Vehicle collision warning system operation and maintenance method fusing gray model and kalman filter

Abstract

The application discloses a vehicle collision warning system operation and maintenance method combining a grey model and Kalman filtering, and the system comprises a signal acquisition module, a vehicle state estimation module, a vehicle trajectory prediction module and a vehicle collision warning module.The vehicle state estimation module is a first vehicle state estimator, and the vehicle trajectory prediction module comprises a grey prediction model and a second vehicle state estimator; the first vehicle state estimator independently operates as the grey prediction model to prepare original data, and then synchronously operates with the vehicle trajectory prediction module; and the collision warning module considers the difference in driving styles of drivers and determines the reaction time according to three types of aggressive, ordinary and conservative when calculating the safe collision warning time.

Description

Technical Field

[0001] This invention belongs to the field of automotive active safety and relates to an operation and maintenance method for a vehicle collision warning system, and more particularly to an operation and maintenance method for a vehicle collision warning system based on information fusion. Background Technology

[0002] With the improvement of people's living standards and the acceleration of the pace of life, automobiles, as a convenient means of transportation, have become the first choice for people's travel. Road traffic accidents have also gradually become a serious threat to people's lives and property. Among them, two-vehicle and multi-vehicle collisions are among the most serious types of road traffic accidents. Some ramp merging points and intersections without traffic lights are extremely prone to vehicle collisions. Once a vehicle collision occurs, it often leads to serious casualties and significant economic losses. Therefore, issuing warnings before a car collision occurs is an important measure for vehicle safety and driver assistance to avoid vehicle collisions. Vehicle collision warning systems mainly have two warning modes: one is based on a safe collision distance model, referencing real-time vehicle speed and acceleration; the other is based on predicting the collision time according to the vehicle's trajectory and referring to a safe collision time model. Both methods require prediction of the vehicle's driving state and trajectory; therefore, accurately estimating the state and predicting the trajectory are key technologies for achieving collision warning.

[0003] To reduce the probability of traffic accidents and the damage they cause, researchers have conducted extensive studies on vehicle collision problems. These studies include automatic emergency braking control using safe time or safe distance algorithms, and collaborative collision warning based on multi-feature collision prediction. While these collision warning methods based on parametric or non-parametric vehicle motion models offer some guidance on vehicle collision problems, the accuracy of collision warnings heavily depends on the accuracy of the trajectory prediction model. If the actual operating conditions of the vehicle do not match the motion model—that is, when there is uncertainty in the operating conditions—the accuracy of collision prediction is difficult to guarantee.

[0004] In practical use, the existing vehicle collision warning system maintenance methods have the following shortcomings:

[0005] 1. Existing technologies do not fully consider the information fusion of grey model prediction and Kalman filtering, and the accuracy of vehicle collision warning needs to be improved.

[0006] 2. Existing technologies do not propose a collaborative method between vehicle state estimators and grey model predictors, and the accuracy of vehicle collision warning needs to be improved.

[0007] 3. Existing technologies do not consider driver style when calculating driver reaction time in the safety collision warning time, resulting in insufficient personalized adaptation energy of the vehicle collision warning system.

[0008] 4. Existing technologies set the number of raw data points for the grey model predictor to a constant, which requires a trade-off between prediction accuracy and computational load, and cannot achieve the optimal balance between the two.

[0009] 5. Existing technologies do not address the synchronous operation relationship between vehicle state estimators and grey model predictors from a collaborative perspective, which can easily lead to data conflicts or data overload. Summary of the Invention

[0010] To overcome the above problems, the present invention proposes a solution that addresses multiple problems simultaneously.

[0011] This paper provides an operation and maintenance method for a vehicle collision warning system that integrates grey model and Kalman filtering. The system includes a signal acquisition module, a vehicle state estimation module, a vehicle trajectory prediction module, and a vehicle collision warning module. The vehicle state estimation module is a first vehicle state estimator. The vehicle trajectory prediction module includes a grey model predictor and a second vehicle state estimator. The signal acquisition module detects the relative position of the target vehicle with respect to the vehicle itself and outputs it as a measurement variable to the first vehicle state estimator. The first vehicle state estimator estimates the relative motion state of the vehicle and uses it as the raw data for the grey model predictor.

[0012] The technical solution adopted by this invention to solve its technical problem is as follows: a gray model predictor performs n-step vehicle relative position prediction as input to a second vehicle state estimator; the second vehicle state estimator performs vehicle relative motion trajectory estimation as input to a vehicle collision warning module; the vehicle collision warning module issues a collision warning to the vehicle when there is a risk of collision within the safe collision warning time; when the gray model predictor is started to perform multi-step vehicle relative position prediction, the first vehicle state estimator first runs m steps independently to prepare raw data for the gray model predictor; the second vehicle state estimator and the gray model predictor work synchronously; the driver can choose one of three categories—aggressive, normal, and conservative—as the current driving style according to their own driving style; the warning system adaptively adjusts the number m of raw data in the gray model predictor by comparing the actual and theoretical information covariance matrices of the second vehicle state estimator.

[0013] Furthermore, the first vehicle state estimator is a first square root capsulated Kalman filter SCKF1, and the second vehicle state estimator is a second square root capsulated Kalman filter SCKF2; at the moment the target appears, let... The signal acquisition module collects measurement variables. SCKF1 is based on the measured variables Perform vehicle relative motion state estimation; when season And continue to collect measurement variables and estimate the relative motion state of the vehicles; when At that time, , ... As the raw data for the grey model predictor; the grey model predictor uses the GM(1,1) model to predict the measurement vector in n steps, obtaining , ... SCKF2 uses the predicted values ​​of the measurement vectors as inputs to recursively obtain the estimated values ​​of the vehicle's relative motion state. , ... .

[0014] Preferably, in the vehicle collision warning system operation and maintenance method, when the warning system is started, the first vehicle state estimator runs independently for m steps, and then works synchronously with the second vehicle state estimator and the gray model predictor.

[0015] Preferably, the vehicle collision warning system operation and maintenance method ensures that the control system fully considers the driver's driving style when calculating the driver's reaction time in the safe collision warning time. A shorter reaction time is used for aggressive driving styles; a medium reaction time is used for normal driving styles; and a longer reaction time is used for conservative driving styles.

[0016] Preferably, the vehicle collision warning system operation and maintenance method adopts the following strategy when adjusting the number m of original data in the gray model predictor:

[0017]

[0018]

[0019] in This is the initial value for the number of raw data points in the grey model predictor. Let m be the value at step k, where k is the discrete time step. and These are the actual and theoretical information covariance matrices, respectively.

[0020] Preferably, the vehicle collision warning system operation and maintenance method uses the following rules to determine the synchronous operation relationship between the vehicle state estimator and the gray model predictor:

[0021] when > At this time, the second vehicle state estimator and the grey model predictor stop working, waiting for the first vehicle state estimator to run independently. After taking a step, work synchronously; when ≤ At the same time, the first vehicle state estimator, the second vehicle state estimator, and the grey model predictor continue to operate synchronously.

[0022] The beneficial effects of this invention are:

[0023] 1. Regarding point 1 in the background technology, the present invention fully considers the shortcomings of collision warning based on vehicle motion models, and improves the warning accuracy by fusing information from gray model prediction and Kalman filtering.

[0024] 2. Regarding point 2 in the background technology, the present invention realizes the collaboration between the vehicle state estimator and the gray model predictor based on the number m of the original data of the gray model predictor.

[0025] 3. Regarding point 3 in the background art, the present invention takes driver style into account when calculating the driver's reaction time in the safety collision warning time, so that the vehicle collision warning system can adapt to different drivers.

[0026] 4. Regarding point 4 in the background technology, the present invention dynamically adjusts the number of original data m of the gray model predictor according to the relationship between the actual information covariance matrix and the theoretical information covariance matrix, thereby achieving the best prediction accuracy and computational load.

[0027] 5. In response to point 5 of the background technology, this invention proposes to realize the collaborative operation relationship between the first vehicle state estimator, the second vehicle state estimator and the gray model predictor, so as to ensure that there is no data conflict or excess.

[0028] Note: The above improvements are not listed in any particular order, and each one makes the present invention different from the prior art and represents a significant advancement. Attached Figure Description

[0029] Figure 1 This is a diagram showing the components of the vehicle collision warning system of the present invention;

[0030] Figure 2 This is a flowchart of the vehicle trajectory prediction process of the present invention;

[0031] Figure 3 This is a flowchart of the vehicle collision warning algorithm of the present invention. Detailed Implementation

[0032] The application system composition of the vehicle collision warning system operation and maintenance method integrating grey model prediction and Kalman filtering is as follows: Figure 1As shown, the early warning system includes a signal acquisition module 20, a vehicle state estimation module 30, a vehicle trajectory prediction module 40, and a vehicle collision warning module 50. The vehicle state estimation module 30 is a first vehicle state estimator 31. The vehicle trajectory prediction module 40 includes a gray model predictor 41 and a second vehicle state estimator 42. The signal acquisition module 20 detects the relative position of the target vehicle 10 relative to the vehicle 60 and outputs it as a measurement variable to the first vehicle state estimator 31. The first vehicle state estimator 31 performs relative motion state estimation of the vehicle and uses it as the raw data for the gray model predictor 41. The gray model predictor 41 performs n-step relative position prediction of the vehicle and uses it as the input for the second vehicle state estimator 42. The second vehicle state estimator 42 performs n-step relative position prediction of the vehicle and vehicle state estimator 50. The relative motion trajectory of the vehicle is estimated and used as input to the vehicle collision warning module 50. When there is a risk of collision within the safe collision warning time, the vehicle collision warning module 50 issues a collision warning to the vehicle 60. When the gray model predictor 41 is started to perform multi-step vehicle relative position prediction, the first vehicle state estimator 31 first runs independently for m steps to prepare raw data for the gray model predictor 41. The second vehicle state estimator 42 and the gray model predictor 41 work synchronously. The driver can choose one of three categories as the current style: aggressive, normal, and conservative, according to his driving style. The warning system adaptively adjusts the number of raw data m of the gray model predictor 41 by comparing the actual and theoretical information covariance matrix of the second vehicle state estimator 42.

[0033] The flowchart for predicting vehicle trajectory is as follows: Figure 2 As shown, the vehicle collision warning algorithm is as follows: Figure 3 As shown, the operation and maintenance method of the vehicle collision warning system that integrates grey model prediction and Kalman filtering is described as follows:

[0034] When the early warning system is activated, the grey model predictor performs multi-step vehicle relative position prediction, while the first vehicle state estimator runs independently for m steps to prepare raw data for the grey model predictor. After the early warning system is activated, the first vehicle state estimator works synchronously with the second vehicle state estimator and the grey model predictor.

[0035] like Figure 2 As shown, the first vehicle state estimator is a first square root capsular Kalman filter SCKF1, and the second vehicle state estimator is a second square root capsular Kalman filter SCKF2; at the moment the target appears, let... The signal acquisition module collects measurement variables. SCKF1 is based on the measured variables Perform vehicle relative motion state estimation; when season And continue to collect measurement variables and estimate the relative motion state of the vehicles; when At that time, , ... As the raw data for the grey model predictor; the grey model predictor uses the GM(1,1) model to predict the measurement vector in n steps, obtaining , ... SCKF2 uses the predicted values ​​of the measurement vectors as inputs to recursively obtain the estimated values ​​of the vehicle's relative motion state. , ... .

[0036] Furthermore, the grey model predictor processes the raw data and builds a grey model to find the development patterns of the system, thereby predicting the future state of vehicle motion. In the collision warning system, the grey model predictor performs n-step vehicle relative position predictions, which serve as input to the second vehicle state estimator.

[0037] Furthermore, when calculating the driver's reaction time during the collision warning period, the control system fully considers the driver's driving style. A shorter reaction time is used for aggressive driving styles; a medium reaction time is used for moderate driving styles; and a longer reaction time is used for conservative driving styles.

[0038] Furthermore, the second vehicle state estimator estimates the relative motion trajectory of the vehicle, which serves as input to the vehicle collision warning module. To improve the accuracy of vehicle state estimation, the warning system adaptively adjusts the number of raw data points, *m*, of the grey model predictor by comparing the actual and theoretical information covariance matrices of the second vehicle state estimator. The strategy for adjusting the number of raw data points, *m*, of the grey model predictor is as follows:

[0039] (1)

[0040] (2)

[0041] Where m0 is the initial number of data points for the grey model predictor, m k Let m be the value at step k, where k is the discrete time step. and C k These are the actual and theoretical information covariance matrices, respectively.

[0042] Furthermore, the following rules are established to determine the synchronous operation relationship between the vehicle state estimator and the grey model predictor:

[0043] when > At this time, the second vehicle state estimator and the grey model predictor stop working, waiting for the first vehicle state estimator to run independently. After taking a step, work synchronously; when ≤ At the same time, the first vehicle state estimator, the second vehicle state estimator, and the grey model predictor continue to operate synchronously.

[0044] Furthermore, after obtaining the predicted relative motion trajectory and motion state estimation results of the target vehicle relative to the vehicle itself, a collision warning algorithm is used to calculate whether there is a risk of collision. If there is a collision risk within the safe collision warning time, the vehicle collision warning module issues a collision warning to the vehicle.

[0045] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. For those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered to fall within the scope of protection of the present invention.

Claims

1. A vehicle collision warning system operation and maintenance method integrating grey model and Kalman filter. The applied warning system includes a signal acquisition module (20), a vehicle state estimation module (30), a vehicle trajectory prediction module (40), and a vehicle collision warning module (50). The vehicle state estimation module (30) is a first vehicle state estimator (31). The vehicle trajectory prediction module (40) includes a grey model predictor (41) and a second vehicle state estimator (42). The signal acquisition module (20) detects the relative position of the target vehicle (10) relative to the vehicle (60) and outputs it as a measurement variable to the first vehicle state estimator (31). The first vehicle state estimator (31) performs relative motion state estimation of the vehicle and uses it as the raw data of the grey model predictor (41). The grey model predictor (41) performs n-step relative position prediction of the vehicle and uses it as the input of the second vehicle state estimator (42). The second vehicle state estimator (42) estimates the relative motion trajectory of the vehicle and uses it as input to the vehicle collision warning module (50); the vehicle collision warning module (50) issues a collision warning to the vehicle (60) when there is a risk of collision within the safe collision warning time, characterized in that: The first vehicle state estimator (31) is the first square root capacitive Kalman filter SCKF1, and the second vehicle state estimator (42) is the second square root capacitive Kalman filter SCKF2. When the gray model predictor (41) is started to perform multi-step vehicle relative position prediction, the first vehicle state estimator (31) first runs m steps independently to prepare raw data for the gray model predictor (41). When the target appears, order The signal acquisition module (20) acquires the measured variables. The first square root capacitive Kalman filter SCKF1 is based on the measured variables. Perform vehicle relative motion state estimation; when season And continue to collect measurement variables and estimate the relative motion state of the vehicles; when At that time, , ... As the raw data for the grey model predictor (41); The grey model predictor (41) uses the GM(1,1) model based on , ... Predict the n-step measurement vector prediction value , ... The second square root capacitive Kalman filter (SCKF2) uses the n-step measurement vector prediction values ​​as inputs to recursively obtain the estimated value of the vehicle's relative motion state. , ... ; When the early warning system is activated, the first vehicle state estimator (31) runs independently for m steps, and then works synchronously with the second vehicle state estimator (42) and the gray model predictor (41). The second vehicle state estimator (42) and the grey model predictor (41) work synchronously; Drivers can choose from three categories—aggressive, normal, and conservative—as their current driving style, based on their own driving style. The driver's reaction time during a collision warning is affected by driving style; an aggressive driving style corresponds to a shorter reaction time; a moderate driving style corresponds to a medium reaction time; and a conservative driving style corresponds to a longer reaction time. The early warning system adaptively adjusts the number of original data m of the grey model predictor (41) by comparing the actual and theoretical information covariance matrices of the second vehicle state estimator (42); The strategy for adjusting the number of original data points m in the grey model predictor (41) is as follows: in The initial value for the number of original data points for the grey model predictor (41) is given. Let m be the value at step k, where k is the discrete time step. and These are the actual and theoretical information covariance matrices, respectively. when > At this time, the second vehicle state estimator (42) and the gray model predictor (41) stop working, waiting for the first vehicle state estimator (31) to run independently. After taking a step, work synchronously; when ≤ At the same time, the first vehicle state estimator (31), the second vehicle state estimator (42), and the gray model predictor (41) continue to work synchronously.

Images