Rear-mounted emergency braking control method and system for a non-steer-by-wire manual transmission vehicle and storage medium

Abstract

The application discloses a rear-mounted emergency braking control method and system of a non-steer-by-wire manual transmission vehicle and a storage medium, and belongs to the field of intelligent automobile active safety technology. The method obtains point cloud data of the front of the vehicle collected by a vehicle-mounted laser radar, determines the longitudinal distance and relative closing speed of a target obstacle through coordinate transformation, region-of-interest cutting, outlier filtering, ground point segmentation and obstacle clustering, constructs a system total time delay from perception, calculation, communication, mechanical execution and continuous frame confirmation time delay, calculates a time delay sliding distance according to the relative closing speed and feeds forward compensates a braking trigger required distance, outputs a braking instruction when the same target continuous N effective point cloud frames all meet the trigger condition, cooperatively controls a rear-mounted braking execution mechanism and a rear-mounted clutch execution mechanism through a five-state finite state machine, and executes a safety exit in case of a fault, and iteratively corrects the subsequent braking trigger required distance according to a terminal residual distance error after braking.

Description

Technical Field

[0001] This invention relates to the field of active safety technology for intelligent vehicles, specifically to a retrofitted emergency braking control method, system, and storage medium for non-drive-by-wire manual transmission vehicles. Background Technology

[0002] Automatic Emergency Braking (AEB) is a crucial component of vehicle active safety technology. Existing AEB solutions typically rely on the vehicle's electronic braking interface or brake-by-wire capability, sending control commands to the braking actuators through the vehicle's control system. For non-drive-by-wire manual transmission vehicles lacking an electronic braking interface, aftermarket AEB requires an external mechanical actuator to drive the brake pedal and, when preset conditions are met, actuate the clutch pedal.

[0003] Compared to the original vehicle's electronic braking actuator chain, the aftermarket mechanical actuator chain adds point cloud perception processing, control calculation, communication transmission, mechanical backlash elimination, pedal action, and braking force establishment. Its response may also be affected by pedal free travel, motion inertia, brake state, and individual differences in the actuator. If the braking trigger boundary does not take into account the above-mentioned additional time delays, it may cause a mismatch between the timing of the braking command generation and the actual timing of braking force establishment.

[0004] The existing technology has at least the following areas that need improvement:

[0005] Some existing braking triggering methods do not uniformly model the perception, calculation, communication, mechanical response and target confirmation delays in the post-installed mechanical execution link, making it difficult to correct the braking triggering distance in advance according to the actual execution link;

[0006] Some manual transmission vehicles' emergency braking schemes only focus on a single aspect of braking control or clutch control, lacking coordinated control over the aftermarket braking actuator, aftermarket clutch actuator, and driver takeover.

[0007] Fixed braking distance parameters are difficult to adapt to changes in brake wear, pedal free travel, and actuator response, which may cause deviations in the final stopping margin as the vehicle's condition changes.

[0008] Furthermore, lidar point clouds can be affected by ground echoes, scattering noise, and short-term target loss. Single-frame judgment is easily interfered with by instantaneous ranging anomalies; when sensors, communication links, or actuators malfunction, fault exit and driver takeover logic that matches the subsequent mechanical execution mode is also required.

[0009] Therefore, a retrofit emergency braking control scheme is needed for non-drive-by-wire manual transmission vehicles. This scheme incorporates the waiting time generated by continuous frame confirmation and the response time of the retrofit mechanical actuator into the entire link delay. It corrects the braking trigger boundary based on the relative proximity of the vehicle to the target obstacle and combines brake-clutch coordinated control, fault exit, and terminal parking margin feedback correction to improve the adaptability and control consistency of the retrofit emergency braking system. Summary of the Invention

[0010] The purpose of this invention is to provide a retrofit emergency braking control method, system and storage medium for non-drive-by-wire manual transmission vehicles, so as to compensate for the additional time delay of the retrofit mechanical execution link and realize the coordinated control of the braking actuator, clutch actuator, fault exit and driver takeover.

[0011] To achieve the above objectives, the present invention adopts the following technical solution:

[0012] This invention provides a retrofitted emergency braking control method for a non-drive-by-wire manual transmission vehicle. The vehicle is equipped with a retrofitted brake actuator and a retrofitted clutch actuator that mechanically act on the brake pedal and clutch pedal, respectively. The method includes the following steps:

[0013] S1. Perception Input and Point Cloud Preprocessing Steps: Acquire 3D point cloud data of the front of the vehicle collected by the vehicle-mounted LiDAR. Perform coordinate transformation, region of interest clipping, outlier filtering, ground point segmentation, and obstacle clustering on the 3D point cloud data in sequence. Determine the longitudinal distance D of the obstacle in the kth valid point cloud frame for the same target obstacle. k Target obstacle velocity v obj,k and the longitudinal speed v of the vehicle k ;

[0014] S2. Full-link delay modeling and delay-based coasting distance calculation steps: The sum of sensing delay, control calculation delay, communication delay, mechanical actuator response delay, and continuous frame confirmation delay is determined as the total system delay. Based on the non-negative relative approach speed between the vehicle and the target obstacle and the total system delay, the reduction in longitudinal distance between the vehicle and the target obstacle within the total system delay is determined as the delay-based coasting distance D. τ,k ;

[0015] S3, Braking Trigger Requirement Distance Construction Steps: Based on the time-delay coasting distance D τ,k The non-negative relative approach speed, the vehicle's equivalent maximum braking deceleration, the target safe stopping margin, and the iterative learning correction amount obtained from the j-th braking learning are used to determine the braking triggering requirement distance D corresponding to the k-th effective point cloud frame. req,k ;

[0016] S4. Continuous Multi-Frame Braking Trigger Verification Step: For the same target obstacle, continuously determine whether the longitudinal distance of the obstacle in N valid point cloud frames is not greater than the corresponding braking trigger requirement distance; when all N consecutive valid point cloud frames meet the condition, output the braking trigger command; when any valid point cloud frame does not meet the condition or the target identifier changes, reset the continuous frame count.

[0017] S5. Desired braking deceleration planning steps: Determine the effective braking distance based on the longitudinal distance of the obstacle, the time delay sliding distance, the target safe stopping margin, and the iterative learning correction amount, and determine the desired braking deceleration, which is no greater than the maximum braking deceleration allowed by the system, based on the effective braking distance and the non-negative relative approach speed.

[0018] S6. Five-state finite state machine collaborative execution control steps: The rear-mounted braking actuator and the rear-mounted clutch actuator are controlled by a finite state machine including normal monitoring state, pre-braking state, emergency braking and clutch disengagement state, parking holding state and fault exit state; when the preset emergency braking condition is reached, the rear-mounted clutch actuator is controlled to disengage the clutch; when the preset fault condition is detected, the finite state machine is controlled to switch from the current state to the fault exit state, and the output of the automatic actuator is reduced by a preset rate of change;

[0019] S7. Iterative learning adaptive correction and update step: After an effective braking process ends, obtain the actual remaining terminal distance when the vehicle stops, and update the iterative learning correction amount according to the error between the actual remaining terminal distance and the target safe stopping margin, so as to correct the braking triggering requirement distance of subsequent braking processes.

[0020] Furthermore, the total system latency is the sum of sensing latency, control calculation latency, communication latency, mechanical actuator response latency, and continuous frame confirmation latency; the continuous frame confirmation latency is determined based on the number of continuous confirmation frames and the lidar frame rate.

[0021] Furthermore, the time-delayed gliding distance is determined based on the non-negative relative approach speed between the vehicle and the target obstacle and the total system time delay; when the target obstacle is stationary, the non-negative relative approach speed is equal to the longitudinal speed of the vehicle.

[0022] Furthermore, consecutive frame confirmation is performed on the same target obstacle; the consecutive frame count is reset when the target tracking identifier changes, the target data becomes invalid, or any valid point cloud frame does not meet the braking triggering conditions.

[0023] Furthermore, the coordinate transformation adopts a rotation matrix around the Y-axis of the vehicle coordinate system, the outlier filtering adopts the K-nearest neighbor average distance criterion, the ground point segmentation adopts a linear ground model in polar coordinate grid and an absolute height residual criterion, and the obstacle clustering adopts Euclidean clustering with minimum point number constraint.

[0024] Furthermore, the mechanical actuator response is equivalent to a model combining a first-order inertial element and a pure time-delay element to characterize the elimination of mechanical backlash, pedal action, and braking force establishment process.

[0025] Furthermore, the iterative learning correction amount is set with a preset upper limit and a preset lower limit, and is updated only when the preset valid sample conditions are met during the braking process.

[0026] Furthermore, when the driver's braking input reaches the preset takeover condition, the automatic braking command is stopped from being increased and the output of the aftermarket braking actuator is smoothly reduced; when the driver's steering input reaches the preset takeover condition, the automatic braking control is reduced or disengaged according to the current collision risk level.

[0027] The present invention also provides a rear-mounted emergency braking control system for a non-drive-by-wire manual transmission vehicle, including a lidar, a processor, a memory, a communication unit, a rear-mounted braking actuator, and a rear-mounted clutch actuator.

[0028] The lidar is used to collect three-dimensional point cloud data in front of the vehicle;

[0029] The memory stores a computer program, and the processor executes the computer program to implement the above-mentioned control method.

[0030] The aftermarket brake actuator is mechanically connected to the original vehicle's brake pedal without calling the original vehicle's electronic brake interface, and the aftermarket clutch actuator is mechanically connected to the original vehicle's clutch pedal.

[0031] The communication unit is used to transmit control commands and status feedback between the processor and the rear-mounted braking actuator and the rear-mounted clutch actuator.

[0032] The present invention also provides a computer-readable storage medium on which a computer program stored is executed by a processor to implement the above-described rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle.

[0033] Compared with the prior art, the present invention has at least the following beneficial effects:

[0034] 1. Integrate the continuous frame confirmation delay with the response delays of sensing, computing, communication and mechanical actuators into the overall link delay, and calculate the delay sliding distance based on the non-negative relative approach speed, so that the braking trigger requirement distance can compensate for the reduced longitudinal spacing during the generation of braking commands and the establishment of mechanical braking force.

[0035] 2. By confirming the same obstacle with N consecutive frames and incorporating the confirmation delay into the overall link delay, we can suppress false triggering caused by single-frame point cloud anomalies and reduce the adverse impact of confirmation waiting on triggering timing.

[0036] 3. The aftermarket brake actuator and aftermarket clutch actuator are coordinated by a five-state finite state machine, and a fault exit state that can be entered from any working state and a driver takeover priority logic are set to improve the determinism of aftermarket control of non-drive-by-wire manual transmission vehicles.

[0037] 4. Update the learning correction amount iteratively based on the terminal remaining distance error at the end of the effective braking process, so that the subsequent braking triggering distance can adapt to changes in brake state, pedal free travel and actuator response.

[0038] 5. Improve the stability of the longitudinal distance of the target obstacle used for braking judgment by using coordinate transformation, region of interest clipping, outlier filtering, polar coordinate ground point segmentation, and Euclidean clustering with minimum point number constraint. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating the overall process of the aftermarket emergency braking control method for non-drive-by-wire manual transmission vehicles provided in this embodiment of the invention.

[0041] Figure 2 A schematic diagram of the three-layer architecture of the rear-mounted emergency braking control system provided in an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of a lidar point cloud preprocessing pipeline provided in an embodiment of the present invention;

[0043] Figure 4 A finite state machine state transition diagram provided for an embodiment of the present invention;

[0044] Figure 5 The timing diagram of the entire link delay of the rear-mounted emergency braking control system provided in the embodiment of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] like Figures 1 to 5 As shown, the non-drive-by-wire manual transmission vehicle described in this embodiment is equipped with an aftermarket brake actuator and an aftermarket clutch actuator that mechanically act on the brake pedal and clutch pedal respectively without calling the original vehicle's electronic brake interface, including:

[0048] S1. Perception Input and Point Cloud Preprocessing Steps: Acquire 3D point cloud data of the front of the vehicle collected by the vehicle-mounted LiDAR. Perform coordinate transformation, region of interest clipping, outlier filtering, ground point segmentation, and obstacle clustering on the 3D point cloud data in sequence. Determine the longitudinal distance D of the obstacle in the kth valid point cloud frame for the same target obstacle. k Target obstacle velocity v obj,k and the longitudinal speed v of the vehicle k ;

[0049] S2. Steps for full-link delay modeling and delay-based coasting distance calculation: The sum of sensing delay, control calculation delay, communication delay, mechanical actuator response delay, and continuous frame confirmation delay is determined as the total system delay. Based on the non-negative relative approach speed between the vehicle and the target obstacle and the total system delay, the reduction in longitudinal distance between the vehicle and the target obstacle within the total system delay is determined as the delay-based coasting distance Dτ,k.

[0050] S3, Braking Trigger Requirement Distance Construction Steps: Based on the time-delay coasting distance D τ,k The non-negative relative approach speed, the vehicle's equivalent maximum braking deceleration, the target safe stopping margin, and the iterative learning correction amount obtained from the j-th braking learning are used to determine the braking triggering requirement distance D corresponding to the k-th effective point cloud frame. req,k ;

[0051] S4. Continuous Multi-Frame Braking Trigger Verification Step: For the same target obstacle, continuously determine whether the longitudinal distance of the obstacle in N valid point cloud frames is not greater than the corresponding braking trigger requirement distance; when all N consecutive valid point cloud frames meet the condition, output the braking trigger command; when any valid point cloud frame does not meet the condition or the target identifier changes, reset the continuous frame count.

[0052] S5. Desired braking deceleration planning steps: Determine the effective braking distance based on the longitudinal distance of the obstacle, the time delay sliding distance, the target safe stopping margin, and the iterative learning correction amount, and determine the desired braking deceleration, which is no greater than the maximum braking deceleration allowed by the system, based on the effective braking distance and the non-negative relative approach speed.

[0053] S6. Five-state finite state machine collaborative execution control steps: The rear-mounted braking actuator and the rear-mounted clutch actuator are controlled by a finite state machine including normal monitoring state, pre-braking state, emergency braking and clutch disengagement state, parking holding state and fault exit state; when the preset emergency braking condition is reached, the rear-mounted clutch actuator is controlled to disengage the clutch; when the preset fault condition is detected, the finite state machine is controlled to switch from the current state to the fault exit state, and the output of the automatic actuator is reduced by a preset rate of change;

[0054] S7. Iterative learning adaptive correction and update step: After an effective braking process ends, obtain the actual remaining terminal distance when the vehicle stops, and update the iterative learning correction amount according to the error between the actual remaining terminal distance and the target safe stopping margin, so as to correct the braking triggering requirement distance of subsequent braking processes.

[0055] The above method specifically includes: the system first processes the lidar point cloud, and obtains the longitudinal distance D of the obstacle in the kth effective point cloud frame for the same target obstacle. k And estimate the longitudinal velocity v of the target obstacle based on the target position in consecutive frames. obj,k The total system delay is determined according to equation (1):

[0056] (1)

[0057] In the formula, τ tot τ is the total system delay. s For sensing delay; τ c To control computational delay; τ m For communication delay; τ a For the response delay of the mechanical actuator; τ conf The consecutive frame acknowledgment delay is determined by equation (2):

[0058] (2)

[0059] In the formula, N is the number of consecutively confirmed valid point cloud frames, and f s The frame rate is the lidar speed. The non-negative relative approach speed v between the vehicle and the target obstacle is... c,k and the delay coasting distance D within the total system delay τ,k Determine according to formula (3):

[0060] (3)

[0061] In the formula, v k v is the longitudinal velocity of the vehicle. obj,k v is the longitudinal velocity of the target obstacle; when the target obstacle is stationary, v obj,k =0, therefore v c,k=v k The physical meaning of the aforementioned time-delay coasting distance is the reduction in longitudinal distance between the vehicle and the target obstacle within the total system time delay.

[0062] The braking triggering distance D for the kth effective point cloud frame is determined based on the time-delayed coasting distance, vehicle braking capacity, and target safe stopping margin. req,k Specifically, as shown in equation (4):

[0063] (4)

[0064] In the formula, a eq D0 is the vehicle's equivalent maximum braking deceleration in positive form; D0 is the target safe stopping margin; ΔD j This is the iterative learning correction value obtained after the j-th effective braking process. When the target obstacle is stationary, v in equation (4) is... c,k It can be determined by the longitudinal speed v of the vehicle k replace.

[0065] The steps for verifying braking triggering across multiple consecutive frames are as follows:

[0066] For the same target obstacle, define a trigger indication function I for the k-th valid point cloud frame. k As shown in equation (5):

[0067] (5)

[0068] When the trigger indication function of N consecutive valid point cloud frames satisfies equation (6), the braking trigger command is output:

[0069] (6)

[0070] When the target tracking identifier changes, the target data becomes invalid, or the I of any valid point cloud frame changes. k When the value is 0, the continuous frame counter is reset and starts counting again. Continuous frame verification is performed only on valid point cloud frames of the same target obstacle to avoid stitching together ranging results from different targets into a continuous confirmation result.

[0071] The desired braking deceleration planning steps are as follows:

[0072] The effective braking distance D is determined based on the longitudinal distance to the obstacle, the time-delayed sliding distance, the target safe stopping margin, and the iterative learning correction. e,k As shown in equation (7):

[0073] (7)

[0074] The desired braking deceleration a is determined based on the effective braking distance and the non-negative relative approach speed. d,k As shown in equation (8):

[0075] (8)

[0076] In the formula, a max The maximum permissible braking deceleration of the system; a d,k Represented by non-negative amplitude. When D e,k When the value is not greater than 0, the system outputs the maximum permissible braking deceleration.

[0077] Vehicle equivalent maximum braking deceleration a eq and the system's maximum allowable braking deceleration a max Calibration is performed based on vehicle type, road surface adhesion conditions, and the capabilities of aftermarket actuators.

[0078] The steps for finite state machine cooperative execution control are as follows:

[0079] The state set of the finite state machine is shown in equation (9):

[0080] (9)

[0081] The normal operating state transitions in the following sequence: normal monitoring state S0, pre-braking state S1, emergency braking and clutch disengagement state S2, and parking holding state S3. When a preset fault condition is detected, the system transitions from any operating state to the fault exit state S1. f .

[0082] The meanings of each state are as follows:

[0083] S0 is the normal monitoring state; S1 is the pre-braking state, in which the rear-mounted brake actuator eliminates mechanical backlash and establishes preload; S2 is the emergency braking and clutch disengagement state, in which the rear-mounted brake actuator outputs braking force and the rear-mounted clutch actuator disengages the clutch; S3 is the parking hold state; S f The system has exited due to a fault.

[0084] When sensor data remains invalid to the point of N fault When a frame or communication link times out, or when the actuator status feedback exceeds a preset range, the finite state machine transitions to state S with high priority. f ; in S f In the middle, the system stops adding automatic braking commands, reduces the output of the controllable actuator by a preset rate of change, issues a fault alarm and requests the driver to take over. After the fault is reset and the restart conditions are met, it returns to S0.

[0085] The iterative learning adaptive correction and update steps are as follows: a braking process is only identified as a valid braking sample when the target obstacle is continuously tracked, the actuator is fault-free, and the vehicle has come to a stop.

[0086] The longitudinal distance D of the obstaclek The acquisition process includes coordinate transformation, region of interest clipping, statistical outlier filtering, polar coordinate grid ground point segmentation, and Euclidean clustering with minimum point number constraint.

[0087] For point p in the lidar coordinate system L,i Through the rotation matrix R around the Y-axis of the vehicle coordinate system y (θ) Obtain point p in the vehicle coordinate system v,i As shown in equation (10):

[0088] (10)

[0089] In the formula, θ represents the installation pitch deviation of the lidar relative to the vehicle coordinate system. y (θ) is a 3×3 rotation matrix; when it is necessary to compensate for the installation translation, a translation vector can be added on the basis of the rotation transformation.

[0090] Point p L,i and p v,i All are represented by three-dimensional column vectors, thus avoiding the mixing of 3×3 rotation matrices and four-dimensional homogeneous coordinates.

[0091] In statistical outlier filtering, point p i The average Euclidean distance of K nearest neighbors is calculated according to equation (11):

[0092] (11)

[0093] (12)

[0094] When equation (12) holds, point p i Identify and remove outliers. In the formula, μ d Let σ be the mean of the K-nearest neighbor distances of all points to be processed. d K represents the standard deviation, and λ is the outlier determination coefficient. K and λ are calibrated based on the lidar point density, ranging range, and road environment.

[0095] Perform Euclidean clustering on the point cloud after removing ground points, determine the connected point set based on a distance threshold, and remove points with a number less than the minimum point count threshold N. min The target cluster is output as the effective obstacle cluster.

[0096] For each valid obstacle cluster, the longitudinal distance D of the obstacle is determined based on the nearest valid point of the obstacle cluster in the longitudinal direction of the vehicle body or the leading edge of the bounding box. k And obtain the target identifiers of consecutive frames through target association.

[0097] The polar coordinate grid ground point segmentation adopts a local linear ground model.

[0098] In each local polar coordinate grid, the ground height z is fitted based on the horizontal distance r of the point. g (r), as shown in equation (13):

[0099] (13)

[0100] In the formula, k g and b g These are the slope and intercept of the linear ground model within the current local grid, respectively.

[0101] When the absolute residual of the point cloud height relative to the linear ground model satisfies equation (14), the corresponding point is determined as a ground point:

[0102] (14)

[0103] In the formula, h th The height residual threshold is used; points that do not satisfy equation (14) are retained as non-ground points.

[0104] The response process of a mechanical actuator includes the elimination of mechanical backlash, pedal action, and the establishment of braking force. It can be equivalently represented by a model combining a first-order inertial element and a pure time-delay element, and its transfer function is shown in equation (15):

[0105] (15)

[0106] In the formula, K b For the equivalent gain of the actuator, T b L is the first-order inertial time constant. b Let be the pure time delay, and s be the complex frequency domain variable in the Laplace transform.

[0107] Mechanical actuator response delay τ a The time required for the actuator output to reach the preset steady-state ratio can be determined and obtained through bench testing or actual vehicle calibration.

[0108] For the j-th valid braking sample, the actual remaining terminal distance is D. end,j The terminal distance error is determined according to formula (16):

[0109] (16)

[0110] When e j A value greater than 0 indicates that the actual remaining distance of the terminal is less than the target safe stopping margin, and the trigger distance needs to be increased in subsequent braking; when e j A value less than 0 indicates that the actual remaining distance of the terminal is greater than the target safe parking margin.

[0111] (17)

[0112] In equation (17), K p and K i These are the proportional gain and integral gain, respectively; 'sat' is the limiting function used to limit ΔD. j Limited to a preset lower limit ΔD min With the preset upper limit ΔD max between.

[0113] In one implementation, K p It can be calibrated in the range of 0.1 to 0.6, K i It can be calibrated within the range of 0.01 to 0.1; ΔD is updated only when the braking sample meets the conditions of continuous target tracking, normal actuator status, and vehicle coming to a stop. j The parameter ranges mentioned above are used to describe the feasible calibration methods and do not constitute a limitation on the scope of protection.

[0114] Region of Interest (ROI) clipping is used to remove near-field points obscured by the vehicle body, far-field points that are outside the detection range, and points that are significantly deviated from the lane area.

[0115] The set of point clouds within the region of interest is determined according to equation (18):

[0116] (18)

[0117] In the formula, P is the point cloud set after coordinate transformation; X min X max Y min Y max Z min and Z max The boundary is set according to the vehicle size, the installation location of the lidar, and the target detection range.

[0118] The main state transition relations of the finite state machine are as shown in equation (19) and Figure 4 As shown.

[0119] When N consecutive valid point cloud frames meet the braking triggering conditions, the system moves from S0 to S1; when the preset emergency braking conditions are met, it moves from S1 to S2; when the vehicle speed is not higher than the parking threshold, it moves from S2 to S3.

[0120] (19)

[0121] Fault Exit Status S f It can be entered via S0, S1, S2 or S3.

[0122] In S0 state, environmental perception and trigger judgment are continuously performed;

[0123] In S1 state, the mechanical backlash of the rear-mounted brake actuator is eliminated and preload is established;

[0124] In S2 state, the rear brake actuator is controlled according to the desired braking deceleration, and the rear clutch actuator is controlled to disengage the clutch.

[0125] In S3 state, the braking force required to maintain the stop is released, and the stop holding is disengaged according to the reset condition;

[0126] S f In this state, the automatic braking command is stopped, the output of the controllable actuator is reduced by a preset rate of change, a fault alarm is issued, and the driver is requested to take over.

[0127] The preset emergency braking conditions for entering S2 from S1 can be determined according to equation (20):

[0128] (20)

[0129] In the formula, a emg For emergency deceleration threshold, v stall The stall threshold is the threshold corresponding to the critical idling condition of the engine. When equation (20) is satisfied, the rear clutch actuator is controlled to disengage the clutch; after the vehicle stops, the braking force and clutch disengagement state are maintained according to the parking holding conditions.

[0130] Driver takeover has higher priority than automatic braking logic. When the pedal force sensor, brake pedal displacement sensor, or other driver braking input signals reach the preset takeover conditions, the system stops increasing automatic braking commands and reduces the output of the aftermarket brake actuators at a preset rate of change based on the driver's braking input and the vehicle's current deceleration. When the steering wheel torque or angle change rate reaches the preset takeover conditions, the automatic braking control is reduced or deactivated based on the current collision risk level.

[0131] Based on the above method, the present invention also provides a retrofitted emergency braking control system for a non-drive-by-wire manual transmission vehicle, including a lidar, a processor, a memory, a communication unit, a retrofitted braking actuator, and a retrofitted clutch actuator. The lidar is used to collect three-dimensional point cloud data in front of the vehicle.

[0132] The memory stores a computer program, and when the processor executes the computer program, it performs point cloud processing, full-link latency calculation, braking triggering distance calculation, continuous frame verification, expected braking deceleration planning, finite state machine cooperative control, and iterative learning correction.

[0133] The aftermarket brake actuator is mechanically connected to the original vehicle's brake pedal without calling the original vehicle's electronic brake interface, and is used to drive the original vehicle's brake pedal according to the control commands output by the processor.

[0134] The aftermarket clutch actuator is mechanically connected to the original vehicle's clutch pedal and is used to drive the original vehicle's clutch pedal when the preset clutch disengagement conditions are met.

[0135] The communication unit is used to transmit control commands and status feedback between the processor, the rear-mounted brake actuator, and the rear-mounted clutch actuator.

[0136] The rear-mounted braking actuator and the rear-mounted clutch actuator are respectively equipped with at least one of the following status feedbacks: position, current, or force feedback. The processor determines whether the actuator is abnormal based on the status feedback.

[0137] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described control method.

[0138] The above technical solution has at least the following effects:

[0139] 1. The latency of sensing, computing, communication, mechanical execution, and continuous frame confirmation are all included in the total system latency. The latency coasting distance is calculated based on the relative proximity speed, and the braking triggering distance is corrected accordingly.

[0140] 2. Perform consecutive N-frame confirmation for the same target obstacle, and reset the count when the target changes or the data is invalid to reduce the impact of instantaneous point cloud anomalies on trigger judgment;

[0141] 3. The rear-mounted braking actuator and rear-mounted clutch actuator are controlled collaboratively by a five-state finite state machine, and fault exit and driver takeover priority logic are set.

[0142] 4. Update the iterative learning correction amount after the limit is based on the terminal remaining distance error during the effective braking process to improve the consistency of the terminal stopping margin under different vehicle states.

[0143] 5. Improve the stability of obstacle distance and target speed information used for braking judgment by multi-stage point cloud preprocessing and continuous tracking of the same target.

[0144] The specific implementation method is as follows:

[0145] Example 1: Overall System Workflow

[0146] This embodiment uses a manual transmission passenger vehicle equipped with a retrofitted brake actuator and a retrofitted clutch actuator as an example to illustrate the execution flow of the method of the present invention. The following parameters are calibration or simulation parameters for the corresponding embodiment. The parameters in different embodiments can be determined according to the vehicle model, actuator, and road conditions.

[0147] 1. Perception Processing

[0148] The lidar acquires point clouds in front of the vehicle at a frame rate of 10 Hz. Based on the lidar installation pitch angle θ, the point cloud is converted to the vehicle coordinate system according to equation (10); the region of interest is clipped according to equation (18), and in this embodiment, the point cloud within the range of longitudinal X∈[2,50] m, lateral Y∈[-2,2] m, and vertical Z∈[-2,0.2] m is retained. Statistical outlier filtering is performed according to equations (11) and (12), with K=50 and λ=1.0; polar coordinate grid ground point segmentation is performed according to equations (13) and (14), with the height residual threshold h being taken. th =0.1 m; finally, Euclidean clustering is performed, and the distance threshold d is taken. th =0.3 m, minimum number of points N min =20, output the effective obstacle clusters and the nearest longitudinal distance D. k .

[0149] 2. Delay Modeling

[0150] The total system delay, consecutive frame confirmation delay, non-negative relative approach speed, and delayed taxiing distance are determined according to equations (1) to (3). In the calibration parameters of this embodiment, the total system delay τ tot Take 0.40 s, which already includes the continuous frame confirmation delay; the mechanical actuator adopts the model of equation (15), with a time constant T. b Take 0.22 s as the pure time delay L b Determined based on the results of the tests conducted by the implementing agency.

[0151] During system operation, the delay components are estimated based on timestamps, communication feedback, and actuator status; when the target obstacle is stationary, according to D... τ,k =v k τ tot Calculate the time delay and coasting distance.

[0152] 3. Trigger judgment and deceleration planning

[0153] The braking triggering distance is calculated according to equation (4), and N consecutive frame confirmations are performed according to equations (5) and (6). The effective braking distance and expected braking deceleration are determined according to equations (7) and (8). In this embodiment, a eq =6.8 m / s², D0 is set within the range of 1.5 to 3.5 m according to the working conditions, N=3.

[0154] Continuous frame confirmation is performed on the same target obstacle; if any frame fails to meet the triggering condition, the target identifier changes, or the target data is invalid, the continuous frame count is reset.

[0155] When the effective braking distance D e,k When the value is not greater than 0, the system outputs the maximum permissible braking deceleration a. maxIn other cases, the output value is no greater than a. max .

[0156] 4. Execution Control

[0157] The finite state machine is initially in the S0 normal monitoring state. After the triggering conditions are met for N consecutive frames, it enters the S1 pre-braking state, and the brake actuator eliminates mechanical backlash and establishes preload; when the conditions shown in equation (20) are met, it enters the S2 emergency braking and clutch disengagement state; after the vehicle speed is not higher than the parking threshold, it enters the S3 parking holding state.

[0158] When sensor data remains invalid for 5 consecutive frames, actuator status feedback exceeds the calibration range, or internal system communication timeout exceeds 200ms, the finite state machine transitions from the current state to state S. f Fault exit status. The system stops adding automatic braking commands, reduces the output of the controllable actuator at a preset rate of change, sends fault information via the vehicle communication bus, and requests driver takeover.

[0159] 5. Adaptive correction

[0160] After the effective braking process is completed, record the actual remaining distance D at the terminal. end,j The terminal distance error is calculated according to equation (16), and ΔD is updated according to equation (17). j In this embodiment, ΔD j Limited to the range of [-1.0, 1.0]m; targets that are tracked out of order, fail to stop, or brake in a vehicle that has not completed its braking process are not used for updates.

[0161] By screening effective samples and limiting the amplitude as described above, the terminal distance error under abnormal operating conditions is avoided from being directly used for subsequent braking trigger distance correction.

[0162] Example 2: Performance Verification under Representative Operating Conditions

[0163] Taking a stationary obstacle condition with an initial vehicle speed of 33.10 km / h and an initial obstacle distance of 35.10 m as an example, the control process of the method of the present invention is verified.

[0164] The system parameters are: total system delay τ tot =0.40s, equivalent maximum braking deceleration a eq =6.80m / s², target safe stopping margin D0=1.5~3.5 m, trigger confirmation frame number N=3, lidar frame rate f s =10 Hz. From equation (2), the consecutive frame confirmation delay is 0.20s.

[0165] During execution, the system first satisfies D. k ≤D req,kAfter the valid point cloud frame is completed, the next two frames are confirmed, and the braking trigger command is output when the condition is met in the third valid point cloud frame; the remaining perception, calculation, communication and mechanical response delays are uniformly compensated by the total delay of equation (1). In the simulation results, the vehicle stops in front of the obstacle, the final speed is 0 km / h, and the remaining distance of the terminal is 1.84 m.

[0166] The experimental results are shown in Table 1:

[0167] Table 1 Comparison of AEB algorithm simulation results

[0168] In the simulation, the TTC algorithm uses the collision time threshold TTC. th =1.6s; The SSD_no_delay algorithm uses the same a as the method of this invention. eq =6.8 m / s² and D0=1.5~3.5 m, but let τ tot =0; The RSS-like algorithm uses a reaction time ρ=0.5s and corresponding longitudinal acceleration / deceleration parameters. Random variables are set as follows: initial vehicle speed v0 follows a uniform distribution over the interval [30,70] km / h, initial obstacle distance D... 0,init Following a uniform distribution over the interval [20, 60]m, the distance-sensing noise ε D It follows a normal distribution with a mean of 0 and a standard deviation of 0.3m. The total system delay τ tot It follows a uniform distribution on the interval [0.35, 0.45]. The collision rate is the proportion of simulation conditions where the longitudinal distance between the vehicle and the obstacle is not greater than 0 before the vehicle stops; the false trigger rate is the proportion of simulation conditions where the braking trigger command is output without meeting the preset danger conditions.

[0169] (15,000 Monte Carlo simulations)

[0170] TTC algorithm 28.12 5.23 0.82 SSD_no_delay algorithm 81.43 2.11 0.15 RSS-like algorithm 9.53 8.76 1.23 Method of the present invention 1.63 1.87 1.84

[0171] As shown in Table 1, under the 15,000 Monte Carlo simulations, the collision rate of the method of this invention is 1.63%, lower than the listed comparative methods; the false trigger rate is 1.87%, and the average remaining terminal distance is 1.84 m. These results are used to illustrate the performance of this embodiment under the same simulation conditions and do not constitute a limitation on results for other vehicle models and operating conditions.

[0172] The results of the ablation experiment are shown in Table 2:

[0173] Table 2 Comparison of Ablation Simulation Results

[0174] Complete method 1.63 — Remove delay compensation 18.83 +17.20 Remove consecutive frame confirmation 3.21 +1.58 Remove ILC Adaptive 2.87 +1.24

[0175] As shown in Table 2, under the simulation settings, removing the delay compensation increases the collision rate by 17.20 percentage points; removing the continuous frame confirmation and iterative learning correction increases the collision rate by 1.58 percentage points and 1.24 percentage points, respectively, indicating that the above components all affect the control results of this embodiment.

[0176] Example 3: Sensor Failure Scenario Verification

[0177] In a sensor failure verification, the LiDAR data remained continuously invalid when the vehicle was in the S1 pre-braking state. The number of consecutive invalid frames reached N. fault Then, the finite state machine transitions to S. f Fault exit status: The system stops adding automatic braking commands, reduces the output of the controllable actuator at a preset rate of change, sends fault information via the CAN bus, and requests driver takeover. The processing time from meeting the fault determination conditions to issuing the fault exit control command does not exceed 100ms.

[0178] The above embodiments are used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Equivalent substitutions or conventional adjustments made by those skilled in the art to parameters, actuator forms, point cloud processing methods, or state transition conditions without departing from the technical solutions defined in the claims shall all fall within the scope of protection of the present invention.

Claims

1. A retrofitted emergency braking control method for a non-drive-by-wire manual transmission vehicle, wherein the non-drive-by-wire manual transmission vehicle is equipped with a retrofitted brake actuator and a retrofitted clutch actuator that mechanically act on the brake pedal and clutch pedal respectively without calling the original vehicle's electronic brake interface, characterized in that, include: S1. Perception Input and Point Cloud Preprocessing Steps: Acquire 3D point cloud data of the front of the vehicle collected by the vehicle-mounted LiDAR; sequentially perform coordinate transformation, region of interest clipping, outlier filtering, ground point segmentation, and obstacle clustering on the 3D point cloud data; determine the longitudinal distance D of the obstacle in the kth valid point cloud frame for the same target obstacle. k Target obstacle velocity v obj,k and the longitudinal speed v of the vehicle k ; S2. Full-link delay modeling and delay-based coasting distance calculation steps: The sum of sensing delay, control calculation delay, communication delay, mechanical actuator response delay, and continuous frame confirmation delay is determined as the total system delay. Based on the non-negative relative approach speed between the vehicle and the target obstacle and the total system delay, the reduction in longitudinal distance between the vehicle and the target obstacle within the total system delay is determined as the delay-based coasting distance D. τ,k ; S3, Braking Trigger Requirement Distance Construction Steps: Based on the time-delay coasting distance D τ,k The non-negative relative approach speed, the vehicle's equivalent maximum braking deceleration, the target safe stopping margin, and the iterative learning correction amount obtained from the j-th braking learning are used to determine the braking triggering requirement distance D corresponding to the k-th effective point cloud frame. req,k ; S4. Continuous Multi-Frame Braking Trigger Verification Step: For the same target obstacle, continuously determine whether the longitudinal distance of the obstacle in N valid point cloud frames is not greater than the corresponding braking trigger requirement distance; when all N consecutive valid point cloud frames meet the condition, output the braking trigger command; when any valid point cloud frame does not meet the condition or the target identifier changes, reset the continuous frame count. S5. Desired braking deceleration planning steps: Determine the effective braking distance based on the longitudinal distance of the obstacle, the time delay sliding distance, the target safe stopping margin, and the iterative learning correction amount, and determine the desired braking deceleration, which is no greater than the maximum braking deceleration allowed by the system, based on the effective braking distance and the non-negative relative approach speed. S6. Five-state finite state machine collaborative execution control steps: The rear-mounted braking actuator and the rear-mounted clutch actuator are controlled by a finite state machine that includes normal monitoring state, pre-braking state, emergency braking and clutch disengagement state, parking holding state and fault exit state. When a preset emergency braking condition is reached, the rear-mounted clutch actuator is controlled to disengage the clutch. When a preset fault condition is detected, the finite state machine is controlled to transition from the current state to the fault exit state, and the output of the automatic actuator is reduced by a preset rate of change. S7. Iterative learning adaptive correction and update step: After an effective braking process ends, obtain the actual remaining terminal distance when the vehicle stops, and update the iterative learning correction amount according to the error between the actual remaining terminal distance and the target safe stopping margin, so as to correct the braking triggering requirement distance of subsequent braking processes.

2. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 1, characterized in that, The total system delay τ tot Continuous frame confirmation delay τ conf Non-negative relative approach velocity v c,k Delayed taxiing distance D τ,k And the braking triggering distance D req,k Determined according to the following relationship: Where, τ s To sense time delay, τ c To control computation delay, τ m For communication delay, τ a For mechanical actuator response delay, N is the number of consecutively acknowledged valid point cloud frames, and f is the response delay. s v is the frame rate of the lidar. obj,k Let a be the longitudinal velocity of the target obstacle. eq Let D0 be the vehicle's equivalent maximum braking deceleration in positive form, and ΔD be the target safe stopping margin. j This is the iterative learning correction amount obtained from the j-th braking learning.

3. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 2, characterized in that, Define a trigger indication function I for the k-th valid point cloud frame for the same target obstacle. k And when the sum of the trigger indication functions of N consecutive valid point cloud frames equals N, a braking trigger command is output: Among them, the target tracking identifier changes, the target data is invalid, or the I of any valid point cloud frame changes. k When the value is 0, the consecutive frame count is reset.

4. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 2, characterized in that, The effective braking distance D e,k and desired braking deceleration a d,k Determined according to the following relationship: Among them, a max The desired braking deceleration is the maximum allowable braking deceleration of the system, expressed as a non-negative amplitude.

5. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 2, characterized in that, The iterative learning correction is updated according to the following relationship: Among them, D end,j K represents the actual remaining terminal distance at the end of the j-th effective braking process. p and K i These are the proportional gain and integral gain, respectively, where sat is the amount of correction for iterative learning that is limited to a preset lower limit ΔD. min With the preset upper limit ΔD max The amplitude limiting function between.

6. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 1, characterized in that, The coordinate transformation uses a rotation matrix around the Y-axis of the vehicle coordinate system; the outlier filtering determines outliers based on the mean and standard deviation of the average Euclidean distance between the K nearest neighbors of each point and the average nearest neighbor distance of all points to be processed; the ground point segmentation fits a linear ground model within a polar coordinate grid, and points whose absolute residual of the point cloud height relative to the linear ground model is not greater than a height threshold are determined as ground points; the obstacle clustering uses Euclidean clustering with a minimum number of points constraint.

7. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 1, characterized in that, The response delay of the mechanical actuator is equivalently represented by a model combining a first-order inertial element and a pure time delay element; the finite state machine enters the pre-braking state from the normal monitoring state when the braking trigger condition is met, enters the emergency braking and clutch disengagement state when the expected braking deceleration reaches the emergency deceleration threshold or the vehicle speed is not higher than the stall threshold, and enters the parking holding state when the vehicle speed is not higher than the parking threshold; when the sensor data is continuously invalid, communication timeout occurs, or the actuator state is abnormal, it transitions from any working state to the fault exit state.

8. The rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle according to claim 1, characterized in that, It also includes a driver takeover priority step: when the driver's braking input reaches the preset takeover conditions, the automatic braking command is stopped from being added, and the output of the aftermarket braking actuator is reduced according to the driver's braking input and the vehicle's current deceleration at a preset rate of change; when the driver's steering input reaches the preset takeover conditions, the automatic braking control is reduced or deactivated according to the current collision risk level, and the driver takeover command has a higher priority than the automatic braking command.

9. A rear-mounted emergency braking control system for a non-drive-by-wire manual transmission vehicle, characterized in that, It includes a lidar, processor, memory, communication unit, rear-mounted braking actuator, and rear-mounted clutch actuator; The lidar is used to collect three-dimensional point cloud data in front of the vehicle; The memory stores a computer program, and when the processor executes the computer program, it causes the control system to perform the method of any one of claims 1 to 8; The aftermarket brake actuator is mechanically connected to the original vehicle's brake pedal without calling the original vehicle's electronic brake interface, and is used to drive the original vehicle's brake pedal according to the brake control commands output by the processor. The aftermarket clutch actuator is mechanically connected to the original vehicle's clutch pedal and is used to drive the original vehicle's clutch pedal when the preset clutch disengagement conditions are met. The communication unit is used to transmit control commands and status feedback between the processor, the rear-mounted braking actuator, and the rear-mounted clutch actuator.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the rear-mounted emergency braking control method for a non-drive-by-wire manual transmission vehicle as described in any one of claims 1 to 8.

Images