Following stop control method based on monocular vision

Abstract

The invention discloses a tracking and parking control method based on monocular vision, which comprises the steps of dividing a tracking and parking control process into a plurality of stages, determining a tracking and parking mode of a vehicle based on the running information of a front vehicle identified by monocular vision, and controlling the running of the vehicle based on the determined tracking and parking mode of the vehicle. The method has the advantages that the pause phenomenon caused by monocular recognition errors is reduced, and the stability and comfort of following stop control are improved. The following parking distance control in the following parking process, the driving comfort in the following parking process and no abnormal secondary starting after following parking are obviously improved, and the overall performance of the intelligent driving system is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of ADAS assisted driving, and in particular to a monocular vision-based front-end and rear-end control method. Background Art

[0002] ADAS assisted driving technology can use a single front camera to detect the movement of vehicles within a certain range in front, and output the motion data (distance, speed, relative acceleration, etc.) of the target vehicle through a visual algorithm. The control algorithm uses this data to control the motion of the vehicle, so that the vehicle can stop after the front vehicle stops, thereby realizing follow-up and stop control based on monocular vision.

[0003] The accuracy of monocular vision stop-and-follow control is not as high as that of LiDAR, so there may be certain detection errors, resulting in a poor experience. The current monocular camera detects the front target, and after being output by the visual algorithm, there is a certain error between the detection value and the true value. Compared with models with front radar, this error is greater than the true value. This recognition error (speed and distance of the front vehicle, etc.) will not have a significant impact on medium and high speeds because the distance and speed of the front vehicle are relatively large. However, it has a great impact on vehicle control in the low-speed stop-and-follow stage. Slight changes in speed and distance can easily cause the whole vehicle control to be jerked. After the stop-and-follow, when the front vehicle is stationary, the camera's recognition of the speed and distance of the front vehicle target also changes slightly. For example, it is easy to mistakenly recognize that the distance to the front vehicle has become slightly farther, and the front vehicle has a certain speed, resulting in a second start.

[0004] Therefore, for the monocular vision follow-up and stop conditions, it is necessary to achieve 1) the follow-up and stop process is comfortable and without setbacks 2) the distance consistency of each follow-up and stop is good 3) the second start phenomenon does not occur; the monocular recognition follow-up and stop control strategy of the existing technology cannot meet the above requirements. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a follow-stop control method based on monocular vision, thereby reducing the frustration caused by monocular recognition errors and improving the stability and comfort of the follow-stop control.

[0006] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a follow-up stop control method based on monocular vision, including dividing the follow-up stop control process into multiple stages, each stage corresponding to a follow-up stop mode, determining the follow-up stop mode of the vehicle based on the preceding vehicle running information recognized by monocular vision, and controlling the driving of the vehicle based on the determined follow-up stop mode of the vehicle.

[0007] The follow-stop mode includes at least one of a follow-motion mode, a follow-stop deceleration mode, and a braking deceleration mode.

[0008] In the medium and high speed stage and when the front vehicle is moving, the following motion mode is used to control the vehicle. In the following motion mode, the MPC algorithm is used to predict the speed of the front vehicle, the speed of the own vehicle, and the distance of the front vehicle within a certain time domain. Under the constraints of the speed of the front vehicle, the speed of the own vehicle, and the distance between the two vehicles, the vehicle speed is controlled to drop to the preset value.

[0009] After the front vehicle stops, the vehicle is controlled in follow-stop deceleration mode, taking the response delay of the actuator into account, predicting the vehicle speed and the distance to the front vehicle within a certain time domain in the future, and calculating the target acceleration after correction based on the kinematic formula, and using this acceleration to control the vehicle's deceleration.

[0010] After the vehicle decelerates to a fixed distance and vehicle speed, it enters the braking deceleration mode. In the braking deceleration mode, a fixed deceleration rate is used to control the vehicle to decelerate to a stop.

[0011] During the following stop process, when the speed of the vehicle and the speed of the preceding vehicle are lower than the set value, the identified movement information of the preceding vehicle is filtered.

[0012] When the detected distance to the vehicle ahead changes but the vehicle speed does not change, a low-pass filter is used to filter the distance to the vehicle ahead for a certain period of time until the difference between the filtered data and the distance to the vehicle ahead detected by the camera is less than a certain value;

[0013] 2) When the speed of the identified front vehicle changes but the distance does not change, a low-pass filter is used to filter the speed of the front vehicle for a certain period of time until the difference between the filtered data and the speed of the front vehicle detected by the camera is less than a certain value;

[0014] 3) When the identified front vehicle speed and front vehicle distance both jump within a certain range, if the jumps are all decreasing, the reduced value is used as input, and the table is looked up to determine whether it is outside the safe area based on the speed of this vehicle, the speed of the front vehicle, and the relative distance. If it is outside the safe distance, low-pass filtering is performed. If it is not outside the safe distance, no processing is performed for safety reasons. If the jumps are all increasing, the final acceleration output is low-pass filtered.

[0015] Set the starting boundary conditions of the follow-stop controller. The vehicle is allowed to start again only when the starting boundary conditions are met.

[0016] The starting boundary conditions include: the speed of the vehicle in front is greater than the set speed threshold and the distance to the vehicle in front is greater than the set distance.

[0017] The advantages of the present invention are: reducing the setback phenomenon caused by monocular recognition errors, improving the stability and comfort of the follow-stop control. There are obvious improvements in the control of the follow-stop distance during the follow-stop process, driving comfort during the follow-stop process, and no abnormal second start after the follow-stop, which effectively improves the overall performance of the intelligent driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The following is a brief description of the contents expressed in the drawings of the present invention and the symbols in the drawings:

[0019] Figure 1 It is a schematic diagram of the principle of the follow-stop control method of the present invention. DETAILED DESCRIPTION

[0020] The specific implementation of the present invention will be further explained in detail below by describing the optimal embodiment with reference to the accompanying drawings.

[0021] The present invention is mainly to reduce the frustration and poor consistency in the existing monocular vision follow-stop control, improve the stability of the assisted driving function of the monocular vision follow-stop control and improve the user experience. The specific scheme is as follows:

[0022] In order to ensure the consistency of the following stop distance each time, a three-stage acceleration calculation will be performed during the following stop process.

[0023] The first stage adopts following motion control, and the MPC algorithm is used to predict the speed of the preceding vehicle, the speed of the own vehicle, and the distance of the preceding vehicle within a certain time domain. During the following and stopping process, the vehicle is controlled to reduce the speed of the own vehicle to the preset value (this value needs to be confirmed by calibration) in advance under the constraints of the speed of the preceding vehicle, the speed of the own vehicle, and the distance between the two vehicles. Specifically, it includes:

[0024] The first stage is when both the front vehicle and the vehicle itself are in motion, that is, the vehicle speeds are greater than 0. At this time, follow-up motion control is adopted, and the MPC algorithm is used to calculate the acceleration.

[0025] The computational constraints are

[0026] (1) The speed of the vehicle in front is the same as that of the vehicle in front;

[0027] (2) The distance between the vehicle in front and the vehicle in front meets the calibrated time interval; (i.e., the following distance, the following distance is obtained by looking up the table according to the lowest speed of the vehicle in front and the vehicle in front.)

[0028] The system input is, based on the distance, speed and acceleration of the front vehicle detected by the current camera, when it is assumed that the acceleration of the front vehicle will remain unchanged in the future, the distance and speed of the front vehicle in each time step (each step is 5s / n, n is a positive integer, which can be calibrated) in the next 5s (5s is the prediction time domain, this value can be calibrated). In addition, considering the response delay of the brake actuator, the speed of the vehicle hundreds of ms later is predicted based on the ACC target acceleration calculated hundreds of ms (the time value that can be calibrated based on the actual vehicle response). If the current step size is k, the ACC target acceleration of the vehicle with a k-step length can be calculated according to the predicted speed of the front vehicle, the speed of the vehicle, and the distance between the two vehicles according to the constraints. Then, based on the acceleration of the vehicle with a k-step length, the speed and distance of the vehicle with a k+1-step length are predicted, and the ACC target acceleration of the subsequent steps is calculated in sequence. During the following stop process, the vehicle can control the vehicle to slow down according to the calculated deceleration (calculated ACC target acceleration). If the vehicle in front does not stop, the vehicle can also follow and accelerate according to the calculated acceleration.

[0029] After the current vehicle stops braking (i.e., the conditions for entering the second stage: the speed of the vehicle in front is 0 and the speed of the vehicle is greater than a certain value), the vehicle will perform the second deceleration calculation; as with the first deceleration calculation, the actuator response delay is taken into account, and the ACC acceleration calculated hundreds of ms ago is used to predict the vehicle speed and the distance to the vehicle in front hundreds of ms in the future, and then the constant acceleration rate formula is used.

[0030]

[0031] S O is the predicted target distance, S3 is the initial required distance between the vehicle and the preceding vehicle in the third stage obtained by calibration, j is the acceleration change rate, V e is the predicted vehicle speed, a e is the target acceleration of the vehicle that needs to be solved; the target acceleration can be calculated according to the above formula.

[0032] The conditions for entering the third stage of the follow-stop control phase are that the speed of the front vehicle is 0 and the speed of the vehicle is less than the set speed threshold, and the distance to the front vehicle is less than a certain threshold. When this condition is met, it is necessary to enter the last stage of acceleration control, that is, terminal braking to stop. The controller will send a fixed deceleration at a fixed distance S3 and the speed of the vehicle V3 (the distance and speed will be given a certain deviation control range). This deceleration is a calibrated value, and at the same time, it will send a braking to stop flag signal to the lower-level brake controller (such as ESP) (ESP will cooperate with ACC to brake to stop at the terminal only after receiving the flag bit). At this point, the entire follow-stop process will be completed.

[0033] In order to reduce the braking jerking caused by the recognition error jump during the following stop process, boundary conditions will be set to filter the recognized movement information of the leading vehicle when the speed of the vehicle and the leading vehicle is low.

[0034] 1) When the identified distance to the vehicle ahead changes (the current data frame changes by a certain distance compared to the previous frame), and the vehicle speed does not change, a low-pass filter is used to filter the distance to the vehicle ahead for a certain period of time until the difference between the filtered data and the distance to the vehicle ahead detected by the camera is less than a certain value. After that, the real-time detected distance to the vehicle ahead by the camera is used to perform stop-and-follow control on the filtered data, thereby reducing data interference and improving the accuracy of the data used for stop-and-follow control.

[0035] 2) When the speed of the identified vehicle ahead changes (the current data frame changes to a certain speed compared to the previous frame), and the distance does not change, a low-pass filter is used to filter the speed of the vehicle ahead for a certain period of time until the difference between the filtered data and the speed of the vehicle ahead detected by the camera is less than a certain value;

[0036] 3) When the identified front vehicle speed and front vehicle distance both jump within a certain range, if the jumps are all decreasing, the reduced value is used as input, and the table is looked up to determine whether it is outside the safe area based on the speed of this vehicle, the speed of the front vehicle, and the relative distance. If it is outside the safe distance, low-pass filtering is performed. If it is not outside the safe distance, no processing is performed for safety reasons. If the jumps are all increasing, the final acceleration output is low-pass filtered.

[0037] After the following stop is completed, in order to ensure that the front vehicle moves slightly, or the front vehicle is actually prohibited, but the camera has a visual recognition error, resulting in the distance of the front vehicle being recognized to be farther, or the front vehicle is recognized to have a small speed, in order to avoid the phenomenon of secondary starting, the boundary conditions for the controller to start again need to be satisfied that the front vehicle speed is greater than a certain value, and the front vehicle distance needs to be greater than a certain value, which is usually about 2m of the following vehicle following stop distance. The above conditions are in a relationship of AND. That is, after the vehicle stops during the following stop process, the vehicle is controlled to start only when the starting conditions are met, otherwise it remains in the braked parking state; the starting conditions include that the front vehicle distance is greater than the set distance threshold and the front vehicle speed is greater than the set threshold.

[0038] Obviously, the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, they are all within the protection scope of the present invention.

Claims

1. A monocular vision-based follow-up control method, characterized in that: include, The following stop control process is divided into multiple stages, each stage corresponds to a following stop mode, the following stop mode of the vehicle is determined based on the running information of the preceding vehicle recognized by monocular vision, and the driving of the vehicle is controlled based on the determined following stop mode of the vehicle.

2. A monocular vision-based follow-up and stop control method as claimed in claim 1, characterized in that: The follow-stop mode includes at least one of a follow-motion mode, a follow-stop deceleration mode, and a braking deceleration mode.

3. A monocular vision-based follow-up stop control method as claimed in claim 2, characterized in that: In the medium and high speed stage and when the front vehicle is moving, the following motion mode is used to control the vehicle. In the following motion mode, the MPC algorithm is used to predict the speed of the front vehicle, the speed of the own vehicle, and the distance of the front vehicle within a certain time domain. Under the constraints of the speed of the front vehicle, the speed of the own vehicle, and the distance between the two vehicles, the vehicle speed is controlled to drop to the preset value.

4. A monocular vision-based follow-up and stop control method as claimed in claim 2, characterized in that: After the front vehicle stops, the vehicle is controlled in follow-stop deceleration mode, taking the response delay of the actuator into account, predicting the vehicle speed and the distance to the front vehicle within a certain time domain in the future, and calculating the target acceleration after correction based on the kinematic formula, and using this acceleration to control the vehicle's deceleration.

5. The monocular vision-based follow-up and stop control method according to claim 2, characterized in that: After the vehicle decelerates to a fixed distance and vehicle speed, it enters the braking deceleration mode. In the braking deceleration mode, a fixed deceleration rate is used to control the vehicle to decelerate to a stop.

6. A monocular vision-based follow-up and stop control method as claimed in claim 5, characterized in that: During the following stop process, when the speed of the vehicle and the speed of the preceding vehicle are lower than the set value, the movement information of the preceding vehicle is filtered.

7. A monocular vision-based follow-up and stop control method as claimed in claim 6, characterized in that: When the detected distance to the vehicle ahead changes but the vehicle speed does not change, a low-pass filter is used to filter the distance to the vehicle ahead for a certain period of time until the difference between the filtered data and the distance to the vehicle ahead detected by the camera is less than a certain value; 2) When the speed of the identified front vehicle changes but the distance does not change, a low-pass filter is used to filter the speed of the front vehicle for a certain period of time until the difference between the filtered data and the speed of the front vehicle detected by the camera is less than a certain value; 3) When the identified front vehicle speed and front vehicle distance both jump within a certain range, if the jumps are all decreasing, the reduced value is used as input, and the table is looked up to determine whether it is outside the safe area based on the speed of this vehicle, the speed of the front vehicle, and the relative distance. If it is outside the safe distance, low-pass filtering is performed. If it is not outside the safe distance, no processing is performed for safety reasons. If the jumps are all increasing, the final acceleration output is low-pass filtered.

8. A monocular vision-based follow-up and stop control method according to any one of claims 1 to 7, characterized in that: Set the starting boundary conditions of the follow-stop controller. The vehicle is allowed to start again only when the starting boundary conditions are met.

9. A monocular vision-based follow-up and stop control method as claimed in claim 8, characterized in that: The starting boundary conditions include: the speed of the vehicle in front is greater than the set speed threshold and the distance to the vehicle in front is greater than the set distance.

Images