A forward collision warning method and device considering the response behavior of truck drivers

Abstract

The present invention relates to a forward collision warning method and device considering the response behavior of truck drivers. The method includes: extracting vehicle collision avoidance event data in the following - vehicle scenario of truck drivers; determining the driver response behavior characteristics and classifying the response behaviors in the vehicle collision avoidance events; dividing the response process into a reaction stage and a braking stage according to the driver response behavior characteristics; predicting the driver's reaction distance based on the support vector regression method and constructing a driver reaction distance prediction model based on the driver response behavior characteristics; constructing a braking distance prediction model for each category based on the long - short - term memory model and respectively predicting the braking distance; determining the safety warning distance for forward collision based on the reaction distance and the braking distance, and sending a collision avoidance warning message to the driver according to the safety warning distance. Compared with the prior art, the present invention has the advantages of high safety, low false alarm rate, and being able to better fit the behavior characteristics of truck drivers, etc.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle active safety, and particularly to a forward collision warning method and device considering the response behavior of truck drivers. Background Art

[0002] To improve the driving safety of freight vehicles, operating enterprises have installed a large number of vehicle active safety devices to remind drivers of the risks existing in road driving. The Forward Collision Warning (FCW) system is an important vehicle active safety function, and its installation rate has exceeded 73%.

[0003] Research on the application evaluation of the FCW system shows that the false alarm rate of the current FCW systems on the market exceeds 18%, and false alarms cause 12% of drivers to distrust the system and 11% of drivers to stop using the system. Research on the evaluation of the FCW system by NHTSA shows that the trust level of commercial truck drivers in the FCW system is only 69.52%. In 2016, the University of Michigan and General Motors conducted a large-scale evaluation of the use of the FCW system in 48 states of the United States. The evaluation results show that there are obvious differences in the driver response behavior characteristics during the use of the FCW. This is mainly reflected in: 1) There are differences in the reaction time of drivers in different collision avoidance scenarios; 2) As the vehicle speed and distance change, the braking force and braking duration of drivers change accordingly.

[0004] The current safety distance algorithm is the core algorithm of the FCW system, and the reaction time and braking characteristics of drivers are the key factors affecting the safety distance. Existing technologies for the algorithm development of the FCW lack the factor of driver response behavior characteristics. Therefore, it is crucial to conduct research on driver response behavior characteristics relying on the natural driving data obtained during the application of the FCW system. Summary of the Invention

[0005] The purpose of the present invention is to provide a forward collision warning method and device considering the response behavior of truck drivers, so as to improve the safety of the algorithm and reduce the false alarm rate.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A forward collision warning method considering the response behavior of truck drivers includes the following steps:

[0008] Step 1: Extract the warning video data and vehicle motion data of vehicle collision avoidance events in the following - vehicle scenario of truck drivers from the natural driving data;

[0009] Step 2: Determine the driver response behavior characteristics according to the warning video data and vehicle motion data, and classify the response behaviors in the vehicle collision avoidance events;

[0010] Step 3: According to the driver response behavior characteristics, divide the response process into a reaction stage and a braking stage;

[0011] Step 4: Based on the support vector regression method, construct a driver reaction distance prediction model based on the driver response behavior characteristics to predict the driver reaction distance;

[0012] Step 5: For the classification of driver response behaviors divided in Step 2, construct a braking distance prediction model for each category based on the long short-term memory model, and predict the braking distance for each response behavior respectively;

[0013] Step 6: Determine the safety warning distance for forward collision based on the reaction distance and the braking distance, and send a collision avoidance warning message to the driver according to the safety warning distance.

[0014] The warning video data is the FCW warning video, and the extracted warning video data includes the video 10s before the warning and the video 10s after the warning.

[0015] The vehicle motion data includes the speed of the truck itself, the speed of the leading vehicle, the relative distance, the acceleration of the truck itself, and the braking moment. Among them, the relative distance represents the distance between the truck itself and the leading vehicle.

[0016] The extraction criteria for vehicle collision avoidance events in Step 1 are that the following two conditions are satisfied simultaneously:

[0017] Condition 1: Relative distance < 70m;

[0018] Condition 2: TTC < 4.5s or THW < 1.5s, where TTC represents the time to collision and THW represents the time headway.

[0019] The response behavior characteristics include:

[0020] a) Speed-related variables: the speed of the truck itself, the relative speed before the warning, the relative speed after the warning, the relative speed change value;

[0021] b) Distance-related variables: the relative distance at the warning moment, the relative distance at the braking moment;

[0022] c) Braking-related variables: the braking deceleration before the warning, the braking deceleration after the warning, the deceleration change amount, the braking duration, the minimum deceleration;

[0023] d) Reaction-related variables: reaction time.

[0024] The k-means cluster method is used to classify the response behaviors in vehicle collision avoidance events, and the truck driving response behaviors are divided into three categories: response before warning, response after warning, and no response.

[0025] The inputs of the driver reaction distance prediction model are the initial relative distance, reaction time, and relative vehicle speed, and the output is the reaction distance.

[0026] The inputs of the braking distance prediction model are the initial relative distance, braking time, braking force, the vehicle speed of the truck itself, and the vehicle speed of the leading vehicle, and the output is the braking distance.

[0027] The safety warning distance is:

[0028] d w = d r + d b

[0029] Wherein, d w is the safety warning distance, d r is the reaction distance during the driver's collision avoidance process, and d b is the braking distance during the driver's collision avoidance process.

[0030] A forward collision warning device considering the response behavior of truck drivers includes a memory, a processor, and a program stored in the memory. When the processor executes the program, the above-mentioned method is implemented.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) High safety: The present invention considers the response characteristics of truck drivers to the system warning information, and determines the safety warning distance in combination with the reaction time and braking characteristics, ensuring the safety in the driver's collision avoidance scenario.

[0033] (2) Low false alarm rate: The present invention can overcome the differences caused by different driver reaction times and braking measures in the scenarios of following closely at low speeds and following at a long distance at high speeds, avoiding the false alarms generated by traditional safety distance models or fixed threshold algorithms, and effectively reducing the false alarm rate.

[0034] (3) Strong adaptability: The present invention comprehensively considers factors such as reaction time, braking force, and braking duration during the use of the FCW system by drivers, and can adapt to the driver response behavior characteristics in different emergency collision avoidance scenarios. Description of the Drawings

[0035] Figure 1 is the method flow chart of the present invention;

[0036] Figure 2 is the type of response behavior obtained in the embodiment of the present invention;

[0037] Figure 3 is the result diagram of the reaction distance prediction model in the embodiment of the present invention;

[0038] Figure 4 This is the result diagram of the braking distance prediction model in the embodiments of the present invention. Among them, (4a) represents the prediction result of the pre-warning response behavior scenario, (4b) represents the prediction result of the post-warning response behavior scenario, and (4c) represents the prediction result of the non-response behavior scenario;

[0039] Figure 5 This is the adaptability result display diagram in the embodiments of the present invention. Among them, (5a) represents the prediction result of the pre-warning response behavior scenario, (5b) represents the prediction result of the post-warning response behavior scenario, and (5c) represents the prediction result of the non-response behavior scenario. Specific Embodiments

[0040] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention, and gives the detailed implementation manner and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0041] This embodiment provides a forward collision warning method considering the response behavior of truck drivers, as Figure 1 shown, including the following steps:

[0042] Step 1: Extract the warning video data and vehicle motion data of vehicle collision avoidance events in the truck driver's following scenario from the natural driving data.

[0043] The warning video data is the FCW warning video. The extracted warning video data includes the 10s video before warning and the 10s video after warning, totaling 20s of FCW warning video.

[0044] The vehicle motion data includes the speed of the truck itself, the speed of the leading vehicle, the relative distance (the distance between the truck itself and the leading vehicle), the acceleration of the truck itself, and the braking moment.

[0045] The extraction criteria for vehicle collision avoidance events are that the following two conditions are met simultaneously:

[0046] Condition 1: The relative distance < 70m;

[0047] Condition 2: TTC < 4.5s or THW < 1.5s, where TTC represents the time to collision and THW represents the time headway.

[0048] In order to avoid inaccurate data, steps such as manual observation of videos, data cleaning, and outlier removal are also set in this embodiment.

[0049] Step 2: Determine the driver response behavior characteristics according to the warning video data and vehicle motion data, and classify the response behaviors in the vehicle collision avoidance events.

[0050] In this embodiment, the response behavior characteristics are divided into four categories:

[0051] a) Vehicle speed variables: the speed of the truck itself, the relative speed before warning (10 s before warning), the relative speed after warning (4 s after warning), and the relative speed change value;

[0052] b) Distance variables: the relative distance at the warning moment and the relative distance at the braking moment;

[0053] c) Braking variables: the braking deceleration before warning (10 s before warning), the braking deceleration after warning (4 s after warning), the deceleration change amount, the braking duration, and the minimum deceleration;

[0054] d) Reaction variables: the reaction time.

[0055] The k-means cluster method is used to classify the response behaviors in the vehicle collision avoidance event. The response behaviors are divided in MATLAB. According to the distance accuracy index of the k-means cluster center point, the response behavior types are determined. The truck driving response behaviors are divided into three categories: pre-warning response, post-warning response, and non-response. The k-means cluster method specifically belongs to the conventional settings in this field. To avoid obscuring the purpose of this application, it will not be elaborated here. The response behavior characteristics obtained in this embodiment are as Figure 2 shown. The pre-warning response behavior occurs in the low-speed and short-distance following scenario. The post-warning response behavior scenario is accompanied by a large braking force. The non-response scenario occurs in the high-speed and long-distance following scenario.

[0056] Step 3: According to the driver response behavior characteristics, the response process is divided into a reaction stage and a braking stage.

[0057] Step 4: Based on the support vector regression method, for the time period from the moment when the driver receives the warning information to the braking moment, a driver reaction distance prediction model is constructed based on the driver response behavior characteristics. The inputs of the model are the initial relative distance, the reaction time, and the relative speed, and the output is the reaction distance, which is used to predict the driver reaction distance. The reaction distance prediction model is as Figure 3 shown. The fitting degree (R 2 ) of the reaction distance prediction model in this embodiment reaches 0.99.

[0058] Step 5: For the driver response behavior classification divided in Step 2, braking distance prediction models corresponding to the pre-warning response behavior, post-warning response behavior, and non-response behavior types are respectively constructed based on the long short-term memory model, and the braking distance is predicted for each response behavior. The inputs of the braking distance prediction model are the initial relative distance, the braking time, the braking force, the speed of the truck itself, and the speed of the leading vehicle, and the output is the braking distance.

[0059] The prediction result of the braking distance prediction model is as follows Figure 4 shown, and the model fitting degree (R 2 ) reaches 0.98.

[0060] Step 6: Determine the safety warning distances for forward collisions in three types of driver scenarios (pre-warning response scenario, post-warning scenario, non-response scenario) based on the reaction distance and braking distance, and send collision avoidance warning messages to the driver according to the safety warning distances.

[0061] In this embodiment, the safety warning distance is:

[0062] d w = d r + d b

[0063] where d w is the safety warning distance, d r is the reaction distance during the driver's collision avoidance process, and d b is the braking distance during the driver's collision avoidance process.

[0064] Specifically, in this embodiment, based on the natural driving data of trucks, 450 collision avoidance events are extracted, and a forward collision warning method considering the response behavior of truck drivers is used for warning. The safety, false alarm rate, and adaptability of the method are verified through simulation respectively, and compared with traditional safety distance algorithms (SDA, Honda algorithm) and safety time threshold methods (TTC = 4.5s, THW = 1.5s).

[0065] Among them, the safety index deceleration rate to avoid a crash (DRAC) is selected for verification, and its formula is:

[0066]

[0067] where v n represents the self-vehicle speed of the truck at the nth moment, and v n-1 represents the self-vehicle speed of the truck at the (n - 1)th moment. The smaller the DRAC, the better the safety. As can be seen from Table 1, the DRAC value of the forward collision warning method considering the driver's response behavior is 0.05, and the safety is improved by 15%.

[0068] Define the following for truck following: dangerous events and safety events. Dangerous events are defined by the Time to Impact (TIT) metric and the driver braking force. The TTC threshold involved in the TIT metric is selected as 4.5 s, and the truck driver braking force is selected as 0.052 g according to previous studies. Table 2 shows the false alarm rate results. The overall accuracy of the proposed forward collision warning method considering the truck driver's response behavior reaches 97.92%, and the false alarm rate is reduced to 1.73%.

[0069] Table 1

[0070]

[0071] Table 2 (1) False alarm rate metrics

[0072]

[0073] (2) False alarm rate of the response behavior scenario before warning

[0074]

[0075] Figure 5 Demonstrate the adaptability of the forward collision warning algorithm considering the behavior of truck drivers. As can be seen from the figure, the proposed algorithm is closer to the characteristics of the driver's following behavior and can better adapt to the collision avoidance characteristics in different collision avoidance scenarios.

[0076] The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those of ordinary skill in the art can make many modifications and variations based on the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art in the technical field based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art should be within the protection scope determined by the claims.

Claims

1. A forward collision warning method considering the response behavior of truck drivers, characterized in that, it includes the following steps: Step 1: Extract the warning video data and vehicle motion data of vehicle collision avoidance events in the following - vehicle scenario of truck drivers from natural driving data; Step 2: Determine the driver response behavior characteristics based on the warning video data and vehicle motion data, and use the k - means cluster method to classify the response behaviors in vehicle collision avoidance events. Divide the truck driving response behaviors into three categories: pre - warning response, post - warning response, and non - response; Step 3: According to the driver response behavior characteristics, divide the response process into a reaction stage and a braking stage; Step 4: Based on the support vector regression method, construct a driver reaction distance prediction model based on the driver response behavior characteristics to predict the driver reaction distance; the inputs of the driver reaction distance prediction model are the initial relative distance, reaction time, and relative vehicle speed, and the output is the reaction distance; Step 5: For the driver response behavior classification divided in Step 2, construct a braking distance prediction model for each category based on the long - short - term memory model to predict the braking distance for each response behavior respectively; Step 6: Determine the safety warning distance for forward collision based on the reaction distance and braking distance, and send a collision avoidance warning message to the driver according to the safety warning distance.

2. A forward collision warning method considering the response behavior of truck drivers according to claim 1, characterized in that, the warning video data is FCW warning video, and the extracted warning video data includes 10 - second video before warning and 10 - second video after warning.

3. A forward collision warning method considering the response behavior of truck drivers according to claim 1, characterized in that, the vehicle motion data includes the speed of the truck itself, the speed of the leading vehicle, the relative distance, the acceleration of the truck itself, and the braking moment, where the relative distance represents the distance between the truck itself and the leading vehicle.

4. A forward collision warning method considering the response behavior of truck drivers according to claim 3, characterized in that, the extraction criteria for vehicle collision avoidance events in Step 1 are that the following two conditions are met simultaneously: Condition 1: Relative distance < 70m; Condition 2: TTC < 4.5s or THW < 1.5s, where TTC represents the time to collision and THW represents the time headway.

5. A forward collision warning method considering the response behavior of truck drivers according to claim 3, characterized in that, the response behavior characteristics include: a) Speed - related variables: the speed of the truck itself, the relative speed before warning, the relative speed after warning, the relative speed change value; b) Distance - related variables: the relative distance at the warning moment, the relative distance at the braking moment; c) Braking - related variables: the braking deceleration before warning, the braking deceleration after warning, the deceleration change amount, the braking duration, the minimum deceleration; d) Reaction - related variables: reaction time.

6. A forward collision warning method considering the response behavior of truck drivers according to claim 1, characterized in that, The inputs of the braking distance prediction model are the initial relative distance, braking time, braking force, the speed of the truck itself, and the speed of the vehicle ahead, and the output is the braking distance.

7. A forward collision warning method considering the response behavior of truck drivers according to claim 1, characterized in that the safety warning distance is: d w = d r + d b Among them, d w is the safety warning distance, d r is the reaction distance during the driver's collision avoidance process, and d b is the braking distance during the driver's collision avoidance process.

8. A forward collision warning device considering the response behavior of truck drivers, including a memory, a processor, and a program stored in the memory, characterized in that when the processor executes the program, it implements the method described in any one of claims 1-7.

Images