A semi-trailer automatic cruise control method under human-machine cooperative driving

Abstract

The application discloses a kind of human-computer collaborative driving under semi-trailer automatic cruise control method, comprising: obtaining vehicle positioning data and environmental perception data by vehicle-mounted sensor, position refinement processing is carried out in combination with the high-precision map established in advance, and refined position information is obtained;Using multi-source data fusion method to optimize vehicle position estimation, to obtain fusion optimization position data;According to the fusion optimization position data, calculate the path deviation, trigger automatic cruise control function to generate trajectory control instruction;Combined with driver intent recognition and system response strategy in human-computer collaborative driving, adjust vehicle travel trajectory;Through vibration detection and environmental perception data, the travel trajectory is smoothed, and the final trajectory execution data is obtained;According to the final trajectory execution data, dynamically update path planning, form closed-loop control mechanism;Not only improve the positioning accuracy and control stability of automatic driving system, also enhance the safety and reliability of human-computer collaborative driving.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to an automatic cruise control method for a semi-trailer truck with human-machine collaborative driving. Background Technology

[0002] The core technological challenge facing autonomous driving systems is achieving high-precision positioning and stable control in complex and ever-changing road environments. Traditional methods rely on data from single sensors, which struggle to cope with various interference factors, leading to insufficient positioning accuracy and unstable control. While multi-source data fusion can improve accuracy, it presents challenges in time synchronization. Changes in vehicle load cause center of gravity shifts, affecting navigation parameter settings. Coordination issues exist between driver intent and the automatic control system; road vibrations disrupt trajectory smoothness, impacting comfort; balancing safety and efficiency in complex traffic environments remains a challenge; fluctuations in positioning accuracy lead to frequent adjustments in control strategies, affecting system stability. These problems are interconnected and mutually influential, forming a complex technical challenge.

[0003] How to build a robust, adaptive, and collaborative autonomous driving system that improves positioning accuracy and control stability while ensuring safety has become a critical issue that urgently needs to be addressed. Summary of the Invention

[0004] The purpose of this invention is to provide a human-machine cooperative driving method for automatic cruise control of semi-trailers, so as to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] A method for automatic cruise control of a semi-trailer truck under human-machine cooperative driving includes: acquiring vehicle positioning data and environmental perception data through onboard sensors, and refining the position by combining it with a pre-established high-precision map to obtain refined position information; optimizing vehicle position estimation using a multi-source data fusion method to obtain fused optimized position data; calculating path deviation based on the fused optimized position data and triggering the automatic cruise control function to generate trajectory control commands; adjusting the vehicle's driving trajectory by combining driver intent recognition and system response strategies in human-machine cooperative driving; smoothing the driving trajectory using vibration detection and environmental perception data to obtain final trajectory execution data; and dynamically updating the path planning based on the final trajectory execution data to form a closed-loop control mechanism.

[0007] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for a semi-trailer described in this invention, the step of acquiring vehicle positioning data and environmental perception data through on-board sensors and performing position refinement processing in conjunction with a pre-established high-precision map to obtain refined position information includes: acquiring environmental perception data through lidar scanning, wherein the environmental perception data consists of point cloud data of the surrounding environment;

[0008] Based on the semi-trailer's work log, the navigation mode of the semi-trailer is identified, wherein the navigation mode includes inertial navigation mode and real-time navigation mode.

[0009] The system acquires the initial position signal from the GPS receiver and obtains the acceleration and angular velocity data of the semi-trailer from the inertial measurement unit. When the quality of the initial position signal is lower than a preset signal quality threshold, it switches to the inertial navigation solution mode to obtain the initial position estimation data and error covariance matrix. Based on a preset inertial navigation model, it inputs the acceleration, angular velocity, and initial position estimation data of the semi-trailer and outputs the position prediction data. It uses an iterative nearest-point matching method to register the point cloud data with a preset high-precision map. If the registration error is greater than the error threshold, it starts a re-matching procedure, performing at least three iterative matchings to obtain the adjusted registration result. Combining historical trajectory data and the vehicle's initial positioning data, it obtains the refined position information and heading angle data of the vehicle in the map coordinate system and updates the position prediction data. It monitors the changes in vehicle load mass in real time using a weighing sensor to obtain the vehicle's center of gravity offset. If the vehicle's center of gravity offset exceeds a preset offset threshold, it adjusts the inertial navigation parameters and updates the position prediction data.

[0010] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for a semi-trailer described in this invention, the method of optimizing vehicle position estimation using a multi-source data fusion approach to obtain fused and optimized position data includes: acquiring GPS position signals, inertial navigation prediction data, and refined position information from lidar; aligning the timestamps of different data sources using a time interpolation method to obtain synchronized multi-source position data; and fusing the multi-source position data using an extended Kalman filter method.

[0011] Initialize the extended Kalman filter based on multi-source location data, and set the initial values ​​of the state vector and the diagonal of the error covariance matrix;

[0012] Using acceleration and angular velocity data under inertial navigation, a state transition matrix is ​​obtained through a motion model, the predicted position and error covariance are updated, and fused optimized position data is generated. Uncertainty information is extracted from the fusion result to obtain the updated error covariance matrix. If the components of the error covariance matrix exceed a preset threshold, the fused optimized position data is then corrected a second time.

[0013] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for a semi-trailer truck according to the present invention, the step of calculating the path deviation based on the fused and optimized position data and triggering the automatic cruise control function to generate trajectory control commands includes: calculating the lateral deviation and heading deviation of the vehicle relative to the preset path based on the fused and optimized position data.

[0014] Construct a path point set H based on the preset path: ,in, Let N represent the coordinates of the nth path point, and N represent the total number of path points in the preset path r. This represents the x-coordinate of the nth path point. This represents the y-coordinate of the nth path point;

[0015] Calculate vehicle position The shortest distance point to the preset path r The specific formula is as follows:

[0016]

[0017] The tangent direction of the shortest distance point is: The angle between the tangent direction of the shortest distance point and the horizontal axis is the reference heading angle. ;

[0018] Based on the shortest distance point The lateral deviation and heading deviation are calculated using the following formulas:

[0019]

[0020]

[0021] in, Indicates lateral deviation. Indicates heading deviation. Indicates the vehicle's current heading angle;

[0022] When the absolute value of the lateral deviation is greater than the lateral deviation threshold or the heading deviation is greater than the heading threshold, the activation signal of the adaptive speed adjustment and lane keeping assist functions is triggered; the steering angle adjustment amount and speed adjustment value are calculated using the vehicle dynamics model; and the trajectory control command is generated and transmitted to the vehicle execution unit.

[0023] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for semi-trailers described in this invention, the step of adjusting the vehicle's driving trajectory by combining driver intention recognition and system response strategies in human-machine collaborative driving includes: obtaining driver intention and cruise status data through the vehicle interactive interface; and analyzing the driving intention tendency using a preset driving behavior model to obtain preliminary intention recognition results.

[0024] By integrating the current vehicle's steering angle and speed data, and outputting the target steering angle adjustment and speed adjustment values ​​according to preset adjustment rules, preliminary control parameters are determined:

[0025] The initial control parameters are compared with the output of the real-time decision allocation model. When the comparison results are inconsistent, the control priority is reallocated through the human-machine collaborative driving decision module to obtain the adjusted control command, form the system response strategy, and adjust the vehicle's driving trajectory.

[0026] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for a semi-trailer described in this invention, the step of smoothing the vehicle's trajectory through vibration detection and environmental perception data to obtain the final trajectory execution data includes: collecting vibration data during vehicle operation and performing preliminary storage and formatting processing to obtain standardized vibration information and extracting the peak vibration acceleration; when the peak vibration acceleration is greater than a preset peak threshold, adjusting the filter window length to obtain the adjusted filter parameters; and using an adaptive filtering method to smooth the fused and optimized position data, obtaining a smoothed position signal based on the adjusted filter parameters.

[0027] By combining position prediction data and heading angle data, trajectory data is obtained;

[0028] The position adjustment parameters are determined by combining driving load monitoring data; the trajectory data is then corrected a second time based on environmental perception data; the optimized trajectory execution command is generated to obtain the final trajectory execution data.

[0029] As a preferred embodiment of the human-machine collaborative driving automatic cruise control method for a semi-trailer described in this invention, the step of dynamically updating the path planning based on the final trajectory execution data includes: acquiring information on the distance to obstacles and road signs ahead through an onboard camera and ultrasonic sensors; triggering automatic braking if the distance to an obstacle is less than a preset obstacle distance threshold or a speed limit sign is detected; generating a warning signal in conjunction with an emergency status indicator; evaluating the driver-system collaboration status through a trust assessment model; adjusting the system response strategy based on the assessment results; and dynamically updating the vehicle's preset path using the path planning optimization function in automatic cruise control to obtain optimized path planning data and transmitting it to the monitoring module.

[0030] Based on the optimized path planning data, combined with the priority of manual intervention in human-machine cooperative driving and the cruise status monitoring in automatic cruise control, if the standard deviation of position accuracy is greater than the preset standard deviation threshold, the LiDAR scanning frequency is adjusted and the timestamp alignment parameters are optimized to obtain updated positioning correction data, which is then fed back to the preliminary position estimation processing.

[0031] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0032] This invention discloses an automatic cruise control method for a semi-trailer truck in human-machine cooperative driving. It achieves high-precision positioning by fusing data from the Global Positioning System (GPS), Inertial Measurement Unit (INS), and LiDAR, and dynamically adjusts navigation parameters by combining onboard weighing sensors to monitor center-of-gravity shift. Extended Kalman filtering is used to fuse multi-source data to solve time synchronization issues. Deviation is calculated based on the fused position, triggering adaptive speed and lane-keeping control. Steering and speed are adjusted collaboratively by recognizing driver intent. Vibration is detected by an accelerometer, and adaptive filtering smooths the trajectory. Obstacle avoidance and traffic sign recognition functions are integrated to achieve safe cruise. Path planning is dynamically updated, and positioning accuracy is continuously monitored, forming a closed-loop adjustment mechanism. This not only improves the positioning accuracy and control stability of the autonomous driving system but also enhances the safety and reliability of human-machine cooperative driving. (See attached figures.)

[0033] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0034] Figure 1 This is a flowchart illustrating an automatic cruise control method for a semi-trailer truck under human-machine collaborative driving according to the present invention. Detailed embodiments are described below.

[0035] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0036] like Figure 1 This embodiment of a human-machine collaborative driving method for automatic cruise control of a semi-trailer includes:

[0037] Vehicle positioning data and environmental perception data are acquired through onboard sensors and refined using a pre-established high-precision map to obtain refined location information. A multi-source data fusion method is then employed to optimize vehicle position estimation, resulting in fused optimized location data. Path deviation is calculated based on this fused optimized location data, triggering the automatic cruise control function to generate trajectory control commands. The vehicle's trajectory is adjusted by incorporating driver intent recognition and system response strategies from human-machine collaborative driving. The trajectory is smoothed using vibration detection and environmental perception data to obtain final trajectory execution data. The path planning is dynamically updated based on this final trajectory execution data, forming a closed-loop control mechanism.

[0038] Specifically, the process of acquiring vehicle positioning data and environmental perception data through onboard sensors and refining the location information by combining it with a pre-established high-precision map includes: acquiring environmental perception data through lidar scanning, wherein the environmental perception data consists of point cloud data of the surrounding environment;

[0039] Based on the semi-trailer's work log, the navigation mode of the semi-trailer is identified, wherein the navigation mode includes inertial navigation solution mode and real-time navigation mode.

[0040] The system acquires the initial position signal from the GPS receiver and obtains the acceleration and angular velocity data of the semi-trailer from the inertial measurement unit. When the quality of the initial position signal is lower than a preset signal quality threshold, it switches to the inertial navigation calculation mode to perform initial position estimation processing, obtaining initial position estimation data and error covariance matrix. Based on a preset inertial navigation model, it inputs the acceleration data, angular velocity data, and initial position estimation data of the semi-trailer and outputs position prediction data. It uses an iterative nearest-point matching method to register the point cloud data with a preset high-precision map. If the registration error is greater than the error threshold, it starts a re-matching procedure, performing at least 3 iterative matchings to obtain the adjusted registration result. Combining historical trajectory data and vehicle initial positioning data, it obtains the refined position information and heading angle data of the vehicle in the map coordinate system and updates the position prediction data. It monitors the change in vehicle load mass in real time through a weighing sensor to obtain the vehicle's center of gravity offset. If the vehicle's center of gravity offset exceeds a preset offset threshold, it adjusts the inertial navigation parameters and updates the position prediction data.

[0041] Basic data is collected within the vehicle's unified coordinate system using an onboard inertial measurement unit (IMU) and a GPS receiver. Initial position signals are obtained from the GPS receiver, and acceleration and angular velocity data of the semi-trailer are acquired from the IMU, resulting in preliminary positioning reference information. Based on this information, the GPS signal carrier-to-noise ratio (CNR), i.e., the initial position signal quality, is detected. If the CNR is below a preset signal quality threshold of 30 dB / Hz, the system switches to an inertial navigation-based solution mode to obtain preliminary position estimation data and the corresponding error covariance matrix. The surrounding environment is scanned using an onboard lidar to acquire point cloud data, resulting in an environmental point cloud dataset containing three-dimensional spatial coordinates. Based on this dataset, an iterative nearest-point matching method is used to initially register the point cloud data with a pre-established high-precision map, determining preliminary vehicle position information and heading angle data. If the registration error exceeds a preset error threshold of 0.1 meters, a re-matching procedure is initiated, performing at least three iterative matching iterations to obtain the adjusted registration result. If the adjusted registration result still has an error exceeding 0.1 meters, the current error value is recorded and passed to the subsequent fusion calculation module to obtain the recorded error dataset. Based on the registration result and combined with historical trajectory data, the refined position information and heading angle data of the vehicle in the map coordinate system are updated to obtain a real-time optimized positioning dataset. Real-time changes in vehicle load mass are collected using onboard weighing sensors to calculate the vehicle's center of gravity offset, resulting in a center of gravity offset dataset. If the center of gravity offset exceeds a preset offset threshold of 0.2 meters, the motion parameters of the inertial navigation system are adjusted based on the offset, resulting in adjusted position prediction data and an uncertainty matrix. Based on the refined position information, heading angle data, and adjusted position prediction data, the error covariance matrix is ​​dynamically updated to obtain real-time optimized positioning reference data. Through multi-source data fusion processing, combined with the real-time optimized positioning reference data and environmental perception input data, the final vehicle state dataset is determined.

[0042] The system utilizes an onboard inertial measurement unit (IMU) to acquire triaxial acceleration data (±16g range) and angular velocity data (±2000° / s range) at a sampling frequency of 100Hz. Simultaneously, it receives a 1Hz position signal from a GPS receiver (RTK positioning mode). A position-velocity-attitude dataset, encompassing the east, north, and sky directions, is established in the vehicle coordinate system. When the GPS carrier-to-noise ratio drops to 28dB-Hz, the inertial navigation calculation mode is activated. Attitude calculation is performed using the quaternion method, and the displacement increment is obtained through Runge-Kutta integration, generating preliminary estimation data including the position error covariance matrix (diagonal element σ²=0.05m²). A Kalman filter (with a 9-dimensional state vector containing position, velocity, and attitude angular errors) fuses the inertial navigation output and GPS residuals. The process noise matrix Q is set as a diagonal matrix. When the standard deviation of the position estimate reaches 0.12 meters, the LiDAR scanning frequency is increased from 10Hz to 15Hz, and the least squares method is used to optimize the multi-sensor timestamp alignment parameters (time deviation converges to ±2ms). The corrected positioning data is then used to update the inertial navigation system error parameters (attitude error compensation) through an extended Kalman filter. The data recorder extracts the standard deviation of the driver's steering wheel angle (historical data window of 60 seconds) as an operational feature quantity, and inputs it into the path planning module synchronously with the positioning data. The path points are regenerated using cubic spline interpolation (the spacing is compressed to 0.5 meters), and the final output is a cruise command set containing lateral control quantity (front wheel angle δ=1.2rad) and longitudinal control quantity (acceleration a=0.3m / s²).

[0043] The vehicle-mounted LiDAR scans the surrounding environment at a frequency of 10Hz, acquiring approximately 100,000 points per frame of point cloud data. Voxel grid filtering is used to downsample this data to 50,000 points, resulting in an environmental point cloud dataset containing XYZ coordinates. Initial registration is performed using the ICP algorithm, with a maximum of 100 iterations and a convergence threshold of 0.05 meters. A re-matching mechanism is triggered when the matching residual reaches 0.12 meters, performing three optimization matches using the NDT algorithm. If the final residual still reaches 0.11 meters, the error value is recorded in the fusion queue. Combining the trajectory data from the past 5 seconds, a Kalman filter is used to update the vehicle position, outputting optimized data with a lateral positioning accuracy of 0.08 meters. The load cell monitors the load at a sampling frequency of 20Hz. When a change in mass exceeding 2 tons is detected, a longitudinal shift of the center of gravity of 0.25 meters is calculated, triggering the IMU parameter adjustment module. An extended Kalman filter is then used to update the attitude angle error by 0.5 degrees. By fusing lidar positioning data (variance 0.01 square meters), IMU heading data (variance 0.2 degrees), and historical trajectory data, the final pose is output through a particle filter algorithm, where the diagonal elements of the position covariance matrix are all less than 0.15 meters.

[0044] Specifically, the method of optimizing vehicle position estimation using multi-source data fusion to obtain fused and optimized position data includes: acquiring GPS position signals, inertial navigation prediction data, and refined LiDAR position information; aligning the timestamps of different data sources using time interpolation to obtain synchronized multi-source position data; and fusing the multi-source position data using an extended Kalman filter.

[0045] Initialize the extended Kalman filter based on multi-source location data, and set the initial values ​​of the state vector and the diagonal of the error covariance matrix;

[0046] Using acceleration and angular velocity data under inertial navigation, a state transition matrix is ​​obtained through a motion model, the predicted position and error covariance are updated, and fused optimized position data is generated. Uncertainty information is extracted from the fusion result to obtain the updated error covariance matrix. If the components of the error covariance matrix exceed a preset threshold, the fused optimized position data is then corrected a second time.

[0047] The system acquires refined position information and heading angle data. Vehicle load mass data is collected in real-time using onboard weighing sensors to obtain the vehicle's current spatial position, heading angle, and load mass distribution. Based on the collected load mass data, combined with the vehicle's geometric model and mass distribution algorithm, the position of the vehicle's center of gravity in the three-dimensional coordinate system is calculated, yielding the center of gravity coordinates. If the deviation between the center of gravity coordinates and the vehicle's standard center of gravity position exceeds a preset threshold of 0.2 meters, a center of gravity offset vector is generated based on the deviation to determine the center of gravity offset. Based on the center of gravity offset vector and the vehicle's kinematic model, the motion parameters of the inertial navigation system are adjusted to obtain an updated set of motion parameters. Using the updated set of motion parameters, the inertial navigation algorithm is run to calculate the adjusted vehicle position prediction data, obtaining the predicted position coordinates. Based on the predicted position coordinates and the error model of the inertial navigation system, an uncertainty matrix is ​​generated, yielding the uncertainty distribution of the position prediction. If the confidence level of the uncertainty matrix is ​​lower than a preset threshold, the position prediction data is optimized using a Kalman filter algorithm to obtain optimized position prediction data. Based on the optimized position prediction data, and combined with a multi-source data fusion algorithm, data from other sensors are integrated to obtain fused position estimation data. This fused position estimation data is then used to update the navigation parameters of the vehicle's automatic cruise control system, resulting in the final navigation control commands.

[0048] The vehicle's refined position information is obtained through a GNSS positioning module with centimeter-level accuracy. Simultaneously, an IMU sensor collects heading angle data at a sampling frequency of 100Hz, combined with an onboard weighing sensor monitoring the load mass of each axle in real time at a frequency of 10Hz (e.g., 5 tons for the front axle and 8 tons for the rear axle), resulting in complete vehicle mass distribution data. Based on the vehicle's three-dimensional geometric model, a weighted center-of-gravity algorithm is used to calculate the center-of-gravity coordinates. If the calculation shows a lateral offset of 0.25 meters exceeding the preset threshold of 0.2 meters, an offset vector [0.25, 0, 0] is generated. The angular velocity compensation parameters of the inertial navigation system are corrected based on this vector, and a quaternion update algorithm is used to adjust the attitude calculation process, outputting the updated position prediction value. A position uncertainty matrix is ​​calculated based on a Kalman filter error model. When the position covariance exceeds 0.1 square meters, a multi-source data fusion process is triggered, fusing the lidar point cloud matching results with visual SLAM data through an extended Kalman filter. Finally, the fused navigation control command is output, controlling the steering motor to perform heading correction at an angular velocity of 0.5 degrees / second.

[0049] The adjusted position prediction data and uncertainty matrix are acquired. Time interpolation is used to align the timestamps of the GPS position signal, inertial navigation prediction data, and refined LiDAR position information, resulting in time-synchronized multi-source position data. Based on this time-synchronized data, the extended Kalman filter (EPF) method is used to initialize the state vector and error covariance matrix, determining the initial fusion state. Through the EPF prediction step, the state vector and error covariance matrix are updated using the inertial navigation prediction data, yielding the predicted position and prediction error covariance. If the GPS position signal is available, its observation model is used to update the EPF state vector and error covariance matrix, resulting in fused GPS position data and error covariance. If the refined LiDAR position information is available, its observation model is used to further update the EPF state vector and error covariance matrix, resulting in fused optimized position data and an updated error covariance matrix. Based on the fused optimized position data and the updated error covariance matrix, the confidence interval of the position data is calculated, determining the optimized position reliability. The smoothed vibration signal is acquired and correlated with the fused and optimized location data through timestamp matching to obtain a time-aligned combination of vibration signal and location data. A comparison processing method is used to calculate the degree of matching between the vibration signal and the fused and optimized location data, determining the consistency of the integrated location data. Based on the consistency judgment result, the fused and optimized location data is updated to obtain the final integrated location data.

[0050] The adjusted position prediction data and uncertainty matrix are obtained. Linear interpolation is used to time-align the Global Positioning System (GPS) signal (e.g., GPS sampling frequency 10Hz), inertial navigation data (100Hz), and lidar data (20Hz), with an interpolation interval of 5ms, resulting in synchronized multi-source data. An extended Kalman filter is initialized based on the synchronized data, with the state vector set to 6 dimensions (3 axes each for position and velocity), and the initial values ​​of the diagonal of the error covariance matrix set to [0.1, 0.1, 0.1, 0.01, 0.01, 0.01]. In the prediction step, inertial navigation acceleration data (e.g., 0.2 m / s² on the X-axis) and gyroscope angular velocity (0.05 rad / s on the Z-axis) are used to calculate the state transition matrix F through a motion model. The process noise Q is set as a diagonal matrix [0.05, 0.05, 0.05, 0.005, 0.005, 0.005], updating the position prediction data and error covariance. If the GPS signal is valid (e.g., signal-to-noise ratio greater than 30dB), its observation matrix H (identity matrix) and measurement noise R (diagonal matrix [1, 1, 1]) are used for correction, reducing the fused position error to 0.5m. If the lidar data is valid (e.g., confidence level higher than 90%), the point cloud data is matched using the ICP algorithm, with the observation noise set to 0.2m, further optimizing the position to 0.3m accuracy. The 3σ confidence interval (e.g., ±0.9m on the X-axis) of the fused optimized position data is calculated to determine reliability. The smoothed vibration signal (sampling rate 1kHz) is matched with the fused optimized position data, and the temporal similarity is calculated using the DTW algorithm. If the matching error is less than 0.1, consistency is determined, and the final fused optimized position data is output.

[0051] Specifically, the step of calculating the path deviation based on the fused and optimized location data and triggering the automatic cruise control function to generate trajectory control commands includes: calculating the lateral deviation and heading deviation of the vehicle relative to the preset path based on the fused and optimized location data.

[0052] Construct a path point set H based on the preset path: ,in, Let N represent the coordinates of the nth path point, and N represent the total number of path points in the preset path r. This represents the x-coordinate of the nth path point. This represents the y-coordinate of the nth path point;

[0053] Calculate vehicle position The shortest distance point to the preset path r The specific formula is as follows:

[0054]

[0055] The tangent direction of the shortest distance point is: The angle between the tangent direction of the shortest distance point and the horizontal axis is the reference heading angle. ;

[0056] Based on the shortest distance point The lateral deviation and heading deviation are calculated using the following formulas:

[0057]

[0058]

[0059] in, Indicates lateral deviation. Indicates heading deviation. Indicates the vehicle's current heading angle;

[0060] When the absolute value of the lateral deviation is greater than the lateral deviation threshold or the heading deviation is greater than the heading threshold, the activation signal of the adaptive speed adjustment and lane keeping assist functions is triggered; the steering angle adjustment amount and speed adjustment value are calculated using the vehicle dynamics model; and the trajectory control command is generated and transmitted to the vehicle execution unit.

[0061] The system acquires the fused and optimized position data X_final. Position information is optimized by fusing navigation prediction information P_INS^update(t) and lidar data P_LiDAR(t) to obtain the precise coordinates and attitude data of the vehicle's current position. Based on the geometric data of the preset path, a coordinate transformation method is used to compare X_final with the preset path to determine the relative position of the vehicle's current position. The vertical distance between the vehicle's current position and the preset path is calculated to obtain the specific value of the lateral deviation. The attitude data of the vehicle's current position is compared with the tangent direction of the preset path to determine the specific value of the heading deviation. If the absolute value of the lateral deviation exceeds 0.15 meters or the heading deviation exceeds 5 degrees, the activation signal of the adaptive speed adjustment and lane keeping assist functions in the automatic cruise control is triggered. Based on the values ​​of the lateral and heading deviations, the required steering angle adjustment is calculated using a control algorithm to obtain the steering control parameters. By combining the vehicle dynamic model with the lateral and heading deviations, the speed adjustment value required for adaptive speed adjustment is calculated to determine the speed control parameters. If the difference between the navigation prediction information P_INS^update(t) and the lidar data P_LiDAR(t) exceeds the deviation threshold δ, then the position of X_final is fine-tuned using the fine-tuning coefficient α to obtain the updated position data. By inputting the steering angle adjustment and speed adjustment values ​​into the trajectory control execution module, the final vehicle control command is generated, completing the trajectory control execution.

[0062] The optimized position data X_final is obtained, and the inertial navigation prediction data of P_INS^update(t) and the lidar point cloud matching results of P_LiDAR(t) are fused using a Kalman filter algorithm. The fine-tuning coefficient α is set to 0.8. When the difference between the two exceeds the threshold δ=0.1 meters, weighted fusion is performed, outputting the vehicle's current latitude and longitude coordinates (118.78°E, 32.04°N) and yaw angle of 45.3°. Based on the high-precision map data of the preset path, a Frenet coordinate system is established, and the global coordinates of X_final are converted into the path longitudinal distance S=120.5 meters and the lateral offset e=-0.12 meters. The perpendicular distance between the vehicle's right front wheel and the path reference line is calculated through geometric projection. After compensation for the RTK positioning error of ±0.02 meters, the lateral offset e=-0.14 meters is obtained. The vehicle's heading angle of 45.3° is obtained using quaternion attitude calculation, and the difference between this and the tangent direction of the path at S=120.5 meters (43.1°) is obtained to obtain the heading offset. When the absolute value of the lateral deviation is detected to be 0.14 meters, which is below the 0.15-meter threshold, but the heading deviation is 2.2°, which is below the 5° threshold, the current control mode is maintained. If either condition is triggered, the PID controller is activated, and the front wheel steering angle increment is calculated based on the lateral deviation e and the heading deviation Δψ. Radius. Based on a two-degree-of-freedom vehicle model, with longitudinal velocity of 20 m / s and lateral acceleration of 0.3g as constraints, the speed adjustment Δv = -0.5 m / s is solved using model predictive control (MPC). During trajectory execution, the steering angle increment of 0.25 radians and the speed adjustment of -0.5 m / s are input to the drive-by-wire system, and a steering motor torque command of 125 N·m and a motor braking pressure of 0.2 MPa are sent via the CAN bus at a frequency of 100 Hz.

[0063] Specifically, the adjustment of the vehicle's driving trajectory by combining driver intent recognition and system response strategies in human-machine collaborative driving includes: obtaining driver intent and cruise status data through the in-vehicle interactive interface; analyzing driving intent tendencies using a preset driving behavior model to obtain preliminary intent recognition results;

[0064] By integrating the current vehicle's steering angle and speed data, and outputting the target steering angle adjustment and speed adjustment values ​​according to preset adjustment rules, preliminary control parameters are determined:

[0065] The initial control parameters are compared with the output of the real-time decision allocation model. When the comparison results are inconsistent, the control priority is reallocated through the human-machine collaborative driving decision module to obtain the adjusted control command, form the system response strategy, and adjust the vehicle's driving trajectory.

[0066] Real-time driver operation data and vehicle cruise status data are acquired through onboard sensors and an interactive interface. This data, combined with a pre-established driving behavior model, is used to analyze driver intent and obtain preliminary intent recognition results. Based on these results, the current vehicle steering angle and speed data are integrated to calculate the target steering angle adjustment and speed adjustment, determining preliminary control parameters. If the preliminary control parameters are inconsistent with the output of the real-time decision allocation model, the human-machine collaborative driving decision module reallocates control priorities, resulting in adjusted control commands. The onboard interactive interface provides feedback on current cruise status monitoring information and manual intervention priority commands, generating control signals to drive the steering actuators and speed controller. These generated control signals drive the steering actuators and speed controller, adjusting the vehicle's trajectory and acquiring actual steering angle and speed data. If the actual steering angle deviates from the target angle by more than a preset threshold of 2 degrees, the control parameters are recalculated by integrating optimized position data, resulting in corrected trajectory control commands. Based on these corrected commands, the drive parameters of the steering actuators and speed controller are updated, the vehicle's trajectory is adjusted, and new vehicle status data is acquired. Vibration data of the vehicle in complex environments is collected by vibration sensors. Combined with the corrected trajectory control commands, the impact of vibration on the trajectory is analyzed to determine vibration compensation parameters. The control signal is optimized using these vibration compensation parameters to update the operating status of the steering actuator and speed controller, maintaining a stable automatic cruise control effect.

[0067] The system collects steering wheel angle signals and accelerator pedal opening data through onboard sensors. Combined with driver operation frequency recorded via the interactive interface, a driving behavior model based on SVM is used to analyze driver intent. If the standard deviation of the steering wheel angle exceeds 5 degrees and the accelerator pedal fluctuation rate is greater than 10%, it is determined to be an intention to actively intervene. The system integrates steering angle data transmitted via the vehicle's CAN bus and GPS speed information, and uses a PID control algorithm to calculate the target steering adjustment. If the target steering angle deviates from the current angle by more than 1 degree, an adjustment command is output. When the control parameters output by the PID differ from the suggested values ​​of the reinforcement learning-based decision model by more than 15%, a human-machine collaborative arbitration mechanism is triggered. A weighted average method is used to reallocate control weights and generate priority commands. The onboard interactive interface displays the current cruise status, including steering deviation and speed difference. If the system detects a manual intervention signal for 500 milliseconds, control authority is switched. The generated PWM control signal drives the electric power steering motor. When the target steering angle is 25 degrees, if the actual angle feedback value deviates by more than 2 degrees, a Kalman filter algorithm is used to fuse IMU and wheel speed sensor data to recalculate the steering compensation. The updated control parameters are sent to the steering ECU to adjust the motor torque output, and simultaneously the braking pressure value of the ESP module is updated synchronously via the CAN bus. Vibration sensors collect chassis vibration data at a frequency of 100Hz, and FFT analysis is used to analyze the vibration spectrum. If the energy proportion in the 10-20Hz frequency band exceeds 30%, a reverse compensation signal is superimposed on the steering control command. The optimized control parameters are written to the steering and power control units in real time to maintain the steering angle error within ±0.5 degrees.

[0068] Specifically, the process of smoothing the vehicle trajectory using vibration detection and environmental perception data to obtain the final trajectory execution data includes: collecting vibration data during vehicle operation and performing preliminary storage and formatting to obtain standardized vibration information and extracting vibration acceleration peak values; when the vibration acceleration peak value is greater than a preset peak threshold, adjusting the filter window length to obtain adjusted filter parameters; and using an adaptive filtering method to smooth the fused and optimized position data, obtaining a smoothed position signal based on the adjusted filter parameters.

[0069] By combining position prediction data and heading angle data, trajectory data is obtained;

[0070] The position adjustment parameters are determined by combining driving load monitoring data; the trajectory data is then corrected a second time based on environmental perception data; optimized trajectory control commands are generated to obtain the final trajectory execution data.

[0071] Vibration data during vehicle operation is collected in real time by vehicle acceleration sensors. This data is initially stored and formatted to obtain standardized vibration information. Based on this standardized vibration information, the peak vibration acceleration is calculated. If the peak acceleration exceeds a preset threshold of 2.5 m / s², the filter window length is dynamically adjusted from an initial 0.2 seconds to a maximum of 0.5 seconds, resulting in adjusted filter parameters. An adaptive filtering method is used to smooth the fused and optimized position data, yielding a smoothed position signal based on the adjusted filter parameters. The vehicle's load mass change is monitored in real time by onboard weighing sensors, and the vehicle's center of gravity offset is calculated. If the offset exceeds a preset threshold of 0.2 meters, the motion parameters of the inertial navigation system are adjusted based on this offset, resulting in adjusted position prediction data. Based on the smoothed position signal and the adjusted position prediction data, combined with heading angle data, multi-source data fusion processing is performed to obtain fused trajectory data. Driver load monitoring data from human-machine cooperative driving is acquired, and combined with environmental adaptation adjustment parameters from automatic cruise control, the fused trajectory data is corrected to determine the trajectory execution data. Through a cyclic processing mechanism, the changing trends of cruise status and trajectory execution data are continuously compared to obtain the trajectory data change rate. Based on the trajectory data change rate, the steering angle and speed control parameters are dynamically adjusted to obtain trajectory control commands. Using the trajectory control commands and obstacle detection data, the driving path data is updated to obtain the final trajectory execution commands.

[0072] Vehicle vibration data is collected in real time using a vehicle accelerometer at a sampling rate of 100Hz. The data is stored in JSON format and normalized to obtain a standardized vibration information matrix. Based on the standardized data, the peak vibration acceleration in the 0-50Hz frequency band is calculated. When the detected peak exceeds 2.5 m / s², a linear interpolation algorithm is used to gradually extend the filter window length from 0.2 seconds to 0.5 seconds. A Kalman filter algorithm is used to smooth the fused position data from GNSS and IMU, with the filter gain coefficient dynamically adjusted according to the window length. Onboard load cells monitor the load on each axle at a frequency of 10Hz. The center of gravity offset is calculated using the torque balance equation. When the offset exceeds 0.2 meters, the least squares method is used to correct the accelerometer bias parameters of the inertial navigation system. The smoothed position signal and the corrected navigation data are input into the extended Kalman filter. When fusing the heading angle data, a 0.1 radian measurement noise compensation is introduced, and the covariance matrix of the fused trajectory is output. The driver load monitoring module analyzes steering wheel torque and pedal travel data, while the environmental adaptation module adjusts safety distance parameters based on obstacle density from millimeter-wave radar. These two parameters are then weighted and fused to perform a quadratic polynomial fitting on the trajectory data. A time-series model of the trajectory data is established, and a sliding window is used to calculate the rate of change of Euclidean distance between trajectory points adjacent to each other within 0.5 seconds. Based on this rate of change, a PID controller generates steering angle commands, and a feedforward control module calculates the target speed based on the radius of curvature. The two are then combined to output a PWM control signal. LiDAR point cloud data is used for obstacle detection via DBSCAN clustering, and combined with control commands to generate a rasterized path cost map. Finally, a trajectory execution command containing the speed curve and steering angle sequence is output.

[0073] Specifically, the step of dynamically updating the path planning based on the final trajectory execution data includes: acquiring information on the distance to obstacles and road signs ahead through an onboard camera and ultrasonic sensors; triggering automatic braking if the distance to an obstacle is less than a preset obstacle distance threshold or a speed limit sign is detected; generating a warning signal in conjunction with an emergency status indicator; evaluating the driver-system coordination status through a trust assessment model; adjusting the system response strategy based on the assessment results; and dynamically updating the vehicle's preset path using the path planning optimization function in automatic cruise control to obtain optimized path planning data and transmitting it to the monitoring module.

[0074] Based on the optimized path planning data, combined with the priority of manual intervention in human-machine cooperative driving and the cruise status monitoring in automatic cruise control, if the standard deviation of position accuracy is greater than the preset standard deviation threshold, the LiDAR scanning frequency is adjusted and the timestamp alignment parameters are optimized to obtain updated positioning correction data, which is then fed back to the preliminary position estimation processing.

[0075] Real-time road environment data is collected via onboard cameras and ultrasonic sensors. Preliminary information processing of obstacle distances and road signs is performed to obtain a raw environmental perception dataset. Based on this dataset, image recognition algorithms and sensor data fusion technology are used to classify and extract features from obstacle distances and traffic signs, determining obstacle distance values ​​and speed limit sign information. If the obstacle distance is less than a preset threshold of 3 meters or a speed limit sign is detected, a braking trigger signal is generated through the automatic cruise control module to obtain braking command data. Based on the braking command data, vehicle speed control parameters are adjusted, and speed adjustment is executed through the vehicle control module to obtain real-time adjusted driving status data. The real-time adjusted driving status data is converted into a warning signal format through the human-machine collaborative driving interface, obtaining feedback data from the interface. Based on this feedback, an emergency state switching algorithm is used to generate a driver warning signal and determine the warning signal transmission command. The onboard data recorder stores the real-time adjusted driving status data and warning signal transmission command, acquiring driver operating habits and system response logs. Based on driver operating habits and system response logs, combined with the path planning optimization function in automatic cruise control, the vehicle's preset path is dynamically updated to obtain optimized path planning data. The vehicle control module transmits optimized route planning data in real time, dynamically updates vehicle operating status, and obtains updated operating feedback data.

[0076] Road image data is acquired via an onboard camera at a sampling rate of 30 frames per second, while an ultrasonic sensor detects obstacle distances at a frequency of 10Hz. A Gaussian filter algorithm is used to reduce noise in the raw data, resulting in a perception dataset containing obstacle coordinates and traffic sign locations. Target detection is performed on the image data using a YOLOv5 model, with a confidence threshold of 0.7 for identifying speed limit signs. A Kalman filter algorithm is used to fuse multi-sensor data, calculating the relative distance error to obstacles to ±0.2 meters. When an obstacle is detected at a distance of 2.8 meters or a 60km / h speed limit sign is identified, the automatic cruise control module uses a PID control algorithm to generate a braking command, outputting a deceleration parameter of -2m / s². The vehicle control module adjusts the motor torque output according to the command and updates the vehicle speed via the CAN bus at 100ms intervals, obtaining status data including the current vehicle speed of 45km / h and acceleration of -1.8m / s². The interactive interface module converts the status data into a PWM signal to drive the HUD display, generating an 800Hz audible and visual alarm. The emergency state switching module determines the driving response delay based on the 500ms alarm trigger duration and generates a warning escalation command with a priority of 3. The data recorder stores vehicle status and warning logs at 1-second intervals, recording operational features such as a 5° steering wheel angle and a 30% brake pedal travel. The path planning module builds a Bayesian probability model based on historical operation data, recalculates the path weights, and outputs a new navigation trajectory. The control module adjusts the steering motor angle by 15° and updates the wheel speed pulse signals based on the new trajectory.

[0077] Real-time driving status data and driver behavior data are acquired by an onboard data recorder and stored in the system database, resulting in driving status datasets and behavior record datasets. Based on these datasets, and combined with a pre-established trust assessment model, analysis is performed to determine the cooperative status between the vehicle and the driver, yielding an initial cooperative trust assessment result. The onboard camera and ultrasonic sensors acquire information on the distance to obstacles and road signs ahead. If an obstacle is detected to be less than a preset threshold of 3 meters or a speed limit sign is identified, automatic braking is triggered, resulting in adjusted vehicle speed data. Based on the adjusted vehicle speed data and the initial cooperative trust assessment result, the path planning optimization function in automatic cruise control dynamically updates the vehicle's preset path, yielding optimized path planning data. Through the emergency state switching function in human-machine cooperative driving, combined with the optimized path planning data, a warning signal is sent to the interactive interface, resulting in real-time adjusted driving status data. Based on the real-time adjusted driving status data and the behavior record dataset, a second analysis is performed using the trust assessment model to determine the updated cooperative trust status, yielding a dynamic trust assessment result. The onboard data recorder stores the dynamic trust assessment result and the real-time adjusted driving status data in the system database, resulting in an updated comprehensive dataset. Based on the updated comprehensive dataset, combined with smoothed position signals and final trajectory execution data, iterative calculations are performed using the path planning optimization function to obtain further optimized path planning data. The automatic cruise control system then applies this further optimized path planning data to vehicle navigation to obtain the final vehicle driving control commands.

[0078] Vehicle speed, steering wheel angle, and brake pedal depth are collected at a sampling rate of 10Hz using an onboard data recorder. Driver input commands, such as turn signal usage frequency and emergency braking count, are also recorded and stored in an SQLite database to form a structured dataset. An LSTM-based trust assessment model is used, inputting lateral control deviation and longitudinal acceleration standard deviation from the driving state dataset, combined with human intervention frequency from the behavior recording dataset, to calculate an initial cooperative trust score and output a normalized result within the range of 0-1. The MobileNet-SSD object detection algorithm is used to process the 1080P video stream from the onboard camera, combined with 0.1-meter accuracy ranging data from an ultrasonic sensor. When an obstacle is detected at a distance of 2.8 meters or a 60km / h speed limit sign is identified, a PID controller is triggered to perform braking with a deceleration of 0.2g. The adjusted vehicle speed data and the initial trust score are input into an A* path planning algorithm. Considering a 5-meter safety distance constraint, the path curvature and speed profile are recalculated to generate an optimized path containing 200 waypoints. Path data is transmitted to the HMI interaction module via the CAN bus at 500ms intervals. When the trust score falls below 0.4, a red visual alarm and a level 3 buzzer alert are triggered. A sliding window mechanism is used to smooth the driving status data of the most recent 30 seconds using Kalman filtering. Combined with updated operation records, an SVM classifier is used to determine whether the current cooperative mode is "dominant" or "auxiliary." The dynamic trust assessment results are aligned with the vehicle's GNSS positioning data by timestamp and written to the time-series database InfluxDB, forming a multidimensional dataset containing location, speed, and trust level. Based on this dataset, a QP optimization solver is invoked, with minimum energy consumption as the objective function, to replan the path elevation profile and acceleration curve within a 100ms interval. Finally, the path point coordinates and desired speed values ​​are sent to the ESP and EPS controllers via the FlexRay bus to generate PWM duty cycle signals to control the actuators.

[0079] The system acquires real-time vehicle position standard deviation data from the position accuracy monitoring module. It checks if the standard deviation exceeds a preset threshold of 0.08 meters. If it does, a LiDAR scanning frequency adjustment command is triggered, resulting in preliminarily adjusted frequency parameters. Based on these parameters, the LiDAR timestamp alignment parameters are optimized, generating updated positioning correction data. The onboard processor feeds this updated data back to the preliminary position estimation processing module, yielding corrected preliminary position estimation data. This corrected preliminary position estimation data is then combined with manual intervention priority data from human-machine cooperative driving to determine the priority adjustment parameters for the current cruise state. Based on these parameters and cruise status monitoring data from automatic cruise control, the vehicle's preset path is dynamically updated, resulting in optimized path planning data. The onboard processor continuously monitors vehicle position accuracy and system stability based on the optimized path planning data, generating real-time position standard deviation data and system stability data. Finally, it checks if the real-time position standard deviation exceeds a preset threshold of 0.08 meters. If so, the latest driving status data is reacquired, resulting in an updated driving status dataset. Based on the updated driving status dataset, combined with driving behavior records and collaborative trust assessment data, dynamic route optimization parameters are generated using the system response logs stored in the vehicle data recorder. These dynamic route optimization parameters are then used to iteratively update and optimize the route planning data, resulting in the final adapted route planning data.

[0080] The system acquires real-time vehicle position standard deviation data from the position accuracy monitoring module. Using a Kalman filter algorithm, the current standard deviation is calculated to be 0.12 meters, exceeding the preset threshold of 0.08 meters. This triggers a scanning frequency adjustment command, increasing the LiDAR scanning frequency from 10Hz to 15Hz. Based on the adjusted frequency parameters, a timestamp alignment algorithm optimizes data synchronization between the LiDAR and the inertial measurement unit (IMU), generating positioning correction data with a time deviation of less than 5 milliseconds. The onboard processor inputs the correction data into an extended Kalman filter, fusing GPS and IMU data to output a corrected position with a diagonal element of the error covariance matrix less than 0.05. Combining the driver's manual intervention signal strength parameter (range 0-1), when the intervention signal strength is greater than 0.7, the path planning update frequency is reduced to 1Hz, generating priority adjustment parameters. Based on the priority parameters and cruise status monitoring data, the A* algorithm is used to recalculate the path, generating an optimized path with a lateral deviation of less than 0.15 meters. The processor calculates the position standard deviation in real time. When three consecutive sample values ​​exceed 0.08 meters, a driving status data update is triggered, collecting the latest IMU angular velocity data (accuracy 0.1 degrees / second) and GPS positioning data (carrier-to-noise ratio 35dB-Hz). Based on the updated driving data, the driver's steering operation history (the most recent 50 operation samples) is analyzed, and a Bayesian probability model is used to calculate a trust score. When the score is below 0.6, a conservative path planning mode is activated. The path planning algorithm is iteratively called by dynamically optimizing parameters until the final adapted path with a curvature change rate of less than 0.05 rad / m is output.

[0081] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0082] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatic cruise control of a semi-trailer under human-machine collaborative driving, characterized in that, include: Vehicle positioning data and environmental perception data are acquired by onboard sensors, and the location is refined by combining them with a pre-established high-precision map to obtain refined location information. A multi-source data fusion method is used to optimize vehicle position estimation, resulting in fused optimized position data. Path deviation is calculated based on the fused optimized position data, triggering the automatic cruise control function to generate trajectory control commands. The vehicle's trajectory is adjusted by combining driver intent recognition and system response strategies in human-machine cooperative driving. The trajectory is smoothed using vibration detection and environmental perception data to obtain final trajectory execution data. The path planning is dynamically updated based on the final trajectory execution data, forming a closed-loop control mechanism.

2. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The process of acquiring vehicle positioning data and environmental perception data through onboard sensors and refining the location information by combining it with a pre-established high-precision map includes: acquiring environmental perception data through lidar scanning, wherein the environmental perception data consists of point cloud data of the surrounding environment; Based on the semi-trailer's work log, the navigation mode of the semi-trailer is identified, wherein the navigation mode includes inertial navigation mode and real-time navigation mode. The system acquires the initial position signal from the GPS receiver and obtains the acceleration and angular velocity data of the semi-trailer from the inertial measurement unit. When the quality of the initial position signal is lower than a preset signal quality threshold, it switches to the inertial navigation solution mode to obtain the initial position estimation data and error covariance matrix. Based on a preset inertial navigation model, it inputs the acceleration, angular velocity, and initial position estimation data of the semi-trailer and outputs the position prediction data. It uses an iterative nearest-point matching method to register the point cloud data with a preset high-precision map. If the registration error is greater than the error threshold, it starts a re-matching procedure, performing at least three iterative matchings to obtain the adjusted registration result. Combining historical trajectory data and the vehicle's initial positioning data, it obtains the refined position information and heading angle data of the vehicle in the map coordinate system and updates the position prediction data. It monitors the changes in vehicle load mass in real time using a weighing sensor to obtain the vehicle's center of gravity offset. If the vehicle's center of gravity offset exceeds a preset offset threshold, it adjusts the inertial navigation parameters and updates the position prediction data.

3. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The method of optimizing vehicle position estimation using multi-source data fusion to obtain fused and optimized position data includes: acquiring GPS position signals, inertial navigation prediction data, and refined LiDAR position information; aligning the timestamps of different data sources using time interpolation to obtain synchronized multi-source position data; and fusing the multi-source position data using an extended Kalman filter. Initialize the extended Kalman filter based on multi-source location data, and set the initial values ​​of the state vector and the diagonal of the error covariance matrix; Using acceleration and angular velocity data under inertial navigation, a state transition matrix is ​​obtained through a motion model, the predicted position and error covariance are updated, and fused optimized position data is generated. Uncertainty information is extracted from the fusion result to obtain the updated error covariance matrix. If the components of the error covariance matrix exceed a preset threshold, the fused optimized position data is then corrected a second time.

4. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The step of calculating the path deviation based on the fused and optimized location data and triggering the automatic cruise control function to generate trajectory control commands includes: calculating the lateral deviation and heading deviation of the vehicle relative to the preset path based on the fused and optimized location data. Construct a path point set H based on the preset path: ,in, Let N represent the coordinates of the nth path point, and N represent the total number of path points in the preset path r. This represents the x-coordinate of the nth path point. This represents the y-coordinate of the nth path point; Calculate vehicle position The shortest distance point to the preset path r The specific formula is as follows: The tangent direction of the shortest distance point is: The angle between the tangent direction of the shortest distance point and the horizontal axis is the reference heading angle. ; Based on the shortest distance point The lateral deviation and heading deviation are calculated using the following formulas: in, Indicates lateral deviation. Indicates heading deviation. Indicates the vehicle's current heading angle; When the absolute value of the lateral deviation is greater than the lateral deviation threshold or the heading deviation is greater than the heading threshold, the activation signal of the adaptive speed adjustment and lane keeping assist functions is triggered; the steering angle adjustment amount and speed adjustment value are calculated using the vehicle dynamics model; and the trajectory control command is generated and transmitted to the vehicle execution unit.

5. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The method of adjusting the vehicle's trajectory by combining driver intent recognition and system response strategies in human-machine collaborative driving includes: acquiring driver intent and cruise status data through the in-vehicle interactive interface; analyzing driving intent tendencies using a preset driving behavior model to obtain preliminary intent recognition results; By integrating the current vehicle's steering angle and speed data, and outputting the target steering angle adjustment and speed adjustment values ​​according to preset adjustment rules, preliminary control parameters are determined: The initial control parameters are compared with the output of the real-time decision allocation model. When the comparison results are inconsistent, the control priority is reallocated through the human-machine collaborative driving decision module to obtain the adjusted control command, form the system response strategy, and adjust the vehicle's driving trajectory.

6. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The process of smoothing the vehicle's trajectory using vibration detection and environmental perception data to obtain the final trajectory execution data includes: collecting vibration data during vehicle operation and performing preliminary storage and formatting to obtain standardized vibration information and extracting the peak vibration acceleration; when the peak vibration acceleration exceeds a preset peak threshold, adjusting the filter window length to obtain adjusted filter parameters; and using an adaptive filtering method to smooth the fused and optimized position data, obtaining a smoothed position signal based on the adjusted filter parameters. By combining position prediction data and heading angle data, trajectory data is obtained; The position adjustment parameters are determined by combining driving load monitoring data; the trajectory data is then corrected a second time based on environmental perception data; the optimized trajectory execution command is generated to obtain the final trajectory execution data.

7. The human-machine collaborative driving automatic cruise control method for a semi-trailer as described in claim 1, characterized in that, The step of dynamically updating the path planning based on the final trajectory data includes: acquiring information on the distance to obstacles and road signs ahead using an onboard camera and ultrasonic sensors; triggering automatic braking if the distance to an obstacle is less than a preset obstacle distance threshold or a speed limit sign is detected; generating a warning signal in conjunction with an emergency status indicator; evaluating the driver-system coordination status using a trust assessment model; adjusting the system response strategy based on the assessment results; and dynamically updating the vehicle's preset path using the path planning optimization function in automatic cruise control to obtain optimized path planning data and transmitting it to the monitoring module. Based on the optimized path planning data, combined with the priority of manual intervention in human-machine cooperative driving and the cruise status monitoring in automatic cruise control, if the standard deviation of position accuracy is greater than the preset standard deviation threshold, the LiDAR scanning frequency is adjusted and the timestamp alignment parameters are optimized to obtain updated positioning correction data, which is then fed back to the preliminary position estimation processing.

Images