A vehicle trajectory prediction method and device, storage medium and electronic equipment

Abstract

The application relates to a vehicle trajectory prediction method and device, a storage medium and an electronic device, which comprises the following steps: acquiring historical trajectory information of a target vehicle, and determining a motion scene of the target vehicle according to the historical trajectory information; predicting a future trajectory of the target vehicle according to the motion scene; wherein when the motion scene is straight driving or lane changing, the future trajectory of the target vehicle is predicted according to the historical trajectory information and a preset rule model; when the motion scene is turning or U-turning, lane line information is acquired, a multi-channel image is obtained according to the historical trajectory information and the lane line information, and the multi-channel image is input into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle, so that the inference time consumption can be effectively reduced and the prediction stability can be improved under the premise of ensuring the prediction accuracy.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and specifically to a vehicle trajectory prediction method and its device, storage medium, and electronic device. Background Art

[0002] In the field of autonomous driving, the vehicle's perception module detects and outputs information about the vehicle's surrounding environment. The vehicle's prediction module receives this information, predicts the trajectory of surrounding target vehicles in the future, and outputs it to the vehicle's planning and decision-making module for path planning.

[0003] However, due to the instability of the perception effect of the ego vehicle's perception module and the great uncertainty of the target vehicle's long-term running trajectory in complex scenarios, the ego vehicle prediction module has poor prediction accuracy and stability for the target vehicle's future trajectory. Summary of the Invention

[0004] The purpose of this application is to propose a vehicle trajectory prediction method and its device, storage medium, and electronic device to improve the trajectory prediction accuracy and stability of the target vehicle.

[0005] To achieve the above objectives, an embodiment of the present application provides a vehicle trajectory prediction method, the method comprising:

[0006] Acquire historical trajectory information of the target vehicle, and determine the movement scene of the target vehicle based on the historical trajectory information; the historical trajectory information includes multiple historical position coordinates of the target vehicle in the world coordinate system and the corresponding time; the historical position coordinates include abscissa and ordinate;

[0007] The future trajectory of the target vehicle is predicted based on the motion scene; when the motion scene is straight driving or lane changing, the future trajectory of the target vehicle is predicted based on the historical trajectory information and a preset rule model; when the motion scene is turning or making a U-turn, lane line information is obtained, and a multi-channel image is obtained based on the historical trajectory information and lane line information. The multi-channel image is input into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle.

[0008] In some embodiments, predicting the future trajectory of the target vehicle based on the historical trajectory information and a preset rule model specifically includes:

[0009] Obtaining the speed and acceleration of the target vehicle according to the historical trajectory information;

[0010] Based on the kinematic model, predicting the trajectory of the target vehicle within a preset future time according to the speed and acceleration of the target vehicle to obtain a first future trajectory;

[0011] According to the speed and acceleration of the target vehicle, a second future trajectory of the target vehicle within a preset future time is predicted based on a Kalman filter model.

[0012] The first future trajectory and the second future trajectory are weighted and smoothed to obtain a future trajectory of the target vehicle.

[0013] In some schemes, the speed and acceleration of the target vehicle are obtained according to the historical trajectory information, specifically including:

[0014] The first speed and the first acceleration of the target vehicle are calculated according to the historical trajectory information, and the second speed and the second acceleration of the target vehicle detected by a self-perception module are obtained;

[0015] The speed of the target vehicle is obtained by weighted summation of the first speed and the second speed, and the acceleration of the target vehicle is obtained by weighted summation of the first acceleration and the second acceleration.

[0016] In some schemes, the first speed and the first acceleration of the target vehicle are calculated according to the historical trajectory information, specifically including:

[0017] According to the plurality of historical position coordinates and their corresponding time points, a quadratic polynomial function is fitted based on a least square method; wherein the input of the quadratic polynomial function is the time point, and the output is the horizontal coordinate or the vertical coordinate;

[0018] Time resampling is performed at a preset frequency to obtain the latest m time points; m is a preset value;

[0019] The horizontal coordinates and the vertical coordinates corresponding to the m time points are calculated according to the m time points and the fitted quadratic polynomial function, and the historical trajectory is obtained according to the horizontal coordinates and the vertical coordinates corresponding to the m time points.

[0020] The first speed and the first acceleration of the target vehicle are calculated according to the historical trajectory.

[0021] In some schemes, the motion scenario of the target vehicle is determined according to the historical trajectory information, specifically including:

[0022] A circle is fitted according to the plurality of historical position coordinates and a least square method fitting circle function.

[0023] If the radius of the fitted circle is greater than a preset first threshold, it is determined that the motion scenario of the target vehicle is straight or lane changing; if the radius of the fitted circle is less than or equal to the preset first threshold, the motion scenario of the target vehicle is turning or U-turn.

[0024] In some embodiments, determining the movement scene of the target vehicle based on the historical trajectory information further includes:

[0025] When the number of historical position coordinates of the target vehicle obtained is less than a preset second threshold, the future trajectory of the target vehicle is not predicted;

[0026] When the number of acquired historical position coordinates of the target vehicle is greater than or equal to a preset second threshold, the future trajectory of the target vehicle is predicted.

[0027] In some embodiments, obtaining a multi-channel image based on the historical trajectory information and lane line information specifically includes:

[0028] Interpolating the multiple historical position coordinates to obtain a historical trajectory, and resampling n historical trajectory points of the historical trajectory at a preset frequency; n is a preset value;

[0029] A multi-channel image is generated according to the n historical trajectory points and the lane line information.

[0030] An embodiment of the present application further provides a vehicle trajectory prediction device, comprising a module for executing the vehicle trajectory prediction method described above.

[0031] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the vehicle trajectory prediction method as described above is implemented.

[0032] An embodiment of the present application also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the vehicle trajectory prediction method as described above when executing the computer program.

[0033] The embodiment of the present application provides a vehicle trajectory prediction method and its device, storage medium, and electronic device. When predicting the future trajectory of other target vehicles on the road, the motion scene of the other target vehicles is judged based on the historical trajectory information of the target vehicle. If the motion scene of the target vehicle is straight or changing lanes, the scene of straight or changing lanes is relatively simple. Therefore, the embodiment of the present application predicts the future trajectory of the target vehicle based on the historical trajectory information and the preset rule model, and can accurately, quickly and stably predict the motion trajectory of the target vehicle; if the motion scene of the target vehicle is turning or turning around, the scene of turning or turning around is relatively complex. Therefore, the embodiment of the present application obtains lane line information, obtains a multi-channel image based on the historical trajectory information and lane line information, and inputs the multi-channel image into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle. The data-driven deep learning neural network learns a large number of real trajectory training samples through large-scale learnable parameters, and can well predict the motion trajectory in complex scenes. In summary, the embodiment of the present application proposes corresponding trajectory prediction algorithms for different motion scenes to optimize prediction accuracy and speed, which can effectively improve the accuracy and robustness of the trajectory prediction solution and reduce the time consumption of reasoning. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 This is a flowchart of a vehicle trajectory prediction method in one embodiment of the present application.

[0036] Figure 2 This is a flowchart of caching and updating perception information in one embodiment of the present application.

[0037] Figure 3 This is a flowchart of a vehicle trajectory prediction method in another embodiment of the present application.

[0038] Figure 4 This is a structural framework diagram of a vehicle trajectory prediction device in one embodiment of the present application. DETAILED DESCRIPTION

[0039] The detailed description of the accompanying drawings is intended to serve as an illustration of the current embodiment of the present application and is not intended to represent the only form in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included in the spirit and scope of the present application.

[0040] See Figure 1 An embodiment of the present application provides a vehicle trajectory prediction method, comprising the following steps:

[0041] Step S10, obtaining historical trajectory information of the target vehicle, and determining the movement scene of the target vehicle based on the historical trajectory information; the historical trajectory information includes multiple historical position coordinates of the target vehicle in the world coordinate system and the corresponding time; the historical position coordinates include abscissa and ordinate;

[0042] Specifically, the target vehicle refers to the vehicle in the surrounding environment of the vehicle. The movement pattern of the target vehicle can be inferred based on the historical trajectory information to determine the movement scene of the target vehicle. The movement scene includes simple scenes and complex scenes. The simple scene is straight driving or changing lanes, and the complex scene is turning or U-turning.

[0043] Step S20, predicting the future trajectory of the target vehicle based on the motion scene; wherein, when the motion scene is straight driving or lane changing, the future trajectory of the target vehicle is predicted based on the historical trajectory information and a preset rule model; when the motion scene is turning or making a U-turn, lane line information is obtained, a multi-channel image is obtained based on the historical trajectory information and the lane line information, and the multi-channel image is input into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle;

[0044] Specifically, two trajectory prediction schemes are provided in step S20. First, if the target vehicle's movement scene is straight or changing lanes, the straight or lane changing scenes are relatively simple. Therefore, this embodiment predicts the future trajectory of the target vehicle based on the historical trajectory information and the preset rule model. The calculation time is short and the prediction accuracy can meet the requirements. Therefore, the movement trajectory of the target vehicle can be accurately, quickly and stably predicted; second, if the target vehicle's movement scene is turning or turning around, the turning or turning around scenes are relatively complex. At this time, if the preset rule model is continued to be used for prediction, the prediction result accuracy is low. Therefore, this embodiment uses a pre-trained deep learning neural network for prediction. The data-driven deep learning neural network learns a large number of real trajectory training samples through large-scale learnable parameters, and can well predict the movement trajectory in complex scenes.

[0045] Among them, the rule model is a rule-based method used to describe and process knowledge and behavior in a specific problem domain. The rule model contains a series of rules that are triggered under specific conditions and perform corresponding operations or reasoning. The rules are usually expressed in the form of conditions-conclusions, where the conditions (also called premises) specify the specific circumstances that trigger the rules, and the conclusions (also called results) specify the operations or reasoning performed under these conditions; the preset rule models are, for example, kinematic models, Kalman filter models, etc.

[0046] It should be noted that in the scenarios of going straight or changing lanes, trajectory prediction based on preset rule models and trajectory prediction based on deep learning neural networks can both achieve relatively good prediction results, but deep learning neural networks require huge computing power and real-time requirements, which have high requirements on system hardware resources and higher requirements on the stability of upstream perception information. Therefore, this embodiment selects a trajectory prediction solution based on a preset rule model in the scenarios of going straight or changing lanes, which saves computing time and improves prediction efficiency. In the scenarios of turning or U-turning, the trajectory prediction based on the preset rule model cannot meet the prediction accuracy requirements. Therefore, in order to ensure prediction accuracy, this embodiment selects a trajectory prediction solution based on a deep learning neural network.

[0047] In summary, this embodiment proposes corresponding trajectory prediction algorithms for different motion scenarios to optimize prediction accuracy and speed, which can effectively improve the accuracy and robustness of the trajectory prediction solution and reduce inference time.

[0048] In some embodiments, the method further comprises:

[0049] Step S30: Acquire the ego vehicle posture, convert the predicted future trajectory of the target vehicle from the world coordinate system to the ego vehicle coordinate system according to the ego vehicle posture, and output it to the ego vehicle planning decision module.

[0050] Specifically, in the field of autonomous driving, the vehicle's perception module detects and outputs perception information about the vehicle's surrounding environment. The vehicle's prediction module receives this perception information, predicts the trajectory of surrounding target vehicles for a period of time in the future, and outputs it to the vehicle's planning and decision-making module for path planning. For example, it calculates the risk of collision between the vehicle and the target vehicle. If there is a risk of collision between the vehicle and the target vehicle, it controls the vehicle to slow down, change lanes, etc.; another example is to issue an early warning prompt. When making decisions and planning, the vehicle's planning and decision-making module generally makes decisions based on the vehicle's coordinate system. However, in this embodiment, the predicted future trajectory of the target vehicle is based on the world coordinate system. Therefore, it needs to be converted to the vehicle's coordinate system before being output to the vehicle's planning and decision-making module for planning and decision-making. Specifically, the predicted trajectory is sent to the vehicle's planning and decision-making module in the form of a point set.

[0051] In some embodiments, the step S20 of predicting the future trajectory of the target vehicle based on the historical trajectory information and a preset rule model specifically includes:

[0052] Step S211, obtaining the speed and acceleration of the target vehicle according to the historical trajectory information;

[0053] In step S212, based on the kinematic model, the target vehicle's trajectory within a preset future time is predicted according to its speed and acceleration to obtain a first future trajectory, for example, a trajectory for the next three seconds, including coordinates of 30 predicted trajectory points. The prediction principle based on the kinematic model is shown in the following formula:

[0054] x=x+(v x +a x ×0.1)×0.1

[0055] v x =v x +a x ×0.1

[0056] y=y+(v y +a y ×0.1)×0.1

[0057] v y =v y +a y ×0.1

[0058] Among them, v x and v y are the lateral component and longitudinal component of the target vehicle's velocity, respectively, and a x and a y are the lateral component and longitudinal component of the acceleration of the target vehicle, respectively; x is the lateral coordinate of the target vehicle, and y is the longitudinal coordinate of the target vehicle;

[0059] Assuming that the X-direction position at the current moment t is x, then at the next moment t+1, the X-direction velocity is v x =v x +a x ×0.1, so the X-direction position at time t+1 is x=x+(v x +a x ×0.1)×0.1, and then update the X-direction velocity to the velocity at time t+1. Similarly, calculate the velocity and X-direction position at times t+2, t+3, ..., t+n. Similarly, the Y-direction position can be predicted.

[0060] Step S213: Based on the Kalman filter model, the target vehicle's trajectory within a preset future time period is predicted based on the target vehicle's speed and acceleration to obtain a second future trajectory, for example, a trajectory for the next three seconds, including coordinates of 30 predicted trajectory points. Specifically, the Kalman filter is an algorithm that uses a linear system state equation to optimally estimate the system state through system input and output observation data. In this embodiment, the target vehicle's speed and acceleration are used as the observation data.

[0061] Step S214 : performing weighted smoothing on the first future trajectory and the second future trajectory to obtain the future trajectory of the target vehicle.

[0062] Specifically, this embodiment combines two rule models, the kinematic model and the Kalman filter model, and performs weighted smoothing on the prediction results of the two rule models to obtain the future trajectory of the target vehicle. This can improve the reliability of the prediction results, effectively alleviate the noise and temporal and spatial jitter problems in the perception results, and improve the generalization and robustness of the algorithm.

[0063] In some embodiments, step S211 specifically includes:

[0064] Calculating a first speed and a first acceleration of the target vehicle according to the historical trajectory information, and obtaining a second speed and a second acceleration of the target vehicle detected by the ego vehicle perception module;

[0065] The speed of the target vehicle is obtained by weighted summing the first speed and the second speed, and the acceleration of the target vehicle is obtained by weighted summing the first acceleration and the second acceleration, as shown in the following formula:

[0066] v=0.5×v s +0.5×v P

[0067] a=0.5×a s +0.5×a p

[0068] Among them, v is the final speed of the target vehicle, a is the final acceleration of the target vehicle, and v is the final acceleration of the target vehicle. s is the first velocity, v P is the second speed, a s is the first acceleration, a p is the second acceleration;

[0069] Specifically, this embodiment calculates the speed v based on historical trajectory information. s , acceleration a s , and the speed v directly output by the vehicle perception module P , acceleration a pThe weighted sum is performed to obtain the speed v and the acceleration a of the target vehicle, which can further effectively alleviate the noise and time-space jitter problems in the perception result, and improve the generalization and robustness of the algorithm.

[0070] In some embodiments, the step S211 of calculating the first speed and the first acceleration of the target vehicle according to the historical trajectory information specifically comprises:

[0071] In step a, a quadratic polynomial function is fitted according to the plurality of historical position coordinates and the corresponding time points based on the least square method, wherein the input of the quadratic polynomial function is the time point, and the output is the horizontal coordinate or the vertical coordinate;

[0072] Specifically, the formula of the quadratic polynomial function is as follows:

[0073] x = a1t 2 +b1t+c1

[0074] y = a2t 2 +b2t+c2

[0075] The parameters a1, b1, c1, a2, b2, and c2 of the quadratic polynomial function can be solved by substituting the plurality of historical position coordinates and the corresponding time points into the formula of the quadratic polynomial function.

[0076] In step b, time resampling is performed at a preset frequency to obtain the latest m time points; m is a preset value.

[0077] Specifically, 10 latest neighboring time points can be sampled at a fixed frequency of 10 hz, that is, m = 10.

[0078] In step c, the horizontal coordinate and the vertical coordinate corresponding to the m time points are calculated according to the m time points and the fitted quadratic polynomial function, and the historical trajectory is obtained according to the horizontal coordinate and the vertical coordinate corresponding to the m time points.

[0079] Specifically, the horizontal coordinate and the vertical coordinate corresponding to the m time points can be calculated by substituting the m time points into the fitted quadratic polynomial function.

[0080] In step d, the first speed and the first acceleration of the target vehicle are calculated according to the historical trajectory.

[0081] Specifically, the historical trajectory fitted by the two points has a fixed time difference of 0.1 s, and the first speed v s and the first acceleration a s can be calculated from the distance between the two points and the time according to the formula of the kinematic model.

[0082] In some embodiments, the step S10 specifically comprises:

[0083] Step S11, fitting a circle according to the multiple historical position coordinates and a least squares fitting circle function;

[0084] Specifically, the least squares method fits the circular function as follows:

[0085] (xa) 2 +(yb) 2 =r 2

[0086] Where x is the horizontal coordinate of the target vehicle's historical position coordinate, and y is the vertical coordinate of the target vehicle's historical position coordinate;

[0087] Step S12: If the radius of the fitted circle is greater than a preset first threshold, it is determined that the target vehicle's movement scene is straight driving or lane changing; if the radius of the fitted circle is less than or equal to the preset first threshold, the target vehicle's movement scene is turning or U-turning;

[0088] Specifically, this embodiment infers the target vehicle motion scene based on the radius r of the fitted circular function, and the first threshold is 100.

[0089] In some embodiments, the step S10 further includes:

[0090] Before determining the motion scene of the target vehicle based on the historical trajectory information, determine whether to predict the future trajectory of the target vehicle. When the number of historical position coordinates of the target vehicle obtained is less than a preset second threshold, the future trajectory of the target vehicle is not predicted because the motion scene of the target vehicle cannot be accurately determined if the number of historical position coordinates is too small. When the number of historical position coordinates of the target vehicle obtained is greater than or equal to the preset second threshold, the motion scene of the target vehicle is determined based on the historical trajectory information, and the future trajectory of the target vehicle is predicted based on the motion scene of the target vehicle. The preset second threshold is, for example, 20.

[0091] In some embodiments, in order to facilitate the management of the perception information output by the ego vehicle perception module (including historical trajectory information, lane line information, ego vehicle posture information, etc.), before executing step S10, the method further includes:

[0092] Create a historical trajectory cache queue for each target vehicle to cache the historical trajectory information of the target vehicle. Each element of the historical trajectory cache queue includes the trajectory ID, the corresponding time, and the position coordinates in the world coordinate system;

[0093] Create a lane line cache queue for each lane line to cache lane line information, including lane line ID, position point set in the world coordinate system, and lane line attribute information;

[0094] Create an RTK information cache queue to cache the vehicle's pose information (for coordinate transformation), including the corresponding time, position information, and orientation information.

[0095] During trajectory prediction, the perception information output by the ego vehicle perception module is cached and updated based on the created historical trajectory cache queue, lane line cache queue, and RTK information cache queue.

[0096] Among them, see Figure 2 , the update of the historical track cache queue is as follows:

[0097] When a new perception information (new element) is received, the trajectory perception information first obtains the corresponding trajectory id. If the historical trajectory cache queue corresponding to the trajectory id already exists and the historical trajectory cache queue is not full, the new element is inserted into the head; if the historical trajectory cache queue corresponding to the trajectory id already exists and the historical trajectory cache queue is full, the tail element of the queue is first removed, and then the new element is inserted into the head; if the historical trajectory cache queue corresponding to the trajectory id does not exist, a new cache queue corresponding to the trajectory id is created and the new element is inserted into the head; generally speaking, the ego vehicle perception module will detect all target vehicles within the ego vehicle detection range and output the corresponding trajectory information. If all historical trajectory cache queues are traversed and the trajectory id of the historical trajectory cache queue does not appear in the received new trajectory perception message, the cache queue will be judged as expired data and deleted.

[0098] Among them, see Figure 2 , the lane line cache queue is updated as follows:

[0099] When a new perception information (new element) is received, the trajectory perception information first obtains the corresponding lane line id. If the lane line cache queue corresponding to the lane line id already exists and the lane line cache queue is not full, the new element is inserted into the head; if the lane line cache queue corresponding to the lane line id already exists and the lane line cache queue is full, the element at the tail of the queue is removed first, and then the new element is inserted into the head; if the lane line cache queue corresponding to the lane line id does not exist, a new cache queue corresponding to the lane line id is created and the new element is inserted into the head; generally speaking, the ego vehicle perception module will detect all target vehicles within the ego vehicle detection range and output the corresponding lane line information. If all lane line cache queues are traversed, if the lane line id of the lane line cache queue does not appear in the new trajectory perception message received, the cache queue will be judged as expired data and deleted.

[0100] The RTK information cache queue is updated as follows:

[0101] When a new vehicle pose information (new element) is received, the new element is inserted into the head of the queue if the RTK information cache queue is not full. When the RTK information cache queue is full, the data at the end of the queue is removed first, and then the new element is inserted into the head of the queue. Specifically, the number of caches in the RTK cache queue can be set to 200.

[0102] In some embodiments, obtaining a multi-channel image based on the historical trajectory information and lane line information in step S20 specifically includes:

[0103] Interpolating the multiple historical position coordinates to obtain a historical trajectory, and resampling n historical trajectory points of the historical trajectory at a preset frequency; n is a preset value; and generating a multi-channel image according to the n historical trajectory points and the lane line information.

[0104] For example, quadratic spline interpolation can be performed on the multiple historical position coordinates, and 10 historical trajectory points (constituting the historical trajectory) can be resampled at a fixed frequency of 10 Hz, that is, n = 10; further, SG filtering (Savitzky-Golay Filter) can be performed on the historical trajectory to smooth the trajectory misalignment and jitter problems existing in the perception input; based on the MotionCNN scheme, the historical trajectory and lane lines are rendered into multi-channel images; for example, the rendered image size is 224×224×23, the lane line rendering image size is 224×224×3, the target vehicle historical trajectory rendering image size is 224×224×10, and the surrounding vehicle historical trajectory rendering image size in the same scene is 224×224×10; the rendered multi-channel image is input into the trained deep convolutional neural network for trajectory prediction, and the corresponding predicted trajectory is output.

[0105] In some embodiments, when generating the multi-channel image, traffic light information at the intersection can also be added, and the historical trajectory, lane lines, and traffic lights at the intersection can be rendered into a multi-channel image, which is then input into a deep convolutional neural network for trajectory prediction, and the corresponding predicted trajectory is output, for example Figure 3 As shown in the flowchart of the vehicle trajectory prediction method, in this embodiment, the first threshold is set to 100, the second threshold is set to 20, the perception module of the ego vehicle performs 3D object detection and tracking, lane line and traffic light detection, and outputs corresponding perception information. The perception information is cached and updated according to the perception information output by the perception module of the ego vehicle ( Figure 2Furthermore, when the historical trajectory length is greater than 20 (the second threshold), a least squares method is used to fit a circular function. If the radius of the fitted circle is greater than 100 (the first threshold), it is determined that the target vehicle is in a straight / lane change scenario. If the radius of the fitted circle is less than or equal to 100 (the first threshold), it is determined that the target vehicle is in a turning / U-turn scenario. In different scenarios, different prediction strategies are used to predict the future trajectory of the target vehicle and then output it.

[0106] See Figure 4 Another embodiment of the present application further provides a vehicle trajectory prediction device, comprising a module for executing the vehicle trajectory prediction method described above, wherein the module comprises:

[0107] An information acquisition module is used to obtain historical trajectory information of a target vehicle and determine a movement scene of the target vehicle based on the historical trajectory information;

[0108] A trajectory prediction module is used to predict the future trajectory of the target vehicle based on the motion scene; when the motion scene is straight driving or lane changing, the future trajectory of the target vehicle is predicted based on the historical trajectory information and a preset rule model; when the motion scene is turning or making a U-turn, lane line information is obtained, and a multi-channel image is obtained based on the historical trajectory information and lane line information. The multi-channel image is input into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle.

[0109] It should be noted that the vehicle trajectory prediction device of the above embodiment corresponds to the vehicle trajectory prediction method of the above embodiment. Therefore, the undescribed parts of the vehicle trajectory prediction device of the above embodiment can be obtained by referring to the contents of the vehicle trajectory prediction method of the above embodiment. That is, the specific steps recorded in the vehicle trajectory prediction method of the above embodiment can be understood as the functions that can be implemented by the information acquisition module and the trajectory prediction module of the above embodiment, and will not be repeated here.

[0110] Furthermore, if the vehicle trajectory prediction device of the above embodiment is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0111] Another embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the vehicle trajectory prediction method as described in the above embodiment is implemented.

[0112] Specifically, the computer-readable storage medium may include: any entity or recording medium that can carry the computer program instructions, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0113] Another embodiment of the present application provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the vehicle trajectory prediction method described in the above embodiment when executing the program.

[0114] Wherein, electronic equipment can also include the bus connecting different components (including memory and processor).Memory can include computer-readable medium in the form of volatile memory, such as random access memory (RAM) and / or cache memory.Memory can also include at least one program product, and this program product has one group (such as at least one) program module, and these program modules are configured to perform the function of each embodiment of the present application.Electronic equipment can also communicate with one or more external devices (such as keyboard, pointing device, display etc.), can also communicate with one or more devices that enable users to interact with this electronic equipment, and / or communicate with any device (such as network card) that enables this electronic equipment to communicate with one or more other computing devices, this communication can be carried out through input / output (I / O) interface, and electronic equipment can also communicate with one or more networks (such as local area network (LAN), wide area network (WAN) and / or public network, such as Internet) through network adapter.

[0115] While various embodiments of the present application have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A vehicle trajectory prediction method, characterized in that: The method comprises: Acquire historical trajectory information of the target vehicle, and determine the movement scene of the target vehicle based on the historical trajectory information; the historical trajectory information includes multiple historical position coordinates of the target vehicle in the world coordinate system and the corresponding time; the historical position coordinates include abscissa and ordinate; The future trajectory of the target vehicle is predicted based on the motion scene; when the motion scene is straight driving or lane changing, the future trajectory of the target vehicle is predicted based on the historical trajectory information and a preset rule model; when the motion scene is turning or making a U-turn, lane line information is obtained, and a multi-channel image is obtained based on the historical trajectory information and lane line information. The multi-channel image is input into a pre-trained deep learning neural network for processing to predict the future trajectory of the target vehicle.

2. The method according to claim 1, characterized in that The predicting of the future trajectory of the target vehicle based on the historical trajectory information and a preset rule model specifically includes: Obtaining the speed and acceleration of the target vehicle according to the historical trajectory information; Based on the kinematic model, predicting the trajectory of the target vehicle within a preset future time according to the speed and acceleration of the target vehicle to obtain a first future trajectory; Based on the Kalman filter model, predicting the trajectory of the target vehicle within a preset future time according to the speed and acceleration of the target vehicle to obtain a second future trajectory; The first future trajectory and the second future trajectory are weighted and smoothed to obtain the future trajectory of the target vehicle.

3. The method according to claim 2, characterized in that The obtaining of the speed and acceleration of the target vehicle according to the historical trajectory information specifically includes: Calculating a first speed and a first acceleration of the target vehicle according to the historical trajectory information, and obtaining a second speed and a second acceleration of the target vehicle detected by the ego vehicle perception module; The speed of the target vehicle is obtained by weighted summing of the first speed and the second speed, and the acceleration of the target vehicle is obtained by weighted summing of the first acceleration and the second acceleration.

4. The method according to claim 3, characterized in that The calculating the first speed and the first acceleration of the target vehicle according to the historical trajectory information specifically includes: Based on the least squares method, a quadratic polynomial function is fitted according to the multiple historical position coordinates and their corresponding times; wherein the input of the quadratic polynomial function is the time, and the output is the horizontal coordinate or the vertical coordinate; Time resampling is performed at a preset frequency to obtain the latest m moments; m is a preset value; Calculating the horizontal coordinates and vertical coordinates corresponding to the m moments according to the m moments and the fitted quadratic polynomial function, and obtaining a historical trajectory according to the horizontal coordinates and vertical coordinates corresponding to the m moments; A first speed and a first acceleration of the target vehicle are calculated according to the historical trajectory.

5. The method according to claim 1, wherein Determining the movement scene of the target vehicle according to the historical trajectory information specifically includes: Fitting a circle according to the plurality of historical position coordinates and a least squares fitting circle function; If the radius of the fitted circle is greater than a preset first threshold, it is determined that the movement scene of the target vehicle is straight driving or lane changing; if the radius of the fitted circle is less than or equal to the preset first threshold, the movement scene of the target vehicle is turning or U-turn.

6. The method according to claim 5, characterized in that The determining of the movement scene of the target vehicle according to the historical trajectory information specifically further includes: When the number of historical position coordinates of the target vehicle obtained is less than a preset second threshold, the future trajectory of the target vehicle is not predicted; When the number of acquired historical position coordinates of the target vehicle is greater than or equal to a preset second threshold, the future trajectory of the target vehicle is predicted.

7. The method according to claim 1, characterized in that The obtaining of a multi-channel image according to the historical trajectory information and lane line information specifically includes: Interpolating the multiple historical position coordinates to obtain a historical trajectory, and resampling n historical trajectory points of the historical trajectory at a preset frequency; n is a preset value; A multi-channel image is generated according to the n historical trajectory points and the lane line information.

8. A vehicle trajectory prediction device, characterized in that: The invention comprises a module for executing the vehicle trajectory prediction method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the vehicle trajectory prediction method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the vehicle trajectory prediction method according to any one of claims 1 to 7 is implemented.

Images