Methods, apparatus, equipment and storage media for generating vehicle trajectory samples
By analytically solving a pre-built mapping table and a set of boundary constraint equations, continuous spatiotemporal trajectory samples were generated, solving the problems of low trajectory generation efficiency and insufficient smoothness in autonomous driving, and achieving efficient trajectory sample generation.
Patent Information
- Application Number
- CN202611621233.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-20
- Publication Date
- 2026-11-17
AI Technical Summary
In existing autonomous driving trajectory planning, trajectory generation efficiency is low under complex behavior sequences, and it is difficult to balance kinematic smoothness and real-time solution efficiency, resulting in insufficient matching between trajectory and behavioral intent and high computational overhead.
Based on a pre-built mapping table between driving behavior and polynomial order, the corresponding polynomial order is determined for each behavior segment. A set of boundary constraint equations is constructed and solved analytically in combination with the desired motion state to generate continuous spatiotemporal trajectory samples.
It improves the efficiency of trajectory generation under complex behavior sequences, ensures the continuity and smoothness of the trajectory, and adapts to the trajectory sample generation needs under complex driving scenarios.
Smart Images

Figure CN123104703A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of autonomous driving trajectory planning and control, and in particular to a method, apparatus, device and storage medium for generating vehicle trajectory samples. Background Technology
[0002] In autonomous driving trajectory planning, the longitudinal and lateral movements of the vehicle are usually parameterized by using polynomial curves based on the behavior sequence output by the decision module, and a continuous spatiotemporal trajectory is output for the controller to execute.
[0003] However, existing solutions often use a uniform order curve to process different driving behaviors. The splicing between segments often only ensures the continuity of position and speed, making it difficult to balance kinematic smoothness and real-time solution efficiency. This can easily lead to insufficient matching between trajectory and behavioral intent and high computational overhead.
[0004] Therefore, how to improve the efficiency of trajectory sample generation while ensuring the continuity of trajectory generation under complex behavioral sequences has become the technical problem that this application needs to solve. Summary of the Invention
[0005] This application provides a method, apparatus, device, and storage medium for generating vehicle trajectory samples to solve the aforementioned technical problems. For a driving behavior sequence corresponding to a driving scenario to be mined, a pre-constructed mapping table of driving behavior and polynomial order is used to determine the corresponding polynomial order for each behavior segment in the sequence. A set of boundary constraint equations is constructed based on the expected motion states at the start and end times of each behavior segment, and the corresponding polynomials are obtained analytically. The polynomials corresponding to each behavior segment are then concatenated in chronological order to generate a continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, which serves as a trajectory sample. This scheme improves the trajectory generation efficiency under complex behavior sequences by matching the order through a lookup table, and ensures trajectory continuity through polynomial modeling with unified boundary constraints and temporal alignment concatenation, achieving a balance between generation efficiency and trajectory quality.
[0006] Firstly, this application provides a method for generating vehicle trajectory samples, including:
[0007] Obtain the driving behavior sequence corresponding to the driving scenario to be explored. The driving behavior sequence contains multiple behavior segments arranged in chronological order.
[0008] Based on the driving behavior sequence, extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment;
[0009] Based on the behavior type of each behavior segment and the pre-built mapping table between driving behavior and polynomial order, determine the polynomial order corresponding to each behavior segment.
[0010] For each behavior segment, a set of boundary constraint equations is constructed based on the polynomial order and the desired motion state. The set of boundary constraint equations is then solved analytically to determine the polynomial corresponding to the behavior segment.
[0011] The polynomials corresponding to each behavior segment are concatenated according to the time sequence of driving behavior to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, and used as the trajectory sample of the driving scenario to be explored.
[0012] In one possible implementation, obtaining the driving behavior sequence corresponding to the driving scenario to be mined includes:
[0013] Obtain the original driving semantic data corresponding to the driving scenario to be explored;
[0014] The original driving semantic data is temporally segmented and semantically labeled to obtain multiple behavior segments arranged in chronological order.
[0015] Each behavior segment is matched with corresponding motion attributes and duration parameters to form a driving behavior sequence.
[0016] In one possible implementation, the mapping table between driving behavior and polynomial order is constructed as follows:
[0017] Obtain the kinematic characteristic parameters of various driving behaviors and the preset vehicle ride comfort control index, analyze the corresponding boundary constraint requirements and continuity smoothness requirements, and transform the continuity smoothness requirements into inter-segment continuity constraint rules.
[0018] Based on the boundary constraint requirements and the requirement for continuous smoothness, a polynomial model of the corresponding order is matched, and the relationship between various driving behaviors, polynomial models and inter-segment continuity constraint rules is stored in a structured manner to obtain a mapping table between driving behaviors and polynomial orders.
[0019] In one possible implementation, determining the polynomial order corresponding to each action segment includes:
[0020] The behavior type of each behavior segment is broken down into lateral movement type and longitudinal movement type;
[0021] The lateral motion type and longitudinal motion type are respectively substituted into the mapping table between driving behavior and polynomial order for matching, and the polynomial order corresponding to the lateral motion type and longitudinal motion type is obtained respectively.
[0022] In one possible implementation, it also includes:
[0023] Obtain the road curvature parameters, vehicle dynamics envelope parameters, and current working condition attachment parameters corresponding to the driving scenario to be explored;
[0024] Based on the road curvature parameters, vehicle dynamics envelope parameters, and current operating condition attachment parameters, the polynomial order corresponding to each behavior segment is adjusted to obtain the adjusted polynomial order.
[0025] In one possible implementation, the polynomial order corresponding to each behavior segment is adjusted based on road curvature parameters, vehicle dynamics envelope parameters, and current operating condition attachment parameters to obtain the adjusted polynomial order, including:
[0026] When the road curvature parameter is higher than the preset curvature threshold, and / or the current working condition adhesion parameter is less than the road surface adhesion envelope limit in the vehicle dynamics envelope parameter, it is determined that there is a high curvature road segment, and / or there is a low adhesion road surface feature. The order matching rules and inter-segment continuity constraint rules corresponding to the behavior type in the mapping relationship table are adjusted.
[0027] Based on the adjusted order matching rules and behavior type, the adjusted polynomial order is obtained.
[0028] In one possible implementation, adjusting the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table includes:
[0029] Based on the road curvature parameters, the corresponding curvature risk levels are classified;
[0030] Determine the upper limit of the rate of change of acceleration under the current operating condition based on the vehicle dynamics envelope parameters;
[0031] Based on the curvature risk level and the upper limit of the rate of change of acceleration, adjust the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table.
[0032] In one possible implementation, the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table are adjusted based on the curvature risk level and the upper limit of the rate of change of acceleration, including:
[0033] Based on the curvature risk level and the upper limit of the rate of change of acceleration, determine the minimum polynomial order and the minimum continuity constraint order required for the corresponding behavior type;
[0034] If the order matching rule is less than the lowest polynomial order, then the polynomial order of the corresponding behavior type is increased.
[0035] If the inter-segment continuity constraint rule is less than the minimum continuity constraint order, then the inter-segment continuity constraint order of the corresponding behavior type is increased.
[0036] In one possible implementation, based on the polynomial order and the desired motion state, a set of boundary constraint equations corresponding to the behavior segment is constructed, including:
[0037] Extract the position, velocity, and acceleration parameters at the initial moment of the desired motion state as the initial boundary conditions;
[0038] Extract the position, velocity, and acceleration parameters at the termination moment in the desired motion state as termination boundary conditions;
[0039] Based on the polynomial order, initial boundary conditions, and final boundary conditions, construct the boundary constraint equation system corresponding to the behavior segment.
[0040] In one possible implementation, the boundary constraint equations are solved analytically to determine the polynomial corresponding to the behavior segment, including:
[0041] The boundary constraint equations are transformed into a standard linear matrix, and the standard linear matrix is then analyzed algebraically to obtain the multiple undetermined coefficients of the polynomial.
[0042] Based on the multiple undetermined coefficients of the polynomial, generate the polynomial corresponding to the behavior segment.
[0043] In one possible implementation, the boundary constraint equations are solved analytically to determine the polynomials corresponding to the behavior segments, including:
[0044] Based on the boundary constraint equations, determine the boundary constraint conditions corresponding to the behavior segment, arrange the boundary constraint conditions in order according to the dimension of the corresponding polynomial order, and generate the boundary constraint vector.
[0045] The boundary constraint vector is projected onto the basis function coefficient space of the pre-constructed normalized time basis function derivative mapping matrix, and multiple coefficients of the polynomial are extracted through matrix-vector multiplication.
[0046] The polynomial corresponding to each behavior segment is determined based on each coefficient.
[0047] In one possible implementation, it also includes:
[0048] Extract the inter-segment continuity constraint rules corresponding to adjacent behavior segments from the mapping relationship table;
[0049] Based on the inter-segment continuity constraint rules, verify the consistency of motion state parameters of adjacent behavior segments at the connection time;
[0050] When the verification is inconsistent, if the degree of freedom of the current polynomial order cannot meet the requirements of the inter-segment continuity constraint rule, then the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirement.
[0051] In one possible implementation, the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement, including:
[0052] Obtain the total degrees of freedom required to acquire the self-boundary constraints of the behavior segment and the corresponding inter-segment continuity constraints, as well as the available degrees of freedom corresponding to the current polynomial order;
[0053] The total degrees of freedom are compared with the available degrees of freedom. When the total degrees of freedom exceed the available degrees of freedom, the polynomial order of the behavior segment is increased. The corresponding boundary constraint equations are reconstructed and solved based on the increased polynomial order to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement.
[0054] In one possible implementation, the polynomials corresponding to each behavior segment are concatenated according to the temporal order of the driving behavior sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, including:
[0055] Obtain the time interval corresponding to each behavior segment;
[0056] Based on the temporal order of the driving behavior sequence, the polynomials corresponding to adjacent behavior segments are sequentially connected, and the time nodes and motion state parameters at the connection moments are aligned according to the time intervals corresponding to each behavior segment, to obtain the polynomial sequence after connection.
[0057] By integrating the polynomial sequences, the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored is obtained.
[0058] In one possible implementation, the polynomial sequence is integrated to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, including:
[0059] Based on the preset time sampling step, numerical calculations are performed on each polynomial in the polynomial sequence within the corresponding time interval to obtain the motion state parameters corresponding to each sampling moment.
[0060] The motion state parameters corresponding to each sampling moment are spliced and integrated in chronological order to form a continuous temporal motion dataset. The temporal motion dataset is then structured and encapsulated to generate continuous spatiotemporal trajectory data corresponding to the driving scenario to be mined.
[0061] Secondly, this application provides a vehicle trajectory sample generation device, comprising:
[0062] The acquisition module is used to acquire the driving behavior sequence corresponding to the driving scenario to be explored. The driving behavior sequence contains multiple behavior segments arranged in chronological order.
[0063] The processing module is used to extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment based on the driving behavior sequence.
[0064] The processing module is also used to determine the polynomial order corresponding to each behavior segment based on the behavior type of each behavior segment and the pre-built mapping table between driving behavior and polynomial order.
[0065] The processing module is also used to construct a set of boundary constraint equations for each behavior segment based on the polynomial order and the desired motion state, and to solve the set of boundary constraint equations analytically to determine the polynomial corresponding to the behavior segment.
[0066] The processing module is also used to splice the polynomials corresponding to each behavior segment according to the time sequence of driving behavior to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, and use it as the trajectory sample of the driving scenario to be explored.
[0067] In one possible implementation, the acquisition module is further configured to:
[0068] Obtain the original driving semantic data corresponding to the driving scenario to be explored;
[0069] The original driving semantic data is temporally segmented and semantically labeled to obtain multiple behavior segments arranged in chronological order.
[0070] Each behavior segment is matched with corresponding motion attributes and duration parameters to form a driving behavior sequence.
[0071] In one possible implementation, the processing module is also used to construct a mapping table between driving behavior and polynomial order:
[0072] Obtain the kinematic characteristic parameters of various driving behaviors and the preset vehicle ride comfort control index, analyze the corresponding boundary constraint requirements and continuity smoothness requirements, and transform the continuity smoothness requirements into inter-segment continuity constraint rules.
[0073] Based on the boundary constraint requirements and the requirement for continuous smoothness, a polynomial model of the corresponding order is matched, and the relationship between various driving behaviors, polynomial models and inter-segment continuity constraint rules is stored in a structured manner to obtain a mapping table between driving behaviors and polynomial orders.
[0074] In one possible implementation, the processing module is further configured to:
[0075] The behavior type of each behavior segment is broken down into lateral movement type and longitudinal movement type;
[0076] The lateral motion type and longitudinal motion type are respectively substituted into the mapping table between driving behavior and polynomial order for matching, and the polynomial order corresponding to the lateral motion type and longitudinal motion type is obtained respectively.
[0077] In one possible implementation, the processing module is further configured to:
[0078] Obtain the road curvature parameters, vehicle dynamics envelope parameters, and current working condition attachment parameters corresponding to the driving scenario to be explored;
[0079] Based on the road curvature parameters, vehicle dynamics envelope parameters, and current operating condition attachment parameters, the polynomial order corresponding to each behavior segment is adjusted to obtain the adjusted polynomial order.
[0080] In one possible implementation, the processing module is further configured to:
[0081] When the road curvature parameter is higher than the preset curvature threshold, and / or the current working condition adhesion parameter is less than the road surface adhesion envelope limit in the vehicle dynamics envelope parameter, it is determined that there is a high curvature road segment, and / or there is a low adhesion road surface feature. The order matching rules and inter-segment continuity constraint rules corresponding to the behavior type in the mapping relationship table are adjusted.
[0082] Based on the adjusted order matching rules and behavior type, the adjusted polynomial order is obtained.
[0083] In one possible implementation, the processing module is further configured to:
[0084] Based on the road curvature parameters, the corresponding curvature risk levels are classified;
[0085] Determine the upper limit of the rate of change of acceleration under the current operating condition based on the vehicle dynamics envelope parameters;
[0086] Based on the curvature risk level and the upper limit of the rate of change of acceleration, adjust the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table.
[0087] In one possible implementation, the processing module is further configured to:
[0088] Based on the curvature risk level and the upper limit of the rate of change of acceleration, determine the minimum polynomial order and the minimum continuity constraint order required for the corresponding behavior type;
[0089] If the order matching rule is less than the lowest polynomial order, then the polynomial order of the corresponding behavior type is increased.
[0090] If the inter-segment continuity constraint rule is less than the minimum continuity constraint order, then the inter-segment continuity constraint order of the corresponding behavior type is increased.
[0091] In one possible implementation, the processing module is further configured to:
[0092] Extract the position, velocity, and acceleration parameters at the initial moment of the desired motion state as the initial boundary conditions;
[0093] Extract the position, velocity, and acceleration parameters at the termination moment in the desired motion state as termination boundary conditions;
[0094] Based on the polynomial order, initial boundary conditions, and final boundary conditions, construct the boundary constraint equation system corresponding to the behavior segment.
[0095] In one possible implementation, the processing module is further configured to:
[0096] The boundary constraint equations are transformed into a standard linear matrix, and the standard linear matrix is then analyzed algebraically to obtain the multiple undetermined coefficients of the polynomial.
[0097] Based on the multiple undetermined coefficients of the polynomial, generate the polynomial corresponding to the behavior segment.
[0098] In one possible implementation, the processing module is further configured to:
[0099] Based on the boundary constraint equations, determine the boundary constraint conditions corresponding to the behavior segment, arrange the boundary constraint conditions in order according to the dimension of the corresponding polynomial order, and generate the boundary constraint vector.
[0100] The boundary constraint vector is projected onto the basis function coefficient space of the pre-constructed normalized time basis function derivative mapping matrix, and multiple coefficients of the polynomial are extracted through matrix-vector multiplication.
[0101] The polynomial corresponding to each behavior segment is determined based on each coefficient.
[0102] In one possible implementation, the processing module is further configured to:
[0103] Extract the inter-segment continuity constraint rules corresponding to adjacent behavior segments from the mapping relationship table;
[0104] Based on the inter-segment continuity constraint rules, verify the consistency of motion state parameters of adjacent behavior segments at the connection time;
[0105] When the verification is inconsistent, if the degree of freedom of the current polynomial order cannot meet the requirements of the inter-segment continuity constraint rule, then the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirement.
[0106] In one possible implementation, the processing module is further configured to:
[0107] Obtain the total degrees of freedom required to acquire the self-boundary constraints of the behavior segment and the corresponding inter-segment continuity constraints, as well as the available degrees of freedom corresponding to the current polynomial order;
[0108] The total degrees of freedom are compared with the available degrees of freedom. When the total degrees of freedom exceed the available degrees of freedom, the polynomial order of the behavior segment is increased. The corresponding boundary constraint equations are reconstructed and solved based on the increased polynomial order to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement.
[0109] In one possible implementation, the processing module is further configured to:
[0110] Obtain the time interval corresponding to each behavior segment;
[0111] Based on the temporal order of the driving behavior sequence, the polynomials corresponding to adjacent behavior segments are sequentially connected, and the time nodes and motion state parameters at the connection moments are aligned according to the time intervals corresponding to each behavior segment, to obtain the polynomial sequence after connection.
[0112] By integrating the polynomial sequences, the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored is obtained.
[0113] In one possible implementation, the processing module is further configured to:
[0114] Based on the preset time sampling step, numerical calculations are performed on each polynomial in the polynomial sequence within the corresponding time interval to obtain the motion state parameters corresponding to each sampling moment.
[0115] The motion state parameters corresponding to each sampling moment are spliced and integrated in chronological order to form a continuous temporal motion dataset. The temporal motion dataset is then structured and encapsulated to generate continuous spatiotemporal trajectory data corresponding to the driving scenario to be mined.
[0116] Thirdly, this application provides an electronic device, including: a memory and a processor;
[0117] The memory stores instructions that the computer executes;
[0118] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.
[0119] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible embodiments of the first aspect.
[0120] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.
[0121] This application provides a method, apparatus, device, and storage medium for generating vehicle trajectory samples. For a driving behavior sequence corresponding to a driving scenario to be mined, it determines the corresponding polynomial order for each behavior segment in the sequence based on a pre-built mapping table between driving behavior and polynomial order. It then constructs a set of boundary constraint equations based on the expected motion states at the start and end times of each behavior segment and solves them analytically to obtain the polynomials corresponding to each behavior segment. Finally, it concatenates the polynomials corresponding to each behavior segment according to the time sequence of the driving behavior sequence to generate a continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, which serves as the trajectory sample. This scheme quickly determines the polynomial order of each behavior segment through table lookup matching, effectively improving the generation efficiency of trajectory samples under complex behavior sequences. Simultaneously, it ensures the continuity and smoothness of the trajectory in the spatiotemporal dimension through polynomial modeling under boundary constraints and temporal alignment concatenation, balancing generation efficiency with trajectory sample quality, and adapting to the trajectory sample generation needs of complex driving scenarios. Attached Figure Description
[0122] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0123] Figure 1 This application provides a schematic diagram of an application data processing system architecture.
[0124] Figure 2 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 1 ;
[0125] Figure 3 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 2 ;
[0126] Figure 4 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 3 ;
[0127] Figure 5 This is a schematic diagram of the vehicle trajectory sample generation device provided in the embodiments of this application;
[0128] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0129] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0130] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0131] Autonomous driving trajectory planning technology belongs to the field of vehicle motion control and path generation, and is typically deployed in the trajectory generation module between perception, decision-making, and control. This module receives temporal action information such as lane changing, following, turning, and stopping from the upstream behavior decision output, and combines it with the vehicle state and road environment to generate a continuous spatiotemporal trajectory that can be directly executed by the underlying controller, supporting the real-time operation of vehicles such as passenger cars and unmanned delivery vehicles on structured roads.
[0132] Existing trajectory generation methods typically use polynomial curves to parameterize the longitudinal and lateral motions of the vehicle, then fit trajectories for each behavior segment and stitch them together on the time axis. In practical applications, systems often use a uniform order curve for different driving behaviors, with multiple constraints on position and velocity continuity between segments. There is a lack of fine-grained correspondence between boundary conditions and behavioral semantics, and complex behavior sequences require repeated adjustments to curve parameters. The resulting trajectories are prone to abrupt acceleration changes, insufficient curve smoothness, and intent matching deviations, and the iterative solution process also incurs high computational costs.
[0133] Therefore, how to improve the efficiency of trajectory sample generation while ensuring the continuity of trajectory generation under complex behavioral sequences has become the technical problem that this application needs to solve.
[0134] To address the aforementioned issues, this application provides a method, apparatus, device, and storage medium for generating vehicle trajectory samples. For the driving behavior sequence corresponding to the driving scenario to be mined, a pre-constructed mapping table between driving behavior and polynomial order is used to determine the corresponding polynomial order for each behavior segment in the sequence. The expected motion states at the start and end times of each behavior segment are combined to construct a corresponding set of boundary constraint equations, which are then analytically solved to obtain the polynomials corresponding to each behavior segment. The polynomials corresponding to each behavior segment are then concatenated according to the temporal order of the driving behavior sequence to generate a continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, serving as a trajectory sample. This scheme quickly determines the polynomial order of each behavior segment through table lookup matching, effectively improving the generation efficiency of trajectory samples under complex behavior sequences. Simultaneously, polynomial modeling under boundary constraints and temporal alignment concatenation ensure the continuity and smoothness of the trajectory in the spatiotemporal dimension, balancing generation efficiency with trajectory sample quality, and adapting to the trajectory sample generation needs of complex driving scenarios.
[0135] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.
[0136] Figure 1 This is a schematic diagram of an application data processing system architecture provided in an embodiment of this application. The application data processing system 100 is a vehicle domain controller. Figure 1 As shown, the above architecture includes a data acquisition unit 101 and a processing unit 102.
[0137] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the architecture of the application data processing system. In other feasible embodiments of this application, the above architecture may include more or fewer components than illustrated, or combine some components, or split some components, or arrange different components, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown can be implemented in hardware, software, or a combination of both.
[0138] In the specific implementation process, the data acquisition unit 101 may include an input / output interface or a communication interface, and the data acquisition unit 101 may be connected to the processing unit through the input / output interface or the communication interface.
[0139] The processing unit 102 can extract the behavior type and expected motion state of each behavior segment, as well as the start and end times, based on the driving behavior sequence containing multiple behavior segments arranged in chronological order corresponding to the driving scenario to be mined. It can determine the polynomial order corresponding to each behavior segment according to the pre-constructed mapping table between driving behavior and polynomial order. It can construct the boundary constraint equation system corresponding to each behavior segment by combining the polynomial order and the expected motion state and solve it analytically to obtain the corresponding polynomial. It can then splice the polynomials corresponding to each behavior segment according to the chronological order of the driving behavior sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be mined and use it as the trajectory sample of the driving scenario.
[0140] It should be understood that the above processing unit can be implemented by a processor reading instructions from memory and executing those instructions, or it can be implemented by a chip circuit.
[0141] Furthermore, the network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0142] Figure 2 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 1 ,like Figure 2 As shown, the vehicle trajectory sample generation method provided in this embodiment includes:
[0143] S201. Obtain the driving behavior sequence corresponding to the driving scenario to be explored. The driving behavior sequence contains multiple behavior segments arranged in chronological order.
[0144] For example, the driving behavior sequence corresponding to the driving scenario to be mined can be output and received by the upstream behavior decision module. The received driving behavior sequence is then processed as necessary to identify multiple behavior segments arranged in chronological order.
[0145] Driving behavior sequence is used to represent the driving behavior input in the driving scenario to be explored. It is the basic data for subsequent segmented modeling and trajectory generation. Its content can include the temporal combination of behaviors such as lane keeping, automatic lane changing, cruise control, turning, parking, and emergency avoidance. Each behavior segment corresponds to a continuous time window, and adjacent behavior segments are ordered according to the order of changes in actual driving intention.
[0146] In one possible embodiment, after receiving the driving behavior sequence, the system can structurally decompose it according to a preset data structure, recording each behavior segment as an entry containing a start time, an end time, and a behavior label, so as to process each segment subsequently.
[0147] If the upstream output already carries the behavior segment boundary, the boundary information can be read directly; if only the entire behavior flow is output, it can be segmented according to the preset segmentation rules to obtain multiple continuous and non-overlapping behavior segments.
[0148] It should be understood that the above examples are for illustrative purposes only and are not intended to be limiting.
[0149] In one possible implementation, obtaining the driving behavior sequence corresponding to the driving scenario to be mined includes:
[0150] Obtain the original driving semantic data corresponding to the driving scenario to be mined; perform temporal segmentation and semantic annotation on the original driving semantic data to obtain multiple behavior segments arranged in chronological order; match the corresponding motion attributes and duration parameters for each behavior segment to form a driving behavior sequence.
[0151] In this embodiment, the original driving semantic data can come from vehicle operation logs, upstream behavior decision outputs, road event records, or manually annotated scene semantic text. Its content is used to characterize the driving intentions, action states, and scene change information in the driving scenario to be mined.
[0152] After receiving the raw driving semantic data, the system can perform temporal segmentation based on preset time boundaries and action switching rules, breaking down the continuous semantic data into multiple independently processable behavioral segments. After segmentation, each segment is semantically labeled to clarify its corresponding driving meaning. The semantic labeling can include markings for behavior types such as lane changing, following, turning, parking, deceleration, or acceleration, and combining these with their temporal position in the scene to form behavioral segments arranged in chronological order.
[0153] Motion attributes describe the kinematic characteristics of each behavior segment, reflecting the control intent of the behavior segment in longitudinal, lateral, or posture changes; duration parameters represent the expected time span of the behavior segment, corresponding to the start and end boundaries of the behavior segment.
[0154] After semantic annotation, the system can match each behavior segment with preset motion attribute templates and time parameter templates, and write the matching results into a unified data structure to form a driving behavior sequence. This driving behavior sequence can be directly called by the subsequent trajectory generation module to extract behavior type, start and end times, and corresponding motion constraint information.
[0155] The semantic parsing model or annotation model used can be either a rule-based parsing engine or a neural network-based sequence labeling model. In practical applications, other models can also be selected for this component, and this application does not limit this.
[0156] By first segmenting the raw driving semantic data into discrete behavior segments, and then supplementing each behavior segment with motion attributes and duration parameters, the temporal changes in the driving scene can be kept consistent with the semantic relationships of actions. This provides structured input for subsequent extraction of motion states by behavior segment, determination of polynomial order, and construction of boundary constraints. The resulting driving behavior sequence has a clear temporal order and semantic correspondence, which can improve the scene parsing and behavior consistency when generating subsequent trajectory samples, and reduce trajectory splicing errors caused by incomplete semantic information.
[0157] S202. Based on the driving behavior sequence, extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment.
[0158] In this embodiment, the behavior segment is used to break down the driving behavior sequence into the smallest time action unit that is easy to model segment by segment. After the system extracts multiple behavior segments from the driving behavior sequence, it determines the behavior type of each behavior segment and records the expected motion state at the start and end times of the behavior segment.
[0159] The behavior type is used to identify the specific driving intention or action category corresponding to each behavior segment, and is an important basis for determining the order of the polynomial. The expected motion state is used to describe the kinematic goal that each behavior segment should achieve at the start and end times, and is the input condition for constructing the boundary constraint equation system.
[0160] In one possible embodiment, the system can parse the driving behavior sequence segment by segment, and associate each behavior segment with its corresponding behavior label; at the same time, the system extracts the target state at the start and end times of the behavior segment as the expected motion state of the behavior segment.
[0161] To generate the input data required for subsequent constraint solving, the system can encapsulate the behavior type, start time boundary conditions, and end time boundary conditions of each behavior segment into structured data records and store them in the order of behavior segments.
[0162] If the starting or ending motion state of a certain behavior segment comes from the upstream planning result, the system can directly reference the state value; if it comes from the perception and prediction result, it can be calculated through the corresponding state information.
[0163] It should be understood that the above examples are for illustrative purposes only and are not intended to be limiting.
[0164] S203. Based on the behavior type of each behavior segment and the pre-built mapping table between driving behavior and polynomial order, determine the polynomial order corresponding to each behavior segment.
[0165] In this embodiment, the polynomial order is used to describe the order of the polynomial model corresponding to the behavior segment, which is used to match the requirements of different driving behaviors for boundary constraints and continuity. The mapping relationship table between driving behavior and polynomial order is used to establish the correspondence between driving behavior and polynomial order, which is the core basis for realizing structured mapping and fast order selection.
[0166] In one possible embodiment, the system pre-establishes a mapping table between driving behavior and polynomial order, and loads it into memory when the system starts up, as the basis rule base for subsequent order selection; when processing a certain behavior segment, the system queries the mapping table according to the behavior type of the behavior segment to obtain the corresponding polynomial order.
[0167] For behavior segments that need to satisfy multiple boundary conditions simultaneously, a higher-order polynomial can be selected; for behavior segments with fewer boundary conditions or lower continuity requirements, a lower-order polynomial can be selected. With this setting, the determination of the order does not require a complex search for each behavior segment individually, but can be obtained by looking up a table to find the polynomial order suitable for the current behavior segment.
[0168] It should be understood that the above examples are for illustrative purposes only and are not intended to be limiting.
[0169] In one possible implementation, the mapping table between driving behavior and polynomial order is constructed as follows:
[0170] The kinematic characteristic parameters of various driving behaviors and the preset vehicle ride comfort control index are obtained. The corresponding boundary constraint requirements and continuous smoothness requirements are analyzed and transformed into inter-segment continuous constraint rules. Based on the boundary constraint requirements and continuous smoothness requirements, a polynomial model of the corresponding order is matched. The relationship between various driving behaviors, polynomial models and inter-segment continuous constraint rules is structured and stored to obtain a mapping table between driving behaviors and polynomial orders.
[0171] In this embodiment, the kinematic characteristic parameters of driving behavior are used to characterize the differences in speed, acceleration, curvature and displacement changes of behaviors such as lane changing, following, turning and parking. The vehicle ride comfort control index is used to limit the constraints such as displacement continuity, speed continuity and acceleration change limitation when the trajectory is connected between segments.
[0172] Boundary constraint requirements are used to describe the state conditions that each type of driving behavior must satisfy at the start and end points. Continuity smoothness requirements are used to describe the transition constraints between adjacent behavior segments. Inter-segment continuity constraint rules are written as a unified constraint expression that can be called for subsequent solutions.
[0173] In practical implementation, a behavior label table can be established according to the driving behavior type, and the typical kinematic parameter range of each behavior can be obtained by combining historical trajectory sample statistics. Then, the preset smoothness control index can be mapped to the corresponding boundary terms and continuous term constraint set.
[0174] Subsequently, based on the requirements of different behaviors on the curve's degrees of freedom and endpoint constraint strength, a polynomial model of the corresponding order is selected, such as a fifth-order or higher-order polynomial to satisfy higher-order continuity. The behavior type, polynomial order, and inter-segment continuity constraint rules are written into a structured storage unit to form a queryable mapping table.
[0175] During operation, the mapping relationship table is called by the trajectory generation module as the basis for order selection. After identifying the behavior segment type, the module can directly read the corresponding order and continuity constraint rules, and construct the boundary constraint equation system accordingly.
[0176] Since a one-to-one correspondence has been established between behavioral semantics, constraint expression, and polynomial model, the subsequent solution process can complete the parameter determination according to the preset rules, thereby outputting a segmented trajectory that meets the smoothness requirements.
[0177] After adopting this implementation method, a stable structured correspondence is formed between driving behavior and polynomial order, which enables different driving behaviors to call the matching polynomial model and inter-segment continuity constraint rules, reducing the problem of state discontinuity during segment splicing, and reducing the dependence on repeated parameter adjustments during trajectory generation.
[0178] S204. For each behavior segment, construct the boundary constraint equation system corresponding to the behavior segment based on the polynomial order and the desired motion state, and solve the boundary constraint equation system analytically to determine the polynomial corresponding to the behavior segment.
[0179] In this embodiment, the boundary constraint equations are used to transform the desired motion state of the behavior segment into polynomial solution conditions, which is the mathematical basis for determining the polynomial corresponding to the behavior segment.
[0180] After obtaining the polynomial order corresponding to each behavior segment, the system establishes a polynomial expression containing undetermined coefficients based on the order, and writes a set of boundary constraint equations in combination with the expected motion state at the start and end times of the behavior segment. The constraint terms in the set of equations are matched with the selected polynomial order so that the number of undetermined coefficients corresponds to the number of independent constraints.
[0181] In one possible embodiment, if the order of the polynomial corresponding to the behavior segment is n, the system constructs a polynomial with n+1 undetermined coefficients and forms a system of linear equations using the conditions at the start and end times respectively; then, the system of equations is analyzed to directly solve for the undetermined coefficients, thereby obtaining the polynomial corresponding to the behavior segment.
[0182] Because it uses an analytical solution method, the system can solve all the undetermined coefficients at once, without having to approximate them step by step through iterative search, thus reducing redundant calculations.
[0183] If the number of boundary conditions matches the polynomial degrees of freedom, then the system of equations can obtain a unique solution, thereby allowing the trajectory expression of the behavioral segment to be directly determined.
[0184] If the initial and final states of a certain behavior segment come from different sources, the system can first perform a unified coordinate system transformation and time reference alignment on the state variables, and then construct constraint equations to ensure that the solution process uses boundary conditions under the same reference system.
[0185] In this way, the motion target of the behavior segment is explicitly mapped as polynomial coefficients, and the subsequent generation results can maintain consistency with the driver's intention. Figure 1 The kinematic characteristics of the structure.
[0186] It should be understood that the above examples are for illustrative purposes only and are not intended to be limiting.
[0187] In one possible implementation, based on the polynomial order and the desired motion state, a set of boundary constraint equations corresponding to the behavior segment is constructed, including:
[0188] Extract the position, velocity, and acceleration parameters at the initial moment of the desired motion state as the initial boundary conditions; extract the position, velocity, and acceleration parameters at the final moment of the desired motion state as the final boundary conditions; construct the boundary constraint equation system corresponding to the behavior segment based on the polynomial order, the initial boundary conditions, and the final boundary conditions.
[0189] In this embodiment, the initial boundary condition is used to characterize the constraint input of the vehicle state at the beginning of the behavior segment, and the termination boundary condition is used to characterize the constraint input of the vehicle state at the end of the behavior segment. Together, they are used to limit the continuity of the position, velocity and acceleration of the polynomial trajectory at the time endpoints.
[0190] The desired motion state can be obtained from the temporal actions output by the upstream behavior decision and the target motion information at the corresponding time, which includes the kinematic parameters at the start and end points of the behavior segment. The polynomial order is used to determine the number of unknown coefficients of the polynomial to be solved and to match the number of endpoint constraints.
[0191] In the specific implementation, the system first receives the expected motion state corresponding to the behavior segment, and then parses the position information, velocity information and acceleration information at the start time and the end time to form the start boundary conditions and the end boundary conditions.
[0192] Subsequently, based on the polynomial order, write out the horizontal and vertical polynomial expressions for the action segments, as well as the corresponding first derivative (velocity) and second derivative (acceleration) expressions.
[0193] The general form of the transverse displacement polynomial is:
[0194]
[0195] In the formula, For lateral displacement, it means The amount of lateral displacement of the vehicle at any given moment; This is a time variable, corresponding to the time within the line segment; The order of the polynomial corresponding to the lateral movement is matched with the lateral movement type of the current behavior segment; The coefficients of each order of the horizontal polynomial are undetermined, totaling... indivual.
[0196] The general form of the longitudinal displacement polynomial is:
[0197]
[0198] In the formula, For longitudinal displacement, it represents The amount of displacement of the vehicle in the longitudinal direction at any given time; The order of the polynomial corresponding to the longitudinal motion; The coefficients of each order of the vertical polynomial are undetermined, totaling... indivual.
[0199] Taking a typical order as an example: the expression for a horizontal cubic polynomial is as follows: There are 4 undetermined coefficients, matching lateral behaviors with fewer boundary constraints, such as lane keeping; the lateral fifth-degree polynomial expression is... There are 6 undetermined coefficients in total, matching lateral behaviors such as lane changes and turns that require continuous acceleration; the longitudinal cubic polynomial expression is... There are 4 undetermined coefficients, matching longitudinal behaviors such as uniform velocity; the longitudinal quartic polynomial expression is: There are 5 undetermined coefficients in total, which are used to match longitudinal behaviors such as acceleration and deceleration.
[0200] Substituting the boundary conditions at the start and end times into the corresponding expressions for displacement, velocity, and acceleration, a linear constraint relationship with respect to the polynomial coefficients is constructed, thereby forming the boundary constraint equation set corresponding to the behavior segment.
[0201] This set of equations can be directly used for subsequent analytical solutions to obtain the behavior segment polynomial trajectory that satisfies the endpoint motion state constraints.
[0202] During operation, the boundary constraint equations transform the start and end states corresponding to the behavioral semantics into computable algebraic constraints, ensuring that the trajectory of the behavioral segment remains consistent with the desired motion state at the endpoints and providing a unified solution basis for segment splicing. Because the equations simultaneously constrain position, velocity, and acceleration, the trajectory of the behavioral segment can form a continuous kinematic connection at the endpoints.
[0203] With this implementation method, the trajectory construction of the behavior segment can be directly constrained by the motion state at the start and end times. The resulting polynomial matches the desired motion state more accurately, the motion continuity at the endpoints is more stable, and the number of subsequent trajectory corrections can be reduced, thereby reducing the solution overhead and improving the smoothness of trajectory generation.
[0204] In one possible implementation, the boundary constraint equations are solved analytically to determine the polynomial corresponding to the behavior segment, including:
[0205] The boundary constraint equations are transformed into a standard linear matrix, and the standard linear matrix is algebraically analyzed to obtain multiple undetermined coefficients of the polynomial. Based on the multiple undetermined coefficients of the polynomial, the polynomial corresponding to the row segment is generated.
[0206] In this embodiment, the standard linear matrix is used to uniformly represent the initial boundary conditions and the final boundary conditions in linear algebraic form. The undetermined coefficients are unknown parameters in the polynomial expansion, and their values are directly determined by the boundary constraints.
[0207] Algebraic analysis refers to performing direct analytical operations on the matrix, eliminating unknowns based on the coefficient relationships of the equations, and obtaining a unique or corresponding coefficient solution, so that the polynomial can satisfy the position, velocity, and acceleration constraints at the start and end times of the behavior segment.
[0208] For example, taking a horizontal fifth-degree polynomial as an example, let the starting time of the line segment be... The termination time is ,in, Let be the duration of the behavior segment, i.e., the length of the time interval corresponding to this behavior segment. Substituting the position, velocity, and acceleration boundary conditions at the start and end times into the polynomial, we obtain the corresponding set of boundary constraint equations:
[0209]
[0210] In the formula, This represents the initial lateral displacement, corresponding to the position parameters in the initial boundary conditions; This represents the initial lateral velocity, corresponding to the velocity parameter in the initial boundary conditions. This represents the initial lateral acceleration, corresponding to the acceleration parameters in the initial boundary conditions. To terminate the lateral displacement, the position parameters in the corresponding termination boundary conditions are used; To terminate the lateral velocity, the velocity parameter in the termination boundary condition is used. To terminate the lateral acceleration, use the acceleration parameters in the termination boundary conditions.
[0211] The coefficient matrix of this system of linear equations is a non-singular matrix (when...). When solving polynomials, all six undetermined coefficients can be efficiently obtained through algebraic elimination, matrix inversion, and other methods. The solution process for cubic and quartic polynomials is similar, as they are all analytical solutions in closed-form, with low computational complexity, and can quickly solve for the coefficients of single-segment polynomials.
[0212] After symbolizing the known terms, coefficient terms, and constraint terms in the matrix, the system uses algebraic elimination, matrix inversion, or Cramer's rule to perform analytical solutions, thereby obtaining the specific values of each undetermined coefficient.
[0213] Subsequently, the obtained undetermined coefficients are substituted into a polynomial expression of a preset order to generate a spatial or temporal polynomial curve corresponding to the behavior segment, so that the polynomial satisfies the continuity constraint at the boundary conditions and is consistent with the behavior semantics.
[0214] The aforementioned polynomials can be used to describe trajectory changes within segments of behaviors such as lane changing, following other vehicles, turning, or parking. In practical applications, the analytical solution form can be adapted according to the polynomial order and the number of constraints, which is not limited in this application.
[0215] By transforming the boundary constraint equations into standard linear matrices and performing algebraic analysis, the undetermined coefficients of the polynomials can be obtained directly, eliminating the need for repeated searches and convergence judgments during numerical iteration, and ensuring that the generated results of the behavior segment polynomials correspond one-to-one with the boundary conditions.
[0216] Therefore, the generated polynomial can stably represent the motion changes within the behavior segment and provide a basic curve that satisfies the constraints for subsequent trajectory splicing, thereby improving the continuity and consistency of trajectory samples.
[0217] In another possible implementation, the boundary constraint equations are solved analytically to determine the polynomials corresponding to the behavior segments, including:
[0218] Based on the boundary constraint equations, determine the boundary constraint conditions corresponding to the behavior segment, arrange the boundary constraint conditions in order of the dimension of the corresponding polynomial order to generate a boundary constraint vector; project the boundary constraint vector onto the basis function coefficient space of the pre-constructed normalized time basis function derivative mapping matrix, and extract multiple coefficients of the polynomial through matrix-vector multiplication; determine the polynomial corresponding to the behavior segment based on each coefficient.
[0219] In this embodiment, the boundary constraint vector is used to organize the boundary constraint conditions at the start and end times into a unified vector form according to the dimensions corresponding to the polynomial order, which facilitates subsequent matrix-based solution.
[0220] The normalized time basis function derivative mapping matrix is used to characterize the mapping relationship between basis functions and their derivatives and coefficients in the normalized time domain. The basis function coefficient space is used to carry the coefficient expression after the mapping, so that the boundary constraints can be directly converted into the coefficients to be found.
[0221] The normalized time basis function derivative mapping matrix mentioned above can be pre-constructed as a fixed-dimensional matrix based on the polynomial order and the normalized time variable. In practical applications, this matrix can also adopt other equivalent expressions, which are not limited in this application.
[0222] In the specific implementation, the position information, velocity information and acceleration information of the start and end points of the behavior segment are first extracted based on the boundary constraint equation set, and then the constraint quantities are arranged in sequence to form the boundary constraint vector according to the constraint dimension corresponding to the selected polynomial order.
[0223] Subsequently, the vector is input into the pre-constructed normalized time basis function derivative mapping matrix, and an intermediate result located in the basis function coefficient space is obtained through matrix-vector multiplication. Then, multiple undetermined coefficients of the polynomial are directly read from the intermediate result.
[0224] Since the coefficients have already been mapped within the same coefficient space, there is no need for iterative fitting or numerical search; the polynomial expression corresponding to the action segment can be written accordingly.
[0225] In this process, the boundary constraint vector provides the input constraints between the boundary state and the polynomial coefficients, the normalized time basis function derivative mapping matrix completes the linear mapping from constraints to coefficients, the matrix-vector multiplication method realizes the analytical extraction of coefficients, and finally the polynomial corresponding to the behavior segment is determined by each undetermined coefficient.
[0226] With this approach, the solution process of the behavior segment polynomial is directly driven by the boundary conditions, which enables the coefficient solution results to maintain a one-to-one correspondence with the constraint conditions. This ensures that the obtained polynomial is consistent with the expected motion state of the behavior segment and reduces the computational overhead in segmented trajectory splicing.
[0227] S205. The polynomials corresponding to each behavior segment are spliced together according to the time sequence of driving behavior to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, and used as the trajectory sample of the driving scenario to be explored.
[0228] In this embodiment, the continuous spatiotemporal trajectory is used to represent the trajectory result that is continuous in both time and space, formed by splicing together polynomials of each row segment, and is the final output trajectory sample.
[0229] After obtaining the polynomials corresponding to each behavior segment, the system splices the polynomials in their respective time intervals according to the original time sequence of the driving behavior sequence, and calculates the trajectory points, velocity, acceleration and heading angle and other state quantities on a unified time axis, thereby forming a continuous spatiotemporal trajectory containing lateral coordinates, longitudinal coordinates and related kinematic parameters.
[0230] In one possible embodiment, during splicing, the local time of each behavior segment is first mapped to the global time, and then the endpoints are aligned according to the termination boundary and the starting boundary of adjacent behavior segments to ensure that the time connection relationship is clear; subsequently, the continuous spatiotemporal trajectory is expressed in a structured manner to generate trajectory sample data.
[0231] Trajectory samples can be output in time series format for direct use by subsequent analysis, training, or control modules. By combining the polynomials obtained from piecewise solutions in time sequence, the trajectory samples retain both the local motion characteristics of each behavior segment and form a complete and continuous trajectory that can be directly used for subsequent mining tasks.
[0232] It should be understood that the above examples are for illustrative purposes only and are not intended to be limiting.
[0233] In one possible implementation, before concatenating the polynomials corresponding to each behavior segment according to the time sequence of driving behaviors, the method further includes:
[0234] Extract the inter-segment continuity constraint rules corresponding to adjacent behavior segments from the mapping relationship table; verify the consistency of motion state parameters of adjacent behavior segments at the connection time according to the inter-segment continuity constraint rules; when the verification is inconsistent, determine that the degree of freedom of the current polynomial order cannot meet the requirements of the inter-segment continuity constraint rules, then increase the polynomial order of the corresponding behavior segment and re-solve to generate the corresponding polynomial, and obtain the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirements.
[0235] In this embodiment, adjacent behavior segments refer to two behavior segments that are immediately adjacent to each other in the time series. The inter-segment continuity constraint rule is used to limit the continuity conditions that the two behavior segments should meet at the connection time. The motion state parameters include at least position, velocity and acceleration, and the parameter consistency at the corresponding connection time is used to characterize the smoothness of the splicing of adjacent behavior segments.
[0236] Degrees of freedom are used to characterize the size of the coefficient space available to satisfy the constraints for the current polynomial order. When the number of constraints is greater than the number of independent coefficients that the current order can provide, the degrees of freedom are considered insufficient.
[0237] In actual implementation, in addition to recording the correspondence between behavior types and polynomial orders, the mapping table also contains allowed continuity constraints between adjacent behavior types. After the system completes the polynomial solution for each behavior segment, it reads the termination state of adjacent behavior segments and the start state of the next behavior segment in chronological order, and compares the two item by item according to the continuity constraint rules between segments.
[0238] When any constraint term in position, velocity, or acceleration fails to meet the consistency requirements, the system upgrades the order of the polynomial corresponding to the current behavior segment to a higher order, and reconstructs the boundary constraint equations under the new order to update the undetermined coefficients and solve them again until the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirements is obtained. The higher-order polynomial used can provide more undetermined coefficients, thereby accommodating higher-order continuity constraints. In practical applications, other values of this order can also be selected, and this application does not limit this.
[0239] During operation, the process first extracts the continuity constraints that match the adjacent behavior segments from the mapping table, and then uses the state parameters at the connection time as the verification object to complete the consistency check. If the check result shows that the current order is insufficient to satisfy all constraints at the same time, the number of adjustable coefficients is expanded by increasing the order and the solution is re-solved so that the adjacent behavior segments form a continuous trajectory expression at the time connection.
[0240] By adopting this implementation method, the motion state constraints at the splicing point of the behavior segment can be adaptively matched with the polynomial order, so that the generated trajectory maintains higher consistency at the connection time, reduces the state abrupt change caused by insufficient order, and reduces the number of times the boundary conditions are manually adjusted repeatedly.
[0241] Based on the above implementation method, optionally, the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement, including:
[0242] Obtain the total degrees of freedom required for the boundary constraints of the behavior segment and the corresponding inter-segment continuity constraints, as well as the available degrees of freedom corresponding to the current polynomial order; compare the total degrees of freedom with the available degrees of freedom, and when the total degrees of freedom exceed the available degrees of freedom, increase the polynomial order of the behavior segment, and reconstruct and solve the corresponding boundary constraint equations based on the increased polynomial order to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirements.
[0243] In this implementation, the total degrees of freedom are used to characterize the number of degrees of freedom occupied by the boundary constraints of the behavior segment itself and the continuity constraints between segments. The available degrees of freedom are used to characterize the coefficient solution space that the current polynomial order can provide.
[0244] The self-boundary constraints of a behavior segment can be composed of the position, velocity, and acceleration conditions at the start and end times, while the continuity constraints between segments can be composed of the consistency requirements of the position, velocity, acceleration, or jerk at the connection time of adjacent behavior segments.
[0245] The available degrees of freedom corresponding to the current polynomial order are usually equal to the number of undetermined coefficients of the polynomial of that order minus the number of degrees of freedom already occupied by boundary conditions.
[0246] In actual implementation, the system first extracts boundary constraint parameters based on the start and end times of the behavior segment and the desired motion state, and calculates the total degrees of freedom in combination with the connection consistency requirements of adjacent behavior segments. Then, it calculates the available degrees of freedom based on the currently allocated polynomial order.
[0247] When the total degrees of freedom are not greater than the available degrees of freedom, the current polynomial order is used directly to construct and solve the boundary constraint equations. When the total degrees of freedom exceed the available degrees of freedom, the polynomial order is increased to meet the degree of freedom coverage requirement, so that the newly added coefficients can support the joint solution of boundary constraints and inter-segment continuity constraints.
[0248] Subsequently, the time basis functions, boundary condition vectors, and constraint matrices are reorganized according to the improved order to form a new set of boundary constraint equations. The undetermined coefficients are obtained through analytical solution, thereby generating the polynomials corresponding to the behavior segments that meet the inter-segment consistency requirements.
[0249] In this process, the current polynomial order can be obtained by the initial matching of the current behavior segment, and the improved order can be determined by gradually increasing the degree of freedom gap. In practical applications, other order determination methods can also be selected, and this application does not limit this.
[0250] This implementation compares the constraint requirements with the order's achievable degrees of freedom, and automatically increases the order and resolves when insufficient. This ensures that the behavior segment polynomial simultaneously satisfies its own boundary conditions and the continuity requirements of adjacent segments, thereby maintaining higher consistency and smoothness in the transition between segments of the spliced trajectory. Since order adjustment and resolving are completed within the same constraint framework, the computational overhead caused by repeated trial calculations is reduced, and the stability and executability of trajectory generation results in complex driving scenarios are improved.
[0251] This application provides a method for generating vehicle trajectory samples. For the driving behavior sequence corresponding to the driving scenario to be mined, a pre-constructed mapping table between driving behavior and polynomial order is used to determine the corresponding polynomial order for each behavior segment in the sequence. The method then constructs a set of boundary constraint equations based on the expected motion states at the start and end times of each behavior segment and solves them analytically to obtain the polynomials corresponding to each behavior segment. Finally, the polynomials corresponding to each behavior segment are concatenated according to the time sequence of the driving behavior sequence to generate a continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, which serves as the trajectory sample. This scheme quickly determines the polynomial order of each behavior segment through table lookup matching, effectively improving the generation efficiency of trajectory samples under complex behavior sequences. Simultaneously, polynomial modeling under boundary constraints and temporal alignment concatenation ensure the continuity and smoothness of the trajectory in the spatiotemporal dimension, balancing generation efficiency with trajectory sample quality, and adapting to the trajectory sample generation needs of complex driving scenarios.
[0252] Figure 3 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 2 ,like Figure 3 As shown, this embodiment, based on the above embodiments, details the process of determining the polynomial order corresponding to each action segment, including:
[0253] S301. Decompose the behavior type of each behavior segment into horizontal movement type and vertical movement type.
[0254] For example, lateral motion types are used to characterize the movement intention of a behavior segment in the lateral direction of the vehicle, specifically corresponding to lane keeping, changing lanes to the left, changing lanes to the right, lateral offset correction, etc.; longitudinal motion types are used to characterize the movement intention of a behavior segment in the longitudinal direction of the vehicle, specifically corresponding to constant speed driving, acceleration, deceleration, following other vehicles, stopping, etc.
[0255] The decomposition of behavior types follows the principle of horizontal and vertical decoupling, breaking down the complete driving behavior into two independent dimensions of motion description. The two types of motion are independent of each other and correspond to different trajectory control requirements, providing a basis for subsequent dimensional order matching.
[0256] S302. Substitute the lateral motion type and the longitudinal motion type into the mapping table between driving behavior and polynomial order to match them, and obtain the polynomial order corresponding to the lateral motion type and the longitudinal motion type respectively.
[0257] For example, the mapping table between driving behavior and polynomial order is a pre-built data table that establishes a correspondence between different motion types and polynomial orders that meet boundary constraints and continuity requirements. The table can be stored using hash indexes or key-value pairs to quickly return matching results during queries.
[0258] In actual matching, when the lateral motion type corresponds to lane keeping, a low-order polynomial can be matched to characterize smooth lateral control; when it corresponds to lane changing, a higher-order polynomial can be matched to meet the lateral displacement, velocity and acceleration constraints.
[0259] For longitudinal motion types corresponding to constant speed or following other vehicles, a polynomial order that describes smooth speed changes can be matched; for stopping or significant deceleration, a polynomial order that can satisfy the termination speed constraint can be matched.
[0260] By matching the order separately along different dimensions, the lateral and longitudinal trajectory parameters can be made to correspond to the behavioral semantics, providing a clear order basis for the construction of the subsequent boundary constraint equation system. This makes the generated polynomial more in line with the specific driving intention and improves the continuity and trajectory solving efficiency when splicing between segments.
[0261] S303. Obtain the road curvature parameters, vehicle dynamics envelope parameters, and current working condition attachment parameters corresponding to the driving scenario to be explored.
[0262] For example, the road curvature parameter is used to characterize the curvature of the road centerline or lane reference line in the driving scenario to be explored, and can be calculated from high-precision maps, visual perception results, or fused positioning results.
[0263] Vehicle dynamics envelope parameters are used to characterize the feasible motion boundaries of a vehicle under constraints of current speed, center of gravity position, steering ability, and braking ability. They are composed of a vehicle parameter calibration table, yaw stability boundary, and longitudinal braking boundary.
[0264] The current working condition adhesion parameters are used to characterize the adhesion level between the tire and the road surface, and can be obtained by jointly estimating wheel speed, acceleration response, road surface identification results, and historical slip characteristics.
[0265] These three types of parameters together constitute the environmental and operating condition constraints of the current driving scenario, providing a quantitative basis for the dynamic adjustment of subsequent order rules.
[0266] S304. When the road curvature parameter is higher than the preset curvature threshold, and / or the current working condition adhesion parameter is less than the road surface adhesion envelope limit in the vehicle dynamics envelope parameter, it is determined that there is a high curvature road segment, and / or there is a low adhesion road surface feature. Adjust the order matching rule and inter-segment continuity constraint rule corresponding to the behavior type in the mapping relationship table.
[0267] For example, a preset curvature threshold is used to characterize the boundary for determining when a road transitions from a regular curve to a high curvature curve, and a road surface adhesion envelope limit is used to characterize the minimum allowable adhesion within the vehicle dynamics envelope.
[0268] When the road curvature parameter exceeds the preset curvature threshold, the current behavior segment is determined to be a high curvature road segment. At this time, the behavior type related to turning, merging, or avoiding will be mapped to a higher-order polynomial model, while improving the continuity constraints of adjacent behavior segments in terms of curvature, heading angle, and rate of change of curvature.
[0269] When the current working condition adhesion parameters are lower than the road surface adhesion envelope limit, it is determined that there are low adhesion road surface characteristics. At this time, the order matching rules corresponding to behaviors such as rapid acceleration and deceleration and sharp turning are tightened into smoother order combinations to reduce the dependence on high-order dynamic changes.
[0270] The mapping relationship table can be pre-stored as associated entries for behavior type, curvature interval, attachment interval and order selection rules. The entries also contain continuity constraints for adjacent behavior segments. After the scene determination is completed, the rules can be adjusted by locally replacing the entries or modifying the parameters.
[0271] In one possible implementation, adjusting the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table includes:
[0272] Based on the road curvature parameters, the corresponding curvature risk levels are classified; based on the vehicle dynamics envelope parameters, the upper limit of the rate of change of acceleration under the current working condition is determined; based on the curvature risk level and the upper limit of the rate of change of acceleration, the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping relationship table are adjusted.
[0273] In this embodiment, the curvature risk level is used to characterize the degree of influence of road curvature on trajectory generation constraints. Typically, the road curvature parameter can be associated with a preset curvature threshold, curvature change gradient, and curve duration to form a classification result of low risk, medium risk, and high risk. The upper limit of acceleration change rate is used to limit the allowable change boundary of longitudinal or lateral acceleration under the current working condition. Its value is determined by the vehicle dynamics envelope parameters, which include at least one or more of the following: tire adhesion limit, center of gravity position, axle load distribution, steering system response capability, and braking capability.
[0274] The mapping table is used to store the correspondence between behavior types and polynomial orders, as well as the correspondence between inter-segment continuity constraint rules and behavior types. In practical applications, it can be stored in the form of database tables, parameter dictionaries, or memory caches. This application does not limit this.
[0275] In the specific implementation, after receiving the road curvature parameters, the system first determines the curvature risk level based on the absolute value of the curvature parameters, the local fluctuation amplitude, and the change trend of the preceding and following sections.
[0276] When the road curvature corresponds to a high risk level, the order matching rule corresponding to the behavior type is adjusted to a higher order model in order to provide higher curve freedom for behaviors such as turning, lane changing and lane changing. At the same time, the inter-segment continuity constraint rule is expanded from only constraining position continuity to constraining position continuity, velocity continuity and acceleration continuity.
[0277] If the curvature risk level is medium risk, then the original order matching relationship remains unchanged or only the order of the local behavior segment is increased, and at least one of velocity continuity and acceleration continuity is imposed on the continuity between segments.
[0278] If the curvature risk level is low, a lower-order matching rule is used, and the continuity constraint is allowed to remain within the range of positional and velocity continuity.
[0279] When the vehicle dynamics envelope parameters indicate that the upper limit of the rate of change of acceleration under the current operating condition is low, the system will increase the target order in the order matching rule and impose stricter acceleration change restrictions on the inter-segment continuity constraint rule so that the acceleration transition between adjacent behavior segments does not exceed the upper limit; when the upper limit of the rate of change of acceleration is high, the original order matching rule can be retained and the restrictions on the rate of change of acceleration in the inter-segment continuity constraint can be appropriately relaxed.
[0280] Therefore, the system can modify the mapping table based on the curvature risk level and the upper limit of the rate of change of acceleration, so that the behavior type and the polynomial order and the inter-segment continuity constraints form a constraint relationship that matches the current road and vehicle state.
[0281] During operation, after determining the behavior type of the behavior segment, the trajectory generation module reads the adjusted mapping table and uses the corresponding order matching rules and continuity constraint rules to construct the boundary constraint equation set, so that the polynomial trajectory obtained by solving is consistent with the current curvature risk and vehicle dynamics capability.
[0282] Because the mapping table is dynamically modified, the order selection and continuity boundary when splicing behavior segments will change synchronously with the working conditions, thus enabling the generated trajectory to maintain more stable smoothness and semantic consistency under complex curves and constrained dynamic conditions.
[0283] With this implementation, the correspondence between behavior type and polynomial order can be adaptively adjusted according to road curvature risk and vehicle dynamics envelope. The inter-segment continuity constraint can also converge or relax synchronously with the upper limit of acceleration change rate, making the kinematic continuity of the spliced trajectory more in line with real-time working conditions, reducing the probability of trajectory mutation in high curvature or strong constraint scenarios, and reducing the solution overhead caused by repeatedly correcting the mapping relationship table.
[0284] Based on the above implementation method, optionally, the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table are adjusted according to the curvature risk level and the upper limit of the rate of change of acceleration, including:
[0285] Based on the curvature risk level and the upper limit of the rate of change of acceleration, determine the minimum polynomial order and the minimum continuity constraint order required for the corresponding behavior type; if the order matching rule is less than the minimum polynomial order, then increase the polynomial order of the corresponding behavior type; if the inter-segment continuity constraint rule is less than the minimum continuity constraint order, then increase the inter-segment continuity constraint order of the corresponding behavior type.
[0286] In this implementation, the lowest polynomial order is used to represent the minimum polynomial order that satisfies trajectory smoothness and boundary solvability under the current curvature risk level and acceleration change rate upper limit constraints; the lowest continuity constraint order is used to represent the minimum continuity level that satisfies the requirements for splicing adjacent behavior segments in the current scenario.
[0287] Curvature risk level can be divided into different risk ranges based on road curvature parameters, and the upper limit of acceleration change rate can be determined based on the allowable longitudinal and lateral acceleration change range in the vehicle dynamics envelope parameters. Both together characterize the constraint strength of the current behavior segment on the order of the trajectory curve and the continuity between segments.
[0288] In the specific implementation, after reading the mapping table, the system compares the default order matching rule and the default inter-segment continuity constraint rule corresponding to the current behavior type with the lowest polynomial order and the lowest continuity constraint order in the current scenario, respectively.
[0289] If the default order matching rule is lower than the required minimum polynomial order, then the polynomial order corresponding to the behavior type is increased to an order not less than the minimum polynomial order; if the default inter-segment continuity constraint rule is lower than the required minimum continuity constraint order, then the continuity constraint level corresponding to the behavior type is increased to a level not less than the minimum continuity constraint order.
[0290] The above adjustments can be applied to the behavior type entries corresponding to lateral motion type, longitudinal motion type, or a combination of both, and simultaneously update the corresponding order and constraint fields in the mapping table, so that the construction of subsequent boundary constraint equation sets can directly adopt the updated order and continuity requirements.
[0291] In practical applications, if a behavior segment corresponds to a high-curvature road segment and has a low upper limit for the rate of change of acceleration, the system will increase the order of the polynomial for the corresponding behavior type to a higher order to provide more degrees of freedom to fit the curvature change during analytical solving. Simultaneously, the order of the inter-segment continuity constraint will be increased to ensure continuous connection between adjacent behavior segments in terms of position, velocity, acceleration, or higher-order derivatives. This approach allows the behavior type parameters in the mapping table to adaptively adjust according to scenario risks and directly constrains the smooth transition of curves when outputting continuous spatiotemporal trajectories.
[0292] With this implementation method, the mapping table no longer uses a fixed uniform order and uniform continuity level. Instead, it can dynamically increase the constraint strength of behavior type based on the curvature risk level and the upper limit of the acceleration change rate. This allows the trajectory sample to maintain more stable continuity and smoothness in high curvature and strong dynamic change scenarios, and reduces the occurrence of insufficient order or insufficient constraints when stitching trajectories in the future.
[0293] S305. Based on the adjusted order matching rules and behavior type, the adjusted polynomial order is obtained.
[0294] For example, after the adjustment is completed, the final polynomial order corresponding to the two dimensions is output according to the adjusted order matching rule, based on the horizontal and vertical motion types of the current behavior segment. This result can be directly used to construct the subsequent boundary constraint equation system.
[0295] Through this dynamic adjustment method, the polynomial order corresponding to the behavior segment is no longer fixed with a uniform configuration, but is adjusted synchronously with road geometry, vehicle dynamics constraints and road surface adhesion status. This ensures that the curve model of each behavior segment is consistent with the road geometric features, vehicle feasible motion boundaries and actual adhesion conditions, thereby improving the consistency between the trajectory and driving intention and reducing curvature abrupt changes and dynamic infeasibility problems that occur in subsequent splicing.
[0296] The vehicle trajectory sample generation method provided in this application decomposes driving behavior into lateral and longitudinal motion types, enabling independent decoupling of different motion dimensions and providing a clear classification basis for dimensional order matching. By matching the polynomial order in each dimension through a pre-constructed mapping table, the order can accurately correspond to the driving intent in each dimension, solidifying the foundation for trajectory solution efficiency and inter-segment continuity. Obtaining road curvature, vehicle dynamics envelope, and operating condition attachment parameters provides a quantitative scenario and operating condition basis for dynamic adjustment of order rules. Adjusting the order matching rules and inter-segment continuity constraints of the mapping table based on scenario features allows the order configuration to adapt to the current road and vehicle boundary conditions. Finally, the output of a polynomial order adapted to the operating conditions improves the correlation between trajectory and driving intent. Figure 1 Consistency, reducing curvature abrupt changes and dynamic infeasibility issues at splicing points.
[0297] Figure 4 A flowchart illustrating the vehicle trajectory sample generation method provided in this application embodiment. Figure 3 ,like Figure 4 As shown, this embodiment, based on the above embodiments, provides a detailed explanation of the specific processing procedure for obtaining the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, including:
[0298] S401. Obtain the time interval corresponding to each behavior segment.
[0299] For example, the time interval is used to characterize the effective duration of each behavior segment in the driving behavior sequence, serving as the time domain boundary basis when splicing adjacent polynomials. The time interval of each behavior segment corresponds to its own start and end times, which can be uniformly mapped onto the global time axis to accurately locate the connection time of adjacent behavior segments.
[0300] S402. Based on the time sequence of driving behavior, sequentially connect the polynomials corresponding to adjacent behavior segments, and align the time nodes and motion state parameters of the connection moment according to the time interval corresponding to each behavior segment to obtain the polynomial sequence after connection.
[0301] For example, during the connection process, time nodes are used to identify the connection time of adjacent behavior segments. Motion state parameters cover the motion information such as position, velocity, and acceleration corresponding to that time. The termination time node of the previous behavior segment is precisely aligned with the start time node of the next behavior segment. At the same time, the motion state parameter values of the two polynomial segments are consistent at the connection time. After alignment, the polynomial segments are arranged in chronological order to form a continuous polynomial sequence, which serves as the intermediate model basis for generating continuous spatiotemporal trajectories.
[0302] S403. Based on the preset time sampling step size, perform numerical calculations on each polynomial in the polynomial sequence within the corresponding time interval to obtain the motion state parameters corresponding to each sampling moment.
[0303] For example, the preset time sampling step is used to limit the time interval for discretization. It can be flexibly set according to the trajectory output accuracy and real-time calculation requirements. Discrete sampling points are sequentially taken in the time interval corresponding to each polynomial according to the sampling step. The position, velocity, acceleration and other motion state parameters corresponding to each sampling moment are calculated by substituting into the polynomial expression, and the values of each state component are recorded according to the time index.
[0304] S404. The motion state parameters corresponding to each sampling moment are spliced and integrated in chronological order to form a continuous temporal motion dataset. The temporal motion dataset is then structured and encapsulated to generate continuous spatiotemporal trajectory data corresponding to the driving scenario to be mined.
[0305] For example, the motion state parameters of all sampling points are arranged sequentially along the time axis to form a continuous temporal motion dataset covering the entire driving scenario. The dataset is then standardized and structured, and organized into data formats that can be directly called, such as trajectory arrays and spatiotemporal sequence tables. The resulting continuous spatiotemporal trajectory data can be directly used as trajectory samples to adapt to subsequent application scenarios such as scene mining, sample labeling, and model training.
[0306] The vehicle trajectory sample generation method provided in this application provides a temporal boundary basis for splicing adjacent polynomials by obtaining the time intervals corresponding to each behavior segment, accurately locating the connection time between segments. By splicing polynomials in chronological order and aligning the time nodes and motion state parameters of the connection time, the continuity of motion states between segments can be ensured, forming an ordered polynomial sequence. Numerical calculations are performed on the polynomial sequence according to a preset time sampling step size to obtain the discrete motion state parameters corresponding to each sampling time. Integrating and structurally encapsulating the discrete parameter time series generates standardized continuous spatiotemporal trajectory data, suitable for subsequent scene mining and sample applications.
[0307] Figure 5 This is a schematic diagram of the vehicle trajectory sample generation device provided in an embodiment of this application. The device in this embodiment can be in the form of software and / or hardware. Figure 5 As shown, the vehicle trajectory sample generation device 500 provided in this application embodiment includes: an acquisition module 501 and a processing module 502.
[0308] The acquisition module 501 is used to acquire the driving behavior sequence corresponding to the driving scenario to be explored. The driving behavior sequence contains multiple behavior segments arranged in chronological order.
[0309] Processing module 502 is used to extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment based on the driving behavior sequence.
[0310] The processing module 502 is also used to determine the polynomial order corresponding to each behavior segment based on the behavior type of each behavior segment and the pre-built mapping table between driving behavior and polynomial order.
[0311] The processing module 502 is also used to construct a set of boundary constraint equations corresponding to each behavior segment based on the polynomial order and the desired motion state, and to solve the set of boundary constraint equations analytically to determine the polynomial corresponding to the behavior segment.
[0312] The processing module 502 is also used to splice the polynomials corresponding to each behavior segment according to the time sequence of the driving behavior sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be mined, and use it as the trajectory sample of the driving scenario to be mined.
[0313] In one possible implementation, the acquisition module 501 is further configured to:
[0314] Obtain the original driving semantic data corresponding to the driving scenario to be explored;
[0315] The original driving semantic data is temporally segmented and semantically labeled to obtain multiple behavior segments arranged in chronological order.
[0316] Each behavior segment is matched with corresponding motion attributes and duration parameters to form a driving behavior sequence.
[0317] In one possible implementation, the processing module 502 is also used to construct a mapping table between driving behavior and polynomial order:
[0318] Obtain the kinematic characteristic parameters of various driving behaviors and the preset vehicle ride comfort control index, analyze the corresponding boundary constraint requirements and continuity smoothness requirements, and transform the continuity smoothness requirements into inter-segment continuity constraint rules.
[0319] Based on the boundary constraint requirements and the requirement for continuous smoothness, a polynomial model of the corresponding order is matched, and the relationship between various driving behaviors, polynomial models and inter-segment continuity constraint rules is stored in a structured manner to obtain a mapping table between driving behaviors and polynomial orders.
[0320] In one possible implementation, the processing module 502 is further configured to:
[0321] The behavior type of each behavior segment is broken down into lateral movement type and longitudinal movement type;
[0322] The lateral motion type and longitudinal motion type are respectively substituted into the mapping table between driving behavior and polynomial order for matching, and the polynomial order corresponding to the lateral motion type and longitudinal motion type is obtained respectively.
[0323] In one possible implementation, the processing module 502 is further configured to:
[0324] Obtain the road curvature parameters, vehicle dynamics envelope parameters, and current working condition attachment parameters corresponding to the driving scenario to be explored;
[0325] Based on the road curvature parameters, vehicle dynamics envelope parameters, and current operating condition attachment parameters, the polynomial order corresponding to each behavior segment is adjusted to obtain the adjusted polynomial order.
[0326] In one possible implementation, the processing module 502 is further configured to:
[0327] When the road curvature parameter is higher than the preset curvature threshold, and / or the current working condition adhesion parameter is less than the road surface adhesion envelope limit in the vehicle dynamics envelope parameter, it is determined that there is a high curvature road segment, and / or there is a low adhesion road surface feature. The order matching rules and inter-segment continuity constraint rules corresponding to the behavior type in the mapping relationship table are adjusted.
[0328] Based on the adjusted order matching rules and behavior type, the adjusted polynomial order is obtained.
[0329] In one possible implementation, the processing module 502 is further configured to:
[0330] Based on the road curvature parameters, the corresponding curvature risk levels are classified;
[0331] Determine the upper limit of the rate of change of acceleration under the current operating condition based on the vehicle dynamics envelope parameters;
[0332] Based on the curvature risk level and the upper limit of the rate of change of acceleration, adjust the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table.
[0333] In one possible implementation, the processing module 502 is further configured to:
[0334] Based on the curvature risk level and the upper limit of the rate of change of acceleration, determine the minimum polynomial order and the minimum continuity constraint order required for the corresponding behavior type;
[0335] If the order matching rule is less than the lowest polynomial order, then the polynomial order of the corresponding behavior type is increased.
[0336] If the inter-segment continuity constraint rule is less than the minimum continuity constraint order, then the inter-segment continuity constraint order of the corresponding behavior type is increased.
[0337] In one possible implementation, the processing module 502 is further configured to:
[0338] Extract the position, velocity, and acceleration parameters at the initial moment of the desired motion state as the initial boundary conditions;
[0339] Extract the position, velocity, and acceleration parameters at the termination moment in the desired motion state as termination boundary conditions;
[0340] Based on the polynomial order, initial boundary conditions, and final boundary conditions, construct the boundary constraint equation system corresponding to the behavior segment.
[0341] In one possible implementation, the processing module 502 is further configured to:
[0342] The boundary constraint equations are transformed into a standard linear matrix, and the standard linear matrix is then analyzed algebraically to obtain the multiple undetermined coefficients of the polynomial.
[0343] Based on the multiple undetermined coefficients of the polynomial, generate the polynomial corresponding to the behavior segment.
[0344] In one possible implementation, the processing module 502 is further configured to:
[0345] Based on the boundary constraint equations, determine the boundary constraint conditions corresponding to the behavior segment, arrange the boundary constraint conditions in order according to the dimension of the corresponding polynomial order, and generate the boundary constraint vector.
[0346] The boundary constraint vector is projected onto the basis function coefficient space of the pre-constructed normalized time basis function derivative mapping matrix, and multiple coefficients of the polynomial are extracted through matrix-vector multiplication.
[0347] The polynomial corresponding to each behavior segment is determined based on each coefficient.
[0348] In one possible implementation, the processing module 502 is further configured to:
[0349] Extract the inter-segment continuity constraint rules corresponding to adjacent behavior segments from the mapping relationship table;
[0350] Based on the inter-segment continuity constraint rules, verify the consistency of motion state parameters of adjacent behavior segments at the connection time;
[0351] When the verification is inconsistent, if the degree of freedom of the current polynomial order cannot meet the requirements of the inter-segment continuity constraint rule, then the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirement.
[0352] In one possible implementation, the processing module 502 is further configured to:
[0353] Obtain the total degrees of freedom required to acquire the self-boundary constraints of the behavior segment and the corresponding inter-segment continuity constraints, as well as the available degrees of freedom corresponding to the current polynomial order;
[0354] The total degrees of freedom are compared with the available degrees of freedom. When the total degrees of freedom exceed the available degrees of freedom, the polynomial order of the behavior segment is increased. The corresponding boundary constraint equations are reconstructed and solved based on the increased polynomial order to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement.
[0355] In one possible implementation, the processing module 502 is further configured to:
[0356] Obtain the time interval corresponding to each behavior segment;
[0357] Based on the temporal order of the driving behavior sequence, the polynomials corresponding to adjacent behavior segments are sequentially connected, and the time nodes and motion state parameters at the connection moments are aligned according to the time intervals corresponding to each behavior segment, to obtain the polynomial sequence after connection.
[0358] By integrating the polynomial sequences, the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored is obtained.
[0359] In one possible implementation, the processing module 502 is further configured to:
[0360] Based on the preset time sampling step, numerical calculations are performed on each polynomial in the polynomial sequence within the corresponding time interval to obtain the motion state parameters corresponding to each sampling moment.
[0361] The motion state parameters corresponding to each sampling moment are spliced and integrated in chronological order to form a continuous temporal motion dataset. The temporal motion dataset is then structured and encapsulated to generate continuous spatiotemporal trajectory data corresponding to the driving scenario to be mined.
[0362] The vehicle trajectory sample generation device 500 provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.
[0363] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 600 provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the electronic device 600 further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus.
[0364] In a specific implementation, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to perform the above-described method.
[0365] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.
[0366] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.
[0367] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.
[0368] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.
[0369] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0370] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.
[0371] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.
[0372] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.
[0373] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.
[0374] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0375] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0376] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0377] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0378] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.
Claims
1. A method for generating vehicle trajectory samples, characterized in that, include: Obtain the driving behavior sequence corresponding to the driving scenario to be explored, wherein the driving behavior sequence contains multiple behavior segments arranged in chronological order; Based on the driving behavior sequence, extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment; Based on the behavior type of each behavior segment and the pre-built mapping table between driving behavior and polynomial order, determine the polynomial order corresponding to each behavior segment; For each behavior segment, a set of boundary constraint equations corresponding to the behavior segment is constructed based on the order of the polynomial and the desired motion state. The set of boundary constraint equations is solved analytically to determine the polynomial corresponding to the behavior segment. The polynomials corresponding to each behavior segment are concatenated according to the time sequence of the driving behavior to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, and this trajectory is used as the trajectory sample of the driving scenario to be explored.
2. The method according to claim 1, characterized in that, The step of obtaining the driving behavior sequence corresponding to the driving scenario to be explored includes: Obtain the original driving semantic data corresponding to the driving scenario to be explored; The original driving semantic data is subjected to temporal segmentation and semantic annotation to obtain multiple behavior segments arranged in chronological order; Each behavior segment is matched with a corresponding motion attribute and duration parameter to form the driving behavior sequence.
3. The method according to claim 1, characterized in that, The mapping table between driving behavior and polynomial order is constructed in the following way: Obtain kinematic characteristic parameters of various driving behaviors and preset vehicle ride comfort control indicators, analyze the corresponding boundary constraint requirements and continuous smoothness requirements, and transform the continuous smoothness requirements into inter-segment continuous constraint rules. Based on the boundary constraint requirements and the continuity and smoothness requirements, a polynomial model of the corresponding order is matched, and the relationship between the various driving behaviors, the polynomial model, and the inter-segment continuity constraint rules is structured and stored to obtain a mapping table between the driving behaviors and the polynomial order.
4. The method according to claim 1, characterized in that, Determining the polynomial order corresponding to each of the action segments includes: The behavior type of each behavior segment is broken down into lateral movement type and longitudinal movement type; The lateral motion type and the longitudinal motion type are respectively substituted into the mapping relationship table between driving behavior and polynomial order for matching, so as to obtain the polynomial order corresponding to the lateral motion type and the longitudinal motion type respectively.
5. The method according to claim 4, characterized in that, Also includes: Obtain the road curvature parameters, vehicle dynamics envelope parameters, and current working condition attachment parameters corresponding to the driving scenario to be explored; Based on the road curvature parameters, the vehicle dynamics envelope parameters, and the current operating condition attachment parameters, the polynomial order corresponding to each behavior segment is adjusted to obtain the adjusted polynomial order.
6. The method according to claim 5, characterized in that, The step of adjusting the polynomial order corresponding to each behavior segment based on the road curvature parameter, the vehicle dynamics envelope parameter, and the current operating condition attachment parameter to obtain the adjusted polynomial order includes: When the road curvature parameter is higher than the preset curvature threshold, and / or the current working condition adhesion parameter is less than the road surface adhesion envelope limit in the vehicle dynamics envelope parameter, it is determined that there is a high curvature road segment, and / or there is a low adhesion road surface feature, and the order matching rule and inter-segment continuity constraint rule corresponding to the behavior type in the mapping relationship table are adjusted. The adjusted polynomial order is obtained based on the adjusted order matching rules and the behavior type.
7. The method according to claim 6, characterized in that, The adjustment of the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table includes: Based on the road curvature parameters, the corresponding curvature risk levels are classified; The upper limit of the rate of change of acceleration under the current operating condition is determined based on the vehicle dynamics envelope parameters. Based on the curvature risk level and the upper limit of the rate of change of acceleration, adjust the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table.
8. The method according to claim 7, characterized in that, The step of adjusting the order matching rules and inter-segment continuity constraint rules corresponding to the behavior types in the mapping table based on the curvature risk level and the upper limit of the rate of change of acceleration includes: Based on the curvature risk level and the upper limit of the rate of change of acceleration, determine the minimum polynomial order and the minimum continuity constraint order required for the corresponding behavior type; When the order matching rule is less than the lowest polynomial order, the polynomial order of the corresponding behavior type is increased. If the inter-segment continuity constraint rule is less than the minimum continuity constraint order, then the inter-segment continuity constraint order of the corresponding behavior type is increased.
9. The method according to claim 1, characterized in that, The step of constructing the boundary constraint equation set corresponding to the behavior segment based on the polynomial order and the desired motion state includes: Extract the position, velocity, and acceleration parameters at the initial moment of the desired motion state as the initial boundary conditions; Extract the position, velocity, and acceleration parameters at the termination moment of the desired motion state as termination boundary conditions; Based on the polynomial order, the initial boundary conditions, and the final boundary conditions, construct the boundary constraint equation set corresponding to the behavior segment.
10. The method according to claim 1, characterized in that, The step of analytically solving the boundary constraint equations to determine the polynomial corresponding to the behavior segment includes: The boundary constraint equations are transformed into standard linear matrices, and algebraic analysis is performed on the standard linear matrices to obtain multiple undetermined coefficients of the polynomials. Based on the multiple undetermined coefficients of the polynomial, the polynomial corresponding to the behavior segment is generated.
11. The method according to claim 1, characterized in that, The step of analytically solving the boundary constraint equations to determine the polynomial corresponding to the behavior segment includes: Based on the set of boundary constraint equations, determine the boundary constraint conditions corresponding to the behavior segment, and arrange the boundary constraint conditions sequentially according to the dimensions of the corresponding polynomial order to generate a boundary constraint vector. The boundary constraint vector is projected onto the basis function coefficient space of the pre-constructed normalized time basis function derivative mapping matrix, and multiple coefficients of the polynomial are extracted by matrix-vector multiplication. The polynomial corresponding to the behavior segment is determined based on each of the coefficients.
12. The method according to claim 1, characterized in that, Also includes: Extract the inter-segment continuity constraint rules corresponding to adjacent behavior segments from the mapping table; Based on the inter-segment continuity constraint rules, verify the consistency of motion state parameters of adjacent behavior segments at the connection time; When the verification is inconsistent, it is determined that the degree of freedom of the current polynomial order cannot meet the requirements of the inter-segment continuity constraint rule. Then, the polynomial order of the corresponding behavior segment is increased and the corresponding polynomial is re-solved to obtain the polynomial corresponding to the behavior segment that meets the inter-segment consistency requirement.
13. The method according to claim 12, characterized in that, The process of increasing the order of the polynomial corresponding to the behavior segment and resolving to generate the corresponding polynomial to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement includes: Obtain the total degrees of freedom required for the self-boundary constraints and corresponding inter-segment continuity constraints of the behavior segment, as well as the available degrees of freedom corresponding to the current polynomial order; The total degrees of freedom are compared with the available degrees of freedom. When the total degrees of freedom exceed the available degrees of freedom, the polynomial order of the behavior segment is increased, and the corresponding boundary constraint equations are reconstructed and solved based on the increased polynomial order to obtain the polynomial corresponding to the behavior segment that satisfies the inter-segment consistency requirement.
14. The method according to any one of claims 1-13, characterized in that, The step of concatenating the polynomials corresponding to each behavior segment according to the time sequence of the driving behavior sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored includes: Obtain the time interval corresponding to each of the aforementioned behavior segments; Based on the temporal order of the driving behavior sequence, the polynomials corresponding to adjacent behavior segments are sequentially connected, and the time nodes and motion state parameters of the connection moment are aligned according to the time interval corresponding to each behavior segment to obtain the polynomial sequence after connection. By integrating the polynomial sequence, the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored is obtained.
15. The method according to claim 14, characterized in that, The step of integrating the polynomial sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored includes: According to the preset time sampling step, each polynomial in the polynomial sequence is numerically calculated in the corresponding time interval to obtain the motion state parameter corresponding to each sampling moment; The motion state parameters corresponding to each sampling moment are spliced and integrated in the time order to form a continuous temporal motion dataset. The temporal motion dataset is then structured and encapsulated to generate continuous spatiotemporal trajectory data corresponding to the driving scenario to be explored.
16. A vehicle trajectory sample generation device, characterized in that, include: The acquisition module is used to acquire the driving behavior sequence corresponding to the driving scenario to be explored, wherein the driving behavior sequence contains multiple behavior segments arranged in chronological order; The processing module is used to extract the behavior type of each behavior segment and the expected motion state at the start and end times of each behavior segment based on the driving behavior sequence. The processing module is further configured to determine the polynomial order corresponding to each behavior segment based on the behavior type of each behavior segment and a pre-built mapping table between driving behavior and polynomial order. The processing module is further configured to, for each behavior segment, construct a set of boundary constraint equations corresponding to the behavior segment based on the order of the polynomial and the desired motion state, perform analytical solution on the set of boundary constraint equations, and determine the polynomial corresponding to the behavior segment. The processing module is further configured to concatenate the polynomials corresponding to each behavior segment according to the time sequence of the driving behavior sequence to obtain the continuous spatiotemporal trajectory corresponding to the driving scenario to be explored, and use it as the trajectory sample of the driving scenario to be explored.
17. An electronic device, characterized in that, include: Memory, processor; The memory stores instructions that the computer executes; The processor executes computer execution instructions stored in memory, causing the processor to perform the method as described in any one of claims 1-15.
18. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-15.
