An anti-collision warning method based on fuzzy logic theory

Abstract

This invention discloses a collision avoidance warning method based on fuzzy logic. First, the uncertainty of the driver's lateral dynamics is modeled using fuzzy logic to obtain the lateral dynamics membership coefficient. Based on this, the membership coefficient is used to weight the one-dimensional pre-collision time to collision to obtain a two-dimensional road safety indicator, FL-iTTC. The proposed measurement is then used as the basis for alarms, and the collision avoidance system alarm threshold is calibrated using natural driving data. The proposed method can accurately identify potential collision events in complex traffic scenarios, especially multi-lane deceleration and lane change scenarios, and has significant positive implications for road traffic safety and the design of collision avoidance warning systems.

Description

Technical Field

[0001] The present invention relates to the field of vehicle active safety, and in particular to an anti-collision warning method based on fuzzy logic theory. Background Art

[0002] Vehicle collision avoidance systems, such as forward collision warning (FCW) systems, have been widely used. These active safety warning systems can effectively reduce traffic accidents and improve road safety. However, most current collision warning systems have significant limitations in their operating range. On the one hand, existing collision warning systems can only identify potential collision events within the same lane and cannot identify potential collision risks posed by vehicles in adjacent lanes. On the other hand, current collision warning algorithms focus only on the one-dimensional dynamics of the vehicle, ignoring the interactions between vehicles in the two-dimensional plane. Therefore, it is urgent to develop a collision warning method that can reflect the two-dimensional dynamic characteristics and interactions of vehicles. This method needs to accurately identify the complex and diverse traffic risks in multi-lane scenarios in real time. In addition, this warning method should be close to the driver's subjective risk perception. Summary of the Invention

[0003] The present invention provides a collision avoidance warning method based on fuzzy logic theory, aiming to solve the problem that the current vehicle collision avoidance warning system only considers potential collisions within the same lane, fails to consider the two-dimensional interaction of vehicles, and has a large number of false positive alarms in multi-lane scenarios (especially lane change scenarios).

[0004] The present invention adopts the following technical solutions to solve the above technical problems:

[0005] A fuzzy logic-based anti-collision warning method comprises the following steps:

[0006] S1. Collect natural driving data of the target vehicle in the target area, summarize the uncertainty of the target vehicle's lateral dynamics based on the natural driving data, and obtain the lateral dynamics membership coefficient γ in the vehicle driving safety modeling based on fuzzy logic theory. s,n ;

[0007] S2, the lateral dynamics membership coefficient γ obtained in step S1 s,n Combined with the longitudinal pre-collision time in the vehicle driving scenario, the driving safety measure FL-iTTC is obtained through weighted calculation;

[0008] S3. Based on the driving safety measure FL-iTTC obtained in step S2, an alarm model of the anti-collision warning system is designed, and alarm threshold parameters are set according to the driver's actual braking behavior;

[0009] S4. Combine the driving safety measure FL-iTTC in step S2 with the warning threshold parameter in step S3 to issue a collision warning to the driver according to different scenarios.

[0010] Furthermore, in step S1, the vehicle lateral dynamics membership coefficient γ s,n The steps to solve are as follows:

[0011] Step 1: Summarize the distribution of displacement uncertainty. Displacement uncertainty is defined as the difference between the vehicle's expected lateral position (calculated from its current lateral position and velocity) and its actual position after a given time step. It is calculated as follows:

[0012]

[0013] Where ΔX n represents the uncertainty of the vehicle's lateral displacement, X n,t represents the lateral position of the vehicle at time t, X n,t+Δt represents the lateral position coordinate of the vehicle after Δt time, represents the lateral velocity of the vehicle at time t.

[0014] According to empirical research on actual natural driving data, the uncertainty of displacement follows a Laplace distribution, and the scale parameter has a linear relationship with the time step Δt. Its default probability distribution is as follows:

[0015]

[0016] Step 2: Use a probabilistic approach to find the probability of a vehicle and its neighbors overlapping in the horizontal direction after a given time step. The solution is as follows:

[0017]

[0018] Among them, t represents the current time, Δt represents the prediction time step, and W s Indicates the width of the vehicle, W n represents the width of the neighboring car, β(Δx) is a function related to the probability distribution of Δx,

[0019] Step 3: Horizontal overlap probability Transformed into the lateral dynamics membership coefficient, the mapping relationship is as follows:

[0020]

[0021] Where k is the normalized correction coefficient, which is calculated as follows:

[0022]

[0023] The lateral dynamics membership coefficient γ described in step S1 is obtained s,n .

[0024] Furthermore, in step S2, the driving safety measure FL-iTTC is solved as follows:

[0025] Step 1: Ignoring lateral dynamics, calculate the longitudinal pre-collision time and use it as the prediction time step Δt. The calculation method is as follows:

[0026]

[0027] Where Y n Indicates the longitudinal coordinate of the neighboring vehicle, Y s Indicates the longitudinal coordinate of the vehicle, L n Indicates the length of the adjacent vehicle, L s Indicates the length of the vehicle. represents the longitudinal speed of the neighboring vehicle, Indicates the longitudinal speed of the vehicle.

[0028] Step 2: Calculate the expected lateral displacement Δx when the vehicle and the neighboring vehicle overlap geometrically.

[0029]

[0030] Where Δt is the prediction time step obtained in step 1, X n Indicates the lateral coordinate of the neighboring car, X s represents the lateral coordinate of the vehicle, represents the lateral speed of the neighboring vehicle, Indicates the lateral speed of the vehicle.

[0031] Step 3: Use the membership coefficient obtained in S1 as the weight to correct the inverse of the longitudinal pre-collision time to obtain the driving safety measure FL-iTTC, that is:

[0032]

[0033] Where, γ s,n is the fuzzy logic membership coefficient obtained in step S1, and Δt is the prediction time step.

[0034] Furthermore, in step S3, the alarm model and the alarm threshold parameters of the anti-collision warning system are designed as follows:

[0035] The alarm threshold parameters are calibrated based on the driver's actual driving behavior, so that the alarm mode is close to the driver's risk perception. The following events are defined:

[0036] True positive TP: Both the driver and the warning system consider the event to be risky;

[0037] True negative TN: Both the driver and the warning system believe that the event is not risky;

[0038] False positive FP: The driver believes that the event is not risky, but the warning system believes that the event is risky;

[0039] False negative FN: The driver believes that the event is risky, but the warning system believes that the event is not risky;

[0040] The warning threshold parameter maximizes the difference between the true positive rate and the false positive rate, that is:

[0041] FL-iTTC * =argmax(TPR(FL-iTTC * )-FPR(FL-iTTC * ));

[0042] TPR and FPR represent the true positive rate and false positive rate, respectively, and are calculated as follows:

[0043]

[0044] Furthermore, in step S4, the manner of issuing anti-collision warnings to the driver according to different scenarios is as follows:

[0045] When the vehicle interacts with surrounding vehicles, if the FL-iTTC exceeds the safety threshold, the system will issue an alarm signal. If the threshold is not exceeded, no alarm signal will be issued. That is:

[0046]

[0047] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0048] (1) Based on fuzzy logic theory and taking into account the two-dimensional interaction of vehicles, this method can identify potential collisions caused by adjacent vehicles in adjacent lanes. This expands the operating range of the anti-collision warning system. Compared with existing methods, the recognition accuracy and versatility are significantly improved.

[0049] (2) By introducing the uncertainty of displacement, the present invention extends the original safety measurement from precise variables or Boolean variables to probabilistic methods, significantly improving the flexibility and fault tolerance of the anti-collision warning system in complex traffic scenarios, and can accurately identify more complex risk events (such as unreasonable lane changes by the preceding vehicle) in real time.

[0050] (3) The present invention uses the driver's actual braking behavior to calibrate the warning threshold. This method is close to the driver's actual risk perception and can significantly improve the driver's acceptance of the anti-collision warning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0053] In one embodiment, a collision avoidance warning method based on fuzzy logic theory is provided. Figure 1 As shown, the content includes solving the lateral dynamics membership coefficient, calculating the two-dimensional safety measure FL-iTTC, designing the early warning system, and solving the alarm threshold:

[0054] S1. Solve the lateral dynamics membership coefficient

[0055] The uncertainty of vehicle lateral dynamics is summarized, and the lateral dynamics membership coefficient γ in vehicle driving safety modeling is obtained based on fuzzy logic theory. s,n .

[0056] S2. Calculate the two-dimensional security measure FL-iTTC

[0057] Based on the lateral dynamics membership coefficient obtained in step S1 and combined with the longitudinal pre-collision time in the vehicle driving scenario, a weighted safety measure, FL-iTTC, is derived to reflect the safety level of the vehicle in two-dimensional interaction. The larger the FL-iTTC, the greater the driving risk level in that scenario.

[0058] S3. Solve the alarm threshold

[0059] The alarm threshold parameters are calibrated based on the driver's actual braking behavior. Based on the driver's perception of the safety level in some driving scenarios, a confusion matrix is developed to classify driving scenarios into four types: true positive, false positive, true negative, and false negative. The optimal alarm threshold is set by maximizing the difference between the true positive rate and the false positive rate (a default alarm threshold is 0.15s -1 ).

[0060] S4. Design early warning system and issue alarms

[0061] The two-dimensional safety measure obtained in step S2 is used to assess the driving safety level. A larger FL-iTTC indicates a greater collision risk, and a smaller FL-iTTC indicates greater safety. When FL-iTTC exceeds a given threshold, the anti-collision system issues a collision warning.

[0062] Furthermore, in step S1, the steps for solving the vehicle lateral dynamics membership coefficient are as follows:

[0063] Step 1: Determine the probability distribution of displacement uncertainty using natural driving data. Displacement uncertainty is defined as the difference between the vehicle's predicted lateral position (calculated from its current lateral position and velocity) and its actual position after a given time step. The calculation is as follows:

[0064]

[0065] Where ΔX n represents the uncertainty of the vehicle's lateral displacement, X n,t represents the lateral position of the vehicle at time t, X n,t+Δt represents the lateral position coordinate of the vehicle after Δt time, represents the lateral velocity of the vehicle at time t.

[0066] Based on empirical research on actual natural driving data, the uncertainty of displacement follows a Laplace distribution, and the scale parameter has a linear relationship with the time step Δt. The recommended default probability distribution is as follows:

[0067]

[0068] In actual use, the distribution can be adjusted according to actual needs. The general form of the distribution is:

[0069]

[0070] The specific calibration method of the parameters is:

[0071]

[0072] Step 2: Use a probabilistic approach to find the probability of a vehicle and its neighbors overlapping in the horizontal direction after a given time step. The solution is as follows:

[0073]

[0074] In this formula, t represents the current time, Δt represents the prediction time step, and W s Indicates the width of the vehicle, W n represents the width of the neighboring car, β(Δx) is a function related to the probability distribution of Δx,

[0075] Step 3: Horizontal overlap probability Transformed into the lateral dynamics membership coefficient, the mapping relationship is as follows:

[0076]

[0077] In this formula, k is the normalized correction coefficient, which is calculated as follows:

[0078]

[0079] In this way, the lateral dynamics membership coefficient described in step S1 is obtained. In practical applications, this step can be omitted and the given recommended parameter values and probability distribution can be used directly. The method for solving the distribution parameters is also given here.

[0080] Furthermore, in step S2, the two-dimensional security measure FL-iTTC is solved as follows:

[0081] Step 1: Ignoring lateral dynamics, calculate the longitudinal pre-collision time and use it as the prediction time step Δt. The calculation method is as follows:

[0082]

[0083] Where Y n Indicates the longitudinal coordinate of the neighboring vehicle, Y s Indicates the longitudinal coordinate of the vehicle, L n Indicates the length of the adjacent vehicle, L s Indicates the length of the vehicle. represents the longitudinal speed of the neighboring vehicle, Indicates the longitudinal speed of the vehicle.

[0084] Step 2: Calculate the expected lateral displacement Δx when the vehicle and the neighboring vehicle overlap geometrically.

[0085]

[0086] Where Δt is the prediction time step obtained in step 1, X n Indicates the lateral coordinate of the neighboring car, X s represents the lateral coordinate of the vehicle, represents the lateral speed of the neighboring vehicle, Indicates the lateral speed of the vehicle.

[0087] Step 3: Use the membership coefficient obtained in S1 as the weight to correct the inverse of the longitudinal pre-crash time to obtain the two-dimensional safety measure FL-iTTC, that is:

[0088]

[0089] Where, γ s,n is the fuzzy logic membership coefficient obtained in step S1, and Δt is the prediction time step.

[0090] Furthermore, in step S3, the alarm threshold is calculated as follows:

[0091] The alarm threshold is updated based on the driver's actual driving behavior (braking behavior) so that the alarm mode can be close to the driver's risk perception. In this step, the following events are defined:

[0092] True Positive (TP): Both the driver and the warning system consider the event to be risky.

[0093] True Negative (TN): Both the driver and the warning system believe that the event does not pose a risk.

[0094] False Positive (FP): The driver believes that the event is not risky, but the warning system believes that the event is risky.

[0095] False negative (FN): The driver believes that the event is risky, but the warning system believes that the event is not risky.

[0096] A reasonable warning threshold should maximize the difference between the true positive rate and the false positive rate, that is:

[0097] FL-iTTC * =argmax(TPR(FL-iTTC * )-FPR(FL-iTTC * ));

[0098] TPR and FPR represent the true positive rate and false positive rate, respectively, and are calculated as follows:

[0099]

[0100] In this formula, TP, FN, TN, and FP refer to the total number of TP, FN, TN, and FP events, respectively, which are calibrated according to driving data.

[0101] Furthermore, in step S4, the alarm issuing method is as follows:

[0102] The alarm system is designed based on FL-iTTC. When the FL-iTTC exceeds the safety threshold during the interaction between the vehicle and the surrounding vehicles, the system will issue an alarm signal. If it does not exceed the threshold, no alarm signal will be issued. That is:

[0103]

[0104] In this example, the alarm signal is a unit step signal θ(t). When θ(t) is 1, the system issues an alarm.

[0105] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or replacements within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A fuzzy logic-based anti-collision warning method, characterized in that: The method comprises the following steps: S1. Collect natural driving data of the target vehicle in the target area, summarize the uncertainty of the target vehicle's lateral dynamics based on the natural driving data, and obtain the lateral dynamics membership coefficient γ in the vehicle driving safety modeling based on fuzzy logic theory. s,n ; In step S1, the vehicle lateral dynamics membership coefficient γ s,n The steps to solve are as follows: Step 1: Summarize the distribution of displacement uncertainty. Displacement uncertainty is defined as the difference between the vehicle's predicted lateral position and its actual position after a given time step. It is calculated as follows: Where ΔX n represents the uncertainty of the vehicle's lateral displacement, X n,t represents the lateral position of the vehicle at time t, X n,t+Δt represents the lateral position coordinate of the vehicle after Δt time, represents the lateral velocity of the vehicle at time t; According to empirical research on actual natural driving data, the uncertainty of displacement follows a Laplace distribution, and the scale parameter has a linear relationship with the time step Δt. Its default probability distribution is as follows: Step 2: Use a probabilistic approach to find the probability of a vehicle and its neighbors overlapping in the horizontal direction after a given time step. The solution is as follows: Among them, t represents the current time, Δt represents the prediction time step, and W s Indicates the width of the vehicle, W n represents the width of the neighboring car, β(Δx) is a function related to the probability distribution of Δx, Step 3: Horizontal overlap probability Transformed into the lateral dynamics membership coefficient, the mapping relationship is as follows: Where k is the normalized correction coefficient, which is calculated as follows: The lateral dynamics membership coefficient γ described in step S1 is obtained s,n ; S2, the lateral dynamics membership coefficient γ obtained in step S1 s,n Combined with the longitudinal pre-collision time in the vehicle driving scenario, the driving safety measure FL-iTTC is obtained through weighted calculation; The solution to the driving safety measure FL-iTTC is as follows: Step 1: Ignoring lateral dynamics, calculate the longitudinal pre-collision time and use it as the prediction time step Δt. The calculation method is as follows: Where Y n Indicates the longitudinal coordinate of the neighboring vehicle, Y s Indicates the longitudinal coordinate of the vehicle, L n Indicates the length of the adjacent vehicle, L s Indicates the length of the vehicle. represents the longitudinal speed of the neighboring vehicle, Indicates the longitudinal speed of the vehicle; Step 2: Calculate the expected lateral displacement Δx when the vehicle and the neighboring vehicle overlap geometrically. Where Δt is the prediction time step obtained in step 1, X n Indicates the lateral coordinate of the neighboring car, X s represents the lateral coordinate of the vehicle, represents the lateral speed of the neighboring vehicle, Indicates the lateral speed of the vehicle; Step 3: Use the membership coefficient obtained in S1 as the weight to correct the inverse of the longitudinal pre-collision time to obtain the driving safety measure FL-iTTC, that is: Where, γ s,n is the fuzzy logic membership coefficient obtained in step S1, Δt is the prediction time step; S3. Based on the driving safety measure FL-iTTC obtained in step S2, an alarm model of the anti-collision warning system is designed, and alarm threshold parameters are set according to the driver's actual braking behavior; S4. Combine the driving safety measure FL-iTTC in step S2 with the warning threshold parameter in step S3 to issue a collision warning to the driver according to different scenarios.

2. The anti-collision warning method based on fuzzy logic according to claim 1, characterized in that: In step S3, the alarm model and alarm threshold parameters of the anti-collision warning system are designed as follows: The alarm threshold parameters are calibrated based on the driver's actual driving behavior, so that the alarm mode is close to the driver's risk perception. The following events are defined: True positive TP: Both the driver and the warning system consider the event to be risky; True negative TN: Both the driver and the warning system believe that the event is not risky; False positive FP: The driver believes that the event is not risky, but the warning system believes that the event is risky; False negative FN: The driver believes that the event is risky, but the warning system believes that the event is not risky; The warning threshold parameter maximizes the difference between the true positive rate and the false positive rate, that is: FL-iTTC * =argmax(TPR(FL-iTTC * )-FPR(FL-iTTC * )); TPR and FPR represent the true positive rate and false positive rate, respectively, and are calculated as follows:

3. The anti-collision warning method based on fuzzy logic according to claim 2, characterized in that: In step S4, the method of issuing anti-collision warning to the driver according to different scenarios is as follows: When the vehicle interacts with surrounding vehicles, if the FL-iTTC exceeds the safety threshold, the system will issue an alarm signal. If the threshold is not exceeded, no alarm signal will be issued. That is:

Images