Cut-in intent prediction method, probabilistic intent inference model training method and device
By analyzing the historical detection information of the target vehicle using the probability intention reasoning model and predicting its intention to enter the bicycle lane, the problem of low prediction accuracy in the prior art is solved and a more accurate warning is achieved.
Patent Information
- Application Number
- CN202311651177.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-04
- Publication Date
- 2025-06-06
AI Technical Summary
In the prior art, when judging whether a vehicle has the intention to enter the bicycle lane, the accuracy is low, which can easily lead to misjudgment and improper warning timing.
By obtaining the historical detection information of the target vehicle, including the lateral distance and lateral speed, and inputting it into the probability intention inference model, the discrete probability distribution is used to calculate the node entry probability and weight to obtain the predicted entry probability. If it is greater than the preset threshold, it is determined that the target vehicle has entry intention.
It improves the accuracy of vehicle entry intention prediction, can give early warnings at the right time, and reduces misjudgment.
Smart Images
Figure CN120096591A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of intelligent driving, and in particular to a method for predicting a cutting-in intention, a method for training a probabilistic intention reasoning model, and a device. Background Art
[0002] With the continuous development of autonomous driving and assisted driving technology, the safety of vehicles on the road has been further guaranteed. However, the roads in cities are complex, and vehicles are overtaking each other, constantly merging, changing lanes and other complex actions, which makes traffic accidents prone to occur.
[0003] At present, it is usually judged whether the target vehicle intends to cut into the own lane by whether the target vehicle continues to drive on the boundary line of the own lane. However, this judgment method requires the target vehicle to drive on the boundary line of the own lane for a certain period of time before triggering the warning. At the same time, when a vehicle cuts out of the own lane, it may cause misjudgment, making the warning accuracy low. Therefore, how to more accurately determine whether the vehicle intends to cut into the own lane and give a warning at the right time has become an urgent problem to be solved. Summary of the invention
[0004] In order to solve the above technical problems, the present disclosure provides a method for predicting a cutting-in intention, a method for training a probabilistic intention reasoning model and a device to more accurately determine whether a vehicle has the intention to cut into its own lane.
[0005] In a first aspect, an embodiment of the present disclosure provides a method for predicting a cut-in intention, comprising:
[0006] Acquire historical detection information of a preset number of consecutive frames of the target vehicle before the current moment, wherein the historical detection information includes a lateral distance and a lateral speed of the target vehicle relative to the own vehicle;
[0007] Inputting the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and using the node cut-in probability and the node weight to obtain the predicted cut-in probability;
[0008] If the predicted cut-in probability is greater than a first preset threshold, it is determined that the target vehicle has a cut-in intention.
[0009] In some embodiments, the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of a preset number of consecutive frames through discrete probability distribution, and uses the node cut-in probability and the node weight to obtain the predicted cut-in probability, including:
[0010] The node probability layer of the model is used to calculate the discrete probability distribution, and the corresponding node entry probability is obtained for each lateral distance and lateral speed in the continuous preset number of frames;
[0011] The weight distribution layer of the model determines the node weight corresponding to each lateral distance and lateral speed according to the sequential position of each lateral distance and lateral speed in the continuous preset number of frames;
[0012] The predicted cut-in probability is calculated by the probability layer of the model based on the node cut-in probability and node weight corresponding to each lateral distance and lateral velocity.
[0013] In some embodiments, before acquiring the historical detection information of the target vehicle for a preset number of consecutive frames before the current moment, the method further includes:
[0014] Determine the adjacent lanes of the lane where the vehicle is located;
[0015] A vehicle located in the adjacent lane and whose distance to the own vehicle is less than a preset distance is taken as a target vehicle.
[0016] In some embodiments, determining an adjacent lane of a lane where the vehicle is located includes:
[0017] If the adjacent lane does not exist, a virtual lane adjacent to the lane where the vehicle is located is created as the adjacent lane, and the width of the virtual lane is the same as the width of the lane where the vehicle is located.
[0018] In some embodiments, if the predicted cut-in probability is greater than a first preset threshold, determining that the target vehicle has a cut-in intention includes:
[0019] If the predicted cut-in probability is greater than a first preset threshold value and the driving state information of the target vehicle meets a preset condition, it is determined that the target vehicle has a cut-in intention.
[0020] In some embodiments, the driving state information of the target vehicle meets a preset condition, including:
[0021] The target vehicle is driving on the boundary line of the lane where the own vehicle is located, the target vehicle is the vehicle closest to the own vehicle, the collision time between the target vehicle and the own vehicle is less than a preset time, and the ratio of the distance between the target vehicle and the own vehicle and / or the speed of the own vehicle is less than a second preset threshold.
[0022] In a second aspect, the present disclosure provides a probabilistic intent reasoning model training method, including:
[0023] Acquire training data, the training data including a preset number of groups of historical data, the historical data including a preset number of consecutive frames of historical detection information of a historical vehicle, and actual results of whether a cut-in behavior occurs corresponding to the preset number of consecutive frames of historical detection information, wherein the preset number of consecutive frames of historical detection information includes a lateral distance and a lateral speed of the historical vehicle relative to the historical data collection vehicle for a preset number of consecutive frames;
[0024] Inputting the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and using the node cut-in probability and the node weight to obtain the predicted cut-in probability;
[0025] By comparing the predicted cut-in probability of historical vehicles with the actual results of whether the cut-in behavior occurred, and adjusting the model parameters according to the comparison results, the probabilistic intention reasoning model to be trained is trained to obtain a trained probabilistic intention reasoning model.
[0026] In a third aspect, an embodiment of the present disclosure provides a device for predicting a cut-in intention, including:
[0027] A first acquisition module is used to acquire a preset number of consecutive frames of historical detection information of the target vehicle before the current moment, wherein the historical detection information includes a lateral distance and a lateral speed of the target vehicle relative to the vehicle;
[0028] A prediction module is used to input the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight;
[0029] The first determination module is used to determine whether the target vehicle has a cut-in intention if the predicted cut-in probability is greater than a first preset threshold.
[0030] In a fourth aspect, an embodiment of the present disclosure provides a probabilistic intention reasoning model training device, including:
[0031] A second acquisition module is used to acquire training data, wherein the training data includes a preset number of groups of historical data, wherein the historical data includes historical detection information of a preset number of consecutive frames of historical vehicles, and actual results of whether a cut-in behavior occurs corresponding to the historical detection information of the preset number of consecutive frames, wherein the historical detection information of the preset number of consecutive frames includes a lateral distance and a lateral speed of the historical vehicle relative to the historical data collection vehicle for a preset number of consecutive frames;
[0032] A training module is used to input the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight;
[0033] The adjustment module is used to compare the predicted cut-in probability of historical vehicles with the actual results of whether the cut-in behavior occurred, and adjust the model parameters according to the comparison results, thereby completing the training of the probabilistic intention reasoning model to be trained and obtaining a trained probabilistic intention reasoning model.
[0034] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, including:
[0035] Memory;
[0036] Processor; and
[0037] Computer programs;
[0038] The computer program is stored in the memory and is configured to be executed by the processor to implement the method as described in the first aspect or the second aspect.
[0039] In a sixth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method described in the first aspect or the second aspect.
[0040] In a seventh aspect, an embodiment of the present disclosure provides a vehicle, comprising the device, electronic device or computer-readable storage medium as described above.
[0041] The cutting-in intention prediction method, probabilistic intention reasoning model training method and device provided in the embodiments of the present disclosure utilize historical detection information of a preset number of consecutive frames of the target vehicle to be predicted before the current moment and the probabilistic intention reasoning model to predict the probability of the target vehicle cutting into the own vehicle lane at a future moment, so that the obtained predicted cutting-in probability can be supported by strong data and theory, thereby improving the accuracy of the cutting-in intention prediction method. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.
[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0044] Figure 1 A flow chart of a method for predicting a cut-in intention provided in an embodiment of the present disclosure;
[0045] Figure 2 A schematic diagram of an application scenario provided by an embodiment of the present disclosure;
[0046] Figure 3 A flow chart of a method for predicting a cut-in intention provided by another embodiment of the present disclosure;
[0047] Figure 4 A schematic diagram of the structure of a device for predicting cutting intention provided by an embodiment of the present disclosure;
[0048] Figure 5 A schematic diagram of the structure of a probabilistic intention reasoning model training device provided in an embodiment of the present disclosure;
[0049] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION
[0050] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.
[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.
[0052] The disclosed embodiment provides a method for predicting a cut-in intention, which is described below in conjunction with a specific embodiment.
[0053] Figure 1 The flow chart of the method for predicting the cutting intention provided by the embodiment of the present disclosure is as follows. The method can be applied to Figure 2 The application scenario shown includes a self-vehicle 21 and a target vehicle 22, wherein the self-vehicle 21 includes an on-board device, which may be a vehicle computer, a smart phone, a PDA, a tablet computer, a laptop, an all-in-one machine, an intelligent driving device, etc. It is understandable that the cut-in intention prediction method provided in the embodiment of the present disclosure may also be applied in other scenarios.
[0054] Combine the following Figure 2 The application scenario shown is Figure 1 The cut-in intention prediction method shown in the figure is introduced, and the specific steps of the method are as follows:
[0055] S101, obtaining a preset number of consecutive frames of historical detection information of a target vehicle before a current moment, wherein the historical detection information includes a lateral distance and a lateral speed of the target vehicle relative to the own vehicle.
[0056] The target vehicle refers to other vehicles other than the ego vehicle, especially other vehicles that may have dangerous driving behaviors, including the behavior of the target vehicle cutting into the lane where the ego vehicle is located at a close distance. Specifically, the target vehicle cutting into the lane where the ego vehicle is located includes the target vehicle overtaking the ego vehicle and cutting into the lane from the front of the ego vehicle, or the target vehicle cutting into the lane where the ego vehicle is located from the rear of the ego vehicle.
[0057] The historical detection information is the information about the driving status of the target vehicle obtained by the vehicle-mounted equipment through the sensor equipment carried by the vehicle. The sensor equipment includes but is not limited to laser radar, camera, millimeter wave radar, etc.
[0058] The lateral distance of the target vehicle relative to the own vehicle may be the distance between the target vehicle and the own vehicle in a direction perpendicular to the lane, and the lateral speed of the target vehicle relative to the own vehicle may be the speed of the target vehicle in a direction perpendicular to the lane. Both the lateral distance and lateral speed of the target vehicle relative to the own vehicle are related to whether the target vehicle has the intention to cut in.
[0059] The target vehicle's historical detection information of a preset number of frames before the current moment may be historical detection information of a preset number of consecutive frames before the current moment, or historical detection information of a preset number of frames sampled from multiple frames of historical detection information before the current moment. For example, the lateral distance and lateral speed corresponding to each frame within 40 frames in the historical detection information of the target vehicle are obtained.
[0060] S102: Inputting the historical detection information of the continuous preset number of frames into a probabilistic intention reasoning model to obtain a predicted cut-in probability.
[0061] Optionally, the historical detection information of the continuous preset number of frames is input into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and uses the node cut-in probability and node weight to obtain the predicted cut-in probability.
[0062] Among them, the probabilistic intention reasoning model includes a node probability layer, a weight distribution layer and a probability layer;
[0063] The node probability layer of the model is used to calculate the discrete probability distribution, and the corresponding node entry probability is obtained for each lateral distance and lateral speed in the continuous preset number of frames;
[0064] The weight distribution layer of the model determines the node weight corresponding to each lateral distance and lateral speed according to the sequential position of each lateral distance and lateral speed in the continuous preset number of frames;
[0065] The probability layer of the model calculates the predicted cut-in probability based on the node cut-in probability and node weight corresponding to each lateral distance and lateral speed. It should be noted that before using the probabilistic intent reasoning model, it is necessary to pre-train the probabilistic intent reasoning model used to obtain the predicted cut-in probability. The following specifically introduces the training method of the probabilistic intent reasoning model.
[0066] First, training data is obtained, and the training data includes a preset number of historical data groups, the historical data includes historical detection information of a preset number of consecutive frames of historical vehicles, and the actual results of whether the cut-in behavior occurs corresponding to the historical detection information of the preset number of consecutive frames, wherein the historical detection information of the preset number of consecutive frames includes the lateral distance and lateral speed of the historical vehicle relative to the historical data collection vehicle for the preset number of consecutive frames; the historical data collection vehicle is a vehicle for collecting the driving state data of the historical vehicle, and the actual results in the training data correspond to whether the historical vehicle has cut into the lane where the historical data collection vehicle is located. In addition, in the process of training the probabilistic intention reasoning model and in the process of using the probabilistic intention reasoning model, the number of preset frames in the two processes should be consistent. Further, the historical detection information of the preset number of consecutive frames of the historical vehicle is used as the input data of the probabilistic intention reasoning model, and the predicted cut-in probability of the historical vehicle is used as the output data of the probabilistic intention reasoning model. By comparing the predicted cut-in probability of the historical vehicle with the actual results of whether the cut-in behavior occurs, and adjusting the model parameters according to the comparison results (i.e., adjusting the node weights corresponding to each lateral distance and lateral speed respectively), the probabilistic intention reasoning model to be trained is trained, and a trained probabilistic intention reasoning model is obtained.
[0067] Optionally, through discrete probability distribution calculation, the corresponding node cut-in probability is obtained for each lateral distance and lateral speed in the continuous preset frame number, wherein, through discrete probability distribution calculation, the actual cut-in probability corresponding to each lateral distance in all historical data can be counted, and the actual cut-in probability corresponding to each lateral speed in all historical data can also be counted, and then the actual cut-in probability corresponding to each lateral distance and each lateral speed in the continuous preset frame number can be obtained. In addition, each lateral distance and each lateral speed in the continuous preset frame number can be regarded as a node, and the actual cut-in probability is used as the node cut-in probability, so that the node cut-in probability corresponding to each lateral distance and lateral speed in the continuous preset frame number is obtained.
[0068] Optionally, the node weight corresponding to each lateral distance and lateral speed is determined according to the sequential position of each lateral distance and lateral speed in the continuous preset frame number, wherein, since each frame in the continuous preset frame number has a corresponding sequential position, the position order can be used to determine the node weights corresponding to the lateral distance and lateral speed, respectively, and then the node weights corresponding to each lateral distance and lateral speed in the continuous preset frame number can be obtained; it should be noted that the node weights are obtained after the probabilistic intent inference model training is completed, and the values of the node weights need to be continuously adjusted during the model training.
[0069] Optionally, the predicted cut-in probability is calculated based on the node cut-in probability and node weight corresponding to each lateral distance and lateral speed, wherein each lateral distance and each lateral speed in a continuous preset number of frames can be regarded as a node, and the sum of the products of the node cut-in probability and the node weight of all nodes is calculated to obtain the predicted cut-in probability.
[0070] Afterwards, the predicted cut-in probability of the historical detection information can be compared with the actual result of whether the cut-in behavior occurs (the actual cut-in can be recorded as a probability of 1, and the actual cut-in not occurring can be recorded as a probability of 0). By adjusting the node weight of each node, the average gap between the predicted cut-in probability of each historical detection information and the actual result of whether the cut-in behavior occurs is narrowed. When the average gap is lower than the end threshold, the training can be ended.
[0071] Take the historical detection information of 5 consecutive preset frames as an example, which includes 5 horizontal distances y 0 ,y 1 ,y 2 ,y 3 , y4 and 5 lateral velocities v 0 , v 1 , v 2 , v 3 , v 4, the calculation formula for predicting the cut-in probability can be:
[0072]
[0073] Where I represents whether the vehicle has cut-in behavior, I = 1 represents the vehicle has cut-in behavior, and I = 0 represents the vehicle has not cut-in behavior. V represents the 5 lateral speeds, namely V 0 =v 0 , V 1 =v 1 , V 2 =v 2 , V 3 =v 4 , V 5 =v 5 , Y represents 5 horizontal distances, that is, Y 0 =y 0 , V 1 =y 1 , Y 2 =y 2 , Y 3 =y 4 , Y 5 =y 5 P(I=1|V,Y) is when the lateral distance of the vehicle is y 0 ,y 1 ,y 2 ,y 3 ,y 4 , the lateral velocity is v 0 , v 1 , v 2 , v 3 , v 4 The conditional probability of the cut-in behavior occurring when .
[0074] Among them, P(I=1, V, Y) can be That is, the sum of the product of the node cut-in probability and the node weight of each node, P(I=1, V, Y)+P(I=0, V, Y) can be the sum of the node cut-in probability and the node non-cut-in probability, P(I=0, V, Y) can be That is, in the sum of the product of the node non-cut-in probability and the node weight of each node, the node non-cut-in probability = 1 - the node cut-in probability.
[0075] After obtaining the trained probabilistic intention reasoning model, the historical detection information of a preset number of consecutive frames before the current moment obtained in the above steps is input into the probabilistic intention reasoning model to obtain the cut-in probability of the target vehicle. The cut-in probability of the target vehicle represents the probability of the target vehicle cutting in within a preset time after the corresponding moment of the last frame in the historical detection information.
[0076] It can be understood that the number of frames of the historical detection information is the same as the number of frames during model training, and the continuous relationship between the frames in the historical detection information is also the same as that during model training.
[0077] S103: If the predicted cut-in probability is greater than a first preset threshold, it is determined that the target vehicle has a cut-in intention.
[0078] When the predicted cut-in probability of the target vehicle is greater than the first preset threshold, it is considered that the target vehicle has a greater probability of cutting into the lane where the vehicle is located, that is, it is determined that the target vehicle has a cut-in intention.
[0079] Furthermore, when the predicted cutting-in probability of the target vehicle is greater than a first preset threshold, whether the cutting-in behavior of the target vehicle will affect the safe driving of the own vehicle is judged by whether the driving status information of the target vehicle meets the preset conditions. For example, when the distance between the target vehicle and the own vehicle is too close, it is considered that the cutting-in behavior of the target vehicle may affect the safe driving of the own vehicle, and the vehicle cutting-in warning is triggered at this time.
[0080] The disclosed embodiment obtains historical detection information of the target vehicle for a preset number of frames before the current moment, wherein the historical detection information includes the lateral distance and lateral speed of the target vehicle relative to the own vehicle; the historical detection information for the preset number of frames is input into a probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information for the preset number of frames through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight; if the predicted cut-in probability is greater than a first preset threshold value, it is determined that the target vehicle has the intention to cut in, and the historical detection information of the target vehicle to be predicted for a preset number of frames before the current moment and the probabilistic intention reasoning model are used to predict the probability of the target vehicle cutting into the own vehicle lane at a future moment, so that the obtained predicted cut-in probability can be supported by strong data and theory, thereby improving the accuracy of the cut-in intention prediction method.
[0081] In addition, since the embodiment of the present disclosure uses a probabilistic intent reasoning model, it is more interpretable and has a faster calculation speed than a neural network.
[0082] Figure 3 This is a flow chart of a method for predicting a cut-in intention provided by another embodiment of the present disclosure. Figure 3 As shown, the method includes the following steps:
[0083] S301. Determine the adjacent lane of the lane where the vehicle is located.
[0084] Specifically, if the adjacent lane does not exist, a virtual lane adjacent to the lane where the vehicle is located is created, and the width of the virtual lane is the same as that of the lane where the vehicle is located.
[0085] The lane where the ego vehicle is located is the lane where the ego vehicle is currently traveling. The adjacent lanes of the ego vehicle's lane include the left lane adjacent to the ego vehicle's lane and the right lane adjacent to the ego vehicle's lane. Vehicles traveling in adjacent lanes are the vehicles most likely to cut into the ego vehicle's lane.
[0086] In some embodiments, lane lines can be identified through visual recognition to determine adjacent lanes to the lane where the vehicle is located; alternatively, adjacent lanes to the lane where the vehicle is located can also be determined through high-precision map data.
[0087] In some embodiments, the adjacent lane of the lane where the vehicle is located cannot be identified due to lane line wear or missing map data, so a virtual lane is created as the adjacent lane of the lane where the vehicle is located according to the width of the lane where the vehicle is located.
[0088] S302: A vehicle located in the adjacent lane and whose distance to the own vehicle is less than a preset distance is taken as a target vehicle.
[0089] Preliminary screening of vehicles in adjacent lanes. If a vehicle in an adjacent lane is far away from the vehicle, even if the vehicle cuts into the lane where the vehicle is located, it will not affect the safe driving of the vehicle. Therefore, there is no need to calculate the predicted cut-in probability of these vehicles, and only the target vehicles that are close to the vehicle need to be considered.
[0090] Specifically, vehicles located in the adjacent lanes and whose distance from the vehicle is less than the preset distance are regarded as target vehicles, including target vehicles located in the adjacent lanes in front of the vehicle and in the adjacent lanes behind the vehicle. In some embodiments, different preset distances can be set for the front of the vehicle and the rear of the vehicle, respectively, or the same preset distance can be set, which is not limited in the embodiments of the present disclosure.
[0091] For example, a vehicle located in the adjacent lane and within a range of 50 m in front and 15 m in rear from the vehicle is taken as the target vehicle.
[0092] In some embodiments, the position information of the vehicle in the odom coordinate system (ego vehicle coordinate system) can be obtained, and the position information in the odom coordinate system can be further converted to the ego coordinate system to determine the distance between the vehicle and the ego vehicle.
[0093] S303: Obtain the distance between the target vehicle and the center line of the lane where the target vehicle is located in each frame of historical detection information to obtain the lateral distance of the target vehicle in each frame.
[0094] The lateral distance is the distance between the target vehicle and the vehicle in the direction perpendicular to the lane. For each frame of historical detection information, the distance between the target vehicle and the vehicle in the direction perpendicular to the lane in the historical detection information of the frame is determined as the lateral distance of the target vehicle in the historical detection information of the frame, until the lateral distance of the target vehicle in each frame of historical detection information is obtained.
[0095] S304: Perform differential calculation on the horizontal distances between two adjacent frames to obtain a calculation result.
[0096] S305: Filter the calculation results to obtain the lateral speed of the target vehicle in each frame.
[0097] The lateral distances of two adjacent frames are differentially calculated, that is, the average speed of the target vehicle during the collection of historical detection information of two adjacent frames is calculated based on the difference in the lateral distances of the two adjacent frames and the frame rate of the historical detection information. Since the time interval between the collection of historical detection information of two adjacent frames is extremely short, this average speed can also be regarded as the instantaneous speed of the target vehicle in each frame.
[0098] The calculation results obtained by differential calculation are further subjected to Butterworth filtering to make the calculation results more stable. The characteristic of Butterworth filter is that the frequency response curve in the passband is maximally flat without fluctuations, and gradually decreases to zero in the stopband. On the Bode plot of the logarithm of the amplitude to the diagonal frequency, starting from a certain boundary angular frequency, the amplitude gradually decreases with the increase of the angular frequency and tends to negative infinity.
[0099] S306: Input the historical detection information into a probabilistic intention reasoning model to obtain a predicted cut-in probability of the target vehicle.
[0100] In some embodiments, after the predicted cut-in probability of the target vehicle is obtained, Butterworth filtering is performed on the predicted cut-in probability of the target vehicle.
[0101] S307: If the predicted cut-in probability of the target vehicle is greater than a first preset threshold, and the driving state information of the target vehicle meets a preset condition, a vehicle cut-in warning is triggered.
[0102] Specifically, the driving status information of the target vehicle meets preset conditions, including: the target vehicle is driving on the boundary line of the lane where the own vehicle is located, the target vehicle is the vehicle closest to the own vehicle, the collision time between the target vehicle and the own vehicle is less than a preset time, and the ratio of the distance between the target vehicle and the own vehicle and / or the speed of the own vehicle is less than a second preset threshold.
[0103] Among them, when the target vehicle is driving on the boundary line of the lane where the self-vehicle is located, it is considered that the target vehicle has caused a certain impact on the normal driving of the self-vehicle; when the target vehicle is the vehicle closest to the self-vehicle, it is considered that the target vehicle is the vehicle most likely to affect the safe driving of the self-vehicle; when the collision time between the target vehicle and the self-vehicle is less than the preset time, and the ratio of the distance between the target vehicle and the self-vehicle and the speed of the self-vehicle is less than the second preset threshold, it is considered that the target vehicle may collide with the self-vehicle in a very short time in the future. If the target vehicle meets the above preset conditions on the basis of the cut-in probability being greater than the first preset threshold, it is considered that the target vehicle is very likely to affect the safe and normal driving of the self-vehicle when cutting into the lane where the self-vehicle is located, and a vehicle cut-in warning needs to be initiated.
[0104] Specifically, the implementation process and principle of S306-S307 are consistent with those of S102-S103, and will not be repeated here.
[0105] The disclosed embodiment determines the adjacent lane of the lane where the own vehicle is located; takes the vehicle located in the adjacent lane and the distance between the own vehicle and the vehicle is less than the preset distance as the target vehicle; obtains the distance between the target vehicle and the center line of the lane where the target vehicle is located in each frame of the historical detection information, and obtains the lateral distance of the target vehicle in each frame; performs differential calculation on the lateral distances of two adjacent frames to obtain a calculation result; performs Butterworth filtering on the calculation result to obtain the lateral speed of the target vehicle in each frame; inputs the historical detection information into a probabilistic intention reasoning model to obtain the predicted cut-in probability of the target vehicle; if the predicted cut-in probability of the target vehicle is greater than a first preset threshold value and the driving state information of the target vehicle meets the preset conditions, a vehicle cut-in warning is triggered, and the accuracy and flexibility of the cut-in intention prediction method are further improved by further judging the degree of danger of the target vehicle to the safe and normal driving of the own vehicle based on the driving state information of the target vehicle on the basis that the cut-in probability of the target vehicle is greater than the first preset threshold value.
[0106] At the same time, the embodiment of the present disclosure uses the lateral distance and lateral speed to predict the predicted cut-in probability of the target vehicle, and at the same time performs Butterworth filtering on the differentially calculated lateral speed and the inferred predicted cut-in probability, thereby increasing the stability of the cut-in intention prediction method.
[0107] Figure 4 The schematic diagram of the structure of the cutting intention prediction device provided in the embodiment of the present disclosure. The cutting intention prediction device may be the vehicle-mounted device as described in the above embodiment, or the cutting intention prediction device may be a component or assembly in the vehicle-mounted device. The cutting intention prediction device provided in the embodiment of the present disclosure may execute the processing flow provided in the cutting intention prediction method embodiment, such as Figure 4As shown, the cutting-in intention prediction device 40 includes: a first acquisition module 41, a prediction module 42, and a first determination module 43; the first acquisition module 41 is used to obtain historical detection information of a preset number of consecutive frames of the target vehicle before the current moment, and the historical detection information includes the lateral distance and lateral speed of the target vehicle relative to the own vehicle; the prediction module 42 is used to input the historical detection information of the preset number of consecutive frames into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cutting-in probability of each lateral distance and lateral speed in the historical detection information of the preset number of consecutive frames through discrete probability distribution, and obtains the predicted cutting-in probability using the node cutting-in probability and the node weight; the first determination module 43 is used to determine that the target vehicle has a cutting-in intention if the predicted cutting-in probability is greater than a first preset threshold.
[0108] Optionally, the prediction module 42 includes a first determination unit 421, a second determination unit 422, and a first calculation unit 423; the first determination unit 421 is used to calculate by the node probability layer of the model through discrete probability distribution, and obtain the corresponding node cut-in probability for each lateral distance and lateral speed in the continuous preset frame number; the second determination unit 422 is used to determine the node weight corresponding to each lateral distance and lateral speed according to the sequential position of each lateral distance and lateral speed in the continuous preset frame number by the weight allocation layer of the model; the calculation unit 423 is used to calculate the predicted cut-in probability based on the node cut-in probability and node weight corresponding to each lateral distance and lateral speed by the probability layer of the model.
[0109] Optionally, the cutting-in intention prediction device 40 also includes a second determination module 44, including a third determination unit 441 and a fourth determination unit 442; the third determination unit 441 is used to determine the adjacent lane of the lane where the own vehicle is located; the fourth determination unit 442 is used to take the vehicle located in the adjacent lane and the distance between the own vehicle and the vehicle is less than a preset distance as the target vehicle.
[0110] Optionally, the third determination unit 441 is further configured to create a virtual lane adjacent to the lane where the vehicle is located as the adjacent lane if the adjacent lane does not exist, and the width of the virtual lane is the same as the width of the lane where the vehicle is located.
[0111] Optionally, the first determination module 43 is specifically configured to determine that the target vehicle has a cut-in intention if the predicted cut-in probability is greater than a first preset threshold and the driving state information of the target vehicle meets a preset condition.
[0112] Optionally, the driving status information of the target vehicle meets preset conditions, including: the target vehicle is driving on the boundary line of the lane where the own vehicle is located, the target vehicle is the vehicle closest to the own vehicle, the collision time between the target vehicle and the own vehicle is less than a preset time, and the ratio of the distance between the target vehicle and the own vehicle and / or the speed of the own vehicle is less than a second preset threshold.
[0113] Figure 4 The cutting-in intention prediction device of the illustrated embodiment can be used to execute the technical solution of the above-mentioned method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.
[0114] Figure 5 This is a schematic diagram of the structure of the probabilistic intention inference model training device provided in the embodiment of the present disclosure. Figure 5 As shown, the probabilistic intention reasoning model training device includes a second acquisition module 51, a training module 52, and an adjustment module 53; the second acquisition module 51 is used to acquire training data, the training data includes a preset number of historical data groups, the historical data includes historical detection information of a preset number of consecutive frames of historical vehicles, and the actual results of whether the cut-in behavior occurs corresponding to the preset number of consecutive frames of historical detection information, wherein the preset number of consecutive frames of historical detection information includes the lateral distance and lateral speed of the historical vehicle relative to the historical data collection vehicle for a preset number of consecutive frames; the training module 52 is used to input the preset number of consecutive frames of historical detection information into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the preset number of consecutive frames of historical detection information through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight. The adjustment module 53 is used to compare the predicted cut-in probability of the historical vehicle with the actual results of whether the cut-in behavior occurs, and adjust the model parameters according to the comparison results, thereby completing the training of the probabilistic intention reasoning model to be trained, and obtaining a trained probabilistic intention reasoning model.
[0115] In addition, an embodiment of the present disclosure also provides a vehicle, which includes a cutting-in intention prediction device or a probabilistic intention reasoning model training device as described in the above embodiments.
[0116] Figure 6 The electronic device provided in the embodiment of the present disclosure can execute the processing flow provided in the embodiment of the cut-in intention prediction method or the probabilistic intention inference model training method, such as Figure 6 As shown, the electronic device 60 includes: a memory 61, a processor 62, a computer program and a communication interface 63; wherein the computer program is stored in the memory 61 and is configured to be executed by the processor 62 to execute the above-mentioned cut-in intention prediction method or probabilistic intention reasoning model training method.
[0117] In addition, an embodiment of the present disclosure also provides a computer-readable storage medium on which a computer program is stored, and the computer program is executed by a processor to implement the cut-in intention prediction method or the probabilistic intention reasoning model training method described in the above embodiments.
[0118] In addition, an embodiment of the present disclosure also provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the above-mentioned cut-in intention prediction method or probabilistic intention reasoning model training method is implemented.
[0119] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).
[0120] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0121] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0122] The above description is only a specific embodiment of the present disclosure, so that those skilled in the art can understand or implement the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting entry intention, It is characterized in that The method comprises: Acquire historical detection information of a preset number of consecutive frames of the target vehicle before the current moment, wherein the historical detection information includes a lateral distance and a lateral speed of the target vehicle relative to the own vehicle; Inputting the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and using the node cut-in probability and the node weight to obtain the predicted cut-in probability; If the predicted cut-in probability is greater than a first preset threshold, it is determined that the target vehicle has a cut-in intention.
2. The method according to claim 1, It is characterized in that The probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the preset number of consecutive frames through discrete probability distribution, and uses the node cut-in probability and node weight to obtain the predicted cut-in probability, including: The node probability layer of the model is used to calculate the discrete probability distribution, and the corresponding node entry probability is obtained for each lateral distance and lateral speed in the continuous preset number of frames; The weight distribution layer of the model determines the node weight corresponding to each lateral distance and lateral speed according to the sequential position of each lateral distance and lateral speed in the continuous preset number of frames; The predicted cut-in probability is calculated by the probability layer of the model based on the node cut-in probability and node weight corresponding to each lateral distance and lateral velocity.
3. The method according to claim 1, It is characterized in that Before obtaining the historical detection information of the target vehicle for a preset number of consecutive frames before the current moment, the method further includes: Determine the adjacent lanes of the lane where the vehicle is located; A vehicle located in the adjacent lane and whose distance to the own vehicle is less than a preset distance is taken as a target vehicle.
4. The method according to claim 3, It is characterized in that The determining of the adjacent lane of the lane where the vehicle is located includes: If the adjacent lane does not exist, a virtual lane adjacent to the lane where the vehicle is located is created as the adjacent lane, and the width of the virtual lane is the same as the width of the lane where the vehicle is located.
5. The method according to claim 1, It is characterized in that If the predicted cut-in probability is greater than a first preset threshold, determining that the target vehicle has a cut-in intention includes: If the predicted cut-in probability is greater than a first preset threshold value and the driving state information of the target vehicle meets a preset condition, it is determined that the target vehicle has a cut-in intention.
6. The method according to claim 5, It is characterized in that The driving state information of the target vehicle meets the preset conditions, including: The target vehicle is driving on the boundary line of the lane where the own vehicle is located, the target vehicle is the vehicle closest to the own vehicle, the collision time between the target vehicle and the own vehicle is less than a preset time, and the ratio of the distance between the target vehicle and the own vehicle and / or the speed of the own vehicle is less than a second preset threshold.
7. A probabilistic intent reasoning model training method, It is characterized in that The probabilistic intention reasoning model is applied to the method according to any one of claims 1 to 6, the method comprising: Acquire training data, the training data including a preset number of groups of historical data, the historical data including a preset number of consecutive frames of historical detection information of a historical vehicle, and actual results of whether a cut-in behavior occurs corresponding to the preset number of consecutive frames of historical detection information, wherein the preset number of consecutive frames of historical detection information includes a lateral distance and a lateral speed of the historical vehicle relative to the historical data collection vehicle for a preset number of consecutive frames; Inputting the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and using the node cut-in probability and the node weight to obtain the predicted cut-in probability; By comparing the predicted cut-in probability of historical vehicles with the actual results of whether the cut-in behavior occurred, and adjusting the model parameters according to the comparison results, the probabilistic intention reasoning model to be trained is trained to obtain a trained probabilistic intention reasoning model.
8. A device for predicting cutting intention, It is characterized in that The device comprises: A first acquisition module is used to acquire a preset number of consecutive frames of historical detection information of the target vehicle before the current moment, wherein the historical detection information includes a lateral distance and a lateral speed of the target vehicle relative to the vehicle; A prediction module is used to input the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight; The first determination module is used to determine whether the target vehicle has a cut-in intention if the predicted cut-in probability is greater than a first preset threshold.
9. A probabilistic intention reasoning model training device, It is characterized in that The device comprises: A second acquisition module is used to acquire training data, wherein the training data includes a preset number of groups of historical data, wherein the historical data includes historical detection information of a preset number of consecutive frames of historical vehicles, and actual results of whether a cut-in behavior occurs corresponding to the historical detection information of the preset number of consecutive frames, wherein the historical detection information of the preset number of consecutive frames includes a lateral distance and a lateral speed of the historical vehicle relative to the historical data collection vehicle for a preset number of consecutive frames; A training module is used to input the historical detection information of the continuous preset number of frames into the probabilistic intention reasoning model, and the probabilistic intention reasoning model calculates the node cut-in probability of each lateral distance and lateral speed in the historical detection information of the continuous preset number of frames through discrete probability distribution, and obtains the predicted cut-in probability using the node cut-in probability and the node weight; The adjustment module is used to compare the predicted cut-in probability of historical vehicles with the actual results of whether the cut-in behavior occurred, and adjust the model parameters according to the comparison results, thereby completing the training of the probabilistic intention reasoning model to be trained and obtaining a trained probabilistic intention reasoning model.
10. An electronic device, It is characterized in that include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the method according to any one of claims 1 to 7.
11. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
12. A vehicle, include: The cutting intention prediction device as claimed in claim 8; or the probabilistic intention reasoning model training device as claimed in claim 9; Or the electronic device as claimed in claim 10; or, the computer-readable storage medium as claimed in claim 11.