Automatic driving simulation track generation method and computer program product

Abstract

The invention discloses an automatic driving simulation track generation method, which comprises the steps of generating an action set for a vehicle track point set, and obtaining the vehicle track point set by downsampling original vehicle track data; on the basis of a word list generation algorithm, sub-action sets meeting distance conditions are selected from the action set, the sub-action sets are coded, a vocabulary set is generated, and the distance conditions include that the distance between any actions is not smaller than a preset threshold value; searching lexical elements corresponding to the target action and codes of the lexical elements in a vocabulary set; and inputting the lexical elements of the target action and the codes of the lexical elements into an autoregression model for track reconstruction to obtain simulation track points. According to the method, the efficient motion word list is generated by sampling a large amount of vehicle motion data for track simulation generation, so that the dynamic simulation degree of the model is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, in particular to the fields of autonomous driving and artificial intelligence, and specifically to a method for generating an autonomous driving simulation trajectory and a computer program product. Background Art

[0002] In order to improve the safety and reliability of autonomous driving, how to provide a large amount of real, available and wide-coverage data for the autonomous driving model has become a top priority. In related technologies, real road data collected from real vehicles provides most of the training data. In order to supplement the training data, a virtual simulation environment is usually used to simulate driving in the hope of providing more data to support the training of large models. Summary of the invention

[0003] The embodiments of the present application provide a method for generating an autonomous driving simulation trajectory and a computer program product. The present application generates an efficient motion vocabulary by sampling a large amount of vehicle motion data, thereby improving the dynamic simulation degree of the model.

[0004] In a first aspect of an embodiment of the present application, a method for generating a simulation trajectory for autonomous driving is disclosed, comprising: generating an action set for a vehicle trajectory point set, wherein the vehicle trajectory point set is obtained by downsampling original vehicle trajectory data; based on a vocabulary generation algorithm, selecting a sub-action set that meets a distance condition from the action set, encoding the sub-action set, and generating a vocabulary set, wherein the distance condition includes that the distance between any actions is not less than a preset threshold; searching for a word element and its encoding corresponding to a target action in the vocabulary set; inputting the word element and its encoding of the target action into an autoregressive model for trajectory reconstruction to obtain a simulation trajectory point.

[0005] In a feasible implementation, the generating of an action set for a vehicle trajectory point set specifically includes: for each discrete trajectory point in the vehicle trajectory point set, calculating a difference vector between a current trajectory point and a next trajectory point to form an initial action set; and performing a coordinate system transformation on the initial action set to generate an action set.

[0006] In a feasible implementation manner, the method of selecting a sub-action set satisfying the distance condition from the action set based on the word list generation algorithm is as follows: randomly selecting an action a from the action set A. i , for any action v in the current sub-action set V j , calculate the distance d(a i ,v j ),like Then V=V∪{a i}, otherwise, skip and reselect the action. If |V| reaches the set value or the number of skips reaches the maximum value, the algorithm stops; wherein the action set is A = {a 1 ,a 2 ,…,a n}, the sub-action set is V = {v 1 ,v 2 ,…,v m}, where the minimum distance threshold is ∈, m≤n.

[0007] In a feasible implementation manner, the vocabulary set establishes a coordinate system with the movement direction of the previous frame trajectory as the positive direction of the X-axis, and word units are selected in the coordinate system.

[0008] In a feasible implementation manner, the word units in the vocabulary set are obtained by screening through a transverse dynamic mask and / or a longitudinal dynamic mask.

[0009] In a feasible implementation manner, the word elements in the vocabulary set are obtained by longitudinal dynamic mask screening, specifically including the position difference d of the previous frame motion. prev And set the maximum acceleration a max and maximum braking force a min and the known time interval Δt, the feasible position difference d for the next movement is calculated based on the following kinematic formula next .

[0010]

[0011] Based on the above formula, word units that meet the feasible position difference interval are selected.

[0012] In a feasible implementation manner, the word units in the vocabulary set are obtained by lateral dynamic mask screening, specifically including:

[0013] Calculate the rear axle center movement distance Direction angle difference Δψ=ψ current -ψ last , Turning radius Front wheel angle Among them, the rear axle center position x in the previous frame last ,ylast, the vehicle direction in the previous frame ψ last , the current frame rear axis center position x current ,y current , the vehicle direction in the current frame is ψ current , and the length of the front and rear axles is L, based on the above formula, the words that meet the front wheel turning angle range are selected.

[0014] In a feasible implementation manner, searching for the word element and its encoding of the target action through the vocabulary set specifically includes: determining the word element most similar to the target action by calculating the similarity, and indexing its encoding.

[0015] In a feasible implementation, determining the word element that is most similar to the target action by calculating similarity specifically includes: calculating the distance between the vector of the target action and the end point of each word element vector in the vocabulary set, selecting the word element with the smallest distance as the word element that is most similar to the target action, and the distance is a straight-line distance or a Manhattan distance.

[0016] In a feasible implementation, searching for the word unit and its encoding of the target action through the vocabulary set includes discretizing the first action using the word unit and taking the end point of the discretized action as the starting point of the next word unit discretization.

[0017] In a feasible implementation, the accuracy of the vocabulary set is evaluated using the average displacement error and / or the endpoint displacement error, specifically including calculating the average value of the distance difference between each point of the original trajectory and the corresponding point of the trajectory after trajectory discretization and reconstruction and / or calculating the distance difference between the endpoint of the original trajectory and the corresponding point of the trajectory after trajectory discretization and reconstruction, and performing the evaluation based on the size of the average value and / or the size of the distance difference.

[0018] According to a second aspect of the embodiments of the present application, a computer program product is disclosed. When the computer program is executed by a processor, the method steps according to any one of the first aspects are implemented.

[0019] Compared with the prior art, the above solution of the embodiment of the present disclosure has at least the following beneficial effects:

[0020] First, the present invention provides an autonomous driving trajectory generation method that can meet the simulation requirements of autonomous driving vehicle development and / or tuning planners and controllers, quickly, flexibly and accurately generate simulated trajectories, and based on the planning and control scheme corresponding to the obstacle trajectory, provide more accurate data support for subsequent intelligent driving technology.

[0021] Second, the trajectory generation method provided by the present disclosure generates an efficient motion vocabulary by sampling a large amount of vehicle motion data, thereby improving the dynamic simulation of the model. At the same time, it also avoids the problem that the traditional trajectory prediction model method is prone to single distribution, resulting in insufficient diversity in trajectory prediction and the inability to grasp the dynamic characteristics well.

[0022] Other features and advantages of the present application will become apparent from the following detailed description of exemplary embodiments of the present application with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the application and, together with the description, serve to explain the principles of the application.

[0024] Figure 1 is an exemplary schematic diagram of an implementation environment of a trajectory generation method provided in an embodiment of the present application;

[0025] Figure 2 is an exemplary flow chart of a trajectory generation method provided in an embodiment of the present application;

[0026] Figure 3-Figure 4 It is an exemplary effect diagram of the word unit discretization process provided in the embodiment of the present application.

[0027] Figure 5 This is a framework diagram of an embodiment of the present application in which the output features of a decoder are classified in an autoregressive model for trajectory prediction, the distribution of word units is output, and then word units are obtained from the distribution by sampling. DETAILED DESCRIPTION

[0028] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0029] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.

[0030] Technologies, methods and equipment known to persons of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods and equipment should be considered part of the specification.

[0031] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0032] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0033] The present application is described below in conjunction with specific embodiments.

[0034] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0035] The trajectory generation method and computer program product according to the embodiments of the present application are described below with reference to the accompanying drawings.

[0036] Artificial intelligence is a discipline that studies the use of computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level technical fields and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies include computer vision technology, speech recognition technology, natural language processing technology, as well as deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0037] Autonomous driving, also known as unmanned driving, computer driving or wheeled mobile robot, is a cutting-edge technology that relies on computers and artificial intelligence technology to complete complete, safe and effective driving without human control.

[0038] In related technologies, real road data collected from real vehicles provides most of the training data. However, the amount of this data is neither sufficient to support the training of end-to-end large models nor can it cover all traffic scenarios. As a method of supplementing data, a virtual simulation environment is usually used to simulate driving in the hope of providing more data to support the training of large models.

[0039] In some embodiments of the prior art, traditional trajectory prediction models can be used in intelligent agents, and most of them adopt direct regression methods, but this method is prone to single distribution problems, resulting in insufficient diversity in trajectory prediction, failure to grasp the dynamic characteristics well, and causing trajectory mutation problems when used in a closed loop.

[0040] Based on this, an embodiment of the present application provides a trajectory generation method, which generates an efficient motion vocabulary by sampling a large amount of vehicle motion data for trajectory simulation, thereby improving the dynamic simulation degree of the model.

[0041] See also Figure 1 , Figure 1 is a schematic diagram of an implementation environment of a trajectory generation method provided in an embodiment of the present application, such as Figure 1 As shown, the implementation environment includes: a terminal device 101 and a server 102 .

[0042] The terminal device 101 may be at least one of an autonomous driving vehicle, a smart phone, a game console, a desktop computer, a tablet computer, an e-book reader, and a laptop computer. The terminal device 101 is used to execute the trajectory generation method provided in the embodiment of the present application.

[0043] The terminal device 101 may generally refer to one of a plurality of terminal devices, and this embodiment is only illustrated by taking the terminal device 101 as an example. Those skilled in the art may know that the number of the terminal devices 101 may be more or less. For example, the terminal device 101 may be only one, or the terminal devices 101 may be dozens or hundreds, or more. The embodiment of the present application does not limit the number and type of terminal devices.

[0044] The server 102 is a single server, or a server cluster consisting of multiple servers, or any one of a cloud computing platform and a virtualization center, which is not limited in the embodiments of the present application. The server 102 is connected to the terminal device 101 through a wired network or a wireless network. The server 102 has a data receiving function, a data processing function, and a data sending function. Of course, the server 102 may also have other functions, which are not limited in the embodiments of the present application.

[0045] Based on the above implementation environment, the present application embodiment provides a trajectory generation method. Figure 2 As an example, the flow chart of a trajectory generation method provided in the embodiment of the present application is shown in FIG. Figure 1 The terminal device 101 or server 102 executes.

[0046] like Figure 2 As shown, the method comprises the following steps:

[0047] Step 201 : generating an action set for a vehicle trajectory point set, wherein the vehicle trajectory point set is obtained by downsampling original vehicle trajectory data.

[0048] In this application, a large amount of vehicle trajectory data needs to be collected and the collected vehicle trajectories are processed. The specific processing method includes but is not limited to downsampling the vehicle trajectory according to a fixed duration to form a discrete vehicle trajectory point data set. For each discrete trajectory point, the difference between the current trajectory point and the next trajectory point is calculated to form a vector, which is used as an action and forms an initial set. For the initial set converted into actions, for each action, the direction of the previous frame action is used as the positive direction of the x-axis, and the vector of the current action in the coordinate system is calculated to form a rotated action set.

[0049] Step 202: Based on a vocabulary generation algorithm, a sub-action set satisfying a distance condition is selected from the action set, and the sub-action set is encoded to generate a vocabulary set, wherein the distance condition includes that the distance between any actions is not less than a preset threshold.

[0050] In the present application, the process of selecting a sub-action set that meets the distance condition from the action set based on the vocabulary generation algorithm is as follows: first, based on the initialization of the skip count to zero, the action is selected cyclically until the selected vocabulary set reaches the expected size. Each time an action is selected, the distance between the action and all actions in the current vocabulary set is calculated. If the distance between the selected action and any action in the set is less than the preset minimum distance, the action is considered too similar, and the skip count is increased and the action is reselected. If the skip count reaches the set maximum value, the algorithm terminates and prompts that sufficiently diverse actions cannot be found. If the distance between the selected action and all existing actions is greater than or equal to the minimum distance, the action is added to the vocabulary set and the skip count is reset. Repeat the above steps until the vocabulary set reaches the expected size and finally returns to the set.

[0051] If described by mathematical formula, specifically, randomly select an action a from the action set A. i , for any action v in the current sub-action set V j , calculate the distance d(a i ,v j ),like Then V=V∪{a i}, otherwise, skip and reselect the action.

[0052] If |V| reaches the set value or the number of skips reaches the maximum value, the algorithm stops; wherein the action set is A = {a 1 ,a 2 ,…,a n}, the sub-action set is V = {v 1 ,v 2 ,…,v m}, where the minimum distance threshold is ∈, m≤n.

[0053] Based on the above algorithm, the rotated action set is sampled to generate a sampling-based vocabulary, thereby ensuring that the distance between any two actions in the selected vocabulary set is not less than ∈, achieving action diversity.

[0054] The process of encoding the sub-action set and generating the vocabulary set is carried out in the trajectory lexicalizer. The goal of the lexicalizer is to establish the connection between the trajectory and the encoding. First, N trajectory points are converted into N-1 actions (each action is a two-dimensional vector), and then each action is encoded. The above lexicalization process is as follows Figure 3shown.

[0055] Usually, the trajectory tokenizer generates an ID based on the action of a single frame. However, this tokenization method will amplify the discretization error. In the trajectory tokenizer of the embodiment of the present application, a greedy algorithm can be incorporated into the discretization process to minimize the trajectory discretization error. That is, after using the word unit to discretize the first action, the end point of the discretized action is used as the starting point for the next discretization, rather than directly using the end point of the action as the starting point. The discretization process is as follows: Figure 4 shown.

[0056] In the trajectory discretization, the embodiment of the present application adopts a relative coordinate system, takes the motion direction of the previous frame trajectory as the positive direction of the x-axis to establish a coordinate system, and selects words in this coordinate system. After using the relative coordinate system, the vocabulary set only needs to include the possible position of the vehicle relative to the previous frame, so that the vocabulary set can be effectively simplified and more in line with dynamics.

[0057] When the embodiment of the present application uses a lexicalizer to discretize a vocabulary set, a mask is used to cover part of the word unit. The specific masking method includes a longitudinal dynamics mask and a transverse dynamics mask. The specific method of the longitudinal dynamics mask is to calculate the position difference d of the previous frame motion. prev , by setting the maximum acceleration a max and maximum braking force a min , and the known time interval Δt, the feasible position difference d for the next movement can be calculated by the kinematic formula next .

[0058]

[0059] After calculating the motion disparity of the previous frame, the corresponding feasible disparity interval can be obtained. In this way, when the model uses the word list, the length of each motion word is calculated, and only the words that meet the feasible disparity interval are selected.

[0060] The specific method of the lateral dynamics mask is to calculate the following parameters:

[0061] Rear axle center moving distance Direction angle difference Δψ=ψ current -ψ last , Turning radius Front wheel angle Among them, the rear axle center position x in the previous frame last ,y last , the vehicle direction in the previous frame is ψ last , the current frame rear axis center position x current ,y current , the vehicle direction in the current frame is ψ current , and the length of the front and rear axles is L.

[0062] Based on the above formula, the word units that meet the front wheel turning angle range are selected. In this way, the above calculation is performed on the next frame position that each moving word unit in the word list may reach to obtain the front wheel turning angle corresponding to each word unit in the next frame under the current state. By setting the maximum front wheel turning angle and the minimum front wheel turning angle, only the word units that meet the feasible front wheel turning angle range can be selected.

[0063] The longitudinal dynamics mask and / or the lateral dynamics mask are used to limit the positions to which the model can move, thereby optimizing the vocabulary set.

[0064] Step 203: Search the vocabulary set for the word element and its encoding corresponding to the target action.

[0065] For the target action, search for the word element closest to the action in the vocabulary generated by sampling, and index its code (ID), expressed by the formula ID = ArgMin (Diff (action, vocabulary)). In the trajectory model with non-discrete values, the calculation of Diff changes from checking the same to calculating the similarity. For the Diff of two two-dimensional vectors, the embodiment of the present application can use the straight-line distance between the two points, Manhattan distance and other distance calculation methods. By calculating the distance between the end points of the two vectors, the similarity between them can be obtained, so as to select the word element and its ID closest to the given action in the vocabulary.

[0066] The trajectory tokenizer in the embodiment of the present application uses the Euclidean distance by calculating the straight-line distance between the action vector and the end point of each word unit vector, and selecting the word unit with the smallest straight-line distance to the end point as the word unit that is most similar to the input action.

[0067] Step 204: input the word unit of the target action and its encoding into the autoregressive model to reconstruct the trajectory to obtain the simulation trajectory points.

[0068] The embodiment of the present application also needs to reconstruct the trajectory, that is, convert the word unit ID into an action again to obtain the trajectory point. The specific method is to use the word unit code (ID) as an index to find the corresponding word unit action in the vocabulary, and then generate the trajectory point by continuously superimposing the initial position through the word unit action generation.

[0069] In the embodiment of the present application, the output features of the decoder are classified in the autoregressive model of trajectory prediction, and the distribution of word units is output, and then word units are obtained from the distribution by sampling. The specific process is as follows: Figure 5 shown.

[0070] Compared with directly regressing the position of the next word, the embodiment of the present application transforms the regression problem into a classification problem, and can model both left and right actions at the same time. Multi-distribution modeling can help generate multiple different trajectories in trajectory prediction, thereby getting rid of the single distribution problem caused by direct regression in the prior art.

[0071] Since not all trajectory actions can find completely corresponding items in the vocabulary, discretization errors will occur in the process of lexicalizing actions. The higher the accuracy of the vocabulary, the smaller the trajectory discretization error, but at the same time, due to the increase in action lemmas, the learning cost of the model is higher. Therefore, the embodiment of the present application provides a trajectory generation evaluation method.

[0072] The evaluation method includes: evaluating the accuracy of the vocabulary set using the average displacement error and / or the endpoint displacement error, specifically including calculating the average value of the distance difference between each point of the original trajectory and the corresponding point of the trajectory after the trajectory is discretized and reconstructed, and / or calculating the distance difference between the endpoint of the original trajectory and the corresponding point of the trajectory after the trajectory is discretized and reconstructed, and performing the evaluation according to the size of the average value and / or the size of the distance difference.

[0073] In the embodiment of the present application, the error of trajectory discretization is evaluated by the average displacement error and the endpoint displacement error.

[0074] The mean displacement error is calculated as the average of the distance differences between each point of the original trajectory and the corresponding point of the trajectory reconstructed after trajectory discretization.

[0075] in:

[0076] T is the total number of time steps.

[0077] is the predicted position at the tth time step.

[0078] y t is the true position at the tth time step.

[0079] |·| represents the Euclidean distance (i.e., L2 norm).

[0080] The final displacement error calculates the distance difference between the end point of the original trajectory and the corresponding point of the trajectory reconstructed by trajectory discretization.

[0081]

[0082] In the embodiment of the present application, in order to more accurately evaluate the accuracy generated by the embodiment of the present application, displacement error is used for evaluation, which makes the evaluation more efficient.

[0083] The trajectory generation method provided in the embodiment of the present application generates an efficient motion vocabulary by sampling a large amount of vehicle motion data, thereby improving the dynamic simulation of the model. At the same time, it also avoids the problem that the traditional trajectory prediction model method is prone to single distribution, resulting in insufficient diversity in trajectory prediction and the inability to grasp the dynamic characteristics well.

[0084] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0085] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can 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 thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0086] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device obtains the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0087] The computer program instructions for performing the operation of the present invention can be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages. Computer-readable program instructions can be executed completely on a user's computer, partially on a user's computer, executed as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). In certain embodiments, by utilizing the state information of computer-readable program instructions to customize electronic circuits, such as programmable logic circuits, field programmable gate arrays (FPGAs) or programmable logic arrays (PLAs), the electronic circuits can execute computer-readable program instructions, thereby realizing various aspects of the present invention.

[0088] Various aspects of the present invention are described herein with reference to the flow charts and / or block diagrams of the methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each box of the flow chart and / or block diagram and the combination of each box in the flow chart and / or block diagram can be implemented by computer-readable program instructions.

[0089] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0090] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0091] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of an instruction, and a part of the module, program segment or instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0092] Embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terms used herein are selected to best explain the principles of the embodiments, practical applications, or technical improvements in the marketplace, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the present invention is defined by the appended claims.

Claims

1. A method for generating an automatic driving simulation trajectory, characterized in that: The method comprises: Generate an action set for a vehicle trajectory point set, wherein the vehicle trajectory point set is obtained by downsampling original vehicle trajectory data; Based on a vocabulary generation algorithm, a sub-action set satisfying a distance condition is selected from the action set, and the sub-action set is encoded to generate a vocabulary set, wherein the distance condition includes that the distance between any actions is not less than a preset threshold; Find the word element and its encoding corresponding to the target action in the vocabulary set; The word unit of the target action and its encoding are input into the autoregressive model to reconstruct the trajectory to obtain the simulation trajectory points.

2. The method according to claim 1, characterized in that: The generating of an action set for the vehicle trajectory point set specifically includes: For each discrete trajectory point in the vehicle trajectory point set, a difference vector between a current trajectory point and a next trajectory point is calculated to form an initial action set; The initial action set is subjected to coordinate system transformation to generate an action set.

3. The method according to claim 1, characterized in that The sub-action set that satisfies the distance condition is selected from the action set based on the word list generation algorithm as follows: Randomly select an action a from the action set A i , for any action v in the current sub-action set V j , calculate the distance d(a i ,v j ),like Then V=V∪{a i }, Otherwise, skip and reselect the action. If |V| reaches the set value or the number of skips reaches the maximum value, the algorithm stops. The action set is A = {a1, a2, ..., a n }, the sub-action set is V = {v1, v2, ..., v m }, where the minimum distance threshold is ∈, m≤n.

4. The method according to claim 1, characterized in that: The vocabulary set is to establish a coordinate system by taking the moving direction of the previous frame trajectory as the positive direction of the X-axis, and select word units under the coordinate system.

5. The method according to claim 1, characterized in that The word units in the vocabulary set are obtained by screening through a transverse dynamic mask and / or a longitudinal dynamic mask.

6. The method according to claim 5, characterized in that The word units in the vocabulary set are obtained by longitudinal dynamic mask screening, specifically including the position difference d of the previous frame motion prev And set the maximum acceleration a max and maximum braking force a min and the known time interval Δt, the feasible position difference d for the next movement is calculated based on the following kinematic formula next . Based on the above formula, word units that meet the feasible position difference interval are selected.

7. The method according to claim 6, characterized in that The word units in the vocabulary set are obtained by lateral dynamic mask screening, specifically including: Calculate the rear axle center movement distance Direction angle difference Δψ=ψ current -ψ last , Turning radius Front wheel angle Among them, the rear axle center position x in the previous frame last ,y last , the vehicle direction in the previous frame is ψ last , the axis center position x after the current frame current ,y current , the vehicle direction in the current frame is ψ current , and the length of the front and rear axles is L, based on the above formula, the words that meet the front wheel turning angle range are selected.

8. The method according to claim 1, characterized in that: The step of searching for the word element and its encoding of the target action through the vocabulary set specifically includes: The word element that is most similar to the target action is determined by calculating the similarity, and its encoding is indexed.

9. The method according to claim 8, characterized in that Determining the word element that is most similar to the target action by calculating the similarity specifically includes: The distance between the vector of the target action and the end point of each word element vector in the vocabulary set is calculated, and the word element with the smallest distance is selected as the word element most similar to the target action, and the distance is a straight-line distance or a Manhattan distance.

10. The method according to claim 1, characterized in that ,The word unit and the encoding of the target action are searched through the vocabulary set, including using the word unit to discretize the first action, and taking the end point of the discretized action as the starting point of the next word unit discretization.

11. The method according to claims 1-10, characterized in that The accuracy of the vocabulary set is evaluated using the average displacement error and / or the endpoint displacement error, specifically including calculating the average value of the distance difference between each point of the original trajectory and the corresponding point of the trajectory after trajectory discretization and reconstruction and / or calculating the distance difference between the endpoint of the original trajectory and the corresponding point of the trajectory after trajectory discretization and reconstruction, and performing evaluation based on the size of the average value and / or the size of the distance difference.

12. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 11 when being executed by an electronic device.

Images