Trajectory Prediction Method, Device, Equipment and Medium for Moving Objects in Autonomous Driving

Abstract

The present application relates to a method, apparatus, device, and medium for predicting the trajectory of a moving object in autonomous driving. The method includes: determining, from the set of moving objects corresponding to the current frame data collected by an autonomous driving vehicle, a first set of moving objects that have been processed by a preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determining a second set of moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted; obtaining the existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data, and determining the predicted movement trajectories of each first moving object in a preset time period after the current frame data based on the existing predicted movement trajectories; and determining the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model. This method improves the real-time performance of trajectory prediction.

Description

Technical Field

[0001] The present application relates to the technical field of autonomous driving, and in particular, to a method, apparatus, device, and medium for predicting the trajectory of a moving object in autonomous driving. Background Art

[0002] During driving, driving safety has always been a very important matter. To ensure the driving safety of an autonomous vehicle, it is necessary to predict the trajectory of a moving object within the field of view of the autonomous vehicle.

[0003] However, when the number of moving objects within the field of view of the autonomous vehicle is large, due to the computing resources of the in-vehicle computing unit of the autonomous vehicle being limited, the real-time performance of predicting the trajectory of the moving object may not meet the actual requirements. Summary of the Invention

[0004] Based on this, in view of the technical problem that the real-time performance of predicting the trajectory of a moving object by the traditional method may not meet the actual requirements, it is necessary to provide a method, apparatus, device, and medium for predicting the trajectory of a moving object in autonomous driving.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting the trajectory of a moving object in autonomous driving, including:

[0006] From the set of moving objects corresponding to the current frame data collected by the autonomous vehicle, determine the first moving objects that have been processed by a preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determine the second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted;

[0007] Obtain the existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data, and based on each existing predicted movement trajectory, determine the predicted movement trajectories of each first moving object in a preset time period after the current frame data;

[0008] Through the preset trajectory prediction model, determine the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data.

[0009] In a second aspect, an embodiment of the present application provides a device for predicting the trajectory of a moving object in autonomous driving, including:

[0010] A determination module, configured to determine, from the set of moving objects corresponding to the current frame data collected by the autonomous vehicle, the first moving objects that have been processed by a preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determine the second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted;

[0011] An acquisition module, configured to acquire the existing predicted motion trajectories of each first moving object in the first set of objects to be predicted before the current frame of data through the preset trajectory prediction model;

[0012] A first prediction module, configured to determine the predicted motion trajectories of each first moving object in a preset time period after the current frame of data based on the existing predicted motion trajectories of each first moving object;

[0013] A second prediction module, configured to determine the predicted motion trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame of data through the preset trajectory prediction model.

[0014] In a third aspect, an embodiment of the present application provides a computing device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the method for predicting the trajectory of a moving object in an autonomous driving provided in the first aspect of the embodiments of the present application are implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for predicting the trajectory of a moving object in an autonomous driving provided in the first aspect of the embodiments of the present application are implemented.

[0016] The technical solution provided by the embodiments of the present application determines the first moving objects and the second moving objects in the set of moving objects corresponding to the current frame of data collected by the autonomous driving vehicle. For each first moving object that has been processed by the preset trajectory prediction model before the current frame of data, the predicted motion trajectories of each first moving object in a preset time period after the current frame of data are determined based on the existing predicted motion trajectories of each first moving object. For each second moving object that has not been processed by the preset trajectory prediction model, the predicted motion trajectories of each second moving object in a preset time period after the current frame of data are determined through the preset trajectory prediction model. That is to say, the number of moving objects processed by the preset trajectory prediction model is reduced, the computing pressure on the computing unit of the autonomous driving vehicle is alleviated, and thus the real-time performance of the trajectory prediction of each moving object corresponding to the current frame of data is improved. At the same time, the predicted motion trajectories of some moving objects can be determined by the existing predicted motion trajectories predicted before the current frame of data. Overall, the number of predicted motion trajectories output by the computing unit is also increased, and the accuracy of the output predicted motion trajectories is ensured. Description of the Drawings

[0017] Figure 1 It is a schematic flowchart of a method for predicting the trajectory of a moving object in an autonomous driving provided by an embodiment of the present application;

[0018] Figure 2A schematic flowchart of a process for determining the predicted motion trajectory of a first moving object provided by an embodiment of the present application;

[0019] Figure 3 A schematic flowchart of a process for determining the predicted motion trajectory of a second moving object provided by an embodiment of the present application;

[0020] Figure 4 Another schematic flowchart of a process for determining the predicted motion trajectory of a second moving object provided by an embodiment of the present application;

[0021] Figure 5 A schematic structural diagram of a trajectory prediction device for moving objects in autonomous driving provided by an embodiment of the present application;

[0022] Figure 6 A schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0023] In the related art, for moving objects within the field of view of an autonomous driving vehicle, the artificial intelligence model is used to predict their motion trajectories. However, the prediction duration of a complex artificial intelligence model may be relatively long. Thus, when the number of moving objects within the field of view of the autonomous driving vehicle is large, limited by the computing resources of the in-vehicle computing unit of the autonomous driving vehicle, the real-time performance of the trajectory prediction of each moving object may not meet the actual requirements.

[0024] In view of the above problems, the technical solution provided by the embodiment of the present application can not only improve the real-time performance of the trajectory prediction of each moving object, but also increase the number of predicted motion trajectories output and the accuracy of the predicted motion trajectories, thereby improving the driving safety of the autonomous driving vehicle.

[0025] It should be noted that the technical solution provided by the embodiment of the present application can be executed by the in-vehicle computing device in the autonomous driving vehicle or by the cloud device, and the embodiment of the present application does not make a limitation thereto.

[0026] In order to make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application are further described in detail below through the following embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0027] Figure 1 A schematic flowchart of a method for predicting the trajectory of a moving object in autonomous driving provided by an embodiment of the present application. As Figure 1 shown, the method may include:

[0028] S101. From the set of moving objects corresponding to the current frame of data collected by the autonomous vehicle, determine the first moving objects that have been processed by the preset trajectory prediction model before the current frame of data to form a first set of objects to be predicted, and determine the second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted.

[0029] One or more environmental perception devices can be set on the autonomous vehicle to collect information about the vehicle's surrounding environment and the vehicle's driving information, etc. during the vehicle's driving process. Among them, the vehicle's driving information includes, but is not limited to, the vehicle's geographical location, driving speed, and driving direction, etc. Optionally, the above environmental perception devices can be cameras and various sensors, etc.

[0030] After collecting the current frame of data, a preset detection algorithm can be used to detect the moving objects existing in the current frame of data. Optionally, the moving objects can be other vehicles or pedestrians, etc., and the other vehicles can be motor vehicles and non-motor vehicles. Optionally, the above detection algorithm can be a YOLO (You Only Look Once) detection network, or can also be networks such as region-based fully convolutional network (R-CNN) and SSD (Single Shot MultiBox Detector).

[0031] After obtaining the set of moving objects corresponding to the current frame of data, the computing device can traverse each moving object one by one to obtain the service tag corresponding to each moving object, and based on the service tag corresponding to each moving object, divide each moving object. Among them, the above service tag is used to identify whether the moving object has been processed by the preset trajectory prediction model before the current frame of data. Optionally, the above preset trajectory prediction model can be a pre-trained recurrent neural network model, that is, a trajectory prediction model can be constructed using a deep learning algorithm, and the established trajectory prediction model can be trained using a pre-obtained training data set. By continuously iterating until convergence, a trained trajectory prediction model is finally obtained, and the trained trajectory prediction model is stored in the computing device. The preset trajectory prediction model can predict the movement trajectory of the moving object within a preset time period in the future based on the historical movement trajectory of the moving object.

[0032] Exemplarily, when the service flag of a moving object is "true", it indicates that the moving object has been processed by a preset trajectory prediction model before this frame of data. Then, the moving object can be determined as a first moving object, and all the first moving objects form a first set of objects to be predicted. When the service flag of a moving object is "false", it indicates that the moving object has not been processed by a preset trajectory prediction model. Then, the moving object can be determined as a second moving object, and all the second moving objects form a second set of objects to be predicted. Of course, other characters or numerical values can also be used to represent the service flag, such as the strings "yes" and "no", or the numerical values "1" and "0", etc.

[0033] It can be understood that as the autonomous vehicle travels and the surrounding environment continuously changes, the moving objects in this frame of data can include the moving objects in the previous frame of data, and can also include newly added moving objects. That is to say, the second moving objects that have not been processed by the preset trajectory prediction model can include the newly added moving objects in this frame of data, and also include the moving objects in the previous frame of data that have not been processed by the preset trajectory prediction model.

[0034] S102. Obtain the existing predicted motion trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before this frame of data, and determine the predicted motion trajectories of each first moving object in a preset time period after this frame of data based on the existing predicted motion trajectories.

[0035] Considering that during the processing of the previous frame of data, the computing device has already predicted the motion trajectory of the first moving object using the preset trajectory prediction model. That is, before the processing of this frame of data, the computing device already has the existing predicted motion trajectory of the first moving object. Based on this, when predicting the motion trajectory of the moving object in this frame of data, in order to relieve the computing pressure of the computing device, for the first moving object with an existing predicted motion trajectory, the computing device can directly obtain the existing predicted motion trajectory of the first moving object obtained by the preset trajectory prediction model before this frame of data, and use the existing predicted motion trajectory as a reference to determine the predicted motion trajectory of the first moving object in a preset time period after this frame of data, rather than using the preset trajectory prediction model to predict the future motion trajectory of the first moving object.

[0036] S103. Determine the predicted motion trajectories of each second moving object in the second set of objects to be predicted in a preset time period after this frame of data through the preset trajectory prediction model.

[0037] For a second moving object that has not been processed by a preset trajectory prediction model, the computing device can obtain the historical motion trajectory data of the second moving object and input the historical motion trajectory data of the second moving object into the preset trajectory prediction model for model prediction through the preset trajectory prediction model, so as to obtain the predicted motion trajectory of the second moving object in a preset time period after the current frame of data.

[0038] In this way, for the computing device, only the preset trajectory prediction model is used to predict the motion trajectories of some moving objects in the current frame of data. That is to say, the number of moving objects predicted by the preset trajectory prediction model is relatively reduced, greatly improving the processing speed of the computing device, thereby enhancing the real-time performance of trajectory prediction.

[0039] The trajectory prediction method for moving objects in autonomous driving provided by the embodiments of this application determines the first moving object and the second moving object in the moving object set corresponding to the current frame of data collected by the autonomous driving vehicle. For each first moving object that has been processed by the preset trajectory prediction model before the current frame of data, the predicted motion trajectory of each first moving object in a preset time period after the current frame of data is determined based on the existing predicted motion trajectories of each first moving object. For each second moving object that has not been processed by the preset trajectory prediction model, the predicted motion trajectory of each second moving object in a preset time period after the current frame of data is determined through the preset trajectory prediction model. That is to say, the number of moving objects processed by the preset trajectory prediction model is reduced, alleviating the computing pressure on the computing unit of the autonomous driving vehicle, thereby improving the real-time performance of the trajectory prediction for each moving object corresponding to the current frame of data. At the same time, the predicted motion trajectories of some moving objects can be determined by the existing predicted motion trajectories predicted before the current frame of data. Overall, the number of predicted motion trajectories output by the computing unit is also increased, and the accuracy of the output predicted motion trajectories is ensured.

[0040] In practical applications, for the same moving object, usually multiple motion trajectories will be predicted by the preset trajectory prediction model. Based on this, optionally, the process of obtaining the existing predicted motion trajectories of each first moving object in the first set of objects to be predicted before the current frame of data through the preset trajectory prediction model in step S102 may be: obtaining multiple existing predicted motion trajectories of each first moving object in the first set of objects to be predicted before the current frame of data through the preset trajectory prediction model.

[0041] Correspondingly, as Figure 2 shown, the process of determining the predicted motion trajectory of each first moving object in a preset time period after the current frame of data based on each existing predicted motion trajectory in step S102 may be:

[0042] S201. Determine a target existing predicted motion trajectory from multiple existing predicted motion trajectories of each first moving object based on the actual position information of each first moving object at the current moment and the predicted position information of the existing predicted motion trajectories at the current moment.

[0043] Among them, the target existing predicted motion trajectory can be considered as the predicted motion trajectory that best matches the actual running situation of the first moving object. After using the preset detection algorithm to process the data of this frame and obtaining the actual position information of the first moving object at the current moment, the computing device can compare the actual position information of the first moving object at the current moment with the predicted position information of each existing predicted motion trajectory obtained during the processing of the previous frame data at the current moment, and select the existing predicted motion trajectory that best matches the current actual running situation of the first moving object from multiple existing predicted motion trajectories based on the comparison result, and determine this existing predicted motion trajectory as the target existing predicted motion trajectory.

[0044] As an optional implementation manner, the computing device can use the following process to determine the target existing predicted motion trajectory. Optionally, the above S201 may include the following steps:

[0045] S2011. For each first moving object, calculate the Euclidean distance between the actual position information of the first moving object at the current moment and the predicted position information of the existing predicted motion trajectory at the current moment.

[0046] Specifically, the computing device can calculate the Euclidean distance trj between the actual position information of the first moving object at the current moment and the predicted position information of each existing predicted motion trajectory at the current moment according to the following formula 1 or a variant of formula 1.

[0047]

[0048] Among them, (X 预测 , Y 预测 ) is the predicted position information of the first moving object in the existing predicted motion trajectory at the current moment, and (X 实际 , Y 实际 ) is the actual position information of the first moving object at the current moment.

[0049] S2012. Determine the existing predicted motion trajectory corresponding to the predicted position information with the smallest Euclidean distance as the target existing predicted motion trajectory.

[0050] Among them, the smaller the Euclidean distance, the higher the matching degree between the corresponding existing predicted motion trajectory and the actual running situation of the first moving object. On the contrary, the larger the Euclidean distance, the lower the matching degree between the corresponding existing predicted motion trajectory and the actual running situation of the first moving object. Therefore, the existing predicted motion trajectory corresponding to the predicted position information with the smallest Euclidean distance can be determined as the target existing predicted motion trajectory.

[0051] S202. Update the timestamps of the trajectory points in each target existing predicted motion trajectory to obtain the predicted motion trajectories of each first moving object in a preset time period after the current frame of data.

[0052] After obtaining the target existing predicted motion trajectory, the first trajectory point of the target existing predicted motion trajectory can be deleted, and the timestamps of the remaining trajectory points in the target existing predicted motion trajectory can be updated at the time interval between the trajectory points. Based on the updated trajectory points, the predicted motion trajectory of the first moving object in a preset time period after the current frame of data is formed.

[0053] For example, assuming that the time interval between the trajectory points is 100 ms, the timestamps of the remaining trajectory points in the target existing predicted motion trajectory can be updated to Told - 100. Where Told is the original timestamp of the remaining trajectory points.

[0054] In this embodiment, the computing device can determine the target existing predicted motion trajectory with the highest accuracy from multiple existing predicted motion trajectories of the first moving object based on the actual position information of the first moving object at the current moment and the predicted position information of the first moving object in the existing predicted motion trajectory, and update the target existing predicted motion trajectory, so as to obtain the predicted motion trajectory of the first moving object in a preset time period after the current frame of data, improving the accuracy of the predicted motion trajectory.

[0055] In practical applications, there is also such a situation. If it is determined that the number of second moving objects that have not been processed by the preset trajectory prediction model is large and has exceeded the processing capacity limit of the computing device, for this situation, the trajectory prediction of the second moving objects can be performed with reference to the process of the following embodiments. On the basis of the above embodiments, as an optional implementation manner, as Figure 3 shown, the process of S103 can be:

[0056] S301. Based on the position information of each second moving object at the current moment, screen out the second moving objects located within the preset range of the autonomous vehicle from the second set of objects to be predicted to form a third set of objects to be predicted.

[0057] Generally, the closer a moving object is to an autonomous vehicle, the greater its impact on the autonomous vehicle, and the farther a moving object is from the autonomous vehicle, the smaller its impact on the autonomous vehicle. Therefore, in order to improve the real-time performance of trajectory prediction, based on the position information of each second moving object at the current moment, second moving objects located within the preset range of the autonomous vehicle can be screened out from the second set of objects to be predicted, and this part of the second moving objects can be preferentially processed through a preset trajectory prediction model. For other second moving objects outside the preset range, they can be processed when the next frame of data arrives. Among them, the preset range can be set based on the actual situation. For example, the preset range can be set within 30 meters.

[0058] S302. Through the preset trajectory prediction model, determine the predicted movement trajectories of each second moving object in the third set of objects to be predicted within a preset time period after this frame of data.

[0059] Among them, the computing device can input the historical movement trajectory data of each second moving object in the third set of objects to be predicted into the preset trajectory prediction model, and perform model prediction through the preset trajectory prediction model to obtain the predicted movement trajectories of each second moving object within a preset time period after this frame of data.

[0060] In this embodiment, the computing device can preferentially use the preset trajectory prediction model to predict the future driving trajectories of each second moving object located within the preset range of the autonomous vehicle, that is, preferentially use the preset trajectory prediction model to predict the future driving trajectories of the second moving objects with a higher impact on the autonomous vehicle, improving the real-time performance and accuracy of the trajectory prediction of the second moving objects with a higher impact on the autonomous vehicle, thereby assisting the driving decision-making of the autonomous vehicle and contributing to improving the safety of the autonomous vehicle.

[0061] As another alternative implementation, as Figure 4 shown, the process of S103 above can be:

[0062] S401. Based on the driving information of each second moving object, determine the priority of each second moving object.

[0063] Among them, the driving information may include, but is not limited to, the geographical location of the second moving object, the driving speed of the second moving object, the driving direction, the driving acceleration, and the distance between the second moving object and the autonomous vehicle. Generally, a second moving object that is farther away from the autonomous vehicle and has relatively slower driving speed and driving acceleration can be considered to have less impact on the autonomous vehicle, so the priority of this second moving object is determined to be lower, and its motion trajectory can be predicted when the next frame of data arrives; on the contrary, a second moving object that is closer to the autonomous vehicle and has relatively faster driving speed and driving acceleration can be considered to have greater impact on the autonomous vehicle, so the priority of this second moving object is determined to be higher, and its motion trajectory needs to be predicted as soon as possible.

[0064] S402. Determine each second moving object whose priority meets the preset requirements as an object to be predicted.

[0065] Among them, after obtaining the priorities of each second moving object, each second moving object whose priority meets the preset requirements can be determined as an object to be predicted. For example, if the priorities of each second moving object are represented by numerical values, each second moving object whose priority exceeds the preset threshold can be determined as an object to be predicted.

[0066] S403. Determine the predicted motion trajectories of each object to be predicted in a preset time period after the current frame of data through a preset trajectory prediction model corresponding to the type of each object to be predicted.

[0067] Among them, each object to be predicted can be a pedestrian or other vehicle. Different types of moving objects can use different preset trajectory prediction models for trajectory prediction. For example, when the object to be predicted is a pedestrian, the computing device can use a pre-trained pedestrian trajectory prediction model for trajectory prediction; when the object to be predicted is a vehicle, the computing device can use a pre-trained vehicle trajectory prediction model for trajectory prediction.

[0068] That is to say, the computing device can input the historical motion trajectory data of each object to be predicted into a preset trajectory prediction model corresponding to the type of each object to be predicted, and perform model prediction through the corresponding preset trajectory prediction model to obtain the predicted motion trajectories of each object to be predicted in a preset time period after the current frame of data.

[0069] In this embodiment, the computing device may determine the priorities of the second moving objects based on the driving information of each second moving object, and preferentially use the corresponding preset trajectory prediction model to predict the future driving trajectories of the second moving objects whose priorities meet the preset requirements, that is, preferentially use the preset trajectory prediction model to predict the future driving trajectories of the second moving objects with a higher impact on the autonomous vehicle, improving the timeliness and accuracy of the trajectory prediction of the second moving objects with a higher impact on the autonomous vehicle, thereby assisting the driving decision-making of the autonomous vehicle and contributing to improving the safety of the autonomous vehicle.

[0070] For the convenience of understanding by those skilled in the art, the following details the trajectory prediction process of moving objects in autonomous driving. Specifically:

[0071] For the first-frame data collected by the autonomous vehicle, when the number of moving objects in the first-frame data exceeds the processing capacity limit of the computing device, the computing device may first determine the moving objects in the first-frame data that have a greater impact on the autonomous vehicle (for example, it can be determined based on the driving information of each moving object), and preferentially use the preset trajectory prediction model to predict the trajectories of these moving objects. Then, when the second-frame data arrives, the computing device detects the moving objects in the second-frame data to obtain the set of moving objects corresponding to the second-frame data, and determines the first moving objects that have been processed by the preset trajectory prediction model before the second-frame data and the second moving objects that have not been processed by the preset trajectory prediction model from this set of moving objects; further, for each first moving object, the computing device obtains the existing predicted moving trajectories of each first moving object obtained through the preset trajectory prediction model before the second-frame data, and determines the predicted moving trajectories of each first moving object in the preset time period after the second-frame data based on the existing predicted moving trajectories. For each second moving object, the computing device uses the preset trajectory prediction model to determine the predicted moving trajectories of each second moving object in the preset time period after the second-frame data. And so on, the computing device uses this method to predict the future moving trajectories of the moving objects in each frame of data, not only improving the timeliness of the trajectory prediction of each moving object, but also improving the number of predicted moving trajectories output by the computing device as a whole, and at the same time ensuring the accuracy of the output predicted moving trajectories.

[0072] Figure 5 It is a schematic structural diagram of a device for predicting the trajectory of a moving object in autonomous driving provided by an embodiment of the present application. As Figure 5 shown, the device may include: a determination module 501, an acquisition module 502, a first prediction module 503, and a second prediction module 504.

[0073] Specifically, the determination module 501 is configured to determine, from the set of moving objects corresponding to the current frame data collected by the autonomous vehicle, the first moving objects that have been processed by the preset trajectory prediction model before the current frame data, to form a first set of objects to be predicted, and to determine the second moving objects that have not been processed by the preset trajectory prediction model, to form a second set of objects to be predicted;

[0074] The acquisition module 502 is configured to acquire the existing predicted movement trajectories predicted by the preset trajectory prediction model for each first moving object in the first set of objects to be predicted before the current frame data;

[0075] The first prediction module 503 is configured to determine the predicted movement trajectories of each first moving object in a preset time period after the current frame data based on the existing predicted movement trajectories;

[0076] The second prediction module 504 is configured to determine the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model.

[0077] The trajectory prediction device for moving objects in autonomous driving provided by the embodiments of the present application determines the first moving objects and the second moving objects in the set of moving objects corresponding to the current frame data collected by the autonomous vehicle. For each first moving object that has been processed by the preset trajectory prediction model before the current frame data, the predicted movement trajectories of each first moving object in a preset time period after the current frame data are determined based on the existing predicted movement trajectories of each first moving object. For each second moving object that has not been processed by the preset trajectory prediction model, the predicted movement trajectories of each second moving object in a preset time period after the current frame data are determined through the preset trajectory prediction model. That is to say, the number of moving objects processed by the preset trajectory prediction model is reduced, the computing pressure on the computing unit of the autonomous vehicle is alleviated, and thus the real-time performance of the trajectory prediction of each moving object corresponding to the current frame data is improved. At the same time, the predicted movement trajectories of some moving objects can be determined by the existing predicted movement trajectories predicted before the current frame data. Overall, the number of predicted movement trajectories output by the computing unit is also increased, and the accuracy of the output predicted movement trajectories is ensured.

[0078] Based on the above embodiments, optionally, the acquisition module 502 is specifically configured to acquire multiple existing predicted movement trajectories obtained by the preset trajectory prediction model for each first moving object in the first set of objects to be predicted before the current frame data;

[0079] Correspondingly, the first prediction module 503 may include: a determination unit and an update unit.

[0080] Specifically, the determination unit is configured to determine a target existing predicted motion trajectory from multiple existing predicted motion trajectories of each first moving object based on the actual position information of each first moving object at the current moment and the predicted position information of the existing predicted motion trajectory at the current moment.

[0081] The update unit is configured to update the timestamps of the trajectory points in each target existing predicted motion trajectory to obtain the predicted motion trajectories of each first moving object in a preset time period after the current frame of data.

[0082] Based on the above embodiments, optionally, the determination unit is specifically configured to, for each first moving object, calculate the Euclidean distance between the actual position information of the first moving object at the current moment and the predicted position information of the existing predicted motion trajectory at the current moment; and determine the existing predicted motion trajectory corresponding to the predicted position information with the smallest Euclidean distance as the target existing predicted motion trajectory.

[0083] Based on the above embodiments, optionally, the second prediction module 504 is specifically configured to screen out second moving objects located within a preset range of the autonomous driving vehicle from the second set of objects to be predicted based on the position information of each second moving object at the current moment to form a third set of objects to be predicted; and determine the predicted motion trajectories of each second moving object in the third set of objects to be predicted in a preset time period after the current frame of data through the preset trajectory prediction model.

[0084] Based on the above embodiments, optionally, the second prediction module 504 is specifically configured to determine the priority of each second moving object based on the driving information of each second moving object; determine the objects to be predicted as the second moving objects whose priorities meet the preset requirements; and determine the predicted motion trajectories of each object to be predicted in a preset time period after the current frame of data through the preset trajectory prediction model corresponding to the type of each object to be predicted.

[0085] Optionally, the preset trajectory prediction model is a recurrent neural network model.

[0086] In one embodiment, a computing device is provided. The computing device may be a device installed in an autonomous driving vehicle or a cloud device. Refer to Figure 6, the computing device may include a processor, a memory, a network interface, and a database connected by a system bus. Among them, the processor of the computing device is used to provide computing and control capabilities. The memory of the computing device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computing device is used to store data during the trajectory prediction process of moving objects in autonomous driving. The network interface of the computing device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for predicting the trajectory of moving objects in autonomous driving.

[0087] Those skilled in the art can understand that Figure 6 the structure shown in

[0088] In one embodiment, a computing device is provided, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the following steps are implemented:

[0089] From the set of moving objects corresponding to the current frame data collected by the autonomous driving vehicle, determine the first moving objects that have been processed by the preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determine the second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted;

[0090] Obtain the existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data, and determine the predicted movement trajectories of each first moving object in a preset time period after the current frame data based on the existing predicted movement trajectories;

[0091] Through the preset trajectory prediction model, determine the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data.

[0092] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0093] From the set of moving objects corresponding to the current frame data collected by the autonomous vehicle, determine the first moving objects that have been processed by the preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determine the second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted;

[0094] Obtain the existing predicted movement trajectories of the first moving objects in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data, and determine the predicted movement trajectories of the first moving objects in a preset time period after the current frame data based on the existing predicted movement trajectories;

[0095] Through the preset trajectory prediction model, determine the predicted movement trajectories of the second moving objects in the second set of objects to be predicted in a preset time period after the current frame data.

[0096] The trajectory prediction device, equipment, and storage medium of the moving object in autonomous driving provided in the above embodiments can execute the trajectory prediction method of the moving object in autonomous driving provided in any embodiment of the present disclosure, and have corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in the above embodiments can be found in the trajectory prediction method of the moving object in autonomous driving provided in any embodiment of the present disclosure.

[0097] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0098] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0099] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A trajectory prediction method for moving objects in autonomous driving, characterized in that, Including: From the set of moving objects corresponding to the current frame data collected by the autonomous vehicle, determining first moving objects that have been processed by a preset trajectory prediction model before the current frame data to form a first set of objects to be predicted, and determining second moving objects that have not been processed by the preset trajectory prediction model to form a second set of objects to be predicted; Obtaining the existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data, and determining the predicted movement trajectories of each first moving object in a preset time period after the current frame data based on the existing predicted movement trajectories; Determining the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model; Wherein, the obtaining the existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data includes: Obtaining multiple existing predicted movement trajectories of each first moving object in the first set of objects to be predicted obtained by the preset trajectory prediction model before the current frame data; Correspondingly, the determining the predicted movement trajectories of each first moving object in a preset time period after the current frame data based on the existing predicted movement trajectories includes: Based on the actual position information of each first moving object at the current moment and the predicted position information at the current moment in the existing predicted movement trajectories, determining the target existing predicted movement trajectory from the multiple existing predicted movement trajectories of each first moving object; Updating the timestamps of the trajectory points in each target existing predicted movement trajectory to obtain the predicted movement trajectories of each first moving object in a preset time period after the current frame data; Wherein, the determining the target existing predicted movement trajectory from the multiple existing predicted movement trajectories of each first moving object based on the actual position information of each first moving object at the current moment and the predicted position information at the current moment in the existing predicted movement trajectories includes: For each first moving object, calculating the Euclidean distance between the actual position information of the first moving object at the current moment and the predicted position information at the current moment in the existing predicted movement trajectories; Determining the existing predicted movement trajectory corresponding to the predicted position information with the minimum Euclidean distance as the target existing predicted movement trajectory.

2. The method according to claim 1, wherein The determining the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model includes: Based on the position information of each second moving object at the current moment, screening out second moving objects within a preset range of the autonomous vehicle from the second set of objects to be predicted to form a third set of objects to be predicted; Determining the predicted movement trajectories of each second moving object in the third set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model.

3. The method according to claim 1, characterized in that The determining the predicted movement trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame data through the preset trajectory prediction model includes: Based on the driving information of each second moving object, determine the priority of each second moving object; Determine each second moving object whose priority meets the preset requirements as an object to be predicted; Through a preset trajectory prediction model corresponding to the type of each object to be predicted, determine the predicted motion trajectories of each object to be predicted in a preset time period after the current frame of data.

4. The method according to claim 1, wherein The preset trajectory prediction model is a recurrent neural network model.

5. A trajectory prediction device for moving objects in autonomous driving, characterized in that, It includes: A determination module, configured to determine, from the set of moving objects corresponding to the current frame of data collected by the autonomous vehicle, first moving objects that have been processed by the preset trajectory prediction model before the current frame of data, to form a first set of objects to be predicted, and determine second moving objects that have not been processed by the preset trajectory prediction model, to form a second set of objects to be predicted; An acquisition module, configured to acquire the existing predicted motion trajectories predicted by the preset trajectory prediction model for each first moving object in the first set of objects to be predicted before the current frame of data; A first prediction module, configured to determine the predicted motion trajectories of each first moving object in a preset time period after the current frame of data based on the existing predicted motion trajectories; A second prediction module, configured to determine the predicted motion trajectories of each second moving object in the second set of objects to be predicted in a preset time period after the current frame of data through the preset trajectory prediction model; Wherein, the acquisition module is specifically configured to acquire multiple existing predicted motion trajectories obtained by the preset trajectory prediction model for each first moving object in the first set of objects to be predicted before the current frame of data; Correspondingly, the first prediction module includes: A determination unit, configured to determine a target existing predicted motion trajectory from the multiple existing predicted motion trajectories of each first moving object based on the actual position information of each first moving object at the current moment and the predicted position information at the current moment in the existing predicted motion trajectories; An update unit, configured to update the time stamps of the trajectory points in each target existing predicted motion trajectory to obtain the predicted motion trajectories of each first moving object in a preset time period after the current frame of data; Wherein, the determination unit includes: An Euclidean distance calculation sub-unit, configured to calculate, for each first moving object, the Euclidean distance between the actual position information of the first moving object at the current moment and the predicted position information at the current moment in the existing predicted motion trajectories; A target existing predicted motion trajectory determination sub-unit, configured to determine the existing predicted motion trajectory corresponding to the predicted position information with the minimum Euclidean distance as the target existing predicted motion trajectory.

6. A computing device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images