WO2023050749A1 - Procédé et appareil de prédiction de voie, dispositif électronique et support de stockage - Google Patents

Procédé et appareil de prédiction de voie, dispositif électronique et support de stockage Download PDF

Info

Publication number
WO2023050749A1
WO2023050749A1 PCT/CN2022/084204 CN2022084204W WO2023050749A1 WO 2023050749 A1 WO2023050749 A1 WO 2023050749A1 CN 2022084204 W CN2022084204 W CN 2022084204W WO 2023050749 A1 WO2023050749 A1 WO 2023050749A1
Authority
WO
WIPO (PCT)
Prior art keywords
grid
sub
running
moment
target
Prior art date
Application number
PCT/CN2022/084204
Other languages
English (en)
Chinese (zh)
Inventor
李樊
孙钢
刘春晓
石建萍
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Publication of WO2023050749A1 publication Critical patent/WO2023050749A1/fr

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • B60W60/001—Planning or execution of driving tasks
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
    • B60W2050/0001—Details of the control system
    • B60W2050/0043—Signal treatments, identification of variables or parameters, parameter estimation or state estimation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Definitions

  • the present disclosure relates to the technical field of prediction, and relates to a trajectory prediction method, device, electronic equipment, and storage medium.
  • Object behavior prediction is an important issue in the field of automatic driving. Specifically, it can be expressed as the prediction of the running trajectory of the objects perceived by the automatic driving system.
  • the above-mentioned objects include but are not limited to cars, two-wheeled electric vehicles, bicycles, and pedestrians.
  • Object behavior prediction is an important sub-module of the automatic driving system, and it is an important basis for the automatic driving system to perform scene understanding, behavior decision-making and motion planning.
  • Embodiments of the present disclosure at least provide a trajectory prediction method, device, electronic equipment, and storage medium.
  • an embodiment of the present disclosure provides a trajectory prediction method, the method is executed by an electronic device, and the method includes:
  • the preset running grid includes at least one running sub-grid; the running sub-grid includes at least one reference object at a plurality of future reference times The position point; the future reference time is the time after the sample time as the starting point, after the first preset duration; the reference object and the target object have the same operation mode;
  • the target operation sub-grid where the target object is located at a second moment; wherein, the second moment is a moment after the first preset time period elapses at the first moment;
  • the running trajectory of the target object from the first moment to the second moment is determined.
  • the running trajectory of the target object is predicted by using the preset running grid that matches the running mode of the target object, which can fully consider the characteristics such as the running speed and mode of the target object, and improve the pertinence and accuracy of trajectory prediction.
  • Accuracy
  • different objects use preset running grids for trajectory prediction, so the scheme in this aspect is generalizable, and there is no need to set different models or methods for different objects, making the scheme in this aspect High efficiency; at the same time, this aspect is not based on the point set when predicting the trajectory, but uses the grid. Compared with the point set, the dimension of the processed data is reduced, so the amount of data processed by the solution in this aspect is effectively reduced. Efficiency is improved.
  • the embodiment of the present disclosure also provides a trajectory prediction device, including:
  • the information acquisition part is configured to acquire the operation information of the target object at the first moment
  • the positioning preprocessing part is configured to determine a preset running grid that matches the running mode of the target object; the preset running grid includes at least one running sub-grid; the running sub-grid includes at least one The position points of the reference object at a plurality of future reference moments; the future reference moment is the moment after the first preset time period starting from the sample moment; the reference object has the same operating mode as the target object;
  • the positioning part is configured to determine the target operation sub-grid where the target object is located at a second moment based on the operation information; wherein, the second moment is after the first preset time period elapses at the first moment moment;
  • the trajectory prediction part is configured to determine the trajectory of the target object from the first moment to the second moment based on the target operation sub-grid and the operation information.
  • an embodiment of the present disclosure further provides an electronic device, including: a processor, a memory, and a bus, the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processing
  • the processor communicates with the memory through a bus, and when the machine-readable instructions are executed by the processor, the above-mentioned first aspect, or the steps in any possible implementation manner of the first aspect are executed.
  • embodiments of the present disclosure further provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned first aspect, or any of the first aspects of the first aspect, may be executed. Steps in one possible implementation.
  • an embodiment of the present disclosure further provides a computer program product, the computer program product includes a computer program or instruction, and when the computer program or instruction is run on a computer, the computer executes the above-mentioned first aspect , or a step in any possible implementation manner in the first aspect.
  • FIG. 1 shows a flowchart of a trajectory prediction method provided by an embodiment of the present disclosure
  • Fig. 2 shows a schematic diagram of road condition information provided by an embodiment of the present disclosure
  • Fig. 3 shows a flow chart of determining a target running sub-grid provided by an embodiment of the present disclosure
  • FIG. 4 shows a flow chart of generating a matching preset operation grid for a certain operation mode provided by an embodiment of the present disclosure
  • FIG. 5 shows a schematic diagram of a preset running grid provided by an embodiment of the present disclosure
  • FIG. 6 shows a flow chart of determining a dithering sub-grid provided by an embodiment of the present disclosure
  • Fig. 7 shows a schematic diagram of a trajectory prediction device provided by an embodiment of the present disclosure
  • Fig. 8 shows a schematic diagram of an electronic device provided by an embodiment of the present disclosure.
  • the present disclosure provides a trajectory prediction method, device, electronic equipment, and storage medium.
  • the disclosure utilizes a preset operating grid that matches the operating mode of the target object to predict the operating trajectory of the target object, which can fully Considering the characteristics of the target object's running speed and mode, the pertinence and accuracy of trajectory prediction are improved; in addition, different objects use the preset running grid for trajectory prediction, so the scheme in this aspect has generalization In addition, there is no need to set different models or methods for different objects, so that the solution in this aspect is more efficient; at the same time, the present disclosure is not based on point sets when predicting trajectory, but uses grids, which are processed relative to point sets The data dimension of is reduced, so the amount of data processed by the solution in this aspect is effectively reduced and the efficiency is improved.
  • trajectory prediction method provided by the embodiments of the present disclosure will be described below by taking the executing subject as a device capable of computing as an example.
  • the embodiment of the present disclosure provides a trajectory prediction method, which may include the following steps:
  • the above-mentioned target object is the object that needs trajectory prediction.
  • the running information may include road condition information of the location of the target object at the first moment and movement state information of the target object within a period of time from the first moment.
  • the above road condition information may include lane lines 201 around the target object, zebra crossing outline 202 , intersection outline 203 and other information.
  • the above road condition information may exist in the form of pictures.
  • the above-mentioned motion state information may include information such as the position, speed, and orientation of the target object within a period of time. For example, when the first moment is the current moment and the aforementioned period of time is 3s, the motion state information includes information such as the position, speed, and orientation of the target object in the past 3s.
  • the position of the target object is recorded as (x, y), the heading is recorded as heading, and the speed is recorded as speed.
  • the latest historical 16-frame data is taken every 0.2s, and each row records the corresponding motion state information of one frame. Then A 16 ⁇ 4 matrix is obtained, which is the motion state information within 3s.
  • the moment of the second preset duration is passed forward to obtain the third moment; after that, acquiring the motion state information of the target object from the third moment to the first moment , and use the acquired running state information as the motion state information of the target object at the first moment; finally, based on the motion state information of the target object at the first moment and the target object at the first
  • the road condition information of the location at the time is used to determine the running information of the target object at the first time.
  • the motion state information of the target object at the first moment and the road condition information of the location of the target object at the first moment may be directly used as the running information of the target object at the first moment.
  • the above-mentioned third moment is earlier than the first moment, and the motion state information between the third moment and the first moment is used as the motion state information of the target object at the first moment for trajectory prediction, which can not only increase the number of participants in trajectory prediction
  • the amount of information can improve the accuracy of trajectory prediction, and the operating status information during the period from the third moment to the first moment can reflect the operation law of the target object to a certain extent.
  • the trajectory prediction can further improve the accuracy of trajectory prediction.
  • S120 Determine a preset running grid that matches the running mode of the target object; the preset running grid includes at least one running sub-grid; the running sub-grid includes at least one reference object in multiple futures The position point of the reference time; the future reference time is the time after the first preset time period starting from the sample time. Wherein, the reference object and the target object have the same running mode.
  • a preset operation grid may correspond to at least one operation mode, and one operation mode may correspond to at least one type of object.
  • the operating modes of different types of objects can be different, for example, a car has a turning radius, and a pedestrian has no turning radius and can turn freely.
  • the above-mentioned reference object and the target object have the same operation mode, so the location points corresponding to the reference object can be used to form a preset operation grid to predict the location information and operation trajectory of the target object.
  • the operation mode of the target object can be determined first and multiple preset operation grids can be obtained; and then from the multiple preset operation grids, the selection matches the operation mode of the target object The default run grid.
  • Different objects have different characteristics such as running speed and mode, so different objects have different running modes, and different running modes directly affect the trajectory prediction, so the target is based on the preset running grid that matches the running mode of the target object Objects can be used for trajectory prediction, which can effectively improve the pertinence and accuracy of prediction.
  • the first preset duration may be determined according to the speed of different objects, the first preset duration corresponding to a faster object may be shorter, and the second preset duration corresponding to a smaller speed object may be longer.
  • the first preset duration may be set to 5s.
  • Each position point in the preset running grid is the reference object with the sample moment as the movement start time, and the movement end point at the future reference time, the target object takes the first moment as the movement start time, and the second moment as the movement end time, The position of the target object at the second moment is the movement end point of the target object.
  • the duration between the sample moment and the future reference moment is equal to the duration between the first moment and the second moment, both of which are the above-mentioned first preset duration.
  • the reference image can be used as a sample to predict the position or running trajectory of the target image.
  • the prediction time interval corresponding to the position point in the preset operation grid that is, the first preset duration
  • the preset operation grid can be used to predict the elapsed prediction time of the target object The position after the interval, and the running trajectory during it.
  • the following steps may be used to determine the target operation sub-grid where the target object is located at the second moment: first, based on the operation information, determine the position information of the target object at the second moment; then, based on The location information determines the target operation sub-grid where the target object is located at the second moment.
  • the step of determining the target running sub-grid in the above steps can be realized by using a pre-trained neural network, as shown in Figure 3, the movement state information of the target object in the running information at the first moment and the target object at the first moment
  • the road condition information of the location is input into the trained neural network, and the neural network processes the input data to determine the target operation sub-grid where the target object is at the second moment, and output the target operation sub-grid identification symbol.
  • the above-mentioned running state information may include information such as the speed, position, and orientation of the target object, and the above-mentioned road condition information may include information such as lane lines and zebra crossing outlines around the target object.
  • the neural network can more accurately determine the location information of the target object at the second moment.
  • the running sub-grid includes the position points of multiple reference objects, and the position points here can be considered as a kind of sample points. According to the position points in the running sub-grid or according to the position of the running sub-grid, combined The position information of the target object at the second moment can more accurately determine the target running sub-grid where the target object is located.
  • the target operation sub-grid includes a plurality of position points. Based on the position information of each position point in the target operation sub-grid, the position information corresponding to the target operation sub-grid can be determined. The position information can be considered as the target object in the second Time-point location information. Afterwards, based on the terminal position information of the target object at the second moment and the running information of the target object at the first moment, the running trajectory of the target object from the first moment to the second moment can be predicted.
  • the coordinate mean value corresponding to each position point may be determined, and the determined coordinate mean value may be used as the position information corresponding to the target operation sub-grid.
  • the center point of the target running sub-grid is calculated, and the position information of the center point is used as the position information corresponding to the target running sub-grid.
  • the destination position information of the target object at the second moment and the running information of the target object at the first moment can be input into a pre-trained neural network, and after processing by the neural network, the target object can be output from the first moment to the The running trajectory between the second moments.
  • the terminal position information of the target object at the second moment can be determined more accurately, and then based on the terminal position information and running information, the target object can be determined more accurately from the first The running trajectory between the second moment and the second moment.
  • the running trajectory of the target object is predicted by using the preset running grid that matches the target object, which can fully consider the characteristics of the target object’s running speed, mode, etc., and improve the pertinence and accuracy of trajectory prediction;
  • Different objects use preset running grids for trajectory prediction, so the scheme in this aspect is generalized, and there is no need to set different models or methods for different objects, making the scheme in this aspect more efficient; at the same time
  • the trajectory prediction is not based on the point set, but the grid is used. Compared with the point set, the dimension of the processed data is reduced. Therefore, the amount of data processed by the solution in this aspect is effectively reduced and the efficiency is improved.
  • the present disclosure also provides a step of generating a matching preset operation grid for a certain operation mode:
  • Reference objects can be cars, pedestrians, etc.
  • a running mode may correspond to a reference object, or may correspond to multiple reference objects.
  • multiple sample moments can be selected by sliding.
  • a future reference time that is a future time relative to the sample time can be determined according to the first preset duration.
  • the position information on the sample trajectory corresponding to the future reference time is used as the sample position information of the reference object.
  • the position point corresponding to the sample position information may be used as the position point of the reference object at the corresponding future reference moment.
  • each location point can be transformed into the same coordinate system and drawn on a picture, and then the area with the location point in the picture is divided to obtain a preset running grid including at least one running sub-grid.
  • a plurality of position points as possible samples can be determined more accurately, and the preset positions matching the operation mode can be generated more accurately by using the position points. Run grid.
  • the above image can also be drawn in combination with the speed of the reference object at each position point. Combined with the speed at each position point, a preset running grid with more comprehensive and rich information can be generated, which in turn helps to improve the accuracy of estimation and prediction.
  • the color of each location point is determined based on the velocity of the corresponding reference object at each location point.
  • a plurality of speed intervals can be set in advance, and the color corresponding to each speed interval can be determined, and then, for each location point, according to the speed of the location point, the speed interval where the location point is located can be determined, and the speed interval The corresponding color is used as the color of the position point.
  • each location point is drawn according to the location information and color of each location point, and a picture including each location point is obtained.
  • the position points on the picture are segmented to obtain a preset running grid including at least one running sub-grid; wherein, the preset conditions At least one of the following is included; the number of location points in any running sub-grid is within a preset value range; the ratio of the location points in the running sub-grid to all the location points is greater than the preset ratio; The speed difference corresponding to different position points in the same running sub-grid is smaller than the preset value.
  • the above condition that the number of location points in any running sub-grid is within a preset value range can make the number of location points in different running sub-grids close to each other without being far apart.
  • the above preset proportion can be set according to the actual application scenario, for example, it can be set to 90%.
  • the preset proportion is set to be relatively large, so that most of the position points are within the preset operating grid, which can be more comprehensive and accurately predict where the target object is located to run the subraster.
  • the speed difference corresponding to different position points in the same running sub-grid is smaller than the preset value, and the position points with the same color or similar colors can be divided into the same running sub-grid. Among them, the positions with similar speeds have similar colors.
  • the preset operating grid in the figure is obtained by segmenting the car as a reference object.
  • the number of position points included in different operating sub-grids 501 in the figure is similar, and the position points with similar colors can be located in the same position. Run within the subgrid.
  • the grid may be fan-shaped, and of course may also be in other shapes, which is not limited in this embodiment of the present disclosure.
  • the speeds of objects in different operating modes are different, so the number of position points included in the operating sub-grids matching different operating modes may be different, that is, the preset value ranges are different.
  • the following steps can be used to determine a The preset value range corresponding to the running sub-grid matching the running mode:
  • Using the preset value range determined by the average speed of the object can make the position of each position point in the same running sub-grid close, and using the same position to represent the position of each position point in the entire running sub-grid will not affect The accuracy of the position. Based on this, a position corresponding to a running sub-grid is used to represent the position of each position point in the running sub-grid. Compared with the point set, in trajectory prediction, the data processing dimension is reduced, which is conducive to prediction Stability postprocessing.
  • the foregoing embodiments can generate a preset operating grid matching the operating mode based on the position information, speed, and preset conditions of each position point, thereby improving the accuracy of trajectory prediction.
  • the modeling method also needs to be conducive to the post-processing of prediction stability, because the object information obtained from the perception system is noisy, and the noise is easy to cause the prediction of the object’s trajectory. "Jitter” (recovery after an unreasonable change in the predicted trajectory), which can negatively affect the comfort and safety of an automated driving system. Therefore, in some embodiments, after the target running sub-grid is determined, it is also necessary to detect whether the target running sub-grid is a dithering sub-grid.
  • the target running sub-grid is a dithering sub-grid
  • the most recently determined non-jittering sub-grid is used as the final target running sub-grid and participates in trajectory prediction
  • the target running sub-grid is a non-jittering sub-grid
  • the following steps can be used to detect whether the target running sub-grid is a dithering sub-grid:
  • a plurality of historical operation sub-grids determined before the target operation sub-grid is determined; a first historical operation sub-grid among the plurality of historical operation sub-grids is a non-jittering sub-grid.
  • the above-mentioned multiple historical operation sub-grids may be a plurality of consecutive target operation sub-grids determined before the target operation sub-grid is determined.
  • the target running sub-grid is a dithering sub-grid.
  • the target running sub-grid is the same sub-grid as the first historical running sub-grid
  • the target running sub-grid is not the same sub-grid as the first historical running sub-grid, and the number of jittering sub-grids among the multiple historical running sub-grids is greater than a preset number , determine that the target running sub-grid is a non-jittering sub-grid.
  • the target running sub-grid is not the same sub-grid as the first historical running sub-grid, and the number of dithering sub-grids among the multiple historical running sub-grids is less than or equal to a preset number of In this case, it is determined that the target running sub-grid is a dithering sub-grid.
  • the target running sub-grid is the same sub-grid as the first historical running sub-grid, indicating that the running status of the target object has not changed.
  • the determined target running sub-grid is considered accurate and is a non-jittering sub-grid. grid; when the number of jittering sub-grids in multiple historical running sub-grids is large, it means that the running state of the target object has changed, which is a change of normal motion state, not jittering. is a non-dithered subraster. When the number of jittering sub-grids in multiple historical running sub-grids is small, it means that the running status of the target object has not changed. If the target running sub-grid is not the same sub-grid as the first historical running A raster indicating that the target running subraster is determined to be inaccurate, and the subraster is dithered.
  • the first identified target running subraster is preset to a non-dithered subraster.
  • the following steps can be used to achieve the above-mentioned purpose of determining whether the target operating sub-grid is a dithering sub-grid, and determining the final target operating sub-grid:
  • Step 1 After determining a target running sub-grid, judge whether the queue used to store the identifier of the target running sub-grid is empty; if it is empty, it means that the determined target running sub-grid is the first determined target Run the sub-grid, consider the target running sub-grid to be a non-jittering sub-grid, store the identifier of the target running sub-grid in the queue, and use the target running sub-grid as the final target running sub-grid .
  • Step 2 When the above queue is not empty, judge whether the identifier of the target running subgrid is the same as the first identifier stored in the queue; if they are the same, it indicates that the target running subgrid is the same as the first identifier stored in the queue.
  • the running sub-grid corresponding to the identifier is the same running sub-grid.
  • the target running sub-grid is a non-jittering sub-grid. There is no need to store the identifier of the target running sub-grid in the queue.
  • This target run subraster can be used as the final target run subraster.
  • Step 3 If the identifier of the target running sub-grid is not the same as the first identifier stored in the queue, store the identifier of the target running sub-grid at the end of the queue, and determine the current stored in the queue. Whether the number of identifiers is greater than the given value; if it is not greater than the given value, it indicates that the target operating sub-grid is a jittering sub-grid. At this time, the operating sub-grid corresponding to the first identifier stored in the queue needs to be used as The final goal is to run subrasters.
  • Step 4 The format of the identifier stored in the current queue is greater than the given value, indicating that the motion state of the target object has changed, and the target running sub-grid is a non-jittering sub-grid. Identifiers other than the tail identifier are cleared. At this time, the identifier at the end of the queue becomes the identifier at the head of the queue, and the target running sub-grid is used as the final target running sub-grid.
  • the running sub-grid corresponding to the identifier stored in the queue is the historical running sub-grid in the above embodiment.
  • a plurality of historical operation sub-grids are determined, and it is possible to more accurately determine whether the target operation sub-grid is a jittering sub-grid;
  • taking the first historical operation sub-grid as the final target operation sub-grid can effectively reduce the impact of jitter on the trajectory prediction results.
  • the dimension of data processing is improved, which can reduce the difficulty of post-processing of prediction stability and improve the efficiency of post-processing of prediction stability.
  • the possible future motion end points of the object i.e. the end points of the running track
  • each grid represents a possible running track
  • the future running of the object Trajectories can be modeled as a grid of preset runs.
  • the size and number of grids and the trajectory represented by the grid can be customized according to the characteristics of the motion mode of different objects, so that the predictions of different objects are targeted and help to improve the accuracy of trajectory prediction.
  • the object trajectory prediction task is defined as a grid classification task.
  • a grid can contain multiple possible motion end points, thereby reducing the task dimension of object behavior prediction, helping to suppress the jitter of the predicted trajectory, and improve the stability of the prediction. sex.
  • the stability post-processing of the above-mentioned embodiment is quite simple and direct. It is only necessary to suppress the jitter of the fan-shaped grid classification prediction to stabilize the prediction of the trajectory of the object. Compared with the method of modeling the predicted trajectory with a point set, the post-processing difficulty Relatively smaller.
  • the pre-trajectory prediction of the object is changed to grid prediction, thereby transforming the high-dimensional regression task into a low-dimensional classification task, reducing task difficulty and improving prediction stability.
  • the above embodiments associate the prediction mode with the motion mode of the object while the scheme is generalizable, so that the prediction mode can be adaptively adjusted for different objects, which helps to improve the prediction accuracy.
  • the prediction method in the above embodiment can be well connected with various stability post-processing methods, and the stability of the grid classification prediction can be improved through post-processing, thereby improving the stability of the object trajectory prediction.
  • the embodiment of the present disclosure also provides a trajectory prediction device corresponding to the trajectory prediction method. Since the problem-solving principle of the device in the embodiment of the disclosure is similar to the above-mentioned trajectory method in the embodiment of the disclosure, the implementation of the device can be See method implementation.
  • FIG. 7 it is a schematic diagram of the structure of a trajectory prediction device provided by an embodiment of the present disclosure, including:
  • the information acquiring part 701 is configured to acquire the running information of the target object at the first moment.
  • the positioning preprocessing part 702 is configured to determine a preset running grid that matches the running mode of the target object; the preset running grid includes at least one running sub-grid; the running sub-grid includes at least one A position point of a reference object at multiple future reference moments; the future reference moment is a moment after a first preset time period starting from the sample moment; the reference object has the same operation mode as the target object.
  • the positioning part 703 is configured to determine the target operation sub-grid where the target object is located at a second moment based on the operation information; wherein, the second moment is the first preset time after the first moment later moments;
  • the trajectory prediction part 704 is configured to determine the running trajectory of the target object from the first moment to the second moment based on the target running sub-grid and the running information.
  • the positioning preprocessing part 702 determines the preset operating grid matching the operating mode of the target object, it is configured to:
  • a preset operation grid matching the operation mode of the target object is screened; wherein, different preset operation grids correspond to different operation modes.
  • the positioning part 703 when determining the target operation sub-grid where the target object is located at the second moment based on the operation information, the positioning part 703 is configured to:
  • the running information includes motion state information of the target object at the first moment and road condition information of the location of the target object at the first moment.
  • the information acquiring part 701 when the information acquiring part 701 acquires the running information of the target object at the first moment, it is configured to:
  • the second preset time interval from the first moment is the third moment
  • the running information of the target object at the first moment is determined.
  • the positioning preprocessing part 702 is also configured to generate a matching preset operating grid for any operating mode:
  • a plurality of sample moments corresponding to the sample trajectory are determined, based on the first preset duration, a future reference moment corresponding to each sample moment is determined, and based on the sample trajectory, it is determined that at each future reference moment , the sample position information of the reference object corresponding to the sample trajectory;
  • the positioning preprocessing part 702 is configured to:
  • a preset operating grid matching the operating mode is determined based on each determined position point and the speed of the corresponding reference object at each position point.
  • the positioning preprocessing part 702 determines a preset operating grid matching the operating mode based on each determined position point and the velocity of the corresponding reference object at each position point, it is configured to :
  • the position points on the picture are segmented to obtain a preset running grid including at least one running sub-grid; wherein the preset conditions include the following At least one item; the number of location points in any running sub-grid is within the preset value range; the proportion of the location points in the running sub-grid in all the location points is greater than the preset proportion; the same running The speed difference corresponding to different position points in the sub-grid is smaller than the preset value.
  • the positioning preprocessing part 702 is further configured to determine the preset value range:
  • the preset value range corresponding to the operation mode is determined.
  • the trajectory prediction part 704 when determining the running trajectory of the target object from the first moment to the second moment based on the target running sub-grid and the running information, is configured to :
  • the running trajectory of the target object from the first moment to the second moment is determined.
  • trajectory prediction part 704 determines the target operation sub-grid where the target object is located at the second moment, and determines that the target object is located between the first moment and the second moment Before running the track, it is also configured as:
  • the first historical running sub-grid among the multiple historical running sub-grids is a non-jittering sub-grid
  • the first historical running sub-grid is used as the final target running sub-grid.
  • the trajectory prediction part 704 is further configured to:
  • the determined target operating sub-grid is used as the final target operating sub-grid.
  • the trajectory prediction part 704 is configured to:
  • the target running sub-grid is the same sub-grid as the first historical running sub-grid, it is determined that the target running sub-grid is a non-jittering sub-grid.
  • the trajectory prediction part 704 is configured to:
  • the target running sub-grid is not the same sub-grid as the first historical running sub-grid, and the number of jittering sub-grids among the multiple historical running sub-grids is greater than a preset number , determine that the target running sub-grid is a non-jittering sub-grid.
  • the trajectory prediction part 704 is further configured to:
  • the target running sub-grid is not the same sub-grid as the first historical running sub-grid, and the number of dithering sub-grids among the multiple historical running sub-grids is less than or equal to a preset number of In this case, it is determined that the target running sub-grid is a dithering sub-grid.
  • an embodiment of the present disclosure also provides an electronic device.
  • FIG. 8 it is a schematic structural diagram of an electronic device 800 provided by an embodiment of the present disclosure, including a processor 81 , a memory 82 , and a bus 83 .
  • the memory 82 is used to store execution instructions, including a memory 821 and an external memory 822; the memory 821 here is also called an internal memory, and is used to temporarily store calculation data in the processor 81 and exchange data with external memory 822 such as a hard disk.
  • the processor 81 exchanges data with the external memory 822 through the memory 821.
  • the processor 81 communicates with the memory 82 through the bus 83, so that the processor 81 executes the following instructions:
  • the operation information of the target object at the first moment determine the preset operation grid matching the operation mode of the target object; the preset operation grid includes at least one operation sub-grid; the operation sub-grid Including the position point of at least one reference object at a plurality of future reference moments; the future reference moment is a moment after a first preset time period starting from the sample moment; based on the operation information, it is determined that the target object is at The target operation sub-grid at the second moment; wherein, the second moment is the moment after the first preset duration at the first moment; the reference object and the target object have the same operation mode; based on The target running sub-grid and the running information determine the running track of the target object from the first moment to the second moment.
  • Embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, the steps of the trajectory prediction method described in the foregoing method embodiments are executed.
  • the storage medium may be a volatile or non-volatile computer-readable storage medium.
  • the computer program product of the trajectory prediction method provided by the embodiments of the present disclosure includes a computer program or an instruction, and when the computer program or instruction is run on a computer, the computer executes the trajectory described in the above method embodiment
  • the steps of the prediction method refer to the above method embodiments.
  • the computer program product can be realized by hardware, software or a combination thereof.
  • the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK) and the like.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor.
  • the technical solution of the embodiments of the present disclosure is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present disclosure.
  • the aforementioned computer-readable storage medium may be a tangible device capable of retaining and storing instructions used by the instruction execution device, and may be a volatile storage medium or a non-volatile storage medium.
  • a computer readable storage medium may be, for example but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above.
  • Non-exhaustive list of computer-readable storage media include: portable computer disk, hard disk, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), erasable Type programmable read-only memory (Erasable Programmable Read Only Memory, EPROM or flash memory), static random-access memory (Static Random-Access Memory, SRAM), portable compact disk read-only memory (Compact Disk Read Only Memory, CD-ROM) , Digital versatile discs (Digital versatile Disc, DVD), memory sticks, floppy disks, mechanically encoded devices, such as punched cards or raised structures in grooves with instructions stored thereon, and any suitable combination of the foregoing.
  • computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., pulses of light through fiber optic cables), or transmitted electrical signals.
  • the embodiment of the present disclosure acquires the operation information of the target object at the first moment; determines the preset operation grid matching the operation mode of the target object; the preset operation grid includes at least one operation sub-grid; the The running sub-grid includes the position points of at least one reference object at a plurality of future reference moments; the future reference moment is the moment after the sample moment as a starting point and after a first preset duration; the reference object and the target object Have the same running mode; use the preset running grid that matches the running mode of the target object to predict the running trajectory of the target object, which can fully consider the running speed and mode of the target object, and improve the pertinence and accuracy of trajectory prediction. accuracy.

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Traffic Control Systems (AREA)

Abstract

Procédé de prédiction de voie consistant : à obtenir des informations de déplacement d'un objet cible à un premier moment ; à déterminer une grille de déplacement prédéfinie correspondant au mode de déplacement de l'objet cible, la grille de déplacement prédéfinie comprenant au moins une sous-grille de déplacement, la sous-grille de déplacement comprenant des points de position d'au moins un objet de référence à une pluralité de moments de référence futurs, et chaque moment de référence futur étant un moment après une première période prédéfinie à partir d'un moment d'échantillon ; à déterminer, sur la base des informations de déplacement une sous-grille de déplacement cible où l'objet cible est situé à un second moment, le second moment étant un moment après la première période prédéfinie à partir du premier moment ; et à déterminer une voie de déplacement de l'objet cible du premier moment au second moment sur la base de la sous-grille de déplacement cible et des informations de déplacement. Le procédé de prédiction de voie est rapide et simple. Sont également fournis un appareil de prédiction de voie, un dispositif électronique et un support de stockage lisible par ordinateur stockant le procédé de prédiction de voie.
PCT/CN2022/084204 2021-09-30 2022-03-30 Procédé et appareil de prédiction de voie, dispositif électronique et support de stockage WO2023050749A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111160491.2A CN113879333B (zh) 2021-09-30 2021-09-30 轨迹预测方法、装置、电子设备及存储介质
CN202111160491.2 2021-09-30

Publications (1)

Publication Number Publication Date
WO2023050749A1 true WO2023050749A1 (fr) 2023-04-06

Family

ID=79004799

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/084204 WO2023050749A1 (fr) 2021-09-30 2022-03-30 Procédé et appareil de prédiction de voie, dispositif électronique et support de stockage

Country Status (2)

Country Link
CN (1) CN113879333B (fr)
WO (1) WO2023050749A1 (fr)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113879333B (zh) * 2021-09-30 2023-08-22 深圳市商汤科技有限公司 轨迹预测方法、装置、电子设备及存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20180130671A (ko) * 2017-05-30 2018-12-10 현대모비스 주식회사 3차원 격자지도를 이용한 자율 주행 제어 장치 및 방법
CN109969172A (zh) * 2017-12-26 2019-07-05 华为技术有限公司 车辆控制方法、设备及计算机存储介质
CN110085056A (zh) * 2019-04-24 2019-08-02 华南理工大学 一种高速公路车路协同环境下车辆换道瞬时风险识别方法
EP3552904A1 (fr) * 2018-04-10 2019-10-16 Bayerische Motoren Werke Aktiengesellschaft Procédé, dispositif et produit_programme informatique permettant de prédire le développement d'une scène de circulation impliquant plusieurs participants
CN111595352A (zh) * 2020-05-14 2020-08-28 陕西重型汽车有限公司 一种基于环境感知和车辆行驶意图的轨迹预测方法
CN113879333A (zh) * 2021-09-30 2022-01-04 深圳市商汤科技有限公司 轨迹预测方法、装置、电子设备及存储介质

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111382768B (zh) * 2018-12-29 2023-11-14 华为技术有限公司 多传感器数据融合方法和装置
RU2723237C1 (ru) * 2019-08-20 2020-06-09 Постников Роман Владимирович Способы и устройства построения точной траектории перемещений объектов
WO2021175434A1 (fr) * 2020-03-05 2021-09-10 Cambridge Enterprise Limited Système et procédé de prédiction d'une carte à partir d'une image
CN111784728B (zh) * 2020-06-29 2023-08-22 杭州海康威视数字技术股份有限公司 轨迹处理方法、装置、设备及存储介质
CN112000756A (zh) * 2020-08-21 2020-11-27 上海商汤智能科技有限公司 轨迹预测的方法、装置、电子设备及存储介质
CN112597822B (zh) * 2020-12-11 2023-08-15 国汽(北京)智能网联汽车研究院有限公司 车辆的轨迹确定方法、装置、电子设备及计算机存储介质
CN112327888B (zh) * 2021-01-07 2021-03-30 中智行科技有限公司 路径规划方法、装置、电子设备和存储介质
CN112923942B (zh) * 2021-01-22 2022-11-25 北京中交兴路信息科技有限公司 起点和终点之间车辆参考行驶路线的方法和装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20180130671A (ko) * 2017-05-30 2018-12-10 현대모비스 주식회사 3차원 격자지도를 이용한 자율 주행 제어 장치 및 방법
CN109969172A (zh) * 2017-12-26 2019-07-05 华为技术有限公司 车辆控制方法、设备及计算机存储介质
EP3552904A1 (fr) * 2018-04-10 2019-10-16 Bayerische Motoren Werke Aktiengesellschaft Procédé, dispositif et produit_programme informatique permettant de prédire le développement d'une scène de circulation impliquant plusieurs participants
CN110085056A (zh) * 2019-04-24 2019-08-02 华南理工大学 一种高速公路车路协同环境下车辆换道瞬时风险识别方法
CN111595352A (zh) * 2020-05-14 2020-08-28 陕西重型汽车有限公司 一种基于环境感知和车辆行驶意图的轨迹预测方法
CN113879333A (zh) * 2021-09-30 2022-01-04 深圳市商汤科技有限公司 轨迹预测方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN113879333B (zh) 2023-08-22
CN113879333A (zh) 2022-01-04

Similar Documents

Publication Publication Date Title
CN111919225B (zh) 使用模拟环境对自主机器进行培训、测试和验证
Ishihara et al. Multi-task learning with attention for end-to-end autonomous driving
US20180121733A1 (en) Reducing computational overhead via predictions of subjective quality of automated image sequence processing
CN111625950B (zh) 自动驾驶仿真场景重建方法、装置、设备和介质
Munigety et al. Towards behavioral modeling of drivers in mixed traffic conditions
US11919545B2 (en) Scenario identification for validation and training of machine learning based models for autonomous vehicles
WO2023207742A1 (fr) Procédé et système de détection de comportement de trafic anormal
WO2019047595A1 (fr) Procédé et dispositif d'évaluation pour un niveau de confort de système de conduite automatique de bout en bout
CN114303177A (zh) 通过迁移学习生成具有不同疲劳程度的视频数据集的系统和方法
McDuff et al. Causalcity: Complex simulations with agency for causal discovery and reasoning
CN112200131A (zh) 一种基于视觉的车辆碰撞检测方法、智能终端及存储介质
WO2023050749A1 (fr) Procédé et appareil de prédiction de voie, dispositif électronique et support de stockage
Siebinga et al. A human factors approach to validating driver models for interaction-aware automated vehicles
CN111079507A (zh) 一种行为识别方法及装置、计算机装置及可读存储介质
WO2023092982A1 (fr) Procédé et appareil de détection d'état, dispositif informatique, support de stockage et produit programme
WO2023273467A1 (fr) Procédé et appareil de détermination de données de valeur vraie, procédé et appareil d'apprentissage de réseau neuronal, et procédé et appareil de commande de déplacement
Krüger et al. Interaction-aware trajectory prediction based on a 3D spatio-temporal tensor representation using convolutional–recurrent neural networks
Srinivasan et al. Comparing merging behaviors observed in naturalistic data with behaviors generated by a machine learned model
Rong et al. Big data intelligent tourism management platform design based on abnormal behavior identification
Müller et al. Safe and psychologically pleasant traffic signal control with reinforcement learning using action masking
CN111062555B (zh) 城市画像的构建方法、装置、介质及电子设备
JP2023548749A (ja) センサデータのアノテーションのための方法およびシステム
US20150160836A1 (en) Extracting demonstrations from in-situ video content
Agarwal et al. Ordered Atomic Activity for Fine-grained Interactive Traffic Scenario Understanding
CN113302589A (zh) 协调部件接口控制框架

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22874150

Country of ref document: EP

Kind code of ref document: A1