Obstacle motion intention prediction model acquisition method, device, equipment and medium

Abstract

The embodiments of the present disclosure disclose an obstacle motion intention prediction model acquisition method, device, equipment and medium, which comprises: acquiring an obstacle prediction basic data set, and determining a first data set and a second data set in the obstacle prediction basic data set; for each to-be-corrected data group in the second data set, determining the motion intention of each obstacle according to the to-be-corrected data group, the pre-labeled yield relationship between the current vehicle and each obstacle, and each yield scene, and updating the to-be-corrected data group based on the motion intention of each obstacle to obtain an updated second data set, and then updating the obstacle prediction basic data set in combination with the first data set, training an obstacle motion intention initial prediction model based on the updated obstacle prediction basic data set to obtain an obstacle motion intention target prediction model, so as to reduce the complexity of the obstacle motion intention prediction model and improve the accuracy and real-time performance of acquiring the motion intention of the obstacle.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of automatic driving, and particularly relates to an obstacle motion intention prediction model acquisition method and device, equipment and a medium. BACKGROUND

[0002] In the process of automatic driving, the surrounding environment is very complex, and vehicles will interact with various obstacles. In the process of interacting with obstacles, the vehicle needs to make logical judgments of taking or yielding. However, in the process of making logical judgments, the motion intention of the obstacle needs to be predicted.

[0003] At present, the method for predicting the motion intention of the obstacle includes an intention estimation method based on game theory, and some intention prediction methods based on long short-term memory neural network, complex machine learning model and other algorithms. With the increase of data volume, the accuracy of these methods will also be higher and higher, but since there is no initial solution, a large amount of multi-dimensional data needs to be input and trained for millions of times to obtain a relatively ideal model. Moreover, with the increase of training times and data sets, the model obtained by training is very complex, and the time consumption will also increase, which is difficult to deploy and use on real vehicles. SUMMARY

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the embodiments of the present disclosure provide an obstacle motion intention prediction model acquisition method, device, equipment and medium, which realizes the modification of the obstacle prediction basic data set in combination with the pre-labeled taking and yielding relationship, reduces the complexity of the obstacle motion intention prediction model, and further improves the accuracy and real-time performance of obtaining the motion intention of the obstacle.

[0005] In a first aspect, the embodiments of the present disclosure provide an obstacle motion intention prediction model acquisition method, which comprises:

[0006] An obstacle prediction basic data set is acquired, and a first data set and a second data set in the obstacle prediction basic data set are determined; wherein the first data set is reasonable data of a current vehicle taking and yielding in the obstacle prediction basic data set, and the second data set is unreasonable data of the current vehicle taking and yielding in the obstacle prediction basic data set;

[0007] For each to-be-modified data group in the second data set, the motion intention of each obstacle is determined according to the to-be-modified data group, the pre-labeled taking and yielding relationship between the current vehicle and each obstacle in the to-be-modified data group, and each taking and yielding scene corresponding to the to-be-modified data group, and the to-be-modified data group is updated based on the motion intention of each obstacle to obtain an updated second data set;

[0008] update the obstacle prediction base dataset based on the first dataset and the updated second dataset, and train an initial prediction model of the motion intention of the obstacle based on the updated obstacle prediction base dataset to obtain a target prediction model of the motion intention of the obstacle.

[0009] In a second aspect, the embodiments of the present disclosure further provide an obstacle motion intention prediction model obtaining device, which comprises:

[0010] an obstacle prediction base dataset obtaining module, configured to obtain an obstacle prediction base dataset, and determine a first dataset and a second dataset in the obstacle prediction base dataset; wherein the first dataset is reasonable data of the current vehicle in the obstacle prediction base dataset, and the second dataset is unreasonable data of the current vehicle in the obstacle prediction base dataset;

[0011] a dataset updating module, configured to, for each to-be-corrected data group in the second dataset, determine the motion intention of each obstacle according to the to-be-corrected data group, the pre-labeled relationship between the current vehicle and each obstacle in the to-be-corrected data group, and each yield line scene corresponding to the to-be-corrected data group, and update the to-be-corrected data group based on the motion intention of each obstacle to obtain an updated second dataset;

[0012] an obstacle motion intention target prediction model training module, configured to update the obstacle prediction base dataset based on the first dataset and the updated second dataset, and train an initial prediction model of the motion intention of the obstacle based on the updated obstacle prediction base dataset to obtain a target prediction model of the motion intention of the obstacle.

[0013] In a third aspect, the embodiments of the present disclosure further provide an electronic device, which comprises: one or more processors; a storage device configured to store one or more programs; and when the one or more programs are executed by the one or more processors, the one or more processors implement the obstacle motion intention prediction model obtaining method as described above.

[0014] In a fourth aspect, the embodiments of the present disclosure further provide a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the obstacle motion intention prediction model obtaining method as described above.

[0015] This disclosure provides a method for obtaining an obstacle motion intention prediction model. It acquires an obstacle prediction base dataset and determines a first dataset and a second dataset within it. This facilitates subsequent correction of unreasonable right-of-way data for current vehicles. Then, for each data group to be corrected in the second dataset, the motion intention of each obstacle is determined based on the data group to be corrected, the pre-labeled right-of-way relationships between the current vehicle and each obstacle within the data group, and the corresponding right-of-way scenarios. The data group to be corrected is then updated based on the motion intentions of each obstacle to obtain an updated second dataset for correction. The obstacle prediction base dataset is updated based on the first dataset and the updated second dataset. Finally, an initial obstacle motion intention prediction model is trained based on the updated obstacle prediction base dataset to obtain a target obstacle motion intention prediction model. This method achieves correction of the obstacle prediction base dataset by incorporating pre-labeled right-of-way relationships, reducing the complexity of the obstacle motion intention prediction model and thus improving the accuracy and real-time performance of obtaining obstacle motion intentions. Attached Figure Description

[0016] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0017] Figure 1 This is a flowchart of a method for obtaining an obstacle motion intention prediction model according to an embodiment of this disclosure;

[0018] Figure 2 This is a schematic diagram of a reverse search tree according to an embodiment of this disclosure;

[0019] Figure 3 This is a schematic diagram illustrating the positions of the current vehicle and the current obstacle in an intersection scenario, where the current right-of-way relationship in this embodiment of the disclosure is the current vehicle overtaking the current obstacle.

[0020] Figure 4 This is a schematic diagram illustrating the positions of the current vehicle and the current obstacle in an intersection scenario, where the current right-of-way relationship is that the current vehicle yields to the current obstacle in one embodiment of this disclosure.

[0021] Figure 5 This is a schematic diagram illustrating the positions of the current vehicle and the current obstacle in a merging scenario, where the current right-of-way relationship in this embodiment of the disclosure is the current vehicle overtaking the current obstacle.

[0022] Figure 6A position schematic diagram of a current vehicle and a current obstacle in a current yield relationship and a current yield scene of an embodiment of the present disclosure is a yield relationship of the current vehicle yielding to the current obstacle, and the current yield scene is a merging scene.

[0023] Figure 7 A structure schematic diagram of an obstacle motion intention prediction model acquisition device of an embodiment of the present disclosure is shown in FIG. 6.

[0024] Figure 8 A structure schematic diagram of an electronic device of an embodiment of the present disclosure is shown in FIG. 7. DETAILED DESCRIPTION

[0025] Embodiments of the present disclosure will be described in more detail with reference to the drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms, and should not be interpreted as being limited to the embodiments set forth herein, but rather, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes, and are not intended to limit the scope of protection of the present disclosure.

[0026] It should be noted that the concepts of "first", "second", and the like mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules or units.

[0027] The names of the messages or information exchanged between the plurality of devices in the embodiments of the present disclosure are only for illustrative purposes, and are not intended to limit the scope of these messages or information.

[0028] To solve the above problems, an obstacle motion intention prediction model acquisition method is provided in the embodiments of the present disclosure, which modifies an obstacle prediction basic data set in combination with a pre-labeled yield relationship, reduces the complexity of an obstacle motion intention prediction model, and further improves the accuracy and real-time performance of obtaining the motion intention of the obstacle.

[0029] Figure 1 A flowchart of an obstacle motion intention prediction model acquisition method of an embodiment of the present disclosure is shown in FIG. 5. The method can be executed by an obstacle motion intention prediction model acquisition device, which can be implemented in software and / or hardware, and can be configured in an electronic device. As shown in FIG. 5, the method can specifically include the following steps: Figure 1

[0030] S110, an obstacle prediction basic data set is acquired, and a first data set and a second data set in the obstacle prediction basic data set are determined.

[0031] ​The obstacle prediction basic data set is a data set collected based on real vehicle deployment for automatic driving of the current vehicle to yield. It can be understood that the current vehicle can automatically complete the yield operation when the automatic driving controls the current vehicle to yield reasonably, and the current vehicle needs to be manually taken over to complete the yield when the automatic driving controls the current vehicle to yield unreasonably. Accordingly, the obstacle prediction basic data set can be divided into a first data set and a second data set. The first data set is data in the obstacle prediction basic data set in which the current vehicle yields reasonably, and the second data set is data in the obstacle prediction basic data set in which the current vehicle yields unreasonably.

[0032] Specifically, the data of the current vehicle in the automatic driving yield stage is obtained based on the information collection device installed on the current vehicle, for example, the current vehicle speed, the current vehicle entering the interaction area distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering the interaction area distance, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle, etc., to constitute the obstacle prediction basic data set. Further, the obstacle prediction basic data set is divided into a first data set and a second data set according to whether the current vehicle yields reasonably.

[0033] S120, for each to-be-corrected data group in the second data set, determining the motion intention of each obstacle according to the to-be-corrected data group, the pre-labeled yield relationship between the current vehicle and each obstacle in the to-be-corrected data group, and each yield scene corresponding to the to-be-corrected data group, and updating the to-be-corrected data group based on the motion intention of each obstacle to obtain an updated second data set.

[0034] The to-be-corrected data group is a data group when the current vehicle performs automatic driving yield. Since the current vehicle has a continuous yield situation, each to-be-corrected data group includes at least one to-be-corrected data, and there is a certain time interval between different to-be-corrected data groups. The obstacle is another object, such as a vehicle, which needs to yield or yield to the current vehicle. The yield relationship includes the current vehicle yielding to the obstacle and the current vehicle yielding to the obstacle. The yield scene includes a crossing scene and a merging scene. The crossing scene is, for example, a "cross" road, and the merging scene is, for example, a "Y" road. The motion intention of the obstacle is the motion mode of the obstacle, which can be understood as the acceleration of the obstacle.

[0035] Specifically, for each to-be-corrected data group in the second data set, the motion intention of each obstacle corresponding to each to-be-corrected data in the to-be-corrected data group is determined through kinematic model analysis and calculation according to each to-be-corrected data in the to-be-corrected data group, the yield relationship between the current vehicle and each obstacle pre-labeled in the to-be-corrected data group, and each yield scene corresponding to each to-be-corrected data in the to-be-corrected data group. Further, the to-be-corrected data group is updated based on the determined motion intention of each obstacle to obtain an updated second data set.

[0036] On the basis of the above examples, the to-be-corrected data group includes at least one to-be-corrected data, and each to-be-corrected data corresponds to one obstacle of the current vehicle. Then, the motion intention of each obstacle can be determined according to the to-be-corrected data group, the yield relationship between the current vehicle and each obstacle pre-labeled in the to-be-corrected data group, and each yield scene corresponding to the to-be-corrected data group by the following obstacle intention correction method:

[0037] A correction order of each to-be-corrected data in the to-be-corrected data group is determined, and the to-be-corrected data in the first position of the correction order is taken as a current to-be-corrected data;

[0038] Whether there is a motion intention corresponding to the current obstacle is judged according to the current to-be-corrected data, the current yield relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yield scene corresponding to the current to-be-corrected data;

[0039] If there is, the motion intention corresponding to the current obstacle is determined, and whether the current to-be-corrected data is the to-be-corrected data in the last position of the correction order is judged;

[0040] In the case that the current to-be-corrected data is the to-be-corrected data in the last position of the correction order, the motion intention of each obstacle is determined;

[0041] In the case that the current to-be-corrected data is not the to-be-corrected data in the last position of the correction order, the to-be-corrected data in the next position of the current to-be-corrected data in the correction order is updated according to the current to-be-corrected data, the current yield relationship, and the current yield scene, the updated to-be-corrected data in the next position of the current to-be-corrected data in the correction order is taken as a new current to-be-corrected data, and the operation of judging whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, the current yield relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yield scene corresponding to the current to-be-corrected data is performed again;

[0042] If not, the to-be-corrected data located at the previous position of the current correction data is updated, the updated to-be-corrected data located at the previous position of the current correction data is taken as new current correction data, and the operation of determining whether there is a motion intention corresponding to the current obstacle according to the current correction data, the current yield relationship between the current vehicle and the current obstacle marked in the current correction data, and the current yield scene corresponding to the current correction data is executed.

[0043] Wherein, the correction order is the order of occurrence and collection of each piece of to-be-corrected data in the to-be-corrected data group.

[0044] Specifically, the to-be-corrected data are sorted according to the occurrence and collection sequence of each to-be-corrected data in the to-be-corrected data set, and a correction sequence is obtained. According to the correction sequence, the to-be-corrected data located at the first position is taken as the current to-be-corrected data to be corrected first. According to the current to-be-corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yielding scene corresponding to the current to-be-corrected data, whether a reasonable acceleration of the current obstacle can be solved is judged through kinematic model analysis and calculation, that is, whether there is a motion intention corresponding to the current obstacle. If there is, it indicates that the motion intention corresponding to the current obstacle can be corrected, therefore, the motion intention corresponding to the current obstacle is determined through kinematic model analysis and calculation, and whether the current to-be-corrected data is the to-be-corrected data located at the last position in the correction sequence is judged, so as to determine whether to stop correcting the current to-be-corrected data set. In the case that the current to-be-corrected data is the to-be-corrected data located at the last position in the correction sequence, it indicates that each to-be-corrected data in the current to-be-corrected data set has been corrected, therefore, the motion intention of each obstacle can be determined, that is, the acceleration of the obstacle in each to-be-corrected data in the current to-be-corrected data set is obtained. In the case that the current to-be-corrected data is not the to-be-corrected data located at the last position in the correction sequence, it indicates that there are other to-be-corrected data after the current to-be-corrected data in the to-be-corrected data set that need to be corrected, therefore, according to the current to-be-corrected data, the current yielding relationship and the current yielding scene, the related data of the current vehicle is corrected, and the to-be-corrected data located at the next position of the current to-be-corrected data in the correction sequence is updated, the updated to-be-corrected data located at the next position of the current to-be-corrected data in the correction sequence is taken as the new current to-be-corrected data, and the operation of judging whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yielding scene corresponding to the current to-be-corrected data is executed again, so as to correct the new current to-be-corrected data. If there is not, it indicates that the motion intention corresponding to the current obstacle cannot be corrected, therefore, there is a possibility that the other to-be-corrected data before the current to-be-corrected data is corrected incorrectly, which leads to the current correction. Further, the related data of the to-be-corrected data located at the position before the current to-be-corrected data in the correction sequence is adjusted, for example, the acceleration can be increased or decreased within a limited range, the updated to-be-corrected data located at the position before the current to-be-corrected data in the correction sequence is taken as the new current to-be-corrected data, and the operation of judging whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yielding scene corresponding to the current to-be-corrected data is executed again, so as to re-correct the to-be-corrected data located at the position before the current to-be-corrected data in the correction sequence.

[0045] The above execution logic is the execution logic of the reverse search tree, such as... Figure 2 The reverse search tree is shown. After obtaining the pre-labeled yielding relationships between the current vehicle (ego) and each obstacle (obj1, obj2, and obj3), a reverse tree search strategy is used. First, obj1 is calculated. Assuming the pre-labeled yielding relationship between ego and obj1 is ego yielding to obj1, it is determined whether the current yielding scenario between obj1 and ego is a cross or merge scenario. Once the current yielding relationship and scenario are confirmed, different kinematic models are used to determine whether the motion intention of obj1 can be obtained. When the motion intention of obj1 is obtained, the state of ego is updated to the decision point of obj1 (the data to be corrected is located next to the currently corrected data). Based on this, the above steps are then used to determine the intention of obj2. When the motion intention of obj2 is obtained, the state of ego is updated to the decision point of obj2, and so on, to obtain the motion intentions of subsequent obstacles. If the movement intention of any obstacle cannot be obtained, the system returns to the data from when it competed for right-of-way with the previous obstacle (the data to be corrected is located one position above the current data in the correction order) to recalculate the movement intention of each obstacle.

[0046] Based on the above example, since the current right-of-way relationship includes the current vehicle overtaking the current obstacle and the current vehicle yielding to the current obstacle, and the current right-of-way scenario is a crossing scenario and a merging scenario, if the motion intention corresponding to the current obstacle is the current obstacle's acceleration, there are four cases. Next, we will explain how to determine whether there is a motion intention corresponding to the current obstacle based on the current correction data, the pre-labeled current right-of-way relationship between the current vehicle and the current obstacle in the current correction data, and the current right-of-way scenario corresponding to the current correction data in each of the four cases:

[0047] Scenario 1, such as Figure 3 As shown, ego represents the current vehicle, obj represents the current obstacle, the bounding box represents the position of the current vehicle and the current obstacle corresponding to the current correction data, and the dashed box represents the position of the current vehicle and the current obstacle after the current right-of-way competition ends. If the current right-of-way relationship is that the current vehicle overtakes the current obstacle, and the current right-of-way scenario is an intersection scenario, then the following method is used to determine whether there is a motion intention corresponding to the current obstacle:

[0048] When the current right-of-way relationship is that the current vehicle is overtaking the current obstacle, and the current right-of-way scenario is an intersection scenario, the maximum acceleration time of the current vehicle is determined based on the current vehicle speed, the current vehicle maximum acceleration, and the current vehicle maximum speed.

[0049] determine the current vehicle leaving the interaction region time according to the current vehicle leaving the interaction region distance, the current vehicle speed and the current vehicle maximum acceleration time;

[0050] determine whether the current obstacle acceleration exists according to the current vehicle leaving the interaction region time, the current obstacle speed and the current obstacle entering the interaction region distance.

[0051] The interaction region is a region in which the current vehicle and the current obstacle have a possibility of collision in the yielding stage. The current correction data includes the current vehicle speed, the current vehicle entering the interaction region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering the interaction region distance, the current obstacle maximum speed and the motion intention corresponding to the current obstacle. The current vehicle maximum acceleration time is the time required for the current vehicle to accelerate from the current vehicle speed to the current vehicle maximum speed at the current vehicle maximum acceleration. The current vehicle leaving the interaction region distance is the distance between the position of the current vehicle and the position corresponding to the current vehicle leaving the interaction region. The current vehicle leaving the interaction region time is the time between the time of the current correction data and the time of the current vehicle passing through. The current obstacle entering the interaction region distance is the distance between the position of the current obstacle and the position corresponding to the current obstacle entering the interaction region.

[0052] Specifically, in the case that the current yielding relationship is the current vehicle yielding the current obstacle and the current yielding scene is the intersection scene, the current vehicle maximum acceleration time is obtained by dividing the difference between the current vehicle maximum speed and the current vehicle speed by the current vehicle maximum acceleration. The current vehicle leaving the interaction region time can be calculated by the kinematic formula combined with the current vehicle leaving the interaction region distance, the current vehicle speed and the current vehicle maximum acceleration time. Further, the current obstacle acceleration can be determined by the kinematic formula combined with the current vehicle leaving the interaction region time, the current obstacle speed and the current obstacle entering the interaction region distance.

[0053] In the case, the current obstacle acceleration can be determined by the following kinematic formulas:

[0054] v_ego+t_max×ego_max_acc=v_ego_max

[0055] Wherein, v_ego is the current vehicle speed, ego_max_acc is the current vehicle maximum acceleration, v_ego_max is the current vehicle maximum speed, and t_max is the current vehicle maximum acceleration time.

[0056]

[0057]

[0058] wherein v_ego is the current vehicle speed, ego_max_acc is the current vehicle maximum acceleration, v_ego_max is the current vehicle maximum speed, t_max is the current vehicle maximum acceleration time, t1 is the current vehicle leaving interaction area time, and ego_leave_s is the current vehicle leaving interaction area distance.

[0059]

[0060] wherein t1 is the current vehicle leaving interaction area time, v_obj is the current obstacle speed, obj_enter_s is the current obstacle entering interaction area distance, and obj_acc is the current obstacle acceleration.

[0061] Case two, as shown in FIG. 2B, ego represents the current vehicle, obj represents the current obstacle, the implementation block represents the position of the current vehicle and the current obstacle corresponding to the current correction data, and the dashed block represents the position of the current vehicle and the current obstacle after the current yielding is over. If the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scene is the intersection scene, then whether there is a motion intention corresponding to the current obstacle is determined by the following manner: Figure 4 In the case that the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scene is the intersection scene, the current vehicle maximum deceleration time is determined according to the current vehicle speed and the current vehicle minimum acceleration.

[0062] The current vehicle entering interaction area time or parking time is determined according to the current vehicle maximum deceleration time, the current vehicle speed, the current vehicle minimum acceleration, and the current vehicle entering interaction area distance.

[0063] The current obstacle maximum acceleration time is determined according to the current obstacle speed, the current obstacle maximum acceleration, and the current obstacle maximum speed.

[0064] Whether there is the current obstacle acceleration is determined according to the current vehicle entering interaction area time or parking time, the current obstacle maximum acceleration time, the current obstacle speed, and the current obstacle leaving interaction area distance.

[0065]

[0066] ​The current vehicle maximum deceleration time is the time required for the current vehicle to decelerate from the current vehicle speed at the current vehicle minimum acceleration to 0. The current vehicle entering interaction region distance is the distance between the position of the current vehicle and the position of the corresponding entering interaction region of the current vehicle. If the current vehicle decelerates at the minimum acceleration and can reach 0 before entering the interaction region, the time of decelerating to 0 is the parking time; if the current vehicle decelerates at the minimum acceleration and can still be greater than 0 before entering the interaction region, the time of the current vehicle traveling the current vehicle entering interaction region distance is the current vehicle entering interaction region time. The current obstacle maximum acceleration time is the time required for the current obstacle to accelerate from the current obstacle speed at the current obstacle maximum acceleration to the current obstacle maximum speed. The current obstacle leaving interaction region distance is the distance between the position of the current obstacle and the position of the corresponding leaving interaction region of the current obstacle.

[0067] Specifically, in the case that the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is the intersection scenario, the current vehicle maximum deceleration time is obtained by dividing the difference between 0 and the current vehicle speed by the current vehicle minimum acceleration. According to the current vehicle speed and the current vehicle minimum acceleration, the current vehicle parking distance is determined by the kinematic formula. If the parking distance is less than or equal to the current vehicle entering interaction region distance, the parking time, i.e., the current vehicle maximum deceleration time, is determined. If the parking distance is greater than the current vehicle entering interaction region distance, the current vehicle entering interaction region time is determined, and the current vehicle entering interaction region time can be calculated by the kinematic formula in combination with the current vehicle speed, the current vehicle minimum acceleration, and the current vehicle entering interaction region distance. The current obstacle maximum acceleration time is obtained by dividing the difference between the current obstacle maximum speed and the current obstacle speed by the current obstacle maximum acceleration. Further, the current obstacle acceleration is determined by the kinematic formula in combination with the current obstacle leaving interaction region distance, the current obstacle speed, the current obstacle maximum acceleration time, and the current vehicle entering interaction region time or the parking time.

[0068] In case two, the current obstacle acceleration can be determined by the following kinematic formulas:

[0069] v_ego+t_still×ego_min_acc=0

[0070]

[0071]

[0072] Wherein, v_ego is the current vehicle speed, ego_min_acc is the current vehicle minimum acceleration, t_still is the current vehicle maximum deceleration time, i.e. parking time, t2 is the current vehicle entering the interaction area time or parking time, ego_enter_s is the current vehicle entering the interaction area distance, and ego_still is the current vehicle parking distance.

[0073] v_obj+t_max*obj_acc=v_obj_max

[0074]

[0075]

[0076] Wherein, v_obj is the current obstacle speed, v_obj_max is the current obstacle maximum speed, t_max is the current obstacle maximum acceleration time, t2 is the current vehicle entering the interaction area time or parking time, obj_leave_s is the current obstacle leaving the interaction area distance, and obj_acc is the current obstacle acceleration.

[0077] Case three, as shown in Figure 5 , ego represents the current vehicle, obj represents the current obstacle, the realization block represents the position of the current vehicle and the current obstacle corresponding to the current correction data, the dashed box of obj represents the position of the current obstacle entering the interaction area, the dashed box of ego in the middle represents the position of the current vehicle entering the interaction area, and the dashed box of ego on the right represents the position of the current vehicle leaving the preset safety distance, if the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scene is the merging scene, then whether there is a motion intention corresponding to the current obstacle is determined by the following manner: Figure 5 Figure 5 In the case that the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scene is the merging scene, the current vehicle maximum acceleration time is determined according to the current vehicle speed, the current vehicle maximum acceleration and the current vehicle maximum speed;

[0078] The current vehicle leaving the preset safety distance time is determined according to the current vehicle leaving the interaction area distance, the preset safety distance, the current vehicle speed and the current vehicle maximum acceleration time;

[0079] The current vehicle leaving the preset safety distance time is determined according to the current vehicle leaving the interaction area distance, the preset safety distance, the current vehicle speed and the current vehicle maximum acceleration time;

[0080] The current vehicle leaving the preset safety distance time is determined according to the current vehicle leaving the interaction area distance, the preset safety distance, the current vehicle speed and the current vehicle maximum acceleration time;​

[0081] In the presence of the current obstacle acceleration, the current obstacle speed, the current vehicle speed, the current vehicle maximum acceleration and the current obstacle entering the interaction region time, the current obstacle entering the interaction region speed and the current vehicle leaving the preset safe distance speed are determined;

[0082] According to the current obstacle entering the interaction region speed, the current obstacle acceleration and the current vehicle leaving the preset safe distance speed, the current vehicle leaving the preset safe distance acceleration range is updated.

[0083] The preset safe distance is a distance preset to ensure safe passing through the interaction region in the intersection scenario. The current vehicle leaving the preset safe distance time is the time between the time of the current correction data and the time of the current vehicle passing through and driving the preset safe distance. The current obstacle speed interval is an interval composed of the lower limit speed and the upper limit speed of the current obstacle, and the upper limit speed and the lower limit speed are boundary speeds. In general, the upper limit speed is the maximum speed of the current obstacle, and the lower limit speed is 0. Optionally, the difference between the upper limit speed or the upper limit speed and the preset speed error can be taken as the first boundary speed, and the sum of the lower limit speed or the lower limit speed and the preset speed error can be taken as the second boundary speed. The current obstacle entering the interaction region speed is the driving speed when the current obstacle drives to the position corresponding to the entering interaction region of the current obstacle. The current vehicle leaving the preset safe distance speed is the driving speed when the current vehicle drives to the position leaving the preset safe distance. The current vehicle leaving the preset safe distance acceleration range is the acceleration range for limiting the driving speed adjustment after the current vehicle leaves the preset safe distance.

[0084] Specifically, in a case where the current take-and-yield relationship is that the current vehicle takes the current obstacle and the current take-and-yield scene is a merging scene, a current vehicle maximum acceleration time is obtained by dividing a difference between a current vehicle maximum speed and a current vehicle speed by a current vehicle maximum acceleration. A current vehicle time of leaving the preset safety distance is calculated by a kinematic formula in combination with a sum of a current vehicle distance of leaving the interaction region and the preset safety distance, the current vehicle speed, and the current vehicle maximum acceleration time. The current vehicle time of leaving the preset safety distance is taken as the current obstacle entering time of the interaction region. Further, a current obstacle acceleration is determined by the kinematic formula in combination with the current obstacle entering time of the interaction region, a boundary speed, and a current obstacle distance of entering the interaction region. In a case where the current obstacle acceleration exists, since the current obstacle will follow the current vehicle after passing through the interaction region, the current vehicle acceleration cannot be too small, and thus the current vehicle acceleration range of leaving the preset safety distance needs to be updated. A current obstacle entering speed of the interaction region is calculated by the kinematic formula according to the current obstacle acceleration, a current obstacle speed, and the current obstacle entering time of the interaction region. A current vehicle speed of leaving the preset safety distance is calculated according to the current vehicle speed, the current vehicle maximum acceleration, and the current obstacle entering time of the interaction region. Further, a minimum acceleration value of the current vehicle after leaving the preset safety distance is obtained according to the current obstacle entering speed of the interaction region, the current obstacle acceleration, and the current vehicle speed of leaving the preset safety distance, and is used to update the current vehicle acceleration range of leaving the preset safety distance.

[0085] In case three, the current obstacle acceleration can be specifically determined by the following kinematic formulas:

[0086] v_ego+t_max×ego_max_acc=v_ego_max

[0087]

[0088]

[0089] obj_enter_t=t3

[0090] wherein v_ego is the current vehicle speed, ego_max_acc is the current vehicle maximum acceleration, v_ego_max is the current vehicle maximum speed, t_max is the current vehicle maximum acceleration time, t3 is the current vehicle time of leaving the preset safety distance, ego_leave_s is the current vehicle distance of leaving the interaction region, safe_dist is the preset safety distance, and obj_enter_t is the current obstacle entering time of the interaction region.

[0091] Assume that the current obstacle acceleration is greater than 0:

[0092] v_obj + t_max * obj_acc = v_obj_max

[0093]

[0094]

[0095] Wherein, v_obj is the current obstacle speed, v_obj_max is the current obstacle maximum speed (first boundary speed), t_max is the current obstacle maximum acceleration time, obj_enter_t is the current obstacle entering the interaction region time, obj_enter_s is the current obstacle entering the interaction region distance, and obj_acc is the current obstacle acceleration.

[0096] If obj_acc is calculated, if obj_acc is greater than 0, the assumption is met, if obj_acc is less than 0 or no solution, the assumption is not met, and the current obstacle acceleration is less than or equal to 0.

[0097] Assume that the current obstacle acceleration is less than or equal to 0:

[0098] v_obj + t_still * obj_acc = 0

[0099]

[0100]

[0101] Wherein, v_obj is the current obstacle speed, 0 is the current obstacle minimum speed (second boundary speed), t_still is the current obstacle maximum deceleration time, obj_enter_t is the current obstacle entering the interaction region time, obj_enter_s is the current obstacle entering the interaction region distance, and obj_acc is the current obstacle acceleration.

[0102] Under case three, the acceleration range of the current vehicle leaving the preset safety distance can be updated through the following kinematic formulas:

[0103] obj_enter_v + t_obj_still * obj_acc = 0

[0104]

[0105] Wherein, obj_enter_v is the speed at which the current obstacle enters the interaction area, obj_acc is the acceleration of the current obstacle, that is, the acceleration of the current obstacle in the subsequent following phase, which is less than 0, t_obj_still is the maximum deceleration time of the current obstacle after entering the interaction area, obj_enter_v is the speed at which the current obstacle enters the interaction area, ego_safedist_v is the speed at which the current vehicle leaves the preset safe distance, and ego_acc is the minimum acceleration of the current vehicle after leaving the preset safe distance.

[0106] Based on the above inequality, the minimum acceleration that the current vehicle needs to maintain after it has rushed past the preset safe distance can be calculated, so as to update the acceleration range of the current vehicle when it leaves the preset safe distance.

[0107] Scenario 4, such as Figure 6 As shown, ego represents the current vehicle, obj represents the current obstacle, the bounding box represents the position of the current vehicle and the current obstacle corresponding to the current correction data, and the dashed box around ego represents the position where the current vehicle enters the interaction area. Figure 6 The dashed box in the middle of the object indicates the current position of the obstacle entering the interaction area. Figure 6 The dashed box on the right side of the object indicates the position of the obstacle away from the preset safe distance. If the current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is a merging scenario, then the following method is used to determine whether there is a movement intention corresponding to the current obstacle:

[0108] Given that the current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is a merging scenario, determine whether the current vehicle's minimum acceleration is a positive number;

[0109] If so, determine the minimum acceleration time of the current vehicle based on the current vehicle speed, the current vehicle minimum acceleration, and the current vehicle maximum speed, and determine the time for the current vehicle to enter the interaction area based on the current vehicle minimum acceleration time, the current vehicle minimum acceleration, the current vehicle speed, and the distance the current vehicle has traveled into the interaction area.

[0110] If not, determine the maximum deceleration time of the current vehicle based on the current vehicle speed and the current vehicle minimum acceleration, and determine the time for the current vehicle to enter the interaction area based on the current vehicle maximum deceleration time, the current vehicle minimum acceleration, the current vehicle speed, and the distance the current vehicle has traveled into the interaction area.

[0111] Based on the current vehicle's entry time into the interaction area, the current obstacle's speed, the current obstacle's distance from the interaction area, and the preset safety distance, determine whether there is current obstacle acceleration;

[0112] In the presence of the current obstacle acceleration, according to the current obstacle acceleration, the current obstacle speed, the current obstacle entering the interaction region distance and the preset safety distance, the speed of the current obstacle leaving the preset safety distance is determined, and according to the current vehicle speed, the current vehicle minimum acceleration and the current vehicle entering the interaction region time, the current vehicle entering the interaction region speed is determined.

[0113] In the case where the current vehicle entering the interaction region speed is greater than the speed of the current obstacle leaving the preset safety distance, according to the current vehicle entering the interaction region speed, the speed of the current obstacle leaving the preset safety distance and the current obstacle acceleration, the acceleration range of the current vehicle entering the interaction region is updated.

[0114] Among them, the current vehicle minimum acceleration time is the acceleration time when the current vehicle travels at the minimum acceleration to the maximum speed of the current vehicle. The current vehicle maximum deceleration time is the deceleration time when the current vehicle travels at the minimum acceleration (maximum deceleration) to the speed of 0. The current vehicle entering the interaction region speed is the running speed when the current vehicle travels to the position of entering the interaction region. The speed of the current obstacle leaving the preset safety distance is the running speed when the current obstacle travels to the position of leaving the preset safety distance. The current vehicle entering the interaction region acceleration range is the acceleration range for limiting the adjustment of the running speed after the current vehicle enters the interaction region.

[0115] Specifically, in the case that the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is the merging scenario, it is determined whether the minimum acceleration of the current vehicle is a positive number. If yes, it indicates that the current vehicle needs to yield to the current obstacle, but is still in an accelerating state, and thus the minimum acceleration time of the current vehicle is obtained by dividing the difference between the maximum speed of the current vehicle and the speed of the current vehicle by the minimum acceleration of the current vehicle. Further, the time at which the current vehicle enters the interaction region is calculated by a kinematic formula in combination with the minimum acceleration time of the current vehicle, the minimum acceleration of the current vehicle, the speed of the current vehicle and the distance at which the current vehicle enters the interaction region. If no, it indicates that the current vehicle can decelerate to yield, and thus the maximum deceleration time of the current vehicle is obtained by dividing the difference between 0 and the speed of the current vehicle by the minimum acceleration of the current vehicle. The time at which the current vehicle enters the interaction region is determined by a kinematic formula in combination with the maximum deceleration time of the current vehicle, the minimum acceleration of the current vehicle, the speed of the current vehicle and the distance at which the current vehicle enters the interaction region. After obtaining the time at which the current vehicle enters the interaction region, it is determined whether there is an acceleration of the current obstacle by kinematic analysis and calculation in combination with the speed of the current obstacle, the distance at which the current obstacle leaves the interaction region and the preset safety distance. In the case that there is an acceleration of the current obstacle, the acceleration of the current vehicle cannot be too large because the current vehicle will follow the current obstacle after passing through the interaction region, and thus the acceleration range of the current vehicle entering the interaction region needs to be updated. The speed at which the current obstacle leaves the preset safety distance is calculated by a kinematic formula in combination with the acceleration of the current obstacle, the speed of the current obstacle, the distance at which the current obstacle enters the interaction region and the preset safety distance, and the speed at which the current vehicle enters the interaction region is calculated in combination with the speed of the current vehicle, the minimum acceleration of the current vehicle and the time at which the current vehicle enters the interaction region. In the case that the speed at which the current vehicle enters the interaction region is not greater than the speed at which the current obstacle leaves the preset safety distance, the acceleration range of the current vehicle entering the interaction region does not need to be limited. Further, in the case that the speed at which the current vehicle enters the interaction region is greater than the speed at which the current obstacle leaves the preset safety distance, the maximum acceleration value of the current vehicle after entering the interaction region can be obtained in combination with the speed at which the current vehicle enters the interaction region, the speed at which the current obstacle leaves the preset safety distance and the acceleration of the current obstacle, and is used to update the acceleration range of the current vehicle entering the interaction region.

[0116] In case four, the acceleration of the current obstacle can be determined by the following kinematic formulas:

[0117] If the minimum acceleration of the current vehicle is a positive number, i.e., ego_min_acc>0, then:

[0118] v_ego+t_min×ego_min_acc=v_ego_max

[0119]

[0120]

[0121] Where v_ego is the current vehicle speed, ego_min_acc is the current vehicle minimum acceleration, v_ego_max is the current vehicle maximum speed, t_min is the current vehicle minimum acceleration time, t4 is the time the current vehicle enters the interaction area, and ego_enter_s is the distance the current vehicle leaves the interaction area.

[0122] If the current minimum acceleration of the vehicle is not positive, i.e., ego_min_acc≤0, then:

[0123] v_ego + t_still × ego_min_ac = 0

[0124]

[0125]

[0126] Where v_ego is the current vehicle speed, ego_min_acc is the current vehicle minimum acceleration, t_still is the current vehicle maximum deceleration time, t4 is the time the current vehicle enters the interaction area, and ego_enter_s is the distance the current vehicle travels into the interaction area.

[0127]

[0128] Where t4 is the time when the current vehicle enters the interaction area, v_obj is the current obstacle speed, obj_leave_s is the current obstacle distance away from the interaction area, safe_dist is the preset safe distance, and obj_acc is the current obstacle acceleration.

[0129] In scenario four, the acceleration range of the vehicle entering the interaction zone can be updated using the following kinematic formulas:

[0130] Assuming that the current obstacle's acceleration remains constant after leaving the intersection area, and that the current obstacle's speed at which it leaves the preset safe distance is obj_safedist_v, and the current vehicle's speed at which it enters the intersection area is ego_enter_v.

[0131] If ego_enter_v ≤ obj_safedist_v, then there is no restriction on the acceleration range of the current vehicle entering the interaction area.

[0132] If ego_enter_v > obj_safedist_v, then:

[0133] ego_enter_v + t_same * ego_max_acc = obj_enter_v

[0134]

[0135] Wherein, the obj_safedist_v is the speed of the current obstacle leaving the preset safety distance, the ego_enter_v is the speed of the current vehicle entering the intersection region, the t_same is the time of the current vehicle and the current obstacle reaching the same speed, the obj_acc is the acceleration of the current obstacle, and the ego_max_acc is the maximum acceleration of the current vehicle after yielding to enter the intersection region.

[0136] According to the above formula, the maximum acceleration that can be executed after the current vehicle yields to enter the intersection region can be calculated to update the acceleration range of the current vehicle entering the intersection region.

[0137] S130, update the obstacle prediction basic data set based on the first data set and the updated second data set, and train the obstacle motion intention initial prediction model based on the updated obstacle prediction basic data set to obtain an obstacle motion intention target prediction model.

[0138] Wherein, the obstacle motion intention initial prediction model can be a model trained by the obstacle prediction basic data set before updating, or an untrained model. The obstacle motion intention target prediction model is a model trained by the updated obstacle prediction basic data set.

[0139] Specifically, the first data set and the updated second data set are combined to obtain a new obstacle prediction basic data set. Further, the new obstacle prediction basic data set is used to train the obstacle motion intention initial prediction model, and the trained model is used as the obstacle motion intention target prediction model for subsequent deployment on a real vehicle to obtain the obstacle motion intention.

[0140] Based on the above example, the obstacle motion intention initial prediction model can be trained based on the updated obstacle prediction basic data set to obtain the obstacle motion intention target prediction model in the following way:

[0141] The current vehicle speed, the current vehicle entering the intersection region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering the intersection region distance and the current obstacle maximum speed of each data in the updated obstacle prediction basic data set are used as the input of the obstacle motion intention initial prediction model;

[0142] the motion intention corresponding to the current obstacle in each piece of data in the updated obstacle prediction basic data set as the output of the initial prediction model of the obstacle motion intention;

[0143] training the initial prediction model of the obstacle motion intention based on the input and the output of the initial prediction model of the obstacle motion intention, to obtain a target prediction model of the obstacle motion intention.

[0144] Specifically, when training the model, the current vehicle speed, the current vehicle entering interaction region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering interaction region distance, and the current obstacle maximum speed of each piece of data in the updated obstacle prediction basic data set are used as the input of the initial prediction model of the obstacle motion intention, and the motion intention corresponding to the current obstacle in each piece of data in the updated obstacle prediction basic data set is used as the output of the initial prediction model of the obstacle motion intention, so as to train the initial prediction model of the obstacle motion intention, and model verification can also be performed. Furthermore, the model obtained after training is used as the target prediction model of the obstacle motion intention, so as to obtain a model that can output the motion intention corresponding to the current obstacle by inputting the current vehicle speed, the current vehicle entering interaction region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering interaction region distance, and the current obstacle maximum speed.

[0145] The obstacle motion intention prediction model acquisition method provided in this embodiment is used to acquire an obstacle prediction basic data set, determine a first data set and a second data set in the obstacle prediction basic data set, so as to subsequently correct the unreasonable data of the current vehicle yielding, and then, for each to-be-corrected data group in the second data set, the motion intention of each obstacle is determined according to the to-be-corrected data group, the pre-labeled yielding relationship between the current vehicle and each obstacle in the to-be-corrected data group, and each yielding scene corresponding to the to-be-corrected data group, and the to-be-corrected data group is updated based on the motion intention of each obstacle to obtain an updated second data set, so as to correct the second data set, update the obstacle prediction basic data set based on the first data set and the updated second data set, train an initial prediction model of the obstacle motion intention based on the updated obstacle prediction basic data set, and obtain a target prediction model of the obstacle motion intention, so as to realize the correction of the obstacle prediction basic data set combined with the pre-labeled yielding relationship, reduce the complexity of the obstacle motion intention prediction model, and then improve the accuracy and real-time performance of acquiring the motion intention of the obstacle.

[0146] Figure 7 FIG. 1 is a structural schematic diagram of an obstacle motion intention prediction model acquisition device in the embodiments of the present disclosure. Figure 7As shown: the device includes: obstacle prediction basic data set acquisition module 710, data set update module 720 and obstacle motion intention target prediction model training module 730.

[0147] Among them, the obstacle prediction basic data set acquisition module 710 is used for acquiring the obstacle prediction basic data set and determining the first data set and the second data set in the obstacle prediction basic data set; wherein the first data set is the data in the obstacle prediction basic data set that is reasonable for the current vehicle to yield, and the second data set is the data in the obstacle prediction basic data set that is unreasonable for the current vehicle to yield; the data set update module 720 is used for determining the motion intention of each obstacle according to each to-be-corrected data group, the pre-labeled yield relationship between the current vehicle and each obstacle in the to-be-corrected data group, and each yield scene corresponding to the to-be-corrected data group, and updating the to-be-corrected data group based on the motion intention of each obstacle to obtain an updated second data set; the obstacle motion intention target prediction model training module 730 is used for updating the obstacle prediction basic data set based on the first data set and the updated second data set, and training an obstacle motion intention initial prediction model based on the updated obstacle prediction basic data set to obtain an obstacle motion intention target prediction model.

[0148] On the basis of the above examples, optionally, the to-be-corrected data set includes at least one to-be-corrected data, and each to-be-corrected data corresponds to one obstacle of the current vehicle; the data set updating module 720 is further configured to determine a correction order of each to-be-corrected data in the to-be-corrected data set, take to-be-corrected data in a first position of the correction order as a current to-be-corrected data, determine whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, a current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and a current yielding scene corresponding to the current to-be-corrected data, if there is, determine the motion intention corresponding to the current obstacle, determine whether the current to-be-corrected data is to-be-corrected data in a last position of the correction order, in the case that the current to-be-corrected data is to-be-corrected data in the last position of the correction order, determine the motion intention of each obstacle, in the case that the current to-be-corrected data is not to-be-corrected data in the last position of the correction order, update to-be-corrected data in a next position of the current to-be-corrected data according to the current to-be-corrected data, the current yielding relationship, and the current yielding scene, take the updated to-be-corrected data in the next position of the current to-be-corrected data as a new current to-be-corrected data, and return to perform the operation of determining whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yielding scene corresponding to the current to-be-corrected data, if not, update to-be-corrected data in a previous position of the current to-be-corrected data, take the updated to-be-corrected data in the previous position of the current to-be-corrected data as a new current to-be-corrected data, and return to perform the operation of determining whether there is a motion intention corresponding to the current obstacle according to the current to-be-corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-labeled in the current to-be-corrected data, and the current yielding scene corresponding to the current to-be-corrected data.

[0149] On the basis of the above examples, optionally, the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scenario is a crossing scenario; the motion intention corresponding to the current obstacle is a current obstacle acceleration; the dataset updating module 720 is further configured to, in a case where the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is a crossing scenario, determine a current vehicle maximum acceleration time according to a current vehicle speed, a current vehicle maximum acceleration, and a current vehicle maximum speed; determine a current vehicle leaving interactive region time according to a current vehicle leaving interactive region distance, the current vehicle speed, and the current vehicle maximum acceleration time; and determine whether the current obstacle acceleration exists according to the current vehicle leaving interactive region time, a current obstacle speed, and a current obstacle entering interactive region distance. The current correction data includes the current vehicle speed, the current vehicle entering interactive region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering interactive region distance, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

[0150] On the basis of the above examples, optionally, the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scenario is a crossing scenario; the motion intention corresponding to the current obstacle is a current obstacle acceleration; the dataset updating module 720 is further configured to, in a case where the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is a crossing scenario, determine a current vehicle maximum acceleration time according to a current vehicle speed, a current vehicle maximum acceleration, and a current vehicle maximum speed; determine a current vehicle leaving interactive region time according to a current vehicle leaving interactive region distance, the current vehicle speed, and the current vehicle maximum acceleration time; and determine whether the current obstacle acceleration exists according to the current vehicle leaving interactive region time, a current obstacle speed, and a current obstacle entering interactive region distance. The current correction data includes the current vehicle speed, the current vehicle entering interactive region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering interactive region distance, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

[0151] On the basis of the above examples, optionally, the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scenario is a merging scenario; the motion intention corresponding to the current obstacle is a current obstacle acceleration; the data set updating module 720 is further configured to, in the case where the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is a merging scenario, determine a current vehicle maximum acceleration time according to a current vehicle speed, a current vehicle maximum acceleration, and a current vehicle maximum speed; determine a time for the current vehicle to leave a preset safety distance according to a distance for the current vehicle to leave an interaction region, the preset safety distance, the current vehicle speed, and the current vehicle maximum acceleration time; take the time for the current vehicle to leave the preset safety distance as a current obstacle entering interaction region time, and determine whether there is a current obstacle acceleration according to the current obstacle entering interaction region time, a current obstacle speed, a boundary speed in a current obstacle speed interval, and a distance for the current obstacle to enter the interaction region; in the case where there is the current obstacle acceleration, determine a current obstacle entering interaction region speed and a speed for the current vehicle to leave the preset safety distance according to the current obstacle acceleration, the current obstacle speed, the current vehicle speed, the current vehicle maximum acceleration, and the current obstacle entering interaction region time; and update an acceleration range for the current vehicle to leave the preset safety distance according to the current obstacle entering interaction region speed, the current obstacle acceleration, and the speed for the current vehicle to leave the preset safety distance; wherein the current correction data includes the current vehicle speed, a distance for the current vehicle to enter the interaction region, the current vehicle maximum speed, the current obstacle speed, the distance for the current obstacle to enter the interaction region, a current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

[0152] On the basis of the above examples, optionally, the current yielding relationship is that the current vehicle yields to the current obstacle, and the current yielding scenario is a merging scenario; the motion intention corresponding to the current obstacle is a current obstacle acceleration; the data set updating module 720 is further configured to, in a case where the current yielding relationship is that the current vehicle yields to the current obstacle and the current yielding scenario is a merging scenario, determine whether the current vehicle minimum acceleration is positive; if yes, determine a current vehicle minimum acceleration time according to the current vehicle speed, the current vehicle minimum acceleration, and a current vehicle maximum speed, and determine a current vehicle entering interactive region time according to the current vehicle minimum acceleration time, the current vehicle minimum acceleration, the current vehicle speed, and a current vehicle entering interactive region distance; if no, determine a current vehicle maximum deceleration time according to the current vehicle speed and the current vehicle minimum acceleration, and determine the current vehicle entering interactive region time according to the current vehicle maximum deceleration time, the current vehicle minimum acceleration, the current vehicle speed, and the current vehicle entering interactive region distance; determine whether there is a current obstacle acceleration according to the current vehicle entering interactive region time, a current obstacle speed, a current obstacle leaving interactive region distance, and a preset safety distance; in a case where there is the current obstacle acceleration, determine a current obstacle speed at which the current obstacle leaves the preset safety distance according to the current obstacle acceleration, the current obstacle speed, a current obstacle entering interactive region distance, and the preset safety distance, and determine a current vehicle entering interactive region speed according to the current vehicle speed, the current vehicle minimum acceleration, and the current vehicle entering interactive region time; in a case where the current vehicle entering interactive region speed is greater than the current obstacle speed at which the current obstacle leaves the preset safety distance, update an acceleration range of the current vehicle entering the interactive region according to the current vehicle entering interactive region speed, the current obstacle speed at which the current obstacle leaves the preset safety distance, and the current obstacle acceleration; wherein the current correction data includes the current vehicle speed, the current vehicle entering interactive region distance, the current vehicle maximum speed, the current obstacle speed, the current obstacle entering interactive region distance, a current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

[0153] Based on the above example, optionally, the obstacle motion intention target prediction model training module 730 is further configured to use the current vehicle speed, current vehicle distance into the interaction area, current vehicle maximum speed, current obstacle speed, current obstacle distance into the interaction area, and current obstacle maximum speed of each data point in the updated obstacle prediction base dataset as input to the obstacle motion intention initial prediction model; use the motion intention corresponding to the current obstacle in each data point of the updated obstacle prediction base dataset as output to the obstacle motion intention initial prediction model; and train the obstacle motion intention initial prediction model based on the input and output of the obstacle motion intention initial prediction model to obtain the obstacle motion intention target prediction model.

[0154] The obstacle motion intention prediction model acquisition device provided in this embodiment can execute the steps in the obstacle motion intention prediction model acquisition method provided in this embodiment, and has the execution steps and beneficial effects, which will not be repeated here.

[0155] Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of this disclosure. See below for details. Figure 8 It shows a schematic diagram of a structure suitable for implementing the electronic device 800 in the embodiments of this disclosure. Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0156] like Figure 8 As shown, the electronic device 800 may include a processing device 801, a read-only memory (ROM) 802, a random access memory (RAM) 803, a bus 804, an input / output (I / O) interface 805, an input device 806, an output device 807, a storage device 808, and a communication device 809. The processing device (e.g., a central processing unit, a graphics processor, etc.) 801 can perform various appropriate actions and processes to implement the methods of the embodiments described herein, based on a program in the ROM 802 or a program loaded from the storage device 808 into the RAM 803. The RAM 803 also stores various programs and data required for the operation of the electronic device 800. The processing device 801, the ROM 802, and the RAM 803 are interconnected via the bus 804. The input / output (I / O) interface 805 is also connected to the bus 804.

[0157] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the methods illustrated by the flowcharts, thereby implementing the obstacle motion intention prediction model acquisition method as described above. In such embodiments, the computer program can be downloaded and installed from a network by the communication device 809, or installed from the storage device 808, or installed from the ROM 802. When the computer program is executed by the processing device 801, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are performed.

[0158] It should be noted that the computer-readable medium described above in the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium, for example, can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination thereof. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, the computer-readable signal medium can include a data signal propagated in a baseband or as a carrier wave in a propagated data signal, which bears computer-readable program code. Such a propagated data signal can take many forms, including but not limited to, an electromagnetic signal, an optical signal, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium that can be used to carry or store a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained in the computer-readable medium can be transmitted by any suitable medium, including but not limited to, a wire, an optical fiber, an RF (radio frequency) or the like, or any suitable combination thereof.

[0159] The computer-readable medium described above can be included in the electronic device described above; or can exist separately from the electronic device and be not assembled in the electronic device. The computer-readable medium described above carries one or more programs, which, when executed by the electronic device, cause the electronic device to:

[0160] Obtain an obstacle prediction base dataset and determine a first dataset and a second dataset within the obstacle prediction base dataset; wherein, the first dataset consists of data in the obstacle prediction base dataset where the current vehicle's right-of-way cutting is reasonable, and the second dataset consists of data in the obstacle prediction base dataset where the current vehicle's right-of-way cutting is unreasonable;

[0161] For each data group to be corrected in the second dataset, the movement intention of each obstacle is determined based on the data group to be corrected, the pre-labeled yielding relationship between the current vehicle and each obstacle in the data group to be corrected, and each yielding scenario corresponding to the data group to be corrected. The data group to be corrected is then updated based on the movement intention of each obstacle to obtain the updated second dataset.

[0162] Based on the first dataset and the updated second dataset, the obstacle prediction base dataset is updated, and the initial prediction model of obstacle movement intention is trained based on the updated obstacle prediction base dataset to obtain the obstacle movement intention target prediction model.

[0163] Optionally, when one or more of the above-described procedures are executed by the electronic device, the electronic device may also perform other steps described in the above embodiments.

[0164] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0165] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for obtaining an obstacle motion intention prediction model, characterized in that, The method includes: Obtain an obstacle prediction base dataset and determine a first dataset and a second dataset within the obstacle prediction base dataset; wherein, the first dataset consists of data in the obstacle prediction base dataset where the current vehicle's right-of-way cutting is reasonable, and the second dataset consists of data in the obstacle prediction base dataset where the current vehicle's right-of-way cutting is unreasonable; For each data group to be corrected in the second dataset, the movement intention of each obstacle is determined based on the data group to be corrected, the pre-labeled yielding relationship between the current vehicle and each obstacle in the data group to be corrected, and each yielding scenario corresponding to the data group to be corrected. The data group to be corrected is then updated based on the movement intention of each obstacle to obtain the updated second dataset. Based on the first dataset and the updated second dataset, the obstacle prediction base dataset is updated, and the initial prediction model of obstacle movement intention is trained based on the updated obstacle prediction base dataset to obtain the obstacle movement intention target prediction model. The data set to be corrected includes at least one piece of data to be corrected, and each piece of data to be corrected corresponds to an obstacle of the current vehicle. The step of determining the movement intention of each obstacle based on the data set to be corrected, the pre-marked yielding relationships between the current vehicle and each obstacle in the data set to be corrected, and the yielding scenarios corresponding to the data set to be corrected includes: Determine the correction order for each piece of data to be corrected in the data group to be corrected; According to the correction order, each piece of data to be corrected in the data group to be corrected is processed sequentially. Based on the data to be corrected, the pre-marked right-of-way relationship between the current vehicle and the obstacle in the data to be corrected, and the right-of-way scenario corresponding to the data to be corrected, the movement intention of the obstacle corresponding to each piece of data to be corrected is determined. After determining the last piece of data to be corrected in the correction order, the movement intention of each obstacle is obtained.

2. The method according to claim 1, characterized in that, The process involves sequentially addressing each piece of data to be corrected in the data group according to the correction order. Based on the data to be corrected, the pre-marked yielding relationship between the current vehicle and the obstacle in the data to be corrected, and the yielding scenario corresponding to the data to be corrected, the movement intention of the obstacle corresponding to each piece of data to be corrected is determined. After determining the last piece of data to be corrected in the correction order, the movement intention of each obstacle is obtained, including: The data to be corrected that is first in the correction order is taken as the current correction data; Based on the current correction data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current correction data, and the current yielding scenario corresponding to the current correction data, it is determined whether there is a movement intention corresponding to the current obstacle; If it exists, determine the motion intention corresponding to the current obstacle, and determine whether the current correction data is the last data to be corrected in the correction order; If the current correction data is the last data to be corrected in the correction order, determine the movement intention of each obstacle; If the current corrected data is not the last corrected data in the correction order, the corrected data next in the correction order is updated according to the current corrected data, the current yielding relationship, and the current yielding scenario. The updated corrected data next in the correction order is then used as the new current corrected data. The process then returns to the operation of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle as pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data. If it does not exist, then update the data to be corrected that is one position above the current corrected data in the correction order, and use the updated data to be corrected that is one position above the current corrected data as the new current corrected data. Then return to execute the operation of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data.

3. The method according to claim 2, characterized in that, The current right-of-way relationship is that the current vehicle cuts off the current obstacle, and the current right-of-way scenario is an intersection scenario; The motion intention corresponding to the current obstacle is the current obstacle acceleration; The step of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data includes: When the current right-of-way relationship is that the current vehicle is overtaking the current obstacle, and the current right-of-way scenario is an intersection scenario, the maximum acceleration time of the current vehicle is determined based on the current vehicle speed, the current vehicle maximum acceleration, and the current vehicle maximum speed. The time it takes for the vehicle to leave the interaction area is determined based on the distance the vehicle has traveled from the interaction area, the vehicle's current speed, and the vehicle's maximum acceleration time. Based on the time the vehicle leaves the interaction area, the speed of the obstacle, and the distance the obstacle travels into the interaction area, determine whether there is acceleration in the obstacle. The current correction data includes the current vehicle speed, the current distance the vehicle has entered the interaction area, the current vehicle maximum speed, the current obstacle speed, the current obstacle distance it has entered the interaction area, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

4. The method according to claim 2, characterized in that, The current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is an intersection scenario; The motion intention corresponding to the current obstacle is the current obstacle acceleration; The step of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data includes: When the current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is an intersection scenario, the maximum deceleration time of the current vehicle is determined based on the current vehicle speed and the current vehicle minimum acceleration. The current vehicle's entry time into the interaction area or its stopping time is determined based on the current vehicle's maximum deceleration time, current vehicle speed, current vehicle minimum acceleration, and current vehicle's distance into the interaction area. Determine the maximum acceleration time of the current obstacle based on its current speed, maximum acceleration, and maximum speed. Based on the current vehicle's entry time into the interaction area or parking time, the current obstacle's maximum acceleration time, the current obstacle's speed, and the current obstacle's distance from the interaction area, determine whether there is current obstacle acceleration; The current correction data includes the current vehicle speed, the current distance the vehicle has entered the interaction area, the current vehicle maximum speed, the current obstacle speed, the current obstacle distance it has entered the interaction area, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

5. The method according to claim 2, characterized in that, The current right-of-way relationship is that the current vehicle overtakes the current obstacle, and the current right-of-way scenario is a merging scenario; The motion intention corresponding to the current obstacle is the current obstacle acceleration; The step of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data includes: When the current right-of-way relationship is that the current vehicle is overtaking the current obstacle, and the current right-of-way scenario is a merging scenario, the maximum acceleration time of the current vehicle is determined based on the current vehicle speed, the current vehicle maximum acceleration, and the current vehicle maximum speed. The time it takes for the current vehicle to leave the preset safe distance is determined based on the current vehicle's distance from the interaction area, the preset safe distance, the current vehicle's speed, and the current vehicle's maximum acceleration time. The time when the current vehicle leaves the preset safe distance is taken as the time when the current obstacle enters the interaction area. Based on the time when the current obstacle enters the interaction area, the current obstacle speed, the boundary speed in the current obstacle speed range, and the distance when the current obstacle enters the interaction area, it is determined whether there is current obstacle acceleration. In the presence of the current obstacle acceleration, the speed at which the current obstacle enters the interaction area and the speed at which the current vehicle leaves the preset safe distance are determined based on the current obstacle acceleration, the current obstacle speed, the current vehicle speed, the current vehicle maximum acceleration, and the time it takes for the current obstacle to enter the interaction area. Update the acceleration range of the current vehicle leaving the preset safe distance based on the speed at which the current obstacle enters the interaction area, the acceleration of the current obstacle, and the speed at which the current vehicle leaves the preset safe distance; The current correction data includes the current vehicle speed, the current distance the vehicle has entered the interaction area, the current vehicle maximum speed, the current obstacle speed, the current obstacle distance it has entered the interaction area, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

6. The method according to claim 2, characterized in that, The current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is a merging scenario; The motion intention corresponding to the current obstacle is the current obstacle acceleration; The step of determining whether there is a movement intention corresponding to the current obstacle based on the current corrected data, the current yielding relationship between the current vehicle and the current obstacle pre-marked in the current corrected data, and the current yielding scenario corresponding to the current corrected data includes: In the case where the current right-of-way relationship is that the current vehicle yields to the current obstacle, and the current right-of-way scenario is a merging scenario, determine whether the minimum acceleration of the current vehicle is a positive number; If so, then the minimum acceleration time of the current vehicle is determined based on the current vehicle speed, the current vehicle minimum acceleration, and the current vehicle maximum speed, and the time for the current vehicle to enter the interaction area is determined based on the current vehicle minimum acceleration time, the current vehicle minimum acceleration, the current vehicle speed, and the distance the current vehicle has traveled into the interaction area. If not, then determine the maximum deceleration time of the current vehicle based on the current vehicle speed and the current vehicle minimum acceleration, and determine the time of the current vehicle entering the interaction area based on the current vehicle maximum deceleration time, the current vehicle minimum acceleration, the current vehicle speed, and the distance of the current vehicle into the interaction area; Based on the current vehicle's entry time into the interaction area, the current obstacle's speed, the current obstacle's distance from the interaction area, and the preset safety distance, determine whether there is current obstacle acceleration; In the presence of the current obstacle acceleration, the speed at which the current obstacle leaves the preset safety distance is determined based on the current obstacle acceleration, the current obstacle speed, the distance the current obstacle enters the interaction area, and the preset safety distance. The speed at which the current vehicle enters the interaction area is determined based on the current vehicle speed, the current vehicle minimum acceleration, and the time the current vehicle takes to enter the interaction area. If the speed at which the current vehicle enters the interaction area is greater than the speed at which the current obstacle leaves the preset safe distance, the acceleration range of the current vehicle entering the interaction area is updated based on the speed at which the current vehicle enters the interaction area, the speed at which the current obstacle leaves the preset safe distance, and the acceleration of the current obstacle. The current correction data includes the current vehicle speed, the current distance the vehicle has entered the interaction area, the current vehicle maximum speed, the current obstacle speed, the current obstacle distance it has entered the interaction area, the current obstacle maximum speed, and the motion intention corresponding to the current obstacle.

7. The method according to claim 1, characterized in that, The process of training an initial obstacle motion intent prediction model based on the updated obstacle prediction dataset to obtain an obstacle motion intent target prediction model includes: The current vehicle speed, current vehicle distance into the interaction area, current vehicle maximum speed, current obstacle speed, current obstacle distance into the interaction area, and current obstacle maximum speed of each data point in the updated obstacle prediction base dataset are used as inputs to the initial prediction model for obstacle motion intent. The motion intent corresponding to the current obstacle in each data point of the updated obstacle prediction base dataset is used as the output of the initial prediction model for the obstacle motion intent; Based on the input and output of the initial prediction model of obstacle movement intention, the initial prediction model of obstacle movement intention is trained to obtain the target prediction model of obstacle movement intention.

8. A device for acquiring an obstacle motion intention prediction model, characterized in that, include: An obstacle prediction basic dataset acquisition module is used to acquire an obstacle prediction basic dataset and determine a first dataset and a second dataset in the obstacle prediction basic dataset; wherein, the first dataset is the data in the obstacle prediction basic dataset in which the current vehicle's right-of-way cutting is reasonable, and the second dataset is the data in the obstacle prediction basic dataset in which the current vehicle's right-of-way cutting is unreasonable; The dataset update module is used to determine the movement intention of each obstacle based on the data group to be corrected in the second dataset, the pre-labeled yielding relationship between the current vehicle and each obstacle in the data group to be corrected, and the yielding scenarios corresponding to the data group to be corrected, and update the data group to be corrected based on the movement intention of each obstacle to obtain the updated second dataset. The obstacle motion intention target prediction model training module is used to update the obstacle prediction base dataset based on the first dataset and the updated second dataset, and to train the obstacle motion intention initial prediction model based on the updated obstacle prediction base dataset to obtain the obstacle motion intention target prediction model. The data set to be corrected includes at least one piece of data to be corrected, and each piece of data to be corrected corresponds to an obstacle of the current vehicle. The dataset update module is further used to determine the correction order of each piece of data to be corrected in the data set to be corrected. According to the correction order, for each piece of data to be corrected in the data set to be corrected, based on the data to be corrected, the yielding relationship between the current vehicle and the obstacle pre-marked in the data to be corrected, and the yielding scenario corresponding to the data to be corrected, the movement intention of the obstacle corresponding to each piece of data to be corrected is determined. After determining the last piece of data to be corrected in the correction order, the movement intention of each obstacle is obtained.

9. An electronic device, characterized in that, The electronic device includes: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the obstacle motion intention prediction model acquisition method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the obstacle motion intention prediction model acquisition method as described in any one of claims 1-7.

Images