Vehicle door opening anti-collision early warning method and system

Abstract

The invention belongs to the technical field of vehicle safety, and discloses a vehicle door opening anti-collision early warning method and system. Comprises: acquiring intention analysis data; analyzing the intention analysis data, and identifying the door opening intention of the passenger; obtaining vehicle feature data; predicting a driving track according to the vehicle feature data; the driving track is analyzed, and the collision risk is evaluated; according to the assessment result of the collision risk and the door opening intention, giving a corresponding alarm to the passenger; the vehicle door opening risk can be accurately recognized, the vehicle passengers can be reminded in time, the accident risk caused by the vehicle blind area is reduced, the vehicle door opening safety is improved, and the traffic safety of pedestrians and riders is practically guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of vehicle safety technology, and more specifically, to a door opening anti-collision warning method and system. Background Art

[0002] With the rapid growth in the number of cars, the urban road traffic environment is becoming increasingly complex. The safety hazards of the side and rear blind spots when vehicles are driving at low speeds or parking are gradually becoming prominent, especially the frequent "door opening kill" accidents caused by the opening of the car door, which seriously threatens the lives of pedestrians and cyclists. Traditional rearview mirror observation is difficult to fully cover the side and rear blind spots, and the driver's judgment is easily limited by the environment and visual blind spots, resulting in uncontrollable risks. Therefore, an intelligent vehicle door opening warning system is urgently needed to effectively improve the safety of the vehicle when opening the door.

[0003] The patent application with publication number CN116238416A discloses a method, device, equipment and medium for anti-collision warning of automobile door opening; including: obtaining perception information around the current vehicle, the perception information at least including image information and ultrasonic echo information; respectively identifying the target object type and distance in the image information and the ultrasonic echo information; fusing the target object type and distance of the image information with the ultrasonic echo information to obtain each target object type and target object distance; determining whether the current vehicle meets the door opening condition according to each target object type and target object distance, if the current vehicle meets the door opening condition, the door is opened; if the current vehicle does not meet the door opening condition, the door is locked and an anti-collision warning is performed; this application avoids the occurrence of collision accidents caused by human negligence and realizes anti-collision warning for manual or automatic door opening;

[0004] Although the above technology can realize anti-collision warning when the car door is opened, it only roughly describes the recognition results of combining radar data, image data, and ultrasonic echo data to identify obstacles with collision risks around the vehicle, and does not explain in detail the specific technical solution. As a result, in actual applications, the ability to recognize obstacles in complex environments is insufficient, and obstacles with collision risks around the vehicle cannot be effectively identified, so that the driver fails to obtain necessary safety information in time when opening the door, which affects the accuracy and reliability of the warning, increases the risk of accidents, and cannot fully protect the safety of pedestrians and cyclists.

[0005] In view of this, the present invention proposes a door opening anti-collision warning method and system to solve the above-mentioned problem. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purpose, the present invention provides the following technical solution: a door opening anti-collision warning method, comprising:

[0007] Collect intent analysis data;

[0008] Analyze the intention analysis data to identify the passenger's intention to open the door;

[0009] Obtain vehicle characteristic data;

[0010] Predict driving trajectory based on vehicle feature data;

[0011] Analyze driving trajectories and assess collision risks;

[0012] Passengers are given corresponding warnings based on the assessment results of the collision risk and the intention to open the door.

[0013] Further, the intention analysis data includes vehicle speed and door spacing; the door spacing is the distance between the passenger and the inner side of the door;

[0014] The method for identifying a passenger's door opening intention comprises:

[0015] Preset the intention threshold, which includes the vehicle speed threshold and the vehicle distance threshold; compare the vehicle speed with the vehicle speed threshold, and compare the door distance with the vehicle distance threshold; if the vehicle speed is less than the vehicle speed threshold, and the door distance is less than the vehicle distance threshold, then generate a door opening intention instruction; if the vehicle speed is greater than or equal to the vehicle threshold, or the door distance is greater than or equal to the vehicle distance threshold, then do not generate the door opening intention instruction;

[0016] The vehicle characteristic data includes a target distance and a relative speed; the target distance is the distance between the vehicle and the moving obstacle, and the relative speed is the moving speed of the moving obstacle relative to the vehicle.

[0017] Furthermore, the step of predicting the driving trajectory includes:

[0018] Step S101: taking vehicle characteristic data corresponding to the same moving obstacle as a set of data sets, where the data sets correspond to the moving obstacles one by one;

[0019] Step S102: predicting the relative speed of each moving obstacle at a future moment according to the relative speed in each data set, marking it as a predicted speed, and adding each predicted speed to the corresponding data set;

[0020] Step S103: predict the target distance of each moving obstacle at a future moment according to the predicted speed and target distance in each data set, mark it as the predicted distance, and add each predicted distance to the corresponding data set;

[0021] Step S104: constructing a driving trajectory corresponding to each moving obstacle based on the predicted distance in each data set.

[0022] Furthermore, in step S102, the method for predicting the relative speed of each moving obstacle at a future moment includes:

[0023] The relative speed in each data set is regarded as a speed set, and the speed set corresponds to the data set one by one; each speed set is input into the trained speed prediction model to predict the future speed of each moving obstacle. The relative speed at each moment, the speed prediction model is an RNN neural network model; the training method of the speed prediction model includes:

[0024] Pre-collection continuous relative speeds and construct a speed training set. is an integer greater than 1; based on the speed training set, a speed prediction model is trained to predict the relative speed at future moments;

[0025] Preset a sliding step size L and a sliding window length D; convert the relative speeds in the speed training set into multiple training samples using a sliding window method, each training sample includes D relative speeds; use each training sample as the input of the speed prediction model, predict the relative speed after the sliding step size L as the output, and use the last L relative speeds of each training sample as the prediction target. Use the mean absolute percentage error MAPE to evaluate the model accuracy of the prediction results. When the calculated MAPE is less than the preset MAPE, the speed prediction model training corresponding to the relative speed is completed; generate a speed prediction model that predicts the relative speed at future moments based on the relative speed.

[0026] Furthermore, in step S103, the method for predicting the target distance of each moving obstacle at a future moment includes:

[0027] Obtain the acquisition time corresponding to each set of vehicle feature data, sort the acquisition time corresponding to the vehicle feature data in the same data set from late to early, and obtain the time sorting table corresponding to each data set; use the vehicle feature data corresponding to the acquisition time ranked first in each time sorting table as the real-time feature data of the corresponding data set; input each set of real-time feature data into the trained distance prediction model respectively, and predict the target distance at a future time; replace the target distance in each set of real-time feature data with the target distance corresponding to a future time, and input the replaced real-time feature data into the trained distance prediction model respectively, and predict the target distance at the next two times; and so on, predict the future distance corresponding to each moving obstacle in total. Target distance at a moment;

[0028] The training process of the distance prediction model includes:

[0029] Pre-collection Set of real-time feature data, Each set of real-time feature data is set with a corresponding future distance, which is the target distance at a future moment. is an integer greater than 1, and the real-time feature data and the corresponding future distance are converted into a corresponding set of feature vectors; each set of feature vectors is used as the input of the distance prediction model, and the distance prediction model uses a set of predicted future distances corresponding to each set of real-time feature data as output, and uses the actual future distance corresponding to each set of real-time feature data as the prediction target, and the actual future distance is the pre-set future distance corresponding to the real-time feature data; minimizing the sum of the prediction errors of all real-time feature data is used as the training target; the distance prediction model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the distance prediction model is a deep neural network model;

[0030] In step S104, the method for constructing the driving trajectory corresponding to each mobile obstacle is: sorting the predicted distance corresponding to each mobile obstacle from early to late according to the corresponding prediction order, and obtaining a distance sorting table corresponding to each mobile obstacle; and constructing the driving trajectory corresponding to each mobile obstacle according to the distance sorting table.

[0031] Furthermore, the method for assessing collision risk includes:

[0032] The collision threshold is set by using an optimization algorithm, and the predicted distance in the driving trajectory corresponding to each moving obstacle is compared with the collision threshold; if there is a predicted distance less than or equal to the collision threshold, the corresponding moving obstacle is marked as a collision obstacle; if all predicted distances are greater than the collision threshold, the corresponding moving obstacle is not marked; the target azimuth at the current moment is obtained and marked as the current azimuth; the first predicted distance in the driving trajectory corresponding to the collision obstacle that is less than or equal to the collision threshold is marked as the collision distance; the future moment corresponding to the collision distance is marked as the collision moment, and a moment interval is constructed based on the current moment and the collision moment; the predicted speed and predicted distance corresponding to the collision obstacle in the moment interval are used as a prediction set, and the prediction set and the current azimuth are input into the trained angle prediction model to predict the target azimuth corresponding to the collision moment and mark it as the collision azimuth;

[0033] The collision azimuth is compared with 0 to determine the spatial position of the collision obstacle; if the collision azimuth is greater than 0, the collision obstacle is on the right side of the vehicle, and if the collision azimuth is less than 0, the collision obstacle is on the left side of the vehicle; the door corresponding to the door opening intention command is marked as the door to be opened, and the spatial position of the door to be opened is obtained; the spatial position of the door to be opened is compared with the spatial position of the collision obstacle; if the door to be opened and the collision obstacle are located on the same side, it is assessed that the corresponding collision obstacle has a collision risk; if the door to be opened and the collision obstacle are not located on the same side, it is assessed that the corresponding collision obstacle does not have a collision risk.

[0034] Furthermore, the steps of setting the collision threshold using the optimization algorithm include:

[0035] Step S201: Preset initialization temperature , minimum temperature , Temperature reduction coefficient And the maximum number of iterations , and let the current temperature ;

[0036] Step S202: preset a threshold range; randomly set a feasible solution , a feasible solution is a value within the threshold range, the feasible solution The range is the threshold range;

[0037] Step S203: determining a fitness function;

[0038] Step S204: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ;

[0039] Step S205: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ;

[0040] Step S206: loop through steps S204 to S205 until the number of loops reaches the maximum number of iterations. When the cycle ends, let the current temperature , let the maximum number of iterations ;

[0041] Step S207: loop through steps S204 to S206 until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding value is used as the collision threshold.

[0042] Furthermore, in the above step S203, the fitness function is expressed as: ;

[0043] In the formula, For fitness, is accuracy; the accuracy is obtained through a preset threshold set, and the threshold set includes the accuracy corresponding to different collision thresholds;

[0044] In the above step S205, the fitness difference The expression is ;

[0045] Probability The expression is: ; In the formula, e is a natural constant.

[0046] Furthermore, the method for issuing corresponding warnings to passengers includes:

[0047] If the vehicle speed is less than the speed threshold and there is a collision obstacle, the warning lights on the vehicle rearview mirror and the door with the same spatial location as the collision obstacle will light up;

[0048] If a door opening intention command is generated and there is a collision obstacle, the buzzer in the vehicle will sound an alarm, and the warning lights on the vehicle's rearview mirror and the door to be opened will flash;

[0049] If the vehicle speed is greater than or equal to the speed threshold and there is a collision obstacle, the warning light on the vehicle's rearview mirror will light up.

[0050] A vehicle door opening anti-collision warning system, implementing the vehicle door opening anti-collision warning method, comprises:

[0051] Data collection module, used to collect intent analysis data;

[0052] Intention recognition module, used to analyze the intention analysis data and identify the passenger's intention to open the door;

[0053] A data acquisition module, used to acquire vehicle characteristic data;

[0054] Trajectory prediction module, predicts the driving trajectory based on vehicle feature data;

[0055] Collision analysis module, used to analyze driving trajectory and assess collision risk;

[0056] The risk warning module issues corresponding warnings to passengers based on the collision risk assessment results and the door opening intention.

[0057] The technical effects and advantages of the door opening anti-collision warning method and system of the present invention are as follows:

[0058] By collecting intention analysis data, the passenger's intention to open the door can be accurately identified; and by obtaining vehicle feature data, the driving trajectory of moving obstacles can be effectively predicted, so as to accurately analyze the collision risk between the vehicle door and the moving obstacle, and then take differentiated warnings, making full use of the on-board sensor data and data analysis algorithms to realize intelligent collision warning when the door is opened; at the same time, the use of technologies such as adaptive adjustment of collision thresholds and precise risk assessment based on spatial relationships can effectively avoid false alarms or missed alarms, and improve the accuracy and reliability of warnings; it can accurately identify the risk of door opening and promptly remind vehicle passengers, reduce the risk of accidents caused by vehicle blind spots, improve the safety of door opening, and effectively protect the traffic safety of pedestrians and cyclists. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a schematic diagram of a door opening anti-collision warning system according to Embodiment 1 of the present invention;

[0060] Figure 2 This is a flow chart of a driving trajectory prediction method according to Embodiment 1 of the present invention;

[0061] Figure 3 This is a flow chart of a door opening anti-collision warning method according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Example 1

[0064] See also Figure 1 As shown, a door opening anti-collision warning system described in this embodiment includes a data collection module, an intention recognition module, a data acquisition module, a trajectory prediction module, a collision analysis module and a risk warning module; each module is connected by wired and / or wireless means to achieve data transmission between modules.

[0065] Data collection module, used to collect intent analysis data.

[0066] The intention analysis data includes vehicle speed and door distance; the vehicle speed is the vehicle's driving speed, which is obtained by a vehicle speed sensor installed at the bottom of the vehicle; the door distance is the distance between the passenger and the inside of the door, which is obtained by a laser ranging sensor installed on the inside of the door.

[0067] The intention recognition module is used to analyze the intention analysis data and identify the passenger's intention to open the door.

[0068] Methods for identifying a passenger's intention to open the door include:

[0069] The intention threshold is preset, and the intention threshold includes a speed threshold and a distance threshold. The speed threshold and the distance threshold are preset by those skilled in the art according to actual conditions. In this embodiment, the preferred speed threshold is , the vehicle distance threshold is ;

[0070] The vehicle speed is compared with the vehicle speed threshold, and the door distance is compared with the vehicle distance threshold; if the vehicle speed is less than the vehicle speed threshold, and the door distance is less than the vehicle distance threshold, a door opening intention instruction is generated, indicating that the passenger has the intention to open the door; if the vehicle speed is greater than or equal to the vehicle threshold, or the door distance is greater than or equal to the vehicle distance threshold, no door opening intention instruction is generated, indicating that the passenger does not have the intention to open the door.

[0071] The data acquisition module is used to acquire vehicle characteristic data.

[0072] Vehicle characteristic data includes target distance and relative speed; target distance is the distance between the vehicle and the moving obstacle, and relative speed is the moving speed of the moving obstacle relative to the vehicle; moving obstacles are moving objects in the traffic environment that can interfere with or hinder the opening of vehicle doors, such as pedestrians, cyclists, other vehicles, etc.; vehicle characteristic data is obtained through the millimeter wave radar installed on the rear bumper of the vehicle.

[0073] The trajectory prediction module predicts the driving trajectory based on the vehicle feature data.

[0074] like Figure 2 As shown, the steps of predicting the driving trajectory include:

[0075] Step S101: The vehicle characteristic data corresponding to the same moving obstacle is taken as a set of data sets, and the data sets correspond to the moving obstacles one by one; by aggregating the vehicle characteristic data of the same moving obstacle, it is helpful to capture the dynamic characteristics of different moving obstacles, reduce the possibility of data mixing, and improve the accuracy and effectiveness of the subsequent prediction process;

[0076] Step S102: predicting the relative speed of each moving obstacle at a future time according to the relative speed in each data set, marking it as a predicted speed, and adding each predicted speed to the corresponding data set; by predicting the relative speed at a future time, it is helpful to timely understand the future relative speed of the moving obstacle, thereby improving the accuracy of subsequent prediction of the future target distance;

[0077] Step S103: predict the target distance of each moving obstacle at the corresponding future moment according to the predicted speed and target distance in each data set, mark it as the predicted distance, and add each predicted distance to the corresponding data set; by combining the predicted speed and the target distance, the predicted distance at the future moment can be predicted more accurately, thereby obtaining the future driving trajectory of the moving obstacle, which is helpful to evaluate the potential collision risk;

[0078] Step S104: constructing a driving trajectory corresponding to each moving obstacle based on the predicted distance in each data set.

[0079] In the above step S102, the method for predicting the relative speed of each moving obstacle at a future time includes:

[0080] The relative speed in each data set is regarded as a speed set, and the speed set corresponds to the data set one by one; each speed set is input into the trained speed prediction model to predict the future speed of each moving obstacle. The relative speed at each moment, the speed prediction model is the RNN neural network model; this embodiment is preferably The time is 5s; the training method of the speed prediction model includes:

[0081] Pre-collection continuous relative speeds and construct a speed training set. is an integer greater than 1; based on the speed training set, a speed prediction model is trained to predict the relative speed at future moments;

[0082] According to the actual experience of those skilled in the art, the sliding step length L and the sliding window length D are preset; the relative speeds in the speed training set are converted into multiple training samples using the sliding window method, and each training sample includes D relative speeds; each training sample is used as the input of the speed prediction model, and the relative speed after the sliding step length L is predicted as the output, and the last L relative speeds of each training sample are used as the prediction target. The prediction results are evaluated for model accuracy using the mean absolute percentage error MAPE. When the calculated MAPE is less than the preset MAPE, the speed prediction model training corresponding to the relative speed is completed; wherein, the calculation formula of the mean absolute percentage error MAPE is: ; In the formula, For the The predicted target corresponding to the predicted relative speed is For the The predicted relative speed, is the number of predicted relative speeds; and generates a speed prediction model for predicting the relative speed at future moments based on the relative speed.

[0083] Exemplarily, the speed training set A includes 10 relative speeds. , For the A relative speed, , define the length of the sliding window as 3, the sliding step L as 1, use the sliding window to construct 8 training samples, each training sample contains 3 consecutive relative speeds, and the next relative speed of the 3 consecutive relative speeds is used as the prediction target; for example:

[0084] Training samples , training samples The corresponding prediction target is ; Training samples , training samples The corresponding prediction target is ; And so on, it is used to train the speed prediction model corresponding to the relative speed.

[0085] In the above step S103, the method for predicting the target distance of each moving obstacle at a future moment includes:

[0086] Obtain the acquisition time corresponding to each set of vehicle feature data, sort the acquisition time corresponding to the vehicle feature data in the same data set from late to early, and obtain the time sorting table corresponding to each data set; use the vehicle feature data corresponding to the acquisition time ranked first in each time sorting table as the real-time feature data of the corresponding data set; input each set of real-time feature data into the trained distance prediction model respectively, and predict the target distance at a future time; replace the target distance in each set of real-time feature data with the target distance corresponding to a future time, and input the replaced real-time feature data into the trained distance prediction model respectively, and predict the target distance at the next two times; and so on, predict the future distance corresponding to each moving obstacle in total. Target distance at a moment;

[0087] The training process of the distance prediction model includes:

[0088] Pre-collection Set of real-time feature data, Each set of real-time feature data is set with a corresponding future distance, which is the target distance at a future moment. is an integer greater than 1, and the real-time feature data and the corresponding future distance are converted into a corresponding set of feature vectors; the future distance corresponding to the real-time feature data is collected by technicians in the process of historically predicting the target distance of each moving obstacle at the future moment. A set of real-time feature data is collected, and the future distance corresponding to each set of real-time feature data is collected in turn. The group of real-time feature data sets the corresponding future distances in turn;

[0089] Each set of feature vectors is used as the input of the distance prediction model. The distance prediction model uses a set of predicted future distances corresponding to each set of real-time feature data as output, and uses the actual future distance corresponding to each set of real-time feature data as the prediction target. The actual future distance is the pre-set future distance corresponding to the real-time feature data. The training target is to minimize the sum of the prediction errors of all real-time feature data. The calculation formula of the prediction error is: ,in is the prediction error, is the group number of the feature vector corresponding to the real-time feature data, For the The predicted future distance corresponding to the set of real-time feature data, For the The actual future distance corresponding to a group of real-time feature data is trained, and the training is stopped when the sum of the prediction errors reaches convergence.

[0090] The above distance prediction model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights, which determine the importance and influence of data transmission in the neural network; an activation function is applied to each neuron between the hidden layer and the output layer, and the activation function introduces nonlinearity, allowing the network to learn more complex patterns and features.

[0091] In the above step S104, the method for constructing the driving trajectory corresponding to each mobile obstacle is: sorting the predicted distance corresponding to each mobile obstacle from early to late according to the corresponding prediction order, and obtaining a distance sorting table corresponding to each mobile obstacle; and constructing the driving trajectory corresponding to each mobile obstacle according to the distance sorting table.

[0092] The collision analysis module is used to analyze the driving trajectory and assess the collision risk.

[0093] Methods for assessing collision risk include:

[0094] The collision threshold is set by using an optimization algorithm, and the predicted distance in the driving trajectory corresponding to each moving obstacle is compared with the collision threshold; if there is a predicted distance less than or equal to the collision threshold, the corresponding moving obstacle is marked as a collision obstacle; if all predicted distances are greater than the collision threshold, the corresponding moving obstacle is not marked; the target azimuth at the current moment is obtained and marked as the current azimuth; the target azimuth is the angle of the collision obstacle relative to the forward direction of the vehicle, and the target azimuth is obtained by the millimeter-wave radar installed on the rear bumper of the vehicle; the first predicted distance in the driving trajectory corresponding to the collision obstacle that is less than or equal to the collision threshold is marked as the collision distance; the future moment corresponding to the collision distance is marked as the collision moment, and a time interval is constructed based on the current moment and the collision moment; the predicted speed and predicted distance corresponding to the collision obstacle in the time interval are used as the prediction set, and the prediction set and the current azimuth are input into the trained angle prediction model to predict the target azimuth corresponding to the collision moment and mark it as the collision azimuth;

[0095] The collision azimuth is compared with 0 to determine the spatial positioning of the collision obstacle; if the collision azimuth is greater than 0, the collision obstacle is on the right side of the vehicle, and if the collision azimuth is less than 0, the collision obstacle is on the left side of the vehicle; the door corresponding to the door opening intention command is marked as the door to be opened, and the spatial positioning of the door to be opened is obtained; the spatial positioning is located on the left or right side of the vehicle; the spatial positioning of the door to be opened is compared with the spatial positioning of the collision obstacle; if the door to be opened and the collision obstacle are located on the same side, it is assessed that the corresponding collision obstacle has a collision risk; if the door to be opened and the collision obstacle are not located on the same side, it is assessed that the corresponding collision obstacle does not have a collision risk.

[0096] The steps of setting the collision threshold using the optimization algorithm include:

[0097] Step S201: Preset initialization temperature , minimum temperature , Temperature reduction coefficient And the maximum number of iterations , and let the current temperature ; Initializing parameters helps control the convergence speed of the algorithm, avoids premature convergence to the local optimal solution, and improves the global search capability;

[0098] Step S202: preset a threshold range. The threshold range is preset by those skilled in the art according to the actual situation to ensure that the feasible solution is within a reasonable boundary and improve the practicality and reliability of the algorithm; randomly set a feasible solution. , a feasible solution is a value within the threshold range, the feasible solution The range of is the threshold range; randomly generating feasible solutions can introduce diversity, increase exploration opportunities, and thus improve the possibility of finding the optimal solution;

[0099] Step S203: determining a fitness function; this helps to quantify the evaluation criteria of feasible solutions, thereby judging the quality of feasible solutions;

[0100] Step S204: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ; It helps to find better solutions and enhance the local search ability of the algorithm;

[0101] Step S205: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ; Allowing the algorithm to dynamically adjust feasible solutions based on the quality of the solution helps to achieve a more flexible search process, and accepting inferior solutions based on probability helps to avoid falling into local optimality and increase the possibility of global search;

[0102] Step S206: loop through steps S204 to S205 until the number of loops reaches the maximum number of iterations. When the cycle ends, let the current temperature , that is, the current temperature in step S201 is cooled down, and the value after cooling is assigned to the current temperature; the maximum number of iterations is set , that is, assign the value of the reduced maximum number of iterations to the maximum number of iterations; if the reduced maximum number of iterations is not an integer, the reduced maximum number of iterations is rounded up to make it an integer; gradually lowering the temperature helps the algorithm converge gradually, reduce the acceptance of inferior solutions, and finally find the optimal solution;

[0103] Step S207: loop through steps S204 to S206 until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding value is used as the collision threshold.

[0104] In the above step S203, the fitness function is expressed as: ;

[0105] In the formula, For fitness, For accuracy; accuracy is obtained through a preset threshold set, the threshold set includes the accuracy corresponding to different collision thresholds, and the threshold set is pre-set by technical personnel in this field through collision tests and in combination with actual conditions.

[0106] In the above step S205, the fitness difference The expression is ;

[0107] Probability The expression is: ; In the formula, e is a natural constant.

[0108] The risk warning module issues corresponding warnings to passengers based on the collision risk assessment results and the door opening intention.

[0109] The method of giving corresponding warning to passengers includes:

[0110] If the vehicle speed is less than the speed threshold and there is a collision obstacle, the warning lights installed on the vehicle rearview mirror and the door with the same spatial location as the collision obstacle will light up;

[0111] If a door opening intention command is generated and there is a collision obstacle, the buzzer installed in the vehicle will sound three times in succession, and the warning lights on the vehicle's rearview mirror and the door to be opened will flash;

[0112] If the vehicle speed is greater than or equal to the speed threshold and there is a collision obstacle, the warning light on the vehicle's rearview mirror will light up.

[0113] This embodiment collects intention analysis data to accurately identify the passenger's intention to open the door; and obtains vehicle feature data to effectively predict the driving trajectory of moving obstacles, thereby accurately analyzing the collision risk between the vehicle door and the moving obstacle, and then taking differentiated warnings, making full use of the on-board sensor data and data analysis algorithms to achieve intelligent collision warning when the door is opened; at the same time, it adopts technologies such as adaptive adjustment of collision thresholds and precise risk assessment based on spatial relationships to effectively avoid false alarms or missed alarms and improve the accuracy and reliability of warnings; it can accurately identify the risk of door opening and promptly remind vehicle passengers, reduce the risk of accidents caused by vehicle blind spots, improve the safety of door opening, and effectively protect the traffic safety of pedestrians and cyclists.

[0114] Example 2

[0115] See also Figure 3As shown, the part not described in detail in this embodiment can be seen from the description of embodiment 1, and a door opening anti-collision warning method is provided, the method comprising:

[0116] Collect intent analysis data;

[0117] Analyze the intention analysis data to identify the passenger's intention to open the door;

[0118] Obtain vehicle characteristic data;

[0119] Predict driving trajectory based on vehicle feature data;

[0120] Analyze driving trajectories and assess collision risks;

[0121] Passengers are given corresponding warnings based on the assessment results of the collision risk and the intention to open the door.

[0122] Example 3

[0123] The present application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable codes, and when the computer-readable codes are executed by the one or more processors, the vehicle door opening anti-collision warning method as described above may be executed.

[0124] The method or system according to the implementation mode of the present application can also be implemented with the help of the architecture of the electronic device shown in the present application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store a door opening anti-collision warning method provided in the present application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in the present application is only exemplary. When implementing different devices, one or more components in the electronic device shown in the present application may be omitted according to actual needs.

[0125] Example 4

[0126] One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by a processor, a door opening anti-collision warning method according to an embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0127] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application, for example: a door opening anti-collision warning method. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0128] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0129] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A door opening anti-collision warning method, characterized in that: include: Collect intent analysis data; Analyze the intention analysis data to identify the passenger's intention to open the door; Obtain vehicle characteristic data; Predict driving trajectory based on vehicle feature data; Analyze driving trajectories and assess collision risks; Passengers are given corresponding warnings based on the assessment results of the collision risk and the intention to open the door.

2. A door opening anti-collision warning method according to claim 1, characterized in that: The intention analysis data includes vehicle speed and door spacing; Door clearance is the distance between the passenger and the inside of the door; The method for identifying a passenger's door opening intention comprises: Preset the intention threshold, which includes the vehicle speed threshold and the vehicle distance threshold; compare the vehicle speed with the vehicle speed threshold, and compare the door distance with the vehicle distance threshold; if the vehicle speed is less than the vehicle speed threshold, and the door distance is less than the vehicle distance threshold, then generate a door opening intention instruction; if the vehicle speed is greater than or equal to the vehicle threshold, or the door distance is greater than or equal to the vehicle distance threshold, then do not generate the door opening intention instruction; The vehicle characteristic data includes a target distance and a relative speed; the target distance is the distance between the vehicle and the moving obstacle, and the relative speed is the moving speed of the moving obstacle relative to the vehicle.

3. A door opening anti-collision warning method according to claim 2, characterized in that: The step of predicting the driving trajectory comprises: Step S101: taking vehicle characteristic data corresponding to the same moving obstacle as a set of data sets, where the data sets correspond to the moving obstacles one by one; Step S102: predicting the relative speed of each moving obstacle at a future moment according to the relative speed in each data set, marking it as a predicted speed, and adding each predicted speed to the corresponding data set; Step S103: predict the target distance of each moving obstacle at a future moment according to the predicted speed and target distance in each data set, mark it as the predicted distance, and add each predicted distance to the corresponding data set; Step S104: constructing a driving trajectory corresponding to each moving obstacle based on the predicted distance in each data set.

4. A door opening anti-collision warning method according to claim 3, characterized in that: In step S102, the method for predicting the relative speed of each moving obstacle at a future time includes: The relative speed in each data set is regarded as a speed set, and the speed set corresponds to the data set one by one; each speed set is input into the trained speed prediction model to predict the future speed of each moving obstacle. The relative speed at each moment, the speed prediction model is an RNN neural network model; the training method of the speed prediction model includes: Pre-collection A continuous relative speed, and construct a speed training set, is an integer greater than 1; based on the speed training set, a speed prediction model is trained to predict the relative speed at future moments; Preset a sliding step size L and a sliding window length D; convert the relative speeds in the speed training set into multiple training samples using a sliding window method, each training sample includes D relative speeds; use each training sample as the input of the speed prediction model, predict the relative speed after the sliding step size L as the output, and use the last L relative speeds of each training sample as the prediction target. Use the mean absolute percentage error MAPE to evaluate the model accuracy of the prediction results. When the calculated MAPE is less than the preset MAPE, the speed prediction model training corresponding to the relative speed is completed; generate a speed prediction model that predicts the relative speed at future moments based on the relative speed.

5. A door opening anti-collision warning method according to claim 4, characterized in that: In step S103, the method for predicting the target distance of each moving obstacle at a future moment includes: Obtain the acquisition time corresponding to each set of vehicle feature data, sort the acquisition time corresponding to the vehicle feature data in the same data set from late to early, and obtain the time sorting table corresponding to each data set; use the vehicle feature data corresponding to the acquisition time ranked first in each time sorting table as the real-time feature data of the corresponding data set; input each set of real-time feature data into the trained distance prediction model respectively, and predict the target distance at a future time; replace the target distance in each set of real-time feature data with the target distance corresponding to a future time, and input the replaced real-time feature data into the trained distance prediction model respectively, and predict the target distance at the next two times; and so on, predict the future distance corresponding to each moving obstacle in total. Target distance at a moment; The training process of the distance prediction model includes: Pre-collection Set of real-time feature data, Each set of real-time feature data is set with a corresponding future distance, which is the target distance at a future moment. is an integer greater than 1, and the real-time feature data and the corresponding future distance are converted into a corresponding set of feature vectors; each set of feature vectors is used as the input of the distance prediction model, and the distance prediction model uses a set of predicted future distances corresponding to each set of real-time feature data as output, and uses the actual future distance corresponding to each set of real-time feature data as the prediction target, and the actual future distance is the pre-set future distance corresponding to the real-time feature data; minimizing the sum of the prediction errors of all real-time feature data is used as the training target; the distance prediction model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the distance prediction model is a deep neural network model; In step S104, the method for constructing the driving trajectory corresponding to each mobile obstacle is: sorting the predicted distance corresponding to each mobile obstacle from early to late according to the corresponding prediction order, and obtaining a distance sorting table corresponding to each mobile obstacle; and constructing the driving trajectory corresponding to each mobile obstacle according to the distance sorting table.

6. A door opening anti-collision warning method according to claim 5, characterized in that: The method for assessing collision risk comprises: The collision threshold is set by using an optimization algorithm, and the predicted distance in the driving trajectory corresponding to each moving obstacle is compared with the collision threshold; if there is a predicted distance less than or equal to the collision threshold, the corresponding moving obstacle is marked as a collision obstacle; if all predicted distances are greater than the collision threshold, the corresponding moving obstacle is not marked; the target azimuth at the current moment is obtained and marked as the current azimuth; the first predicted distance in the driving trajectory corresponding to the collision obstacle that is less than or equal to the collision threshold is marked as the collision distance; the future moment corresponding to the collision distance is marked as the collision moment, and a moment interval is constructed based on the current moment and the collision moment; the predicted speed and predicted distance corresponding to the collision obstacle in the moment interval are used as a prediction set, and the prediction set and the current azimuth are input into the trained angle prediction model to predict the target azimuth corresponding to the collision moment and mark it as the collision azimuth; The collision azimuth is compared with 0 to determine the spatial position of the collision obstacle; if the collision azimuth is greater than 0, the collision obstacle is on the right side of the vehicle, and if the collision azimuth is less than 0, the collision obstacle is on the left side of the vehicle; the door corresponding to the door opening intention command is marked as the door to be opened, and the spatial position of the door to be opened is obtained; the spatial position of the door to be opened is compared with the spatial position of the collision obstacle; if the door to be opened and the collision obstacle are located on the same side, it is assessed that the corresponding collision obstacle has a collision risk; if the door to be opened and the collision obstacle are not located on the same side, it is assessed that the corresponding collision obstacle does not have a collision risk.

7. A door opening anti-collision warning method according to claim 6, characterized in that: The steps of setting the collision threshold using the optimization algorithm include: Step S201: Preset initialization temperature , minimum temperature , Temperature reduction coefficient and the maximum number of iterations , and let the current temperature ; Step S202: preset a threshold range; randomly set a feasible solution , a feasible solution is a value within the threshold range, the feasible solution The range is the threshold range; Step S203: determining a fitness function; Step S204: Calculate feasible solutions The corresponding fitness ; with feasible solutions For the current point, perform random perturbations in the neighborhood of the current point to obtain a new feasible solution , and calculate the new feasible solution The corresponding fitness ; Step S205: Calculate the fitness difference ; If the fitness difference , then let , the new feasible solution The value of is given to the feasible solution ; If the fitness difference , then calculate the probability , according to the probability make ; Step S206: loop through steps S204 to S205 until the number of loops reaches the maximum number of iterations. When the cycle ends, let the current temperature , let the maximum number of iterations ; Step S207: loop through steps S204 to S206 until the current temperature reaches When , the loop ends and a feasible solution is obtained The corresponding value is used as the collision threshold.

8. The door opening anti-collision warning method according to claim 7, characterized in that: In the above step S203, the fitness function is expressed as: ; In the formula, For fitness, is accuracy; the accuracy is obtained through a preset threshold set, and the threshold set includes the accuracy corresponding to different collision thresholds; In the above step S205, the fitness difference The expression is ; Probability The expression is: ; In the formula, e is a natural constant.

9. A door opening anti-collision warning method according to claim 8, characterized in that: The method for providing corresponding warning to passengers includes: If the vehicle speed is less than the speed threshold and there is a collision obstacle, the warning lights on the vehicle rearview mirror and the door with the same spatial location as the collision obstacle will light up; If a door opening intention command is generated and there is a collision obstacle, the buzzer in the vehicle will sound an alarm, and the warning lights on the vehicle's rearview mirror and the door to be opened will flash; If the vehicle speed is greater than or equal to the speed threshold and there is a collision obstacle, the warning light on the vehicle's rearview mirror will light up.

10. A door opening anti-collision warning system, applied to a door opening anti-collision warning method according to any one of claims 1 to 9, characterized in that: include: Data collection module, used to collect intent analysis data; Intention recognition module, used to analyze the intention analysis data and identify the passenger's intention to open the door; A data acquisition module, used to acquire vehicle characteristic data; Trajectory prediction module, predicts the driving trajectory based on vehicle feature data; Collision analysis module, used to analyze driving trajectory and assess collision risk; The risk warning module issues corresponding warnings to passengers based on the collision risk assessment results and the door opening intention.

Images