Trajectory prediction method and device, computer device and storage medium

Abstract

The present application relates to the technical field of automatic driving, and discloses a trajectory prediction method and device, computer equipment and a storage medium, wherein after obtaining the predicted trajectories output by different trajectory prediction algorithms, the total deviation values of different predicted trajectories of each target vehicle relative to a reference trajectory are obtained based on the vehicle operating parameters of each trajectory point of each target vehicle in different predicted trajectories, the first prior probability matrix is updated based on the obtained total deviation values, the evaluation scores corresponding to different predicted trajectories are obtained based on the updated second prior probability matrix and the total deviation values corresponding to different predicted trajectories, the predicted trajectories in the trajectory set are screened, and the target predicted trajectory of the target vehicle is determined. Thus, the predicted trajectories output by multiple trajectory prediction algorithms are evaluated without losing real-time, and the most likely trajectory of the target vehicle is screened in different road scenes.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and specifically to trajectory prediction methods, devices, computer equipment, and storage media. Background Technology

[0002] Autonomous vehicles rely on the collaborative efforts of artificial intelligence, computer vision, radar, monitoring devices, and GPS to enable computers to operate motor vehicles safely and automatically without any human intervention. The importance of predictive algorithms within the entire autonomous driving software system is self-evident. A superior predictive algorithm can accurately predict the future trajectories of surrounding vehicles and determine their interactions with the vehicle's future trajectory, which plays a crucial role in controlling the vehicle's future behavior patterns.

[0003] Currently, various vehicle trajectory prediction algorithms exist, and these algorithms can achieve good prediction results in single vehicle interaction scenarios. However, for some special urban scenarios, vehicles may make complex left and right turns and enter and exit ramps, and using traditional prediction methods will result in lower trajectory prediction accuracy. Summary of the Invention

[0004] In view of this, the present invention provides a trajectory prediction method, apparatus, computer device and storage medium to solve the problem that the prediction accuracy of traditional prediction algorithms is low when making predictions in complex scenarios.

[0005] In a first aspect, the present invention provides a trajectory prediction method, the method comprising:

[0006] Obtain the first prior probability matrix, the elements of which are the confidence scores of multiple preset trajectory prediction algorithms in different road scenarios;

[0007] In the current road scenario, multiple trajectory prediction algorithms are used to predict the trajectories of multiple target vehicles, resulting in a trajectory set and a trajectory information set. The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories.

[0008] Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory is obtained;

[0009] The first prior probability matrix is ​​updated based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix.

[0010] For each target vehicle, the evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories;

[0011] The trajectory set is filtered based on the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle.

[0012] By measuring the deviation of different predicted trajectories from the reference trajectory, the confidence of different prediction algorithms is optimized in a data-driven manner, continuously improving their performance. Simultaneously, without sacrificing real-time performance, the predicted trajectories output by various trajectory prediction algorithms are evaluated, and the most probable trajectory for the target vehicle is selected under different road scenarios.

[0013] In one optional implementation, the vehicle operating parameters include the vehicle's velocity, attitude, and heading angle at the trajectory points; the characteristic parameters of the reference trajectory include lateral acceleration reference values, velocity reference values, attitude change reference values, and heading angle change reference values.

[0014] Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory is obtained, including:

[0015] For each predicted trajectory of each target vehicle, based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration characteristic value, velocity characteristic value, pose change characteristic value, and heading angle change characteristic value corresponding to the predicted trajectory are obtained;

[0016] For each predicted trajectory of each target vehicle, the total deviation of the predicted trajectory from the reference trajectory is obtained based on the first deviation of the lateral acceleration characteristic value from the lateral acceleration reference value, the second deviation of the velocity characteristic value from the velocity reference value, the third deviation of the pose change characteristic value from the pose change reference value, and the fourth deviation of the heading angle change characteristic value from the heading angle change reference value.

[0017] By measuring the deviations of multiple feature values ​​of different predicted trajectories from the feature parameters of the reference trajectory, the confidence of different prediction algorithms is optimized in a data-driven manner, continuously improving their performance. Simultaneously, without sacrificing real-time performance, the predicted trajectories output by various trajectory prediction algorithms are evaluated, and the most probable trajectory for the target vehicle is selected under different road scenarios.

[0018] In one optional implementation, based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration characteristic value, velocity characteristic value, pose change characteristic value, and heading angle change characteristic value corresponding to the predicted trajectory are obtained, including:

[0019] Based on the velocity and pose of the target vehicle at multiple consecutive moments in the predicted trajectory, a first preset number of lateral accelerations are obtained.

[0020] Calculate the average value among the first preset number of lateral accelerations to obtain the lateral acceleration characteristic value.

[0021] Based on the velocity and pose of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration feature value is obtained, and the lateral acceleration feature of the predicted trajectory is evaluated.

[0022] In one optional implementation, based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration characteristic value, velocity characteristic value, pose change characteristic value, and heading angle change characteristic value corresponding to the predicted trajectory are obtained, including:

[0023] Based on the heading angles corresponding to the trajectory points of the target vehicle at multiple consecutive moments in the predicted trajectory, a second preset number of heading angle change rates is obtained;

[0024] Calculate the average value among the heading angle change rates of the second preset number to obtain the heading angle change characteristic value.

[0025] Therefore, based on the heading angle of the target vehicle at each trajectory point in the predicted trajectory, the heading angle change characteristic value is obtained, and the heading angle change characteristic of the predicted trajectory is evaluated.

[0026] In one optional implementation, the first prior probability matrix is ​​updated based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain a second prior probability matrix, including:

[0027] Based on the total deviation of the different predicted trajectories of each target vehicle, the confidence scores of different trajectory prediction algorithms in the first prior probability matrix are updated in the current road scenario to obtain the updated second prior probability matrix.

[0028] Therefore, the prior probability matrix is ​​used to constrain the predicted trajectory, the accuracy of the predicted trajectory algorithm is evaluated based on vehicle operating parameters, and the prior probability matrix is ​​optimized through a data-driven approach to improve the correctness of trajectory selection.

[0029] In one optional implementation, for each target vehicle, an evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total bias, including:

[0030] For each predicted trajectory of the current target vehicle, calculate the difference between the preset score and the total deviation value corresponding to that predicted trajectory;

[0031] The evaluation score corresponding to the predicted trajectory is obtained by multiplying the difference and the confidence level of the trajectory prediction algorithm corresponding to the predicted trajectory in the second prior probability matrix.

[0032] Therefore, based on the confidence level corresponding to different prediction algorithms, the preset score is subtracted from the total deviation value and then multiplied by the confidence level to obtain the final evaluation score, thereby evaluating the accuracy of different prediction trajectories.

[0033] In one optional implementation, the trajectory set is filtered according to the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle, including:

[0034] The maximum evaluation score is determined based on the evaluation scores corresponding to the different predicted trajectories of the current target vehicle.

[0035] The predicted trajectory corresponding to the maximum evaluation score in the trajectory set is used as the target predicted trajectory for the current target vehicle.

[0036] The evaluation scores are then sorted and the predicted trajectory with the highest score is output to provide guidance for controlling the behavior pattern of the target vehicle over a future period of time.

[0037] In a second aspect, the present invention provides a trajectory prediction device, the device comprising:

[0038] The acquisition module is used to acquire the first prior probability matrix. The elements of the first prior probability matrix are the confidence scores of multiple preset trajectory prediction algorithms in different road scenarios.

[0039] The first processing module is used to predict the trajectories of multiple target vehicles in the current road scenario using multiple trajectory prediction algorithms to obtain a trajectory set and a trajectory information set. The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories.

[0040] The second processing module is used to obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory based on the vehicle operation parameters of each trajectory point in different predicted trajectories of each target vehicle.

[0041] The third processing module is used to update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, so as to obtain the second prior probability matrix.

[0042] The fourth processing module is used to obtain the evaluation score corresponding to different predicted trajectories for each target vehicle based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories.

[0043] The fifth processing module is used to filter the trajectory set according to the evaluation scores corresponding to different predicted trajectories and determine the target predicted trajectory of the target vehicle.

[0044] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the trajectory prediction method of the first aspect or any corresponding embodiment described above.

[0045] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the trajectory prediction method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a schematic flowchart of a trajectory prediction method according to an embodiment of the present invention;

[0048] Figure 2 This is a flowchart illustrating another trajectory prediction method according to an embodiment of the present invention;

[0049] Figure 3 This is a flowchart illustrating another trajectory prediction method according to an embodiment of the present invention;

[0050] Figure 4A This is an implementation block diagram of the trajectory prediction method according to an embodiment of the present invention;

[0051] Figure 4B This is a schematic diagram illustrating the evaluation of trajectory prediction results according to an embodiment of the present invention;

[0052] Figure 4C This is a schematic diagram illustrating the evaluation of another trajectory prediction result according to an embodiment of the present invention;

[0053] Figure 5 This is a structural block diagram of a trajectory prediction device according to an embodiment of the present invention;

[0054] Figure 6 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Autonomous vehicles, also known as driverless cars, computer-driven cars, or wheeled mobile robots, are intelligent vehicles that achieve driverless operation through computer systems. They have existed for decades in the 20th century and showed a trend toward practical application in the early 21st century.

[0057] Autonomous vehicles rely on the collaborative efforts of artificial intelligence, computer vision, radar, monitoring devices, and GPS to enable computers to operate motor vehicles safely and automatically without any human intervention. The importance of predictive algorithms within the entire autonomous driving software system is self-evident. A superior predictive algorithm can accurately predict the future trajectory of the target vehicle and determine the interactions with the vehicle's future trajectory. This is a crucial input for determining the vehicle's future behavior patterns and speed control.

[0058] To further improve the accuracy and comprehensiveness of prediction results, multiple prediction algorithms may exist within the system. For example, rule-based algorithms and deep learning algorithms can be combined to leverage their respective strengths in their preferred prediction scenarios or target types. Suppose multiple algorithms are activated for a specific target, resulting in multiple possible trajectories. In this case, evaluating the quality of these trajectories and selecting the optimal one as the final predicted trajectory becomes a crucial problem to solve.

[0059] Therefore, this invention provides a trajectory prediction scheme. After obtaining the predicted trajectories output by different trajectory prediction algorithms, based on the vehicle operation parameters of each target vehicle at each trajectory point in the different predicted trajectories, the total deviation value of each target vehicle's different predicted trajectories relative to the reference trajectory is obtained. The first prior probability matrix is ​​updated based on the obtained total deviation value. Based on the updated second prior probability matrix and the total deviation value corresponding to different predicted trajectories, the evaluation score corresponding to each predicted trajectory is obtained. This allows for the filtering of predicted trajectories in the trajectory set to determine the target vehicle's target predicted trajectory. Thus, without sacrificing real-time performance, the predicted trajectories output by multiple trajectory prediction algorithms are evaluated, and the most likely trajectory for the target vehicle is selected under different road scenarios.

[0060] According to an embodiment of the present invention, a trajectory prediction method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0061] This embodiment provides a trajectory prediction method that can be used in vehicle controllers, such as MCUs and microcontrollers. Figure 1 This is a flowchart of a trajectory prediction method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0062] Step S101: Obtain the first prior probability matrix.

[0063] Specifically, the elements of the first prior probability matrix are the confidence scores of multiple preset trajectory prediction algorithms under different road scenarios. During real-vehicle testing, test data is collected to obtain the vehicle's actual trajectory and the prediction performance of different trajectory prediction algorithms under various road scenarios. By analyzing the large amount of collected data, the prediction metrics of different trajectory prediction algorithms are evaluated to obtain the first prior probability matrix, which measures the confidence scores of different trajectory prediction algorithms under different road scenarios. For example, the trajectory prediction algorithms may include rule-based trajectory prediction algorithms, Raster trajectory prediction algorithms, and Cslstm trajectory prediction algorithms; this invention is not limited to these.

[0064] Step S102: In the current road scenario, multiple trajectory prediction algorithms are used to predict the trajectories of multiple target vehicles to obtain a trajectory set and a trajectory information set.

[0065] The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories.

[0066] Specifically, in this embodiment of the invention, the predicted object is a vehicle. In practical applications, the predicted object can also be other objects that move according to certain rules, such as bicycles or pedestrians. This invention is not limited to this.

[0067] In a given scenario, there are often multiple target vehicles that need to be predicted. Data such as the location, speed, and historical trajectory of these vehicles can be obtained. Each target vehicle is assigned a unique ID by a pre-defined perception fusion module in the controller for binding and tracking. The trajectory prediction algorithm module obtains relevant information about the target vehicle using its ID and performs inference to predict and output a predicted trajectory. When multiple prediction algorithms are used to predict the same target vehicle, different algorithm modules will output predicted trajectory information containing the same target vehicle ID. Alternatively, a single algorithm may also output multiple possible predicted trajectories for the same target vehicle.

[0068] Specifically, after different trajectory prediction algorithms output predicted trajectories, some corrections are made to the output predicted trajectories to remove obviously unreasonable trajectory points (such as obvious anomalies in trajectory point speed and heading angle). The predicted trajectory is compared with the pre-acquired high-precision map. If a part of the predicted trajectory is outside the map boundary line, the trajectory segment is extracted and combined with the boundary line information of the high-precision map to correct the trajectory points that deviate from the map back to the lane line, so as to facilitate better screening of these predicted trajectories in subsequent steps.

[0069] Next, all target vehicles in the current road scene are traversed to obtain their bound IDs and activated trajectory prediction algorithms (i.e., the algorithms that predict the trajectory of the target vehicle corresponding to that ID). All trajectories are managed by a dictionary, where the key of the dictionary is the ID of the target vehicle, and the value of the dictionary corresponds to several trajectories of the target vehicle, thus obtaining the trajectory set and trajectory information set.

[0070] Step S103: Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory.

[0071] Specifically, without sacrificing real-time performance, a custom evaluation function can be designed. This function includes multiple evaluation items and evaluates different predicted trajectories from multiple aspects such as kinematics, scene information, and target vehicle information based on vehicle operating parameters, thereby obtaining the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory.

[0072] Step S104: Update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix.

[0073] Specifically, although the first prior probability matrix is ​​obtained by analyzing a large amount of empirical data in advance, the prediction performance of different algorithms varies under different road scenarios. Therefore, it is necessary to adjust the confidence level of different algorithms under different scenarios according to the deviation of different predicted trajectories from the reference trajectory, so as to improve the accuracy of the results.

[0074] Step S105: For each target vehicle, the evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories.

[0075] Specifically, based on the updated confidence level of the second prior probability matrix and the deviation of different predicted trajectories from the reference trajectory, different predicted trajectories are evaluated to obtain the evaluation score corresponding to each predicted trajectory.

[0076] Step S106: Filter the trajectory set according to the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle.

[0077] Specifically, based on the evaluation score corresponding to the different predicted trajectories of each target vehicle, multiple predicted trajectories of the target vehicle in the trajectory set are filtered to determine the target predicted trajectory of the target vehicle. In this way, the trajectory points and vehicle operation parameters of the target vehicle in the current road scenario in the future period are determined, and the possible future behavior patterns of the target vehicle are reasonably controlled.

[0078] The trajectory prediction method provided in this embodiment, after obtaining the predicted trajectories output by different trajectory prediction algorithms, obtains the total deviation value of the different predicted trajectories of each target vehicle relative to the reference trajectory based on the vehicle operation parameters of each trajectory point in the different predicted trajectories of each target vehicle. The first prior probability matrix is ​​updated based on the obtained total deviation value. The evaluation score corresponding to the different predicted trajectories is obtained according to the updated second prior probability matrix and the total deviation value corresponding to the different predicted trajectories. In this way, the predicted trajectories in the trajectory set are filtered to determine the target predicted trajectory of the target vehicle.

[0079] By measuring the deviation of different predicted trajectories from the reference trajectory, the confidence of different prediction algorithms is optimized in a data-driven manner, continuously improving their performance. Simultaneously, without sacrificing real-time performance, the predicted trajectories output by various trajectory prediction algorithms are evaluated, and the most probable trajectory for the target vehicle is selected under different road scenarios.

[0080] This embodiment provides a trajectory prediction method that can be used in vehicle controllers, such as MCUs and microcontrollers. Figure 2 This is a flowchart of a trajectory prediction method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0081] Step S201: Obtain the first prior probability matrix. See details below. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0082] Step S202: In the current road scenario, multiple trajectory prediction algorithms are used to predict the trajectories of multiple target vehicles, resulting in a trajectory set and a trajectory information set. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0083] Step S203: Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory.

[0084] Specifically, the vehicle operating parameters include information such as the vehicle's velocity, attitude, and heading angle at the trajectory points, and the characteristic parameters of the reference trajectory include lateral acceleration reference values, velocity reference values, attitude change reference values, and heading angle change reference values. Step S203 above includes:

[0085] Step S2031: For each predicted trajectory of each target vehicle, based on the velocity, pose, and heading angle of each trajectory point of the target vehicle in the predicted trajectory, obtain the lateral acceleration feature value, velocity feature value, pose change feature value, and heading angle change feature value corresponding to the predicted trajectory.

[0086] Specifically, based on the velocity and pose of the target vehicle at multiple consecutive time points along the predicted trajectory, a first preset number of lateral accelerations are obtained. The average value among these first preset number of lateral accelerations is then calculated to obtain the lateral acceleration feature value. It should be noted that the calculation method for lateral acceleration can be found in the descriptions of relevant technologies, and will not be repeated here.

[0087] For example, the predicted trajectory is a set of 80 discrete trajectory points, with a prediction time interval of 0.1s between each discrete point. The adjacent points of a point refer to the trajectory points located 0.1s before and after that point in the prediction time. A lateral acceleration can be calculated for every three adjacent trajectory points. The average value is obtained by dividing the sum of all lateral accelerations by the number of lateral accelerations. This average value is used as the lateral acceleration feature value of the predicted trajectory.

[0088] Based on the velocity and pose of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration feature value is obtained, and the lateral acceleration feature of the predicted trajectory is evaluated.

[0089] Specifically, based on the heading angles corresponding to trajectory points of the target vehicle at multiple consecutive moments in the predicted trajectory, a second preset number of heading angle change rates are obtained, and the average value among the second preset number of heading angle change rates is calculated to obtain the heading angle change characteristic value.

[0090] For example, the predicted trajectory is a set of 80 discrete trajectory points, each of which corresponds to a heading angle. A heading angle change rate (yaw_rate) can be calculated for every two predicted trajectory points. All heading angle change rates are calculated and their average value is taken as the heading angle change feature value of the predicted trajectory.

[0091] Therefore, based on the heading angle of the target vehicle at each trajectory point in the predicted trajectory, the heading angle change characteristic value is obtained, and the heading angle change characteristic of the predicted trajectory is evaluated.

[0092] Specifically, for a given predicted trajectory, the velocity characteristic value of the predicted trajectory is the average velocity of all trajectory points in the predicted trajectory; the pose change characteristic value of the predicted trajectory is the sum of the distances of the trajectory points in the previous few moments relative to the current position of the target vehicle. For example, the first five trajectory points of a target vehicle's predicted trajectory represent the predicted position of the target vehicle in the next 0.5 seconds. If these five trajectory points are far from the current position of the target vehicle or have a large change in driving angle, the target vehicle is unlikely to travel or turn to the predicted position corresponding to these trajectory points within 0.5 seconds. Therefore, the pose change characteristic value corresponding to the predicted trajectory can be used to evaluate the predicted trajectory.

[0093] Step S2032: For each predicted trajectory of each target vehicle, the total deviation of the predicted trajectory from the reference trajectory is obtained based on the first deviation of the lateral acceleration feature value from the lateral acceleration reference value, the second deviation of the velocity feature value from the velocity reference value, the third deviation of the pose change feature value from the pose change reference value, and the fourth deviation of the heading angle change feature value from the heading angle change reference value.

[0094] For example, for a given predicted trajectory, the reference value for the lateral acceleration of the corresponding reference trajectory can range from 0.05 to 0.1 m / s². 2 The velocity reference value can be the average velocity of all trajectory points with non-zero velocities in the predicted trajectory. The pose change reference value can be 0, and the heading angle change characteristic value can be 1-1.5°. The specific values ​​can be set according to the actual scene and the predicted object, and this invention is not limited thereto.

[0095] Specifically, the total deviation of the predicted trajectory from the reference trajectory, offset, can be the first deviation value, offset. accSecond deviation value offset v Third deviation value offset pose and the fourth deviation value offset yaw sum.

[0096] Step S204: Update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0097] Step S205: For each target vehicle, an evaluation score is obtained for each predicted trajectory based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories. See details below. Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0098] Step S206: The trajectory set is filtered based on the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle. For details, please refer to [link to relevant documentation]. Figure 1 Step S106 of the illustrated embodiment will not be described again here.

[0099] The trajectory prediction method provided in this embodiment, after obtaining the predicted trajectories output by different trajectory prediction algorithms, obtains multiple feature values ​​corresponding to different predicted trajectories of each target vehicle based on the vehicle operation parameters of each trajectory point in different predicted trajectories. According to the total deviation of these feature values ​​relative to the feature parameters of the reference trajectory, the first prior probability matrix is ​​updated. According to the updated second prior probability matrix and the total deviation of different predicted trajectories, the evaluation score corresponding to different predicted trajectories is obtained, thereby filtering the predicted trajectories in the trajectory set and determining the target predicted trajectory of the target vehicle.

[0100] By measuring the deviations of multiple feature values ​​of different predicted trajectories from the feature parameters of the reference trajectory, the confidence of different prediction algorithms is optimized in a data-driven manner, continuously improving their performance. Simultaneously, without sacrificing real-time performance, the predicted trajectories output by various trajectory prediction algorithms are evaluated, and the most probable trajectory for the target vehicle is selected under different road scenarios.

[0101] This embodiment provides a trajectory prediction method that can be used in vehicle controllers, such as MCUs and microcontrollers. Figure 3 This is a flowchart of a trajectory prediction method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps:

[0102] Step S301: Obtain the first prior probability matrix. See details below. Figure 2Step S201 of the illustrated embodiment will not be described again here.

[0103] Step S302: In the current road scenario, multiple trajectory prediction algorithms are used to predict the trajectories of multiple target vehicles, resulting in a trajectory set and a trajectory information set. For details, please refer to [link to relevant documentation]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0104] Step S303: Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory. For details, please refer to [link to relevant documentation]. Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0105] Step S304: Update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix.

[0106] Specifically, based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, the confidence scores of different trajectory prediction algorithms in the first prior probability matrix under the current road scenario are updated to obtain the updated second prior probability matrix. For example, the confidence scores in the first prior probability matrix can be updated according to the following formula:

[0107]

[0108] in, This represents the confidence level of different trajectory prediction algorithms in the second prior probability matrix under the corresponding Intersection Scenario (IS). The first prior probability matrix represents the confidence level of different trajectory prediction algorithms in the corresponding road scene IS, and sigmoid represents the activation function. This represents the total deviation of different trajectory prediction algorithms under the corresponding road scene IS.

[0109] Specifically, road scenarios can include scenarios with different interaction relationships, such as straight road interaction scenarios, curved road interaction scenarios, fork road interaction scenarios, merging road interaction scenarios, and intersection interaction scenarios.

[0110] For example, the confidence scores of different prediction algorithms under different interaction scenarios are shown in Table 1 and Table 2 below. Table 1 below represents the confidence score in the first prior probability matrix, and Table 2 below represents the confidence score in the updated second prior probability matrix. The horizontal columns of the tables represent different scenarios: straight road Scenario_straight, curved road Scenario_curve, branching road Scenario_branch, merging road Scenario_merge, and intersection Scenario_intersection. The vertical columns of the tables represent different interaction types between other vehicles and the vehicle: unknown Intersection_unknown, cutting in Intersection_cutin, merging road Scenario_merge, and following Intersection_follow.

[0111] Table 1 First Prior Probability Matrix

[0112]

[0113] Table 2 Second Prior Probability Matrix

[0114]

[0115] For example, in Table 2 above, the confidence levels of the first trajectory prediction algorithm, the second trajectory prediction algorithm, and the third trajectory prediction algorithm in the curve merging scenario are 0.4, 0.511, and 0.544, respectively.

[0116] Therefore, the prior probability matrix is ​​used to constrain the predicted trajectory, the accuracy of the predicted trajectory algorithm is evaluated based on vehicle operating parameters, and the prior probability matrix is ​​optimized through a data-driven approach to improve the correctness of trajectory selection.

[0117] Step S305: For each target vehicle, the evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories.

[0118] Specifically, step S305 includes:

[0119] Step S3051: For each predicted trajectory of the current target vehicle, calculate the difference between the preset score and the total deviation value corresponding to the predicted trajectory.

[0120] For example, the preset score is Score initial It can be 100.

[0121] Step S3052: Based on the product of the difference and the confidence level of the trajectory prediction algorithm corresponding to the predicted trajectory in the second prior probability matrix, the evaluation score corresponding to the predicted trajectory is obtained.

[0122] Specifically, the deviation of different predicted trajectories from the reference trajectory is not directly used as the evaluation criterion for the predicted trajectory. As mentioned earlier, each algorithm has different prediction capabilities for different target vehicles in different scenarios. Therefore, given a certain scenario and a certain type of predicted object, there will be a confidence level for different algorithms. The final evaluation score is obtained by subtracting the total deviation value from the preset score and then multiplying it by the confidence level, thereby evaluating the accuracy of different predicted trajectories.

[0123] For example, the evaluation score can be calculated according to the following formula:

[0124] Score i =(Score) initial -offset)*P i

[0125] Among them, Score i P represents the evaluation score corresponding to the i-th predicted trajectory. i This represents the confidence level corresponding to the i-th predicted trajectory.

[0126] Step S306: Filter the trajectory set according to the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle.

[0127] Specifically, step S306 includes:

[0128] Step S3061: Determine the maximum evaluation score based on the evaluation scores corresponding to the different predicted trajectories of the current target vehicle.

[0129] Specifically, the different predicted trajectories corresponding to each target vehicle are sorted in descending order of evaluation score.

[0130] Step S3062: The predicted trajectory corresponding to the maximum evaluation score in the trajectory set is taken as the target predicted trajectory of the current target vehicle.

[0131] Specifically, for each target vehicle, the predicted trajectory corresponding to its maximum evaluation score is taken as the predicted trajectory of the target vehicle. After sorting the evaluation scores, the predicted trajectory with the highest evaluation score is output to provide guidance for the control of the target vehicle's behavior pattern in the future.

[0132] The trajectory prediction method provided in this embodiment, after obtaining the predicted trajectories output by different trajectory prediction algorithms, obtains multiple feature values ​​corresponding to different predicted trajectories of each target vehicle based on the vehicle operation parameters of each trajectory point in different predicted trajectories. The first prior probability matrix is ​​updated according to the total deviation of these feature values ​​relative to the feature parameters of the reference trajectory. The evaluation score corresponding to different predicted trajectories is obtained according to the updated second prior probability matrix and the total deviation of different predicted trajectories. The predicted trajectories in the trajectory set are filtered according to the maximum value of the evaluation score, and the predicted trajectory corresponding to the maximum evaluation score is determined as the target predicted trajectory of the target vehicle.

[0133] By measuring the deviations of multiple feature values ​​of different predicted trajectories from the feature parameters of the reference trajectory, the confidence of different prediction algorithms is optimized in a data-driven manner, continuously improving their performance. Simultaneously, without sacrificing real-time performance, the predicted trajectories output by various trajectory prediction algorithms are evaluated, and the most probable trajectory for the target vehicle is selected under different road scenarios.

[0134] The trajectory prediction method in this embodiment of the invention will be described in detail below with reference to a practical application example. The specific steps of this application example are as follows.

[0135] like Figure 4A As shown, the initial prior probability matrix corresponding to the scene and interaction relationship is obtained. Then, rule-based prediction algorithm, Raster trajectory prediction algorithm, and Cslstm trajectory prediction algorithm are used to predict the trajectory of the target vehicle in the current scene over a future period, outputting different predicted trajectories for each target vehicle. Next, the deviations corresponding to evaluation items such as the average lateral acceleration, global velocity deviation, position deviation of the first two points of the trajectory, and global heading angle change rate of the trajectory are calculated to obtain the evaluation scores corresponding to different predicted trajectories. Finally, the predicted trajectory corresponding to the maximum evaluation score is taken as the final target predicted trajectory of the target vehicle.

[0136] First, based on the initial prior probability matrix, predictions are made using different trajectory prediction algorithms, and the corresponding results are as follows: Figure 4B As shown, different prediction algorithms have different prediction performance. Figure 4BIn this code, 'cslstm', 'graph', and 'rule_veh' correspond to three trajectory prediction algorithms, 'p1' represents the final selected predicted trajectory, 'ade' (Average Displacement Error) represents the average error of the predicted trajectory, 'fde' represents the error of the last trajectory point in the predicted trajectory, 'var' represents the velocity of the predicted trajectory, 'lat' represents the latitude of the predicted trajectory, 'lon' represents the longitude of the predicted trajectory, and 'count' represents the number of target vehicles in the current scene.

[0137] Furthermore, the initial prior probability matrix is ​​updated based on the vehicle operating parameters in the predicted trajectory. Based on the updated prior probability matrix, different trajectory prediction algorithms are used for prediction, and the corresponding results are as follows: Figure 4C As shown, after updating the initial prior probability matrix, the index of p1 is improved to a certain extent, which means that a more accurate predicted trajectory is selected as the final target predicted trajectory of the output target vehicle.

[0138] This embodiment also provides a trajectory prediction device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0139] This embodiment provides a trajectory prediction device, such as... Figure 5 As shown, it includes:

[0140] The acquisition module 501 is used to acquire the first prior probability matrix, the elements of which are the confidence scores of multiple preset trajectory prediction algorithms in different road scenarios.

[0141] The first processing module 502 is used to predict the trajectories of multiple target vehicles in the current road scenario using multiple trajectory prediction algorithms to obtain a trajectory set and a trajectory information set. The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories.

[0142] The second processing module 503 is used to obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory based on the vehicle operation parameters of each trajectory point in different predicted trajectories of each target vehicle.

[0143] The third processing module 504 is used to update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, so as to obtain the second prior probability matrix.

[0144] The fourth processing module 505 is used to obtain the evaluation score corresponding to different predicted trajectories for each target vehicle based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories.

[0145] The fifth processing module 506 is used to filter the trajectory set according to the evaluation scores corresponding to different predicted trajectories and determine the target predicted trajectory of the target vehicle.

[0146] In some optional implementations, vehicle operating parameters include the vehicle's velocity, pose, and heading angle at trajectory points; characteristic parameters of the reference trajectory include lateral acceleration reference values, velocity reference values, pose change reference values, and heading angle change reference values. The second processing module 503 includes:

[0147] The first processing unit, for each predicted trajectory of each target vehicle, obtains the lateral acceleration characteristic value, velocity characteristic value, pose change characteristic value and heading angle change characteristic value corresponding to the predicted trajectory based on the velocity, pose and heading angle of each trajectory point of the target vehicle in the predicted trajectory.

[0148] The second processing unit is used to, for each predicted trajectory of each target vehicle, obtain the total deviation value of the predicted trajectory relative to the reference trajectory based on the first deviation value of the lateral acceleration characteristic value relative to the lateral acceleration reference value, the second deviation value of the velocity characteristic value relative to the velocity reference value, the third deviation value of the pose change characteristic value relative to the pose change reference value, and the fourth deviation value of the heading angle change characteristic value relative to the heading angle change reference value.

[0149] In some optional implementations, the first processing unit includes:

[0150] The first processing subunit is used to obtain a first preset number of lateral accelerations based on the velocity and pose of the target vehicle at multiple consecutive time points in the predicted trajectory.

[0151] The second processing subunit is used to calculate the average value among the first preset number of lateral accelerations to obtain the lateral acceleration characteristic value.

[0152] In some optional implementations, the first processing unit includes:

[0153] The third processing subunit is used to obtain a second preset number of heading angle change rates based on the heading angles corresponding to the trajectory points of the target vehicle at multiple consecutive moments in the predicted trajectory.

[0154] The fourth processing subunit is used to calculate the average value between the second preset number of heading angle change rates to obtain the heading angle change characteristic value.

[0155] In some alternative implementations, the third processing module 504 includes:

[0156] The third processing unit is used to update the confidence levels of different trajectory prediction algorithms in the first prior probability matrix under the current road scenario based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, so as to obtain the updated second prior probability matrix.

[0157] In some alternative implementations, the fourth processing module 505 includes:

[0158] The fourth processing unit is used to calculate the difference between the preset score and the total deviation value corresponding to the predicted trajectory for each predicted trajectory of the current target vehicle.

[0159] The fifth processing unit is used to obtain the evaluation score corresponding to the predicted trajectory based on the product between the difference and the confidence level of the trajectory prediction algorithm corresponding to the predicted trajectory in the second prior probability matrix.

[0160] In some alternative implementations, the fifth processing module 506 includes:

[0161] The sixth processing unit is used to determine the maximum evaluation score based on the evaluation scores corresponding to the different predicted trajectories of the current target vehicle.

[0162] The seventh processing unit is used to take the predicted trajectory corresponding to the maximum evaluation score in the trajectory set as the target predicted trajectory of the current target vehicle.

[0163] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0164] In this embodiment, the trajectory prediction device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0165] This invention also provides a computer device having the above-described features. Figure 5 The trajectory prediction device shown.

[0166] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 6As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 6 Take a processor 10 as an example.

[0167] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0168] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0169] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0170] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0171] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0172] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0173] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A trajectory prediction method, characterized in that, The method includes: Obtain a first prior probability matrix, the elements of which are the confidence scores of multiple preset trajectory prediction algorithms in different road scenarios; In the current road scenario, the trajectories of multiple target vehicles are predicted by the multiple trajectory prediction algorithms to obtain a trajectory set and a trajectory information set. The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories. Based on the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories, the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory is obtained; The first prior probability matrix is ​​updated based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix. The calculation formula is as follows: in, This represents the confidence level of different trajectory prediction algorithms in the second prior probability matrix under the corresponding road scenario. This represents the confidence level of different trajectory prediction algorithms in the first prior probability matrix under the corresponding road scenario. sigmoid This represents the activation function. This represents the total deviation of different trajectory prediction algorithms in the corresponding road scenario; For each target vehicle, the evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories. The calculation formula is as follows: in, This represents the evaluation score corresponding to the i-th predicted trajectory. The preset score is represented by offset, which indicates the total deviation corresponding to different predicted trajectories. This represents the confidence level corresponding to the i-th predicted trajectory; The trajectory set is filtered based on the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle.

2. The method according to claim 1, characterized in that, The vehicle operating parameters include the vehicle's velocity, attitude, and heading angle at the trajectory points; the characteristic parameters of the reference trajectory include lateral acceleration reference values, velocity reference values, attitude change reference values, and heading angle change reference values. The method of obtaining the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory based on the vehicle operation parameters of each trajectory point in different predicted trajectories includes: For each predicted trajectory of each target vehicle, based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory, the lateral acceleration characteristic value, velocity characteristic value, pose change characteristic value, and heading angle change characteristic value corresponding to the predicted trajectory are obtained. For each predicted trajectory of each target vehicle, the total deviation of the predicted trajectory from the reference trajectory is obtained based on the first deviation of the lateral acceleration feature value relative to the lateral acceleration reference value, the second deviation of the velocity feature value relative to the velocity reference value, the third deviation of the pose change feature value relative to the pose change reference value, and the fourth deviation of the heading angle change feature value relative to the heading angle change reference value.

3. The method according to claim 2, characterized in that, The step of obtaining lateral acceleration feature values, velocity feature values, pose change feature values, and heading angle change feature values ​​corresponding to the predicted trajectory based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory includes: Based on the velocity and pose of the target vehicle at multiple consecutive time points in the predicted trajectory, a first preset number of lateral accelerations are obtained. Calculate the average value among the first preset number of lateral accelerations to obtain the lateral acceleration characteristic value.

4. The method according to claim 2, characterized in that, The step of obtaining lateral acceleration feature values, velocity feature values, pose change feature values, and heading angle change feature values ​​corresponding to the predicted trajectory based on the velocity, pose, and heading angle of the target vehicle at each trajectory point in the predicted trajectory includes: Based on the heading angles corresponding to trajectory points of the target vehicle at multiple consecutive moments in the predicted trajectory, a second preset number of heading angle change rates are obtained; Calculate the average value among the second preset number of heading angle change rates to obtain the heading angle change characteristic value.

5. The method according to any one of claims 1-4, characterized in that, The first prior probability matrix is ​​updated based on the total deviation value corresponding to the different predicted trajectories of each target vehicle to obtain the second prior probability matrix, including: Based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, the confidence scores of different trajectory prediction algorithms in the first prior probability matrix are updated in the current road scenario to obtain the updated second prior probability matrix.

6. The method according to claim 5, characterized in that, For each target vehicle, the evaluation score corresponding to different predicted trajectories is obtained based on the second prior probability matrix and the total deviation value, including: For each predicted trajectory of the current target vehicle, calculate the difference between the preset score and the total deviation value corresponding to the predicted trajectory; The evaluation score corresponding to the predicted trajectory is obtained by multiplying the difference and the confidence level of the trajectory prediction algorithm corresponding to the predicted trajectory in the second prior probability matrix.

7. The method according to claim 6, characterized in that, The step of filtering the trajectory set based on the evaluation scores corresponding to different predicted trajectories to determine the target predicted trajectory of the target vehicle includes: The maximum evaluation score is determined based on the evaluation scores corresponding to the different predicted trajectories of the current target vehicle. The predicted trajectory corresponding to the maximum evaluation score in the trajectory set is taken as the target predicted trajectory of the current target vehicle.

8. A trajectory prediction device, characterized in that, The device includes: The acquisition module is used to acquire a first prior probability matrix, wherein the elements of the first prior probability matrix are the confidence scores of multiple preset trajectory prediction algorithms in different road scenarios. The first processing module is used to predict the trajectories of multiple target vehicles in the current road scenario using the multiple trajectory prediction algorithms to obtain a trajectory set and a trajectory information set. The trajectory set includes the predicted trajectories of each target vehicle under different trajectory prediction algorithms, and the trajectory information set includes the vehicle operation parameters of each target vehicle at each trajectory point in different predicted trajectories. The second processing module is used to obtain the total deviation of each target vehicle's different predicted trajectories relative to the reference trajectory based on the vehicle operation parameters of each trajectory point in different predicted trajectories of each target vehicle. The third processing module is used to update the first prior probability matrix based on the total deviation value corresponding to the different predicted trajectories of each target vehicle, to obtain the second prior probability matrix. The calculation formula is as follows: in, This represents the confidence level of different trajectory prediction algorithms in the second prior probability matrix under the corresponding road scenario. This represents the confidence level of different trajectory prediction algorithms in the first prior probability matrix under the corresponding road scenario. sigmoid This represents the activation function. This represents the total deviation of different trajectory prediction algorithms in the corresponding road scenario; The fourth processing module is used to calculate the evaluation score for each target vehicle based on the second prior probability matrix and the total deviation value corresponding to different predicted trajectories. The calculation formula is as follows: in, This represents the evaluation score corresponding to the i-th predicted trajectory. The preset score is represented by offset, which indicates the total deviation corresponding to different predicted trajectories. This represents the confidence level corresponding to the i-th predicted trajectory; The fifth processing module is used to filter the trajectory set according to the evaluation scores corresponding to different predicted trajectories, and determine the target predicted trajectory of the target vehicle.

9. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the trajectory prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the trajectory prediction method according to any one of claims 1 to 7.

Images