A longitudinal control method and system for vehicle-assisted driving based on self-learning of driving habits

Abstract

This invention relates to a longitudinal control method and system for vehicle assisted driving based on self-learning of driving habits, comprising the following steps: acquiring following target information; acquiring longitudinal control data for heavy commercial vehicle assisted driving based on the following target information; constructing a Transformer self-learning training model, using the longitudinal control data for heavy commercial vehicle assisted driving as input data for the Transformer training model, and acquiring the target following distance; performing limit correction on the target following distance based on a full-speed inter-vehicle distance algorithm to obtain a safe target following distance; and adjusting acceleration and deceleration control based on a PID algorithm according to the safe target following distance. This invention provides a longitudinal control solution for commercial vehicle assisted driving, with good comfort, collision safety, and braking safety.

Description

Technical Field

[0001] This invention relates to the field of commercial vehicle driver assistance technology, and in particular to a vehicle driver assistance longitudinal control method and system based on self-learning of driving habits. Background Technology

[0002] As one of the main functions of automotive driver assistance systems (ADAS), longitudinal control can take over the control of longitudinal throttle and brakes during assisted driving, thereby reducing the driver's workload and improving safety. ADAS is particularly suitable for heavy commercial vehicles, which are driven for profit, involve long driving hours, high labor intensity, and a high proportion of high-speed driving scenarios. The quality of longitudinal control is of great significance to ADAS.

[0003] Existing technologies generally determine target vehicle spacing through inter-vehicle time-distance methods, including headway, collision event prediction, and secondary optimization of vehicle spacing. The main purpose is to minimize the vehicle spacing for following control while ensuring safety. Then, vehicle power and braking control are performed through PID, fuzzy control, etc. For heavy commercial vehicles, this means throttle opening and braking pressure. However, these technologies have the following drawbacks:

[0004] (1) Insufficient braking safety and comfort: The longitudinal control function generally has the same speed control characteristics as the driver's driving. However, in some following conditions, due to the frequent acceleration and deceleration control of the system, the driver's driving experience is not comfortable enough. When heavy vehicles brake downhill, it will also bring some safety hazards.

[0005] (2) Does not conform to the driver's driving habits: Existing technologies generally analyze the habitual driving behavior of some drivers and then summarize a set of vehicle spacing control rules and formulas to set and control vehicle spacing. However, this may not conform to the habits of every user, causing the driver to subconsciously intervene in the system control frequently, resulting in the system exiting. Therefore, complaints and refusal to use the assisted driving function are generated, which is not conducive to the popularization of assisted driving systems. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a vehicle-assisted driving longitudinal control method and system based on self-learning of driving habits.

[0007] This invention is achieved using the following technical solution:

[0008] The vehicle assisted driving longitudinal control method based on driving habit self-learning includes the following steps:

[0009] Obtain information about the target to follow;

[0010] Based on the target information, obtain longitudinal control data for assisted driving of heavy commercial vehicles; construct a Transformer self-learning training model, and use the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model to obtain the target following distance;

[0011] Based on the full-speed vehicle distance algorithm, the target following distance is adjusted to obtain a safe target following distance.

[0012] Based on the target following distance, the acceleration and deceleration control is adjusted using a PID algorithm.

[0013] As a further explanation of the invention, the acquisition of the following target information includes:

[0014] Obtain information about the vehicle ahead; this information includes the type, position, and speed of the vehicle ahead.

[0015] Obtain the target based on information about the vehicle ahead;

[0016] Acquire road scene information; road scene information includes the angle of lane lines relative to vehicles, the number of lanes, and the lane width;

[0017] Target filtering is performed based on road scene information to identify the vehicle in front of this vehicle in its lane;

[0018] Based on the relative distance between this vehicle and the vehicle in front in its lane, key targets are selected, and the closest vehicle in front is taken as the key target to follow.

[0019] As a further explanation of the invention, the step of obtaining the target based on the preceding vehicle information includes:

[0020] Install millimeter-wave radar to obtain the position and speed of the vehicle in front;

[0021] Set up a camera to obtain the position, speed, and type of the vehicle in front;

[0022] The target location detected by millimeter-wave radar is calibrated with the target location detected by camera, and target fusion is generated to acquire the target.

[0023] As a further explanation of the invention, the step of acquiring longitudinal control data for assisted driving of heavy commercial vehicles based on the following target information; constructing a Transformer self-learning training model, using the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model, and obtaining the target following distance; includes:

[0024] Acquire longitudinal control data for assisted driving of heavy commercial vehicles; the longitudinal control data for assisted driving of heavy commercial vehicles includes the relative speed between the vehicle and the vehicle in front, the relative distance between the vehicle and the vehicle in front, the vehicle speed, and the acceleration of the vehicle in front;

[0025] A self-learning model is trained based on the Transformer deep learning model, and a Transformer self-learning training model is constructed. The relative speed and relative distance between the vehicle and the vehicle in front, the speed of the vehicle and the acceleration of the vehicle in front are used as input data, and the target following distance is used as output data.

[0026] As a further explanation of the invention, the calculation formula for obtaining the safe target following distance by adjusting the target following distance based on the full-speed inter-vehicle distance algorithm includes:

[0027] The formula for calculating the standard workshop time interval is:

[0028] d=v*τ+s

[0029] Where v is the vehicle speed, τ is the minimum inter-vehicle time distance, s is the minimum braking distance, and d is the target following distance;

[0030] The formula for calculating dynamic workshop time intervals is:

[0031] τ=τ0+c*v

[0032] Where τ0 is the fixed time interval between the two workshops, and c is a constant coefficient;

[0033] The formula for calculating the minimum safe following distance is:

[0034] d min =v*(τ0+c*v)+s

[0035] Where, d min To maintain the minimum safe following distance;

[0036] The formula for calculating the target following distance is:

[0037]

[0038] As a further explanation of the invention, the calculation formula for adjusting acceleration and deceleration control based on the PID algorithm according to the safe target following distance includes:

[0039] Δd=dd rel

[0040] Δv=v set -v ego

[0041]

[0042] Where Δd is the following distance error, d rel The actual distance to the vehicle in front, Δv is the following speed error, v set The target speed set by the driver, v egov is the speed of this vehicle. rel The relative speed between this vehicle and the vehicle in front is given by k1, k2, and k3, which are gain coefficients.

[0043] When a>0, acceleration is achieved, and a PID algorithm is used for real-time torque adjustment.

[0044] v exp =a*Δt+v ego

[0045] e(t) = v exp +v ego (t)

[0046]

[0047] Among them, v exp Let v be the desired vehicle speed, Δt be the control period, and v be the speed of the vehicle. ego (t) represents the vehicle speed at a certain moment, e(t) represents the speed control error at a certain moment, T represents the control torque, and K represents the speed control error. p K is the proportionality coefficient. i The integral coefficient;

[0048] When a≤0, it is deceleration; it is divided into:

[0049] When a1 < a ≤ 0, no deceleration control is performed;

[0050] When a2 < a ≤ a1, vehicles with retarders use retarder braking, while vehicles without retarders use engine braking.

[0051] When a < a2, a deceleration request a is sent to the EBS system to initiate service braking; where a1 and a2 are calibration values.

[0052] A vehicle-assisted driving longitudinal control system based on self-learning of driving habits, including the vehicle-assisted driving longitudinal control method based on self-learning of driving habits as described in any one of the above claims, characterized in that it includes:

[0053] The perception module is used to acquire information about the target being followed.

[0054] Driving habit self-learning module; Based on the following target information, obtain longitudinal control data of heavy commercial vehicle assisted driving; Construct a Transformer self-learning training model, and use the longitudinal control data of heavy commercial vehicle assisted driving as the input data of the Transformer training model to obtain the target following distance;

[0055] The following distance limit module is used to correct the target following distance based on the full-speed inter-vehicle time distance algorithm to obtain a safe target following distance.

[0056] The acceleration / deceleration control module adjusts the acceleration / deceleration control based on the safe following distance target using a PID algorithm.

[0057] Compared with the prior art, the present invention has the following beneficial technical effects:

[0058] (1) Comfort: Driving habits are learned online using deep learning algorithms to achieve longitudinal control that is more in line with the current driving habits of drivers;

[0059] (2) Collision safety: The following distance is limited based on the self-learning of driving habits to ensure the collision safety performance of the system;

[0060] (3) Braking safety: The combined auxiliary braking ensures that the brake pads will not overheat under long downhill conditions, thus improving braking safety performance. Attached Figure Description

[0061] Figure 1 This is a system schematic diagram of the present invention;

[0062] Figure 2 This is a schematic diagram of the method flow of the present invention;

[0063] Figure 3 This is a schematic diagram of the self-learning method of the present invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0065] Example 1

[0066] like Figure 1-3 As shown, the vehicle assisted driving longitudinal control method based on self-learning of driving habits includes the following steps:

[0067] Obtain information about the target to follow;

[0068] Based on the target information, obtain longitudinal control data for assisted driving of heavy commercial vehicles; construct a Transformer self-learning training model, and use the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model to obtain the target following distance;

[0069] Based on the full-speed vehicle distance algorithm, the target following distance is adjusted to obtain a safe target following distance.

[0070] Based on the target following distance, the acceleration and deceleration control is adjusted using a PID algorithm.

[0071] Obtain target information; including:

[0072] Obtain information about the vehicle ahead; this information includes the type, position, and speed of the vehicle ahead.

[0073] Obtain the target based on information about the vehicle ahead;

[0074] Acquire road scene information; road scene information includes the angle of lane lines relative to vehicles, the number of lanes, and the lane width;

[0075] Target filtering is performed based on road scene information to identify the vehicle in front of this vehicle in its lane;

[0076] Based on the relative distance between this vehicle and the vehicle in front in its lane, key targets are selected, and the closest vehicle in front is selected as the key target to follow.

[0077] Obtain the target based on information about the vehicle ahead; including:

[0078] Install millimeter-wave radar to obtain the position and speed of the vehicle in front;

[0079] Set up a camera to obtain the position, speed, and type of the vehicle in front;

[0080] The target location detected by millimeter-wave radar is calibrated with the target location detected by camera, and target fusion is generated to acquire the target.

[0081] Based on the target information, longitudinal control data for assisted driving of heavy commercial vehicles is obtained; a Transformer self-learning training model is constructed, using the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model to obtain the target following distance; including:

[0082] Acquire longitudinal control data for assisted driving of heavy commercial vehicles; the longitudinal control data for assisted driving of heavy commercial vehicles includes the relative speed between the vehicle and the vehicle in front, the relative distance between the vehicle and the vehicle in front, the vehicle speed, and the acceleration of the vehicle in front;

[0083] A self-learning model is trained based on the Transformer deep learning model, constructing a Transformer self-learning training model. The input data includes the relative speed and distance between the vehicle and the vehicle in front, the vehicle's speed, and the acceleration of the vehicle in front; the output data is the target following distance. Preferably, the relative distance, relative speed, and acceleration data between the vehicle and the vehicle in front are collected by the driver using external cameras and millimeter-wave radar while following the vehicle in front. The vehicle speed is obtained from vehicle status information provided by internal speed and steering sensors.

[0084] In the Transformer self-learning training model, the input data is preprocessed by applying a moving average filter to smooth out instantaneous fluctuations. Then, the continuous variables of the preprocessed input data are normalized to eliminate the influence of units. The input data is then constructed as time-series data, and the output data is used as the label. A three-layer Transformer training model is adopted, including an input layer, an encoding layer, and an output layer. The training, validation, and test sets are divided in an 8:1:1 ratio to ensure that each set contains different road scenarios and captures long-term dependencies in driving habits. During the driver's actual driving process, the Transformer self-learning training model is continuously updated and iterated, learning and fine-tuning online through real-time data to form an expected following distance that adapts to the driver's driving habits.

[0085] The full-speed vehicle distance algorithm applies a limit correction to the target following distance, and the formula for calculating the safe target following distance includes:

[0086] The formula for calculating the standard workshop time interval is:

[0087] d=v*τ+s

[0088] Where v is the vehicle speed, τ is the minimum inter-vehicle time distance, s is the minimum braking distance, and d is the target following distance;

[0089] The formula for calculating dynamic workshop time intervals is:

[0090] τ=τ0+c*v

[0091] Where τ0 is the fixed time interval between the two workshops, and c is a constant coefficient;

[0092] The formula for calculating the minimum safe following distance is:

[0093] d min =v*(τ0+c*v)+s

[0094] Where dmin is the minimum safe following distance;

[0095] The formula for calculating the target following distance is:

[0096]

[0097] To ensure that the target following distance during self-learning remains within a safe range, the full-speed vehicle distance algorithm is used to limit and correct the target following distance. The full-speed vehicle distance algorithm uses dynamic vehicle distance to meet the requirements of full-speed longitudinal control.

[0098] Based on the target following distance, the calculation formula for acceleration / deceleration control adjustment using the PID algorithm includes:

[0099] Δd=dd rel

[0100] Δv=v set -v ego

[0101]

[0102] Where Δd is the following distance error, d rel The actual distance to the vehicle in front, Δv is the following speed error, v set The target speed set by the driver, v ego v is the speed of this vehicle. rel The relative speed between this vehicle and the vehicle in front is given by k1, k2, and k3, which are gain coefficients.

[0103] When a>0, acceleration is achieved, and a PID algorithm is used for real-time torque adjustment.

[0104] v exp =a*Δt+v ego

[0105] e(t) = v exp +v ego (t)

[0106]

[0107] Among them, v exp Let v be the desired vehicle speed, Δt be the control period, and v be the speed of the vehicle. ego (t) represents the vehicle speed at a certain moment, e(t) represents the speed control error at a certain moment, T represents the control torque, and K represents the speed control error. p K is the proportionality coefficient. i The integral coefficient;

[0108] When a≤0, it is deceleration; it is divided into:

[0109] When a1 < a ≤ 0, no deceleration control is performed;

[0110] When a2 < a ≤ a1, vehicles with retarders use retarder braking, while vehicles without retarders use engine braking.

[0111] When a < a2, a deceleration request a is sent to the EBS system to initiate service braking; where a1 and a2 are calibration values.

[0112] The above acceleration and deceleration control uses a PID algorithm for adjustment. Acceleration is achieved using VCU torque control. Deceleration is achieved by controlling engine braking, retarder braking, and service braking to jointly execute the target deceleration based on the deceleration request value. Engine braking is controlled by VCU torque control, and retarder braking is controlled by RCU controller. The execution values ​​of engine braking, retarder braking, and service braking are achieved through braking calibration.

[0113] As described above, this invention's vehicle-assisted driving system, based on self-learning of driving habits, automatically follows the vehicle in front or travels at a target speed in high-speed scenarios, reducing driver workload. Compared to ordinary adaptive cruise control, the self-learning following mode better matches the driver's own following habits. In long-distance use scenarios for commercial vehicles, it reduces the probability of driver intervention, improves driving comfort, and reduces driver fatigue. Simultaneously, the following distance safety limit ensures a safe distance from the vehicle in front at all speeds. On the other hand, the system's acceleration and deceleration control module addresses the potential for brake pad overheating and reduced braking capacity when heavy commercial vehicles continuously apply EBS service braking on downhill sections. It integrates auxiliary braking for deceleration control, and calibration ensures the stability and continuity of deceleration control, improving braking safety on long downhill sections.

[0114] Example 2

[0115] like Figure 1-3 As shown, the vehicle assisted driving longitudinal control system based on driving habit self-learning includes the vehicle assisted driving longitudinal control method based on driving habit self-learning as described in any of the above claims, including...

[0116] The perception module is used to acquire information about the target being followed.

[0117] Driving habit self-learning module; Based on the following target information, obtain longitudinal control data of heavy commercial vehicle assisted driving; Construct a Transformer self-learning training model, and use the longitudinal control data of heavy commercial vehicle assisted driving as the input data of the Transformer training model to obtain the target following distance;

[0118] The following distance limit module is used to correct the target following distance based on the full-speed inter-vehicle time distance algorithm to obtain a safe target following distance.

[0119] The acceleration / deceleration control module adjusts the acceleration / deceleration control based on the safe following distance target using a PID algorithm.

[0120] As described above, the overall architecture of this invention is divided into four core modules: a perception module, a driving habit self-learning module, a following distance limit module, and an acceleration / deceleration control module. The longitudinal control function receives the current state of the vehicle interior and the external environment from the perception layer. After decision-making through perception fusion, driving habit self-learning, following distance limit determination, and acceleration / deceleration control calculation, it controls the engine, service braking system, and retarder in the execution layer to accelerate and decelerate, thereby achieving longitudinal control.

[0121] Those skilled in the art will readily understand that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A vehicle assisted driving longitudinal control method based on self-learning of driving habits, characterized in that, Includes the following steps: Obtain information about the target to follow; Based on the target information, obtain longitudinal control data for assisted driving of heavy commercial vehicles; construct a Transformer self-learning training model, and use the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model to obtain the target following distance; Based on the full-speed vehicle distance algorithm, the target following distance is adjusted to obtain a safe target following distance. Based on the safe following distance target, the acceleration and deceleration control is adjusted using a PID algorithm. The calculation formula for acceleration / deceleration control adjustment based on the PID algorithm according to the safe following distance target is described above. include: Where, Δ d To account for following distance error, d rel Δ is the actual distance to the vehicle in front. v To account for the following vehicle speed error, v set The target speed set for the driver v ego For the vehicle's speed, v rel This represents the relative speed between this vehicle and the vehicle in front. k1, k2, k3 This is the gain coefficient; when a When the torque is >0, acceleration is achieved, and a PID algorithm is used for real-time torque adjustment. in, v exp For the desired vehicle speed, Δ t To control the cycle, v ego (t) The speed of this vehicle at a certain moment. e(t) To control vehicle speed error at a certain moment, T To control torque, K p This is the proportionality coefficient. K i The integral coefficient; when a ≤0 indicates deceleration; it is divided into: when a 1 < a When the value is ≤0, no deceleration control is performed; when a 2< a ≤ a At time 1, vehicles equipped with retarders will use the retarder for braking, while vehicles without retarders will use engine braking. when a < a At 2 o'clock, a deceleration request is sent to the EBS system. a Apply the service brakes; among them a 1. a 2 is the calibration value.

2. The vehicle assisted driving longitudinal control method based on self-learning of driving habits as described in claim 1, characterized in that, The acquisition of the following target information includes: Obtain information about the vehicle ahead; this information includes the type, position, and speed of the vehicle ahead. Obtain the target based on information about the vehicle ahead; Acquire road scene information; road scene information includes the angle of lane lines relative to vehicles, the number of lanes, and the lane width; Target filtering is performed based on road scene information to identify the vehicle in front of this vehicle in its lane; Based on the relative distance between this vehicle and the vehicle in front in its lane, key targets are selected, and the closest vehicle in front is selected as the key target to follow.

3. The vehicle assisted driving longitudinal control method based on self-learning of driving habits as described in claim 2, characterized in that, The step of obtaining the target based on the preceding vehicle information includes: Install millimeter-wave radar to obtain the position and speed of the vehicle in front; Set up a camera to obtain the position, speed, and type of the vehicle in front; The target location detected by millimeter-wave radar is calibrated with the target location detected by camera, and target fusion is generated to acquire the target.

4. The vehicle assisted driving longitudinal control method based on self-learning of driving habits as described in claim 3, characterized in that, The process includes: acquiring longitudinal control data for assisted driving of heavy commercial vehicles based on target information; constructing a Transformer self-learning training model, using the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model, and obtaining the target following distance; including: Acquire longitudinal control data for assisted driving of heavy commercial vehicles; the longitudinal control data for assisted driving of heavy commercial vehicles includes the relative speed between the vehicle and the vehicle in front, the relative distance between the vehicle and the vehicle in front, the vehicle speed, and the acceleration of the vehicle in front; A self-learning model is trained based on the Transformer deep learning model, and a Transformer self-learning training model is constructed. The relative speed and relative distance between the vehicle and the vehicle in front, the speed of the vehicle and the acceleration of the vehicle in front are used as input data, and the target following distance is used as output data.

5. The vehicle assisted driving longitudinal control method based on self-learning of driving habits as described in claim 4, characterized in that, The formula for calculating the safe target following distance by adjusting the target following distance based on the full-speed inter-vehicle distance algorithm includes: The formula for calculating the standard workshop time interval is: in, v For the vehicle's speed, τ Minimum workshop time interval, s Minimum braking distance, d common This is the standard workshop time interval; The formula for calculating dynamic workshop time intervals is: in, τ 0 Set a time interval between the two workshops. c The coefficients are constants. The formula for calculating the minimum safe following distance is: in, d min To maintain the minimum safe following distance; The formula for calculating the target following distance is: In the formula, The target following distance is output by the Transformer deep learning model.

6. A vehicle-assisted driving longitudinal control system based on self-learning of driving habits, comprising the vehicle-assisted driving longitudinal control method based on self-learning of driving habits as described in any one of claims 1-5, characterized in that, include: Sensing module; Used to obtain information about the target being followed; Driving habit self-learning module; Based on the target information, obtain longitudinal control data for assisted driving of heavy commercial vehicles; construct a Transformer self-learning training model, and use the longitudinal control data for assisted driving of heavy commercial vehicles as input data for the Transformer training model to obtain the target following distance; The following distance limit module is used to correct the target following distance based on the full-speed inter-vehicle time distance algorithm to obtain a safe target following distance. The acceleration / deceleration control module adjusts the acceleration / deceleration control based on the safe following distance target using a PID algorithm.

Images