angle correction value and the braking pressure threshold; Set up a timing constraint module, which aligns the start time of command execution based on the inter-vehicle clock synchronization protocol, and monitors the entire delay from command reception to execution completion through a hardware timer, strictly controlling the delay within the range of 20-25ms; An error tolerance verification module is set up, which feeds back the actual displacement in real time through the positioning unit after trajectory correction or braking is completed. If the deviation from the target displacement exceeds 0.2 meters, the emergency braking locking protocol is triggered.

[0047] Strict timing control of instruction execution is achieved through hardware-level interrupts and preloaded parameter mechanisms. The hardware interrupt module bypasses software layer delays and directly controls the underlying controller to ensure that the interrupt response time is ≤1ms; the preloaded parameters are written to the controller before the instruction is issued to reduce the processing time of the execution stage; the clock synchronization protocol and hardware timer control the entire delay error within ±1ms; the error tolerance check module provides real-time feedback on the execution results, and triggers secondary braking lock when the limit is exceeded. This solution improves the overall reliability of instruction execution to more than 99.9%, which is especially suitable for millisecond-level collaborative collision avoidance scenarios.

[0048] The working process of the positioning unit of the present invention is as follows: 1. GPS initial positioning: When the vehicle is started, the GPS module obtains the initial position coordinates (longitude and latitude) with an accuracy of centimeters.

[0049] 2. LiDAR point cloud matching: The LiDAR scans the surrounding environment 20 times per second, generates point cloud data, and matches it with the pre-stored high-precision site map in real time. The position deviation is corrected through the ICP (Iterative Closest Point) algorithm, and the positioning accuracy is improved to the millimeter level.

[0050] 3. Visual SLAM heading calibration: The on-board camera collects environmental images at 30fps, runs the visual SLAM algorithm to extract ORB feature points, builds a local dense map, and performs tight coupling registration with the lidar point cloud to calculate the heading angle deviation. When the deviation exceeds 0.5°, the IMU (inertial measurement unit) angular velocity data is integrated and the calibrated heading angle is output through the extended Kalman filter (EKF).

[0051] The communication unit workflow of the present invention is as follows: 1. Data sharing: Each vehicle broadcasts its own position, speed and predicted reversing trajectory data within the next 3 seconds (generated based on a polynomial fitting algorithm) through C-V2X.

[0052] 2. Potential conflict detection: After receiving all the data, the central decision node calculates the future minimum distance between vehicles and between vehicles and the target. If the distance is less than the dynamic threshold D (D = 1.3~1.5 max(W,L), W, L are the length and width of the target), it is marked as a potential conflict.

[0053] 3. Collaborative collision avoidance path generation: A distributed optimization algorithm (such as distributed model predictive control DMPC) is used to generate a global collision avoidance path based on vehicle position, speed, and conflict markers. The algorithm constraints include path curvature ≥ vehicle minimum turning radius, and path spacing ≥ threshold D.

[0054] 4. Instruction issuance and execution: The dynamic priority algorithm calculates the collision risk value Risk of each vehicle in real time. The vehicle with the highest risk value seizes the dedicated communication channel and issues deceleration, heading adjustment or braking instructions to the associated vehicles. The instructions are triggered by hardware-level interrupts to ensure a response within 20-25ms.

[0055] Structural details of the present invention: Communication module hardware: Adopts C-V2X chipset (such as Qualcomm 9150), supports direct communication via PC5 interface, and transmission delay ≤10ms.

[0056] Path optimization computing node: deployed in the local ECU of the vehicle, using a multi-core processor (such as ARM Cortex-A72) to calculate the collision avoidance path in parallel.

[0057] The hierarchical control logic of deceleration, heading adjustment and braking instructions of the present invention is as follows: 1. Coordinated deceleration command: When a potential conflict is detected, the central decision node broadcasts a deceleration command via V2X, with a target speed of v safe =0.5·v current (50% of the current speed). After receiving the command, each vehicle motor controller adjusts the reverse motor torque through the PID control algorithm to smoothly reduce the speed to v within 1 second. safe .

[0058] 2. Heading angle adjustment command: If there is still a risk of conflict after deceleration, generate a heading angle correction value Δθ based on the optimized path. For example, if the path requires the vehicle to deviate 0.5 meters to the right, then Δθ=arctan(0.5 / D p ), D p is the current distance to the target object. The steering system implements Δθ through EPS (Electric Power Steering), and the stepper motor drives the steering rack with an angle error of ≤0.1°.

[0059] 3. Cascade braking protocol: When the path conflict cannot be eliminated, braking is triggered according to the distance between the vehicles and the conflict point. The closest vehicle is given priority to perform emergency braking (deceleration -5m / s²), and the following vehicles decelerate at -3m / s² and -1m / s² in turn. The braking command is sent to ESP (Electronic Stability Program) via the CAN bus, and the hydraulic unit builds up braking pressure within 50ms.

[0060] The structure of the present invention realizes: Deceleration control module: integrated in the vehicle VCU (vehicle control unit), including the speed planning submodule (generating deceleration gradient based on polynomial curve) and the motor drive interface.

[0061] Steering actuator: A brushless DC motor (such as Maxon EC45) is used to drive the steering system, and the encoder provides real-time feedback of the steering angle.

[0062] Brake hydraulic unit: Bosch iBooster electro-hydraulic brake system, supporting brake-by-wire, pressure response time ≤20ms.

[0063] The constraints of the collaborative collision avoidance path of the present invention are achieved by the following steps: 1. Minimum spacing constraint: When planning a path, the future minimum spacing between multiple vehicle paths is calculated using a geometric collision detection algorithm (such as the GJK algorithm). If the spacing is less than D = 1.3 max(W, L), the path is replanned to ensure that the spacing is ≥ D.

[0064] 2. Curvature radius constraint: According to the vehicle's minimum turning radius R min (For example, R min =3.5 meters), Clothoid curve is used for smooth transition in the path generation stage to ensure that the path curvature 1 / R≤1 / R min The curvature k is calculated using the derivative formula in the Frenet coordinate system: k=(d²y / ds²) / (1+(dy / ds) 2 ) (3 / 2) .

[0065] Path planning process of the present invention: Input: vehicle current position (x, y), heading angle θ, speed v.

[0066] Output: A sequence of path points that satisfies the constraints {(x i ,y i )}.

[0067] Algorithm: A hybrid A* algorithm is used to search for feasible paths, combined with the Dijkstra algorithm to optimize the global cost (path length, curvature).

[0068] The hardware of the present invention supports: Computing unit: NVIDIA Xavier chip, providing 30TOPS computing power and solving path optimization problems in real time.

[0069] Storage module: pre-stores vehicle dynamics parameters (such as wheelbase and steering ratio) for curvature constraint calculation.

[0070] The hardware structure of the multi-source fusion positioning submodule of the present invention includes a GPS receiver (U-blox ZED-F9P), a 16-line laser radar (Hesai Pandar16) and a binocular camera (Basler acA2440). Its workflow is as follows: 1. GPS positioning: ZED-F9P outputs RTK (real-time kinematic positioning) data with a frequency of 10Hz and a horizontal accuracy of ±1cm+1ppm.

[0071] 2. LiDAR point cloud matching: Pandar16 scans the environment to generate a point cloud, which is matched with a high-precision map through the NDT (normal distribution transform) algorithm to output position corrections (Δx, Δy) with an accuracy of ±2cm.

[0072] 3. Visual SLAM mapping: The binocular camera collects images, runs the ORB-SLAM3 algorithm to extract feature points, and builds a local map. The visual coordinate system is converted to the vehicle coordinate system through the hand-eye calibration matrix, and the heading angle θ is output with an accuracy of ±0.3°.

[0073] Data fusion logic of the present invention: Loosely coupled fusion: The location data of GPS, LiDAR and visual SLAM are fused through Kalman filter, and the state vector is [x, y, θ, v m ], the observation equation is Z=HX+v m , where Z is the observation vector of the Kalman filter observation equation, H is the observation matrix, X is the state vector, and v m is the observation noise vector.

[0074] Tightly coupled calibration: The positions of the lidar and visual SLAM are aligned through the ICP algorithm, and the final positioning result is output after the residual is minimized.

[0075] The sensor layout of the present invention: the laser radar is installed in the center of the roof, the GPS antenna is located on the top of the trunk, and the binocular camera is fixed on the inside of the front windshield. Heat dissipation design: the positioning unit housing has built-in heat dissipation fins and fans to ensure that the sensor works stably in an environment of -20℃~60℃.

[0076] The dynamic anti-interference submodule of the positioning unit consists of collaborative positioning logic, visual inertial odometer (VIO) switching module and real-time self-calibration unit, which are specifically implemented as follows: Co-location mode: 1. GPS signal loss detection: When the GPS module does not output valid data for 5 consecutive seconds (data packet loss rate>90%), the collaborative positioning mode is triggered.

[0077] 2. Neighboring vehicle data acquisition: Receive positioning data (including position, heading angle and confidence) of neighboring vehicles within a radius of 50 meters through C-V2X communication.

[0078] 3. Relative position calculation: Use the least square method to calculate the own position. Assume the position of the neighboring vehicle is {(x i ,y i )}, its relative position (x, y) satisfies the equation group: (x i -x) 2 +(y i -y) 2 =d i2 , (d i is the V2X ranging value), and the accuracy can reach ±5cm by solving (x,y) through iterative optimization.

[0079] VIO mode switch: 1. LiDAR failure judgment: If the LiDAR point cloud matching fails (residual > 10cm) for three consecutive frames (150ms interval), switch to VIO mode.

[0080] 2. Data fusion: IMU (model: ADIS16470) outputs angular velocity and acceleration, which are fused with binocular visual feature tracking results (ORB feature points) through tightly coupled filtering. Visual inertial odometer calculation method (VINS-Fusion) solves posture in real time, with an output frequency of 100Hz and a position drift rate of <0.1% / m.

[0081] Real-time self-calibration logic: 1. Multi-source data verification: Compare the positioning data of GPS, LiDAR and visual SLAM every 30 seconds. If the horizontal deviation between GPS and LiDAR is >2cm, global repositioning is triggered.

[0082] 2. Relocalization process: The vehicle stops moving, the lidar scans the environment point cloud, matches it with the preset map, and reinitializes the positioning coordinates, which takes ≤ 2 seconds.

[0083] Co-location module: integrated in the vehicle communication gateway (model: NXP S32G), with built-in ARM Cortex-A53 processor, running the least squares solution algorithm.

[0084] VIO processing unit: Xilinx Zynq UltraScale+ MPSoC is used to achieve hardware-level synchronous acquisition and filtering of IMU and visual data.

[0085] Self-calibration trigger circuit: Based on FPGA (model: Intel Cyclone V), the timeout detection logic is designed to monitor the validity of each sensor data.

[0086] The dynamic feature weight allocation and tight coupling calibration of the visual SLAM algorithm are achieved by the following steps: Dynamic feature weight assignment: 1. Light intensity detection: The camera has a built-in ambient light sensor (model: TI OPT3001) to monitor the light intensity Lux value in real time.

[0087] When Lux>10,000 (strong light) or Lux<50 (low light), the edge feature extraction mode is enabled, and the Canny operator is used to detect high-contrast edges. The weight is increased to 0.8, and the texture feature weight is reduced to 0.2.

[0088] Under normal lighting (50≤Lux≤10,000), the weights of edge and texture features are evenly distributed (0.5 each).

[0089] 2. Target density judgment: Through LiDAR point cloud density analysis, if the number of point clouds per cubic meter is >1000, it is determined to be a target-dense area, and the semantic segmentation network (based on MobileNetV3) is started.

[0090] The segmentation network outputs a passable area mask, retaining only ground plane feature points (such as lane lines and landmarks) and filtering out obstacle features with a height > 0.5 meters.

[0091] Multi-sensor tight-coupled calibration: 1. Point cloud and visual map registration: Align the local dense map (point cloud format) generated by visual SLAM with the single-frame point cloud of the lidar through the ICP algorithm, and calculate the rotation and translation matrix [T].

[0092] 2. Heading angle deviation correction: If the heading angle deviation Δθ=|θ visual -θ lidar |>0.5°, start the extended Kalman filter (EKF), the state vector is [θ, ω] (ω is the IMU angular velocity), and the observation equation is θ measure =θ+ν k , the prediction-update loop outputs the calibrated θ with an accuracy of ±0.1°, where θ visual Represents the heading angle obtained by the visual SLAM system, θ lidar is the heading angle measured by the laser radar, θ measure is the heading angle measured by the equation, ν k is the observation noise.

[0093] Dynamic weight allocation module: deployed on the GPU (model: NVIDIA Jetson AGX Xavier), running the OpenCV library to achieve real-time feature extraction and weight adjustment.

[0094] Semantic segmentation accelerator: uses Google Coral Edge TPU, with an inference speed of ≥30fps.

[0095] ICP acceleration unit: parallel point cloud matching based on FPGA, single frame processing time ≤ 10ms.

[0096] The workflow of the visual SLAM abnormal state self-repair mechanism is as follows: Anomaly Detection: 1. Continuous tracking failure judgment: If the visual SLAM fails to output a valid pose (number of feature point matches <10) for 5 consecutive frames (about 166ms), the self-repair process is triggered.

[0097] Co-location mode switching: 1. Weighted average of neighboring vehicle heading angles: Obtain the heading angles {θ1, θ2, θ3} of the three nearest neighboring vehicles within a radius of 30 meters through V2X, and calculate the weighted average value θ avg =w1θ1+ w2θ2+ w3θ3 (weight w i =1 / D i 2 , D i for distance).

[0098] 2. Reverse pose estimation: Based on the outer contour feature points of the risk target (provided by the laser radar), the PnP (Perspective-n-Point) algorithm is used to solve its own pose. Assume that the coordinates of the corner point of the target are {P i}, by minimizing the reprojection error ∑||u i - K[R T |t]P i || 2 , solve the rotation matrix R T and the translation vector t, where u i is the actual projection point coordinate of the i-th corner point on the risk target on the image plane, and K is the intrinsic parameter matrix inherent in the camera.

[0099] SLAM parameter reset: 1. Initialization parameter overload: set θ avg As the initial heading angle and t as the initial position, reset the ORB-SLAM3 tracker.

[0100] 2. Local map reconstruction: Generate an initial map based on the lidar point cloud for subsequent tracking by visual SLAM, with a recovery time of ≤100ms.

[0101] Anomaly detection circuit: integrated in the visual processing unit, monitors the validity of SLAM output through a counter.

[0102] PnP solution module: Intel Movidius Myriad X VPU is used to accelerate matrix operations, with a single solution time of ≤5ms.

[0103] Parameter reset interface: communicates with the SLAM master controller through the SPI bus and supports fast writing of initialization parameters.

[0104] The dynamic channel optimization and multi-hop relay transmission of the C-V2X communication unit are implemented as follows: Dynamic channel optimization: 1. Channel status monitoring: A spectrum analysis module (model: Analog Devices AD9361) is used to monitor channel load (RSSI) and interference intensity (SINR) in real time.

[0105] 2. Reinforcement learning resource allocation: Based on the DQN (Deep Q Network) algorithm, the state space is [channel load, interference intensity, vehicle risk level], and the action space is the time-frequency resource block allocation scheme. The training phase simulates high-density scenarios (vehicles ≥ 20 / 100 meters), and the optimization goal is to minimize the transmission delay of high-risk vehicles (Risk ≥ 3).

[0106] 3. Priority for high-risk vehicles: When the vehicle Risk ≥ 3, a dedicated resource block (bandwidth 10MHz) is allocated, and the remaining vehicles share the remaining resources.

[0107] Multi-hop relay transmission: 1. Relay node selection: The central decision node (model: Huawei MH5000 module) periodically broadcasts a neighbor list. The relay candidate nodes must meet the following requirements: Signal strength ≥-90dBm (measured distance ≤50 meters); The remaining power is greater than 30% (to prevent the relay node from being disconnected).

[0108] 2. Segmented transmission protocol: The data packet is split into multiple segments, and each segment is forwarded by a relay node. The relay node adds a timestamp to ensure that the delay of each hop is ≤10ms. For example, the vehicle data 250 meters away from the central node is transmitted through 2 hops (hop distance 125 meters), and the total delay is ≤25ms.

[0109] Data encryption mechanism: 1. Quantum Key Distribution (QKD): The collision warning data uses the BB84 protocol to generate a key, which is transmitted through a fiber optic channel with a key update frequency of 1 time per second.

[0110] 2. Lightweight national encryption algorithm: The position and speed data use the SM4 algorithm (packet length 128 bits), the encryption engine is integrated into the communication module, and the throughput is ≥100Mbps.

[0111] Reinforcement learning accelerator: uses Google Edge TPU to run TensorFlow Lite models with inference latency ≤ 2ms.

[0112] Relay control module: Based on the STM32H7 microcontroller, it implements relay logic and power consumption management.

[0113] Encryption chip: Integrates the SJ-A100 chip with national encryption certification and supports SM4 hardware acceleration.

[0114] The implementation steps of the dynamic priority algorithm and communication preemption mechanism are as follows: Value at Risk calculation: 1. Parameter collection: real-time acquisition of vehicle TTC (collision time), speed v, heading angle deviation θ, and path curvature radius R.

[0115] 2. Adaptive weight adjustment: The environmental coefficients α, β, and γ are adjusted dynamically according to the scene. For example, during nighttime testing (light <50 Lux), the TTC weight α is increased to 0.6; when there are dense targets, the heading angle weight γ is increased to 0.3.

[0116] 3. Formula solution: Risk = 0.6 / TTC + 0.2·v / D + 0.2·θ / R, and the calculation result is normalized to the range of 0-10.

[0117] Channel preemption classification: 1. Dedicated emergency channel: When Risk ≥ 8, 10MHz bandwidth is allocated, an emergency flag is added to the data packet, the routing priority is the highest, and the transmission delay is ≤ 15ms.

[0118] 2. High priority shared channel: When 5≤Risk<8, 5MHz bandwidth is allocated and the CSMA / CA mechanism is used to avoid low priority data.

[0119] Instruction strong synchronization mechanism: 1. Hardware interrupt trigger: After the command receiving module (model: Xilinx AXI Interrupt Controller) detects the emergency command, it immediately sends a hardware interrupt signal (priority IRQ0) to the reversing motor controller (model: TI DRV8305) and the steering ECU (model: InfineonAurix).

[0120] 2. Parameter preloading: The coordination parameters (such as the target deceleration gradient -3m / s², the heading correction value Δθ=2°) are written into the controller's ring buffer before the command is issued, and are directly loaded after the interrupt is triggered, reducing the software parsing time.

[0121] 3. Timing monitoring: The hardware timer (model: Microchip MCP7940N) records the delay from instruction reception to execution completion. When the timeout (>25ms) occurs, an alarm is triggered and redundant control is started.

[0122] Interrupt controller: integrated in the vehicle central gateway, supporting 16 levels of interrupt priority.

[0123] Parameter buffer: Dual-port RAM (model: Cypress CY7C024) is used to achieve high-speed reading and writing (access time ≤ 10ns).

[0124] The specific implementation of the hardware-level interrupt and preload coordination parameters is as follows: Hardware interrupt module: 1. Interrupt signal generation: After receiving the instruction packet from the communication module, the programmable logic device (model: Lattice iCE40UP5K) analyzes the risk level. If Risk ≥ 8, a priority interrupt signal (high level 3.3V) is generated and directly connected to the interrupt input port of the motor controller and steering system through the GPIO pin.

[0125] 2. Interrupt response: The motor controller interrupt service routine (ISR) immediately suspends the current control loop and reads the coordination parameters from the preload buffer. The response time is ≤1ms.

[0126] Collaborative parameter preloading: 1. Parameter writing: The parameters generated by the dynamic priority algorithm (such as target speed and heading correction) are transferred to the controller's preload area (address 0x8000-0x8FFF) via DMA (direct memory access) to avoid CPU intervention delays.

[0127] 2. Parameter verification: The CRC verification module (using the polynomial 0x04C11DB7 to generate the verification code) verifies the data integrity and requests retransmission if the verification fails.

[0128] Timing and error control: 1. Clock synchronization: Using IEEE 1588 Precision Time Protocol (PTP), the clock deviation between vehicles is ≤1μs, ensuring the alignment of the start time of command execution.

[0129] 2. Hardware timer monitoring: From the moment the command is received (T0) to the moment the execution is completed (T1), the timer records Δt=T1-T0. If Δt>25ms, the emergency brake lock is triggered (cutting off the power output).

[0130] 3. Error check: The positioning unit feedbacks the actual displacement Δs. If |Δs target -Δs|>0.2m, the brake lock protocol forces the vehicle to stop and broadcasts the fault code via V2X, Δs target Represents the target displacement expected by the positioning unit.

[0131] Interrupt interface circuit: Optocoupler isolation (model: Sharp PC817) is used to prevent electrical interference.

[0132] DMA controller: integrated in MCU (model: NXP S32K144), supporting multi-channel concurrent transmission.

[0133] Clock synchronization module: Based on the Marvell 88X3310P PHY chip, it achieves nanosecond-level time synchronization.

[0134] It should be noted that, although the above describes the various steps in a specific order, it does not mean that the various steps must be performed in the above specific order. In fact, some of these steps can be performed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention are obvious to those skilled in the art.

[0135] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A safety device for reducing the risk of a reverse collision during an autonomous driving test, characterized in that: include: Positioning units are deployed on all test vehicles and risk targets. The positioning units obtain the millimeter-level position coordinates and heading angles of each vehicle in real time. The communication unit based on the C-V2X protocol builds a real-time data link between vehicles and between vehicles and objects based on the position coordinates and heading angles obtained by the positioning unit to share the following information: the real-time position, speed, and reverse trajectory prediction data of each vehicle, as well as the coordinates of the outer contour feature points and safety boundary parameters of the risky object; The communication unit receives the reverse trajectory prediction data uploaded by all vehicles, the coordinates of the outer contour feature points of the risk target and the safety boundary parameters, and performs the following operations: Based on the predicted reversing trajectory data, the future minimum distance between vehicles and between vehicles and risk targets is calculated. If the future minimum distance is less than the dynamic threshold D, it is marked as a potential conflict. The dynamic threshold D ranges from 1.3 to 1.5 times the larger value of the width and length of the risk target. Through a distributed optimization algorithm, a collaborative collision avoidance path is dynamically generated based on the position and speed data of each vehicle. At the same time, a dynamic priority algorithm is used to calculate the collision risk value of the vehicle in real time based on the dynamically generated collaborative collision avoidance path and potential conflict data, and communication channel resources are preferentially allocated to vehicles with the highest collision risk value. Synchronous adjustment instructions are sent to their adjacent vehicles. The synchronous adjustment instructions include deceleration, adjustment of heading angle and braking, and force adjacent vehicles to respond to and execute the synchronous adjustment instructions within 20-25ms.

2. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 1, characterized in that: The specific response method for issuing synchronous adjustment instructions is: When a deceleration control command is issued, a coordinated deceleration command is sent to all associated vehicles through V2X communication, so that the driving speed of each vehicle is synchronously reduced to a preset safety value; When issuing a control command to adjust the heading angle, the heading angle of each vehicle is dynamically adjusted based on the optimized path to ensure that its trajectory deviates from the safety boundary of the risk target; When issuing a braking control command, if the path conflict cannot be eliminated, the cascade braking protocol is triggered, and emergency braking is performed in sequence from the nearest to the conflict point.

3. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 1, characterized in that: The constraints used when dynamically generating the cooperative collision avoidance path for each vehicle include: The minimum distance between the reversing paths of each vehicle ≥ the dynamic threshold D; and The curvature radius of the collision avoidance path ≥ the minimum turning radius of the vehicle.

4. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 1, characterized in that: The positioning unit includes a multi-source fusion positioning submodule, which integrates GPS, laser radar point cloud matching and visual SLAM algorithm, and performs millimeter-level positioning through the following steps: The initial position coordinates are obtained through GPS, and the feature point cloud of the surrounding environment is scanned by lidar, which is matched with the high-precision preset site map in real time to correct the position deviation; based on the image data collected by the on-board camera, the visual SLAM algorithm is used to generate a local dense map to calibrate the heading angle.

5. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 4, characterized in that: The positioning unit also includes a dynamic anti-interference submodule that performs the following operations when the signal is interfered with: When it detects that the GPS signal is blocked, it starts the collaborative positioning mode based on vehicle-to-vehicle V2X communication, receives the positioning data of adjacent vehicles, and calculates its own relative position through the least squares method; If the LiDAR point cloud fails to match for at least three consecutive frames, it switches to pure visual inertial odometry (VIO) mode, fusing IMU data with visual feature tracking results to maintain positioning accuracy. Real-time self-calibration is performed and multi-source data consistency check is executed every 30 seconds. When the positioning deviation between GPS and lidar exceeds 2 cm, the global repositioning process is triggered to correct the positioning deviation.

6. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 5, characterized in that: The way to use the visual SLAM algorithm includes the following steps: Set up a dynamic feature weight allocation submodule, which dynamically adjusts the extraction weight of visual feature points according to the ambient light intensity and the density of risk targets. The specific operation mode is as follows: If the ambient light intensity is strong or low, high-contrast edge feature points are extracted first, and the extraction ratio and weight of texture features are reduced to reduce texture dependence; when the density of risk targets reaches the high-density threshold, the semantic segmentation network is enabled to identify the passable area, and only the ground plane feature points are retained for SLAM mapping; Set up a multi-sensor tight coupling calibration submodule, which does the following: The local dense map generated by visual SLAM is aligned with the lidar point cloud frame by frame, and the heading angle deviation is calculated using the ICP algorithm. When the heading angle deviation exceeds 0.5°, the extended Kalman filter EKF is constructed by integrating the IMU angular velocity data for coupling calibration, and the calibrated heading angle is output.

7. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 6, characterized in that: The operation logic of the visual SLAM algorithm also includes an abnormal state self-repair mechanism, which operates as follows: When visual SLAM tracking fails for 5 consecutive frames, it switches to the collaborative positioning mode based on V2X communication, receives the heading angle data of adjacent vehicles and performs weighted averaging, reversely infers its own posture through the outer contour feature points of the risk target, and resets the SLAM initialization parameters.

8. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 1, characterized in that: The communication unit comprises: The dynamic channel optimization submodule performs the following operations: monitor the load status and interference intensity of the communication channel in real time, dynamically allocate time-frequency resource blocks based on the reinforcement learning algorithm, and prioritize the data transmission of vehicles with collision risk values ​​above the risk threshold according to the collision risk values ​​of the vehicles calculated in real time; start the multi-hop relay transmission mode in the scenario where there are ≥20 test vehicles and / or ≥5 risk targets within a radius of 100 meters: set the central decision node, and transmit the vehicle data that is more than 200 meters away from the central decision node in segments through the relay nodes of adjacent vehicles, with a maximum delay of ≤10ms per hop; the relay node selection criteria are signal strength ≥-90dBm and remaining power >30%; The data priority encryption submodule encrypts the transmitted data in a hierarchical manner: quantum key distribution QKD is used to encrypt collision warning data, and the lightweight national encryption algorithm SM4 is used to encrypt position and speed status data; Among them, the central decision node dynamically adjusts the number of iterations of the path optimization algorithm according to the communication delay and packet loss rate; if the packet loss rate is greater than 5%, the local collision avoidance mode is triggered so that each vehicle only relies on local sensor data to perform emergency braking.

9. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 1, characterized in that: The method of calculating the collision risk value of the vehicle and allocating communication channel resources is as follows: Calculate the collision risk value Risk based on dynamic weighting: Among them, θ is the vehicle heading angle deviation, R is the path curvature radius, α, β, γ are the environment adaptation weight coefficients, and the values ​​are: α=8, β=2, γ=1, TTC is the collision time, v is the vehicle speed, and D is the dynamic threshold; Communication channel resources are allocated through a channel preemption hierarchical strategy, specifically: Vehicles with a collision risk value of Risk ≥ 8 occupy a dedicated emergency channel, the bandwidth of which is set to 10 MHz, and the instruction transmission delay is ≤ 15 ms; vehicles with a collision risk value of 5 ≤ Risk < 8 share a high priority channel, the bandwidth of which is set to 5 MHz, and the delay is ≤ 25 ms; The command strong synchronization mechanism is used when forcing adjacent vehicles to execute synchronous adjustment commands. The operation mode of the mechanism is: after the adjacent vehicle receives the command, the control signal is triggered through the hardware-level interrupt, the coordination parameters are preloaded in the reversing motor controller and the steering system, the trajectory correction or braking response time is set within 20-25ms, and the error tolerance is ≤0.2 meters.

10. The safety device for reducing the risk of reverse collision during automatic driving testing according to claim 9, characterized in that: The specific operation modes of the instruction strong synchronization mechanism include: The hardware-level interrupt trigger module is directly connected to the underlying hardware interface between the reversing motor controller and the steering system, and a priority interrupt signal is generated through a programmable logic device to forcibly interrupt the current control instruction and load the coordination parameters; The coordination parameters are preloaded in the following manner: the coordination parameters generated by the dynamic priority algorithm are written into the controller and the steering system by the coordination parameter preloading module before the command is issued, wherein the coordination parameters include the target deceleration gradient, the heading angle correction value and the braking pressure threshold; Set up a timing constraint module, which aligns the start time of command execution based on the inter-vehicle clock synchronization protocol, and monitors the entire delay from command reception to execution completion through a hardware timer, strictly controlling the delay within the range of 20-25ms; An error tolerance verification module is set up, which feeds back the actual displacement in real time through the positioning unit after trajectory correction or braking is completed. If the deviation from the target displacement exceeds 0.2 meters, the emergency braking locking protocol is triggered.

Images