Vehicle motion control method, electronic device, and program product

Abstract

The application relates to the field of intelligent driving, and in particular to a vehicle motion control method, an electronic device and a program product. The method comprises the following steps: obtaining a plurality of candidate trajectories, operation parameters and navigation data of a vehicle; checking the candidate trajectories to obtain a plurality of effective trajectories; performing intention clustering on the effective trajectories to obtain a plurality of candidate intention sets, each candidate intention set corresponding to an intention label; determining the values of the candidate intention sets on a plurality of factors according to the intention label, the operation parameters, the navigation data and the effective trajectories in each candidate intention set; performing weighted summation on the values according to preset weights corresponding to the factors to obtain calculation values of the candidate intention sets; determining a target set as the candidate intention set with the highest calculation value; performing smoothness screening on the effective trajectories in the target set to determine a target trajectory; and outputting the target trajectory to a vehicle control module to control the motion of the vehicle. The method is used to improve the accuracy of target trajectory determination.

Description

Technical Field

[0001] This application relates to the field of intelligent driving, and in particular to a vehicle motion control method, electronic device, and software product. Background Technology

[0002] With the development of intelligent driving technology, vehicle motion control needs to generate multiple candidate trajectories in real time based on multi-source information in complex and dynamic traffic environments, and then determine the target trajectory for vehicle control. This process belongs to the field of intelligent driving motion planning and control technology, and has high requirements for real-time performance, safety, and interpretability.

[0003] Existing technologies employ methods such as unified neural network scoring, target-guided selection, or pre-trained semantic model judgment to process candidate trajectories. These approaches concentrate trajectory selection, intent judgment, and comfort optimization into the same model or processing step, with evaluation dimensions being mutually coupled and lacking an independent intent judgment step.

[0004] The above scheme does not classify and distinguish candidate trajectories at the intent level before directly evaluating or ranking them uniformly. This makes it easy for trajectories with different intents to be mixed and compared in the same evaluation system, affecting the rationality and traceability of the target trajectory determination, and making it difficult to meet the requirements of vehicle motion control for real-time performance and determinism. Summary of the Invention

[0005] This application provides a vehicle motion control method, electronic device, and program product to solve the problem that the accuracy of target trajectory determination is insufficient due to the coupling between intent judgment and trajectory selection and the mixed evaluation of candidate trajectories with different driving intentions.

[0006] In a first aspect, embodiments of this application provide a vehicle motion control method, including:

[0007] Acquire multiple candidate trajectories, operating parameters, and navigation data for the vehicle;

[0008] The candidate trajectories are validated to obtain multiple valid trajectories;

[0009] The valid trajectories are clustered to obtain multiple candidate intent sets, and each candidate intent set corresponds to an intent label.

[0010] Based on intent labels, operation parameters, navigation data, and valid trajectories in each candidate intent set, determine the values ​​of each candidate intent set on multiple factors;

[0011] The values ​​are weighted and summed according to the preset weights corresponding to each factor to obtain the calculated value of each candidate intent set;

[0012] The set of candidate intents with the highest calculated values ​​is determined as the target set;

[0013] The smoothness of valid trajectories in the target set is used to determine the target trajectory;

[0014] The target trajectory is output to the vehicle control module to control the vehicle's movement.

[0015] In one possible embodiment, the candidate trajectory is validated, including:

[0016] The candidate trajectories are subjected to security checks to obtain multiple safe trajectories;

[0017] The safety trajectory was verified for compliance, resulting in multiple valid trajectories.

[0018] In one possible embodiment, security verification of the candidate trajectory includes:

[0019] Obtain the predicted motion trajectories of dynamic obstacles around the vehicle;

[0020] Collision verification is performed between the candidate trajectory and the motion prediction trajectory to obtain multiple safe trajectories.

[0021] In one possible embodiment, collision verification between the candidate trajectory and the predicted motion trajectory includes:

[0022] Obtain the collision time between the candidate trajectory and the predicted motion trajectory;

[0023] Candidate trajectories with collision times greater than a preset time threshold are extracted to obtain multiple safe trajectories.

[0024] In one possible embodiment, compliance verification of the safety trajectory includes:

[0025] Get the lane markings of the lane where the vehicle is located and the status of the traffic lights in front of the vehicle;

[0026] The safety trajectory is verified based on lane markings and traffic light status, resulting in multiple valid trajectories.

[0027] In one possible embodiment, the safety trajectory is verified based on lane markings and traffic light status, including:

[0028] Determine whether the safe trajectory involves crossing solid lane lines or violating traffic light status;

[0029] If a valid track exists, delete the corresponding safe track to obtain multiple valid tracks.

[0030] In one possible embodiment, determining the values ​​of each candidate intent set on multiple factors includes:

[0031] Driver operation signals and navigation expectations are obtained based on operating parameters and navigation data;

[0032] The intent labels are matched with the driver's operation signals and navigation expectations respectively to obtain the matching results;

[0033] The values ​​of the driver intent factor and navigation factor are obtained by querying the preset mapping table based on the matching results.

[0034] In one possible embodiment, it also includes:

[0035] The total deceleration of surrounding vehicles during the execution of the effective trajectory is predicted to obtain a predicted value, and the negative value of the predicted value is taken as the value of the interaction factor.

[0036] Calculate the ratio of the average speed of the effective trajectory to the road speed limit, and use this ratio as the value of the efficiency factor.

[0037] In one possible embodiment, intention clustering of valid trajectories includes:

[0038] Extract the feature vector of the effective trajectory;

[0039] The feature vectors are matched with preset intent labels to obtain the intent labels corresponding to each valid trajectory.

[0040] Clustering of valid trajectories based on intent labels yields multiple candidate intent sets.

[0041] In one possible embodiment, extracting the feature vector of the effective trajectory further includes:

[0042] Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs;

[0043] Add the trajectory curvature distribution features and associated features to the feature vector.

[0044] In one possible embodiment, matching the feature vector with a preset intent label further includes:

[0045] If the associated feature indicates a trajectory that conflicts with a traffic sign, the intent label of the corresponding valid trajectory will be determined as an invalid intent.

[0046] In one possible embodiment, smoothness screening of valid trajectories in the target set includes:

[0047] Calculate the smoothing value and style matching value of each valid trajectory in the target set;

[0048] The smoothness value is obtained by weighted summation of the smoothness value and style matching value based on the preset harmonic coefficient;

[0049] The effective trajectory with the smallest smoothness value is determined as the target trajectory.

[0050] In one possible embodiment, calculating the smoothing value of each valid trajectory in the target set includes:

[0051] Obtain the longitudinal and lateral jerk of each valid trajectory;

[0052] Discretely sum the squares of the longitudinal and lateral jerks to obtain a smooth value.

[0053] In one possible embodiment, discrete summation includes:

[0054] Obtain the spatiotemporal weighting factor, which includes the weight of the starting segment and the weight of the middle and later segments;

[0055] The trajectory points in the effective trajectory are divided into the starting segment and the middle and rear segments according to a preset ratio;

[0056] The smoothed value is obtained by weighting the sum of the squares of the starting segment and the middle and later segments according to the weights of the starting segment and the middle and later segments.

[0057] In one possible embodiment, calculating the style matching value includes:

[0058] Obtain the average acceleration of the preset driving style and each effective trajectory;

[0059] Match the preset driving style with the preset style reference acceleration mapping table to obtain the style reference acceleration;

[0060] The style matching value is obtained by calculating the square of the difference between the average acceleration and the style reference acceleration.

[0061] In one possible embodiment, it also includes:

[0062] Obtain current scene information, including traffic light status and remaining distance of the navigation path;

[0063] Adjust the preset weights of each factor based on the current scenario information.

[0064] Secondly, embodiments of this application provide a vehicle motion control device, comprising:

[0065] The acquisition module is used to acquire multiple candidate trajectories, operating parameters, and navigation data of the vehicle.

[0066] The verification module is used to verify the candidate trajectories and obtain multiple valid trajectories;

[0067] The clustering module is used to cluster the valid trajectories to obtain multiple candidate intent sets, each of which corresponds to an intent label;

[0068] The analysis module is used to determine the values ​​of each candidate intent set on multiple factors based on the intent tags, operation parameters, navigation data, and valid trajectories in each candidate intent set;

[0069] The analysis module is also used to perform a weighted summation of the values ​​according to the preset weights corresponding to each factor to obtain the calculated value of each candidate intent set;

[0070] The determination module is used to identify the set of candidate intents with the highest calculated values ​​as the target set;

[0071] The filtering module is used to perform smoothness filtering on the valid trajectories in the target set to determine the target trajectory;

[0072] The control module is used to output the target trajectory to the vehicle control module to control the vehicle's movement.

[0073] In one possible embodiment, the verification module is specifically used for:

[0074] The candidate trajectories are subjected to security checks to obtain multiple safe trajectories;

[0075] The safety trajectory was verified for compliance, resulting in multiple valid trajectories.

[0076] In one possible embodiment, the verification module is further specifically used for:

[0077] Obtain the predicted motion trajectories of dynamic obstacles around the vehicle;

[0078] Collision verification is performed between the candidate trajectory and the motion prediction trajectory to obtain multiple safe trajectories.

[0079] In one possible embodiment, the verification module is further specifically used for:

[0080] Obtain the collision time between the candidate trajectory and the predicted motion trajectory;

[0081] Candidate trajectories with collision times greater than a preset time threshold are extracted to obtain multiple safe trajectories.

[0082] In one possible embodiment, the verification module is further specifically used for:

[0083] Get the lane markings of the lane where the vehicle is located and the status of the traffic lights in front of the vehicle;

[0084] The safety trajectory is verified based on lane markings and traffic light status, resulting in multiple valid trajectories.

[0085] In one possible embodiment, the verification module is further specifically used for:

[0086] Determine whether the safe trajectory involves crossing solid lane lines or violating traffic light status;

[0087] If a valid track exists, delete the corresponding safe track to obtain multiple valid tracks.

[0088] In one possible embodiment, the analysis module is specifically used for:

[0089] Driver operation signals and navigation expectations are obtained based on operating parameters and navigation data;

[0090] The intent labels are matched with the driver's operation signals and navigation expectations respectively to obtain the matching results;

[0091] The values ​​of the driver intent factor and navigation factor are obtained by querying the preset mapping table based on the matching results.

[0092] In one possible embodiment, the analysis module is further specifically used for:

[0093] The total deceleration of surrounding vehicles during the execution of the effective trajectory is predicted to obtain a predicted value, and the negative value of the predicted value is taken as the value of the interaction factor.

[0094] Calculate the ratio of the average speed of the effective trajectory to the road speed limit, and use this ratio as the value of the efficiency factor.

[0095] In one possible embodiment, the clustering module is specifically used for:

[0096] Extract the feature vector of the effective trajectory;

[0097] The feature vectors are matched with preset intent labels to obtain the intent labels corresponding to each valid trajectory.

[0098] Clustering of valid trajectories based on intent labels yields multiple candidate intent sets.

[0099] In one possible embodiment, the clustering module is further specifically used for:

[0100] Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs;

[0101] Add the trajectory curvature distribution features and associated features to the feature vector.

[0102] In one possible embodiment, the clustering module is further specifically used for:

[0103] If the associated feature indicates a trajectory that conflicts with a traffic sign, the intent label of the corresponding valid trajectory will be determined as an invalid intent.

[0104] In one possible embodiment, the filtering module is specifically used for:

[0105] Calculate the smoothing value and style matching value of each valid trajectory in the target set;

[0106] The smoothness value is obtained by weighted summation of the smoothness value and style matching value based on the preset harmonic coefficient;

[0107] The effective trajectory with the smallest smoothness value is determined as the target trajectory.

[0108] In one possible embodiment, the filtering module is further specifically used for:

[0109] Obtain the longitudinal and lateral jerk of each valid trajectory;

[0110] Discretely sum the squares of the longitudinal and lateral jerks to obtain a smooth value.

[0111] In one possible embodiment, the filtering module is further specifically used for:

[0112] Obtain the spatiotemporal weighting factor, which includes the weight of the starting segment and the weight of the middle and later segments;

[0113] The trajectory points in the effective trajectory are divided into the starting segment and the middle and rear segments according to a preset ratio;

[0114] The smoothed value is obtained by weighting the sum of the squares of the starting segment and the middle and later segments according to the weights of the starting segment and the middle and later segments.

[0115] In one possible embodiment, the filtering module is further specifically used for:

[0116] Obtain the average acceleration of the preset driving style and each effective trajectory;

[0117] Match the preset driving style with the preset style reference acceleration mapping table to obtain the style reference acceleration;

[0118] The style matching value is obtained by calculating the square of the difference between the average acceleration and the style reference acceleration.

[0119] In one possible embodiment, the analysis module is further configured to:

[0120] Obtain current scene information, including traffic light status and remaining distance of the navigation path;

[0121] Adjust the preset weights of each factor based on the current scenario information.

[0122] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0123] The memory stores computer-executed instructions;

[0124] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0125] Fourthly, embodiments of this application provide 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 implementations of the first aspect.

[0126] Fifthly, embodiments of this application provide 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.

[0127] The vehicle motion control method, electronic device, and program product provided in this application first acquire multiple candidate trajectories, operation parameters, and navigation data of the vehicle and verify them to obtain valid trajectories. Then, the valid trajectories are clustered by intent to obtain multiple candidate intent sets with intent labels. The values ​​of each factor are determined according to the intent label, operation parameters, navigation data, and valid trajectories in each candidate intent set, and weighted summation is performed. The candidate intent set with the highest calculated value is determined as the target set. The valid trajectories in the target set are screened for smoothness to determine the target trajectory and output to the vehicle control module. By validating candidate trajectories and further clustering intentions, different decision-making levels, such as trajectory invalidity, driving intentions, and trajectory selection, can be differentiated and processed, thereby reducing the computational complexity and decision coupling caused by multi-dimensional evaluation. By combining intention tags, operation parameters, and navigation data to perform factor values ​​and weighted summation on each candidate intention set, the overall adaptability of candidate trajectories in the current scenario can be more comprehensively reflected, thereby improving the accuracy and scenario adaptability of trajectory decisions. By further screening the smoothness of valid trajectories in the target set, the comfort and executability of the output trajectory can be further improved, thereby improving the efficiency of candidate trajectory decision-making and enhancing the stability and interpretability of the results while ensuring safety and compliance. Attached Figure Description

[0128] 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.

[0129] Figure 1 Flowchart of the vehicle motion control method provided in this application Figure 1 ;

[0130] Figure 2 Flowchart of the vehicle motion control method provided in this application Figure 2 ;

[0131] Figure 3 Flowchart of the vehicle motion control method provided in this application Figure 3 ;

[0132] Figure 4 Flowchart of the vehicle motion control method provided in this application Figure 4 ;

[0133] Figure 5 Flowchart of the vehicle motion control method provided in this application Figure 5 ;

[0134] Figure 6 Flowchart of the vehicle motion control method provided in this application Figure 6 ;

[0135] Figure 7 Flowchart of the vehicle motion control method provided in this application Figure 7 ;

[0136] Figure 8 Flowchart of the vehicle motion control method provided in this application Figure 8 ;

[0137] Figure 9 A schematic diagram of the vehicle motion control device provided in this application;

[0138] Figure 10 A schematic diagram of the structure of the electronic device provided in this application.

[0139] 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

[0140] 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.

[0141] First, let me explain the terms used in this application:

[0142] Candidate trajectory: refers to the vehicle driving trajectory that can be selected from the trajectory generation model or planning module.

[0143] Operating parameters: These are parameters that reflect the driver's operating intentions or the vehicle's operating status, including driver operating signals, etc.

[0144] Navigation data refers to data provided by the navigation system that is related to the vehicle's driving path, including navigation expectations and remaining distance of the navigation path.

[0145] Valid trajectory: refers to the trajectory that is retained after the candidate trajectory has been verified.

[0146] Security verification: refers to performing security-related screening on candidate trajectories to obtain safe trajectories.

[0147] Safe trajectory: refers to the trajectory that is retained after the candidate trajectory has passed the security verification.

[0148] Compliance verification: This refers to screening security tracks for compliance with regulations to obtain valid tracks.

[0149] Dynamic obstacles: These are traffic participants that are in motion around a vehicle.

[0150] Predicted motion trajectory: refers to the trajectory obtained by predicting the future motion state of a dynamic obstacle.

[0151] Collision verification refers to the process of determining whether a collision has occurred based on the candidate trajectory and the predicted motion trajectory.

[0152] Collision time: refers to the time required for the candidate trajectory to collide with the predicted motion trajectory.

[0153] Preset time threshold: refers to a pre-set threshold used to determine whether the collision time meets the requirements.

[0154] Lane markings: These are the markings on the lane where a vehicle is located used to indicate the boundaries of the lane.

[0155] Traffic light status: refers to the current indication status of the traffic light in front of the vehicle.

[0156] Intent clustering refers to the process of classifying the intent of valid trajectories based on their characteristics.

[0157] Candidate intent set: refers to the set of trajectories corresponding to the same intent label obtained after clustering the valid trajectories by intent.

[0158] Intent label: refers to a label used to identify the driving intent corresponding to the candidate intent set.

[0159] Factor: refers to the dimension used to evaluate the set of candidate intentions.

[0160] Preset weights: These are weights that are set in advance and used to sum the values ​​of each factor.

[0161] Calculated value: refers to the result obtained by weighting and summing the values ​​of each factor according to their corresponding preset weights.

[0162] Target set: refers to the set of candidate intents with the highest calculated value.

[0163] Smoothness filtering: refers to performing smoothness-related filtering on valid trajectories in the target set to determine the target trajectory.

[0164] Target trajectory: refers to the trajectory determined after smoothness screening and used to output to the vehicle control module.

[0165] Vehicle control module: This refers to the module that receives the target trajectory and controls the movement of the vehicle accordingly.

[0166] Feature vector: refers to the vector composed of features extracted from valid trajectories for intended clustering.

[0167] Preset intent labels: These are pre-defined intent labels used for matching feature vectors.

[0168] Trajectory curvature distribution characteristics: refers to the characteristics that reflect how the trajectory curvature changes with the trajectory.

[0169] The correlation characteristics between trajectory and traffic signs refer to the characteristics that reflect the relationship between trajectory and traffic signs.

[0170] Invalid intent: refers to an intent label that is determined to be invalid due to a conflict between the trajectory and traffic signs.

[0171] Driver operation signals: These are signals that reflect the driver's operating behavior.

[0172] Navigation expectation: refers to the expectation given by the navigation system that is related to the vehicle's driving actions.

[0173] Matching result: refers to the result obtained after matching the intent label with the driver's operation signal or navigation expectation.

[0174] Preset mapping table: refers to a pre-defined mapping table used to determine factor values ​​based on matching results.

[0175] Driver intention factor: refers to a factor used to evaluate the degree of matching between the candidate intention set and the driver's operational intention.

[0176] Navigation factor: refers to a factor used to evaluate the degree of matching between the set of candidate intentions and navigation expectations.

[0177] Interaction factor: refers to a factor used to evaluate the degree of influence of the candidate intent set on the movement of surrounding vehicles.

[0178] Efficiency factor: refers to a factor used to evaluate the passage efficiency of the trajectory corresponding to the candidate intent set.

[0179] Total deceleration: refers to the total amount of deceleration caused by surrounding vehicles trying to avoid this vehicle.

[0180] Road speed limit: refers to the maximum speed allowed for vehicles on the road they are on.

[0181] Smoothing value: refers to a numerical value used to evaluate the smoothness of a trajectory.

[0182] Style matching value: This refers to the numerical value used to evaluate the degree of matching between the trajectory and the preset driving style.

[0183] Preset harmonic coefficient: refers to a pre-set coefficient used to weight and sum the smoothing value and the style matching value.

[0184] Smoothness value: refers to the value obtained by weighting and summing the smoothness value and style matching value according to a preset harmonic coefficient.

[0185] Longitudinal jerk: refers to the derivative of the vehicle's longitudinal acceleration with respect to time.

[0186] Lateral jerk: refers to the derivative of a vehicle's lateral acceleration with respect to time.

[0187] Discrete summation refers to the process of accumulating and summing the values ​​at discrete trajectory points.

[0188] Spatiotemporal weighting factor: refers to the factor used to assign different weights to the sum of squares of acceleration at different positions of the trajectory, including the weight of the starting segment and the weight of the middle and later segments.

[0189] Starting segment weight: refers to the weight assigned to the sum of squares of accelerations of the starting segment of the trajectory.

[0190] Mid-to-late stage weight: refers to the weight assigned to the sum of squared accelerations in the mid-to-late stage of the trajectory.

[0191] Preset ratio: refers to the pre-set ratio used to divide the trajectory points into the starting segment and the middle and later segments.

[0192] Preset driving style: refers to the driving style that is set in advance and used to determine the style reference acceleration.

[0193] Average acceleration: refers to the average acceleration of all points on the trajectory.

[0194] Preset Style Reference Acceleration Mapping Table: This refers to a pre-set mapping table used to determine the style reference acceleration based on the preset driving style.

[0195] Style reference acceleration: refers to the reference acceleration determined based on the preset driving style and used to calculate the style matching value.

[0196] Current scene information: refers to information reflecting the current scene of the vehicle, including the status of traffic lights and the remaining distance of the navigation path.

[0197] Vehicle motion control and intelligent driving trajectory planning technologies are applied to autonomous vehicles equipped with perception, localization, navigation, and control modules. When vehicles are driving on urban roads, intersections, or highways, the planning system typically receives multi-source perception data and navigation information, generates multiple candidate trajectories, and filters the trajectories based on the vehicle's status before outputting control commands to drive the vehicle to complete actions such as cruising, lane changing, steering, and obstacle avoidance.

[0198] Current trajectory decision-making typically combines safety verification, intent judgment, and smoothness evaluation. Candidate trajectories are first scored uniformly, and then the optimal result is selected. While this approach can filter trajectories, in complex and dynamic traffic environments, the number of candidate trajectories is large, and different trajectories have varying degrees of ineffectiveness, driving intent, and comfort. Centralizing the processing of multiple factors can lead to lengthy computational chains, with some clearly ineffective trajectories still entering subsequent evaluation stages, increasing the burden on real-time planning. Furthermore, the lack of a clear boundary between intent judgment and trajectory selection makes the decision difficult to interpret when traffic rule constraints, driver operating parameters, and navigation requirements are intertwined, easily leading to selection biases and affecting the stability and scene adaptability of vehicle motion control.

[0199] In existing technologies, candidate trajectories are evaluated and ranked by means of unified neural network scoring, target-guided selection, or pre-trained semantic model judgment. However, there are technical problems such as trajectory screening, intent judgment, and comfort optimization being concentrated in the same model or the same processing step, and candidate trajectories with different driving intentions being evaluated in a mixed manner.

[0200] The vehicle motion control method provided in this application obtains a candidate intent set with intent labels by performing intent clustering on effective trajectories, determines the value of each factor according to the intent label, operation parameters, navigation data and effective trajectories in each candidate intent set, calculates the calculated value by weighted summation according to preset weights and determines the target set, and then performs smoothness screening on the effective trajectories in the target set to determine the target trajectory. This technical means solves the technical problem of insufficient accuracy in determining the target trajectory due to the mutual coupling of intent judgment and trajectory selection and the mixed evaluation of candidate trajectories with different driving intentions.

[0201] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are 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 now be described with reference to the accompanying drawings.

[0202] Figure 1 Flowchart of the vehicle motion control method provided in this application Figure 1 ,like Figure 1 As shown, the method includes:

[0203] S101. Obtain multiple candidate trajectories, operating parameters, and navigation data for the vehicle.

[0204] Specifically, during vehicle operation, the trajectory generation module outputs multiple candidate trajectories for selection, while simultaneously acquiring operational parameters reflecting the driver's actions and navigation data provided by the navigation system. Candidate trajectories refer to the selectable vehicle driving trajectories output by the trajectory generation model or planning module. Operational parameters refer to parameters reflecting the driver's operational intentions or the vehicle's operating status.

[0205] Navigation data refers to data related to the vehicle's driving path provided by the navigation system. By collecting trajectory data, operation data, and navigation data from the trajectory generation module, the driver's operating terminal, and the navigation system respectively, a complete data source is provided for subsequent verification, intent clustering, and factor value selection. This ensures that subsequent processing has complete input data and avoids processing interruptions or result deviations due to missing data.

[0206] S102. Verify the candidate trajectories to obtain multiple valid trajectories.

[0207] Specifically, verification refers to the screening of candidate trajectories to remove those that do not meet the requirements. A valid trajectory is a candidate trajectory that is retained after verification. Verification of candidate trajectories removes those that do not meet safety or regulatory requirements. By setting verification conditions, candidate trajectories are screened one by one, trajectories that do not meet the conditions are excluded, and those that do meet the conditions are retained. This ensures that subsequent processing is only performed on the retained trajectories, reducing the impact of invalid trajectories on intended clustering and weighted summation, and improving processing efficiency and result reliability.

[0208] For example, the vehicle can be an autonomous vehicle equipped with perception, localization, navigation and control modules. The candidate trajectory can be generated by the planning system based on the current road environment, lane geometry information, the state of surrounding traffic participants and the vehicle's own motion state, and sent to the processing flow of this method through the vehicle communication link or local control link.

[0209] Operating parameters may include steering wheel input, accelerator pedal opening, brake pedal status, steering request, driving mode parameters, target speed constraints, and current control priority. Navigation data may include target road information, path segment sequence, lane keeping requirements, turning instructions, speed limit information, and driving constraints corresponding to the navigation destination.

[0210] Candidate trajectories are validated before entering subsequent processing. If a candidate trajectory has a breakpoint, exceeds the range of vehicle dynamics, or has a collision risk exceeding a threshold, the trajectory is discarded and will not enter the subsequent intent clustering process. Candidate trajectories that meet integrity and safety constraints are retained as valid trajectories.

[0211] Candidate trajectories are used to carry multiple possible vehicle driving schemes. Valid trajectories are a set of trajectories that can continue to participate in intent judgment and decision calculation after verification. Operational parameters are used to characterize the constraint information and control preferences required for the current vehicle motion control. Navigation data is used to characterize the vehicle's current driving path and target driving direction.

[0212] Based on the above processing method, trajectories that clearly do not meet the constraints are filtered out early on, and subsequent judgments are only made on executable trajectories, thereby reducing invalid computation links and maintaining the consistency of subsequent decision inputs.

[0213] S103. Perform intent clustering on the valid trajectories to obtain multiple candidate intent sets, each candidate intent set corresponding to an intent label.

[0214] Specifically, intent clustering refers to the process of classifying the intents of valid trajectories based on their characteristics. A candidate intent set is a set of trajectories corresponding to the same intent label, obtained after intent clustering of valid trajectories. An intent label is a label used to identify the driving intent corresponding to the candidate intent set.

[0215] By extracting the features of valid trajectories, the valid trajectories are categorized into sets corresponding to different driving intentions according to the features. Trajectories with the same driving intention are concentrated in the same set. The valid trajectories are distinguished according to driving intention, which provides a structural basis for the subsequent determination of factor values ​​according to intention. This avoids the mixing of trajectories with different driving intentions, and allows the determination of subsequent factor values ​​and calculated values ​​to be carried out on an intention-by-intention basis.

[0216] For example, intent clustering can be performed based on the trajectory feature information of valid trajectories. The system can map each valid trajectory to a corresponding feature representation and match it with a preset intent label. When the similarity of several valid trajectories in the feature space reaches a preset requirement and their motion directions are consistent, these valid trajectories are divided into the same candidate intent set and the set is assigned a corresponding intent label.

[0217] If some valid trajectories, although similar in motion pattern, conflict with traffic signs, lane rules, or navigation routes, they can be included in the set associated with the conflict state or directly excluded from the intent clustering results to ensure that subsequent factor values ​​are based on consistent intent. Intent labels are used to identify the driving intent corresponding to the candidate intent set, and the candidate intent set is used to aggregate valid trajectories with similar motion targets.

[0218] Through the clustering method described above, multiple candidate results that were originally at the trajectory level are elevated to grouped results at the intent level, so that the subsequent calculation objects for different decision dimensions are transformed from single trajectories into intent sets, which facilitates the unified processing of similar moving targets and preserves scene semantics.

[0219] S104. Based on the intent label, operation parameters, navigation data, and valid trajectories in each candidate intent set, determine the values ​​of each candidate intent set on multiple factors.

[0220] Specifically, a factor refers to a dimension used to evaluate a set of candidate intentions. For each set of candidate intentions, values ​​are determined for multiple factors. The values ​​of some factors are determined based on the matching relationship between the intention tags and operation parameters / navigation data, while the values ​​of others are calculated based on valid trajectories within the candidate intention set.

[0221] By processing discrete and continuous factors separately, the discrete factors are determined based on matching relationships, while the continuous factors are calculated based on trajectory data. This allows each candidate intent set to obtain values ​​under the same set of factor dimensions, providing data for subsequent weighted summation and enabling different intents to be calculated separately under the same evaluation system.

[0222] For example, multiple factors may include several factors used to comprehensively evaluate multiple decision dimensions of the candidate intent set, and the values ​​of each factor are jointly determined by different information sources. The more the effective trajectory in each candidate intent set matches the movement direction indicated by the intent label, and the more it matches the control requirements represented by the current operation parameters and navigation data, the higher the corresponding factor value; if the effective trajectory in a candidate intent set is significantly unsuitable in the current scenario, the corresponding factor value will decrease.

[0223] Factors are used to comprehensively evaluate multiple decision dimensions of a candidate intent set, and their values ​​characterize the degree of fit of the set in the corresponding dimension. By jointly mapping intent labels, operation parameters, navigation data, and trajectory features to multiple factors, subsequent weighted decisions can be mapped to interpretable scenario dimensions, and driver operation information, navigation constraints, and environmental relationships can be incorporated into the same processing chain.

[0224] S105. The values ​​are weighted and summed according to the preset weights corresponding to each factor to obtain the calculated value of each candidate intent set.

[0225] Specifically, preset weights refer to the pre-defined weights used to weight and sum the values ​​of each factor. The calculated value is the result obtained by weighting and summing the values ​​of each factor according to their corresponding preset weights. The weighted summation is performed according to the preset weights corresponding to each factor.

[0226] By using weights to reflect the importance of different factors in the evaluation, the values ​​of each factor are accumulated according to their weights to obtain the calculated value of each candidate intent set, which serves as the quantitative basis for subsequently determining the target set. This allows different factors to participate in the calculation according to their respective importance, improving the rationality of the calculated value in evaluating the candidate intent set.

[0227] S106. The set of candidate intentions with the highest calculated value is determined as the target set.

[0228] Specifically, the target set refers to the set of candidate intentions with the highest calculated values. After obtaining the calculated values ​​of each candidate intention set, the candidate intention set with the highest calculated value is selected as the target set. By comparing the calculated values ​​of each candidate intention set and selecting the highest one, intention-level comparison and selection are performed based on the completed intention differentiation. This ensures that subsequent processing is only carried out on the valid trajectories in the target set, avoiding the mixing and comparison of trajectories of different driving intentions in the same evaluation system.

[0229] For example, preset weights are used to characterize the decision priority of different factors in vehicle motion control, and the weights can be pre-configured in the control parameters. During weighted summation, the value of each factor can be multiplied by its corresponding weight and then summed to obtain the calculated value of the candidate intent set. The higher the calculated value, the higher the degree of matching between the set and the control requirements in the current scenario.

[0230] The calculated values ​​of all candidate intent sets are compared, and the set with the highest value is selected as the target set. The valid trajectories corresponding to the target set then enter the smoothness screening process. The calculated values ​​are used to compare the overall score of each candidate intent set, and the target set is used to carry out subsequent trajectory refinement screening.

[0231] This processing method unifies and quantifies multiple heterogeneous factors into a single comparable result, enabling the selection at the candidate intent level to form a clear priority relationship between different scenario constraints, and ensuring that the final set entering the trajectory-level screening has high scenario consistency.

[0232] S107. Perform smoothness screening on the valid trajectories in the target set to determine the target trajectory.

[0233] Specifically, smoothness screening refers to performing smoothness-related filtering on valid trajectories in the target set to determine the target trajectory. The target trajectory is the trajectory determined after smoothness screening and used for output to the vehicle control module. Under the same driving intention corresponding to the target set, the smoothness of valid trajectories is compared, and the trajectory that meets the smoothness requirements is selected. The target trajectory is determined under the same driving intention, ensuring that the screening results are consistent with the driving intention and improving the accuracy of target trajectory determination.

[0234] S108. Output the target trajectory to the vehicle control module to control the vehicle's movement.

[0235] Specifically, the vehicle control module is the module that receives the target trajectory and controls the vehicle's movement accordingly. The determined target trajectory is output to the vehicle control module, which then controls the vehicle's movement based on it. By transmitting the determined target trajectory to the vehicle control module, which then executes vehicle motion control accordingly, the process from candidate trajectory to target trajectory is completed. The processing result is then used for vehicle motion control, achieving a seamless connection between trajectory determination and control execution.

[0236] For example, smoothness screening can evaluate each valid trajectory within the target set separately and determine the target trajectory based on preset screening rules. First, the smoothness of the valid trajectories in the target set can be evaluated. Then, each valid trajectory can be compared according to preset conditions, and the valid trajectory that meets the requirements can be selected as the target trajectory. The target trajectory is then sent to the vehicle control module, which can generate steering, acceleration, braking, or torque distribution control quantities based on the target trajectory and apply these control quantities to the actuators to make the vehicle adjust its movement according to the target trajectory.

[0237] Smoothness filtering is used to further select the most suitable trajectory for actual control execution within the same target set. The target trajectory is the final output control trajectory, which the vehicle control module receives and uses to control vehicle movement. By performing trajectory-level smoothness filtering after completing the set selection at the intent level, it is possible to maintain the intended smoothness of the vehicle. Figure 1 Under the premise of consistency, further constraints are placed on the dynamic continuity and comfort of the trajectory, so that the final output can be directly used for vehicle control closed loop.

[0238] This invention first acquires and verifies multiple candidate trajectories, operation parameters, and navigation data to obtain valid trajectories. Then, it clusters these valid trajectories according to driving intentions to obtain multiple candidate intention sets with intention labels, thus distinguishing trajectories with different driving intentions. Next, it determines the values ​​of each factor based on the intention label, operation parameters, navigation data, and valid trajectories in each candidate intention set, and calculates the calculated value by weighting them according to preset weights. The candidate intention set with the highest calculated value is determined as the target set. Finally, under the same driving intention, the valid trajectories in the target set are screened for smoothness to determine the target trajectory, which is then output to the vehicle control module. Because trajectories with different driving intentions are distinguished before determining factor values ​​and calculated values, and the target trajectory is determined under the same driving intention, trajectories with different driving intentions are no longer mixed for comparison, thus improving the accuracy of target trajectory determination. It should be understood that the above example is merely illustrative and not limiting.

[0239] Figure 2Flowchart of the vehicle motion control method provided in this application Figure 2 ,like Figure 2 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0240] S201. Perform security checks on the candidate trajectories to obtain multiple safe trajectories.

[0241] Specifically, security verification refers to performing security-related screening on candidate trajectories to obtain safe trajectories. Safe trajectories are those candidate trajectories that are retained after security verification. By performing security verification on candidate trajectories, those that do not meet security requirements are excluded, while those that do meet the requirements are retained. This ensures that subsequent processing is only performed on the retained trajectories, reducing the impact of insecure trajectories on subsequent compliance verification and intent clustering, and improving the reliability of the processing results.

[0242] S202. Perform compliance verification on the safety trajectory to obtain multiple valid trajectories.

[0243] Specifically, compliance verification refers to screening security tracks for regulatory compliance to obtain valid tracks. Valid tracks are those security tracks that are retained after compliance verification. By verifying the compliance of security tracks, tracks that do not meet regulatory requirements are excluded, while those that do are retained. This allows subsequent intent clustering and factor selection to be based on these compliant tracks, improving the compliance of target track determination.

[0244] In this application, the candidate trajectory is a set of trajectories generated by the planning module based on the vehicle's current position, environmental perception results, and navigation constraints. The trajectory set can consist of multiple discrete trajectory points and includes corresponding timestamps, speeds, and curvature information.

[0245] The safety check is used to determine the collision between the candidate trajectory and the predicted motion trajectory of dynamic obstacles. Dynamic obstacles include vehicles, pedestrians, non-motorized vehicles and static occupants. The predicted motion trajectory can be extrapolated based on the target's current speed, heading angle and acceleration. When the spatial distance between the candidate trajectory and any dynamic obstacle is less than the safety threshold within a preset time window, or when there is a risk of intersection, the candidate trajectory is judged as unsafe and is removed. The remaining trajectories constitute multiple safe trajectories.

[0246] Compliance verification is used to determine road rules for safe trajectories based on lane markings and traffic light status. Lane marking information can be obtained by matching high-precision positioning results with lane boundaries in high-precision maps, and traffic light status can be obtained by roadside sensing units or vehicle-mounted visual recognition results. When a safe trajectory crosses a solid lane line, crosses a prohibited lane-changing area, or crosses a stop line during a red light, the trajectory is deleted, resulting in multiple valid trajectories.

[0247] The thresholds for determining safe and effective trajectories can be set based on vehicle wheelbase, braking capacity, and road type. In practical applications, other configurations can also be used for the relevant parameters, which are not limited in this application.

[0248] In this implementation, safety checks first exclude trajectories with collision risks, and compliance checks then remove trajectories that violate road rules. This ensures that the trajectory selection process converges sequentially based on safety and regularity, and the output valid trajectories can be directly used for subsequent intent clustering and control decisions.

[0249] Since the candidate trajectories have undergone two levels of verification before entering subsequent calculations, the retained trajectories are consistent with the dynamic targets around the vehicle and road constraints, thereby improving the stability and interpretability of trajectory decisions and reducing the processing burden caused by invalid trajectories participating in subsequent calculations.

[0250] This invention provides a method to first perform security verification on candidate trajectories to obtain safe trajectories, and then perform compliance verification on the safe trajectories to obtain valid trajectories. This allows for phased screening based on security and regulatory requirements. Trajectories retained in the previous phase serve as input for the next phase. Trajectories that do not meet security requirements are no longer subject to compliance verification, and trajectories that do not meet regulatory requirements are no longer subject to intent clustering. This improves the efficiency and reliability of trajectory verification results.

[0251] Figure 3 Flowchart of the vehicle motion control method provided in this application Figure 3 ,like Figure 3 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0252] S301. Obtain the predicted motion trajectory of dynamic obstacles around the vehicle.

[0253] Specifically, dynamic obstacles refer to moving traffic participants around a vehicle. Motion prediction trajectories are trajectories obtained by predicting the future motion of dynamic obstacles. Acquiring these motion prediction trajectories provides comparison objects for subsequent collision verification, allowing candidate trajectories to be compared with the future motion of dynamic obstacles, thus improving the targeting of collision verification.

[0254] S302. Collision verification is performed between the candidate trajectory and the motion prediction trajectory to obtain multiple safe trajectories.

[0255] Specifically, collision detection refers to the process of determining whether a collision has occurred based on the candidate trajectory and the predicted motion trajectory. Collision detection is performed between the candidate trajectory and the predicted motion trajectory to eliminate candidate trajectories that are likely to collide, while retaining those that are not. This ensures that subsequent processing is only performed on safe trajectories, reducing the impact of colliding trajectories on subsequent processing.

[0256] Dynamic obstacles represent external traffic participants around the vehicle who may interact with it. These external traffic participants include pedestrians, non-motorized vehicles, motorized vehicles, or other moving targets. The motion prediction trajectory represents the pose changes of the dynamic obstacle at future moments. It can be predicted by combining the current state information output by the perception module with velocity, acceleration, heading angle, and historical motion sequences. Collision verification determines whether candidate trajectories spatially overlap with dynamic obstacles at each prediction moment or whether the minimum distance is less than a preset safety threshold, thereby filtering out trajectories with collision risks.

[0257] In practical implementation, after obtaining candidate trajectories, the current position, velocity, and direction of motion of dynamic obstacles are extracted based on the target detection results around the vehicle. These are then combined with a trajectory tracking model, a Kalman filter model, or a prediction model based on a temporal neural network to generate the predicted motion trajectory for each dynamic obstacle. The candidate trajectory can be represented as a series of reference pose points for the vehicle in the discrete programming time domain, while the predicted motion trajectory corresponds to the set of predicted pose points of the dynamic obstacle in the same time domain.

[0258] After aligning the two in time and space, the geometric relationship between the vehicle outline and the obstacle outline is calculated at each time step, or the distance relationship between the vehicle's outer rectangle and the obstacle's safety envelope is used to determine whether the collision-free condition is met. When the collision risk determination condition is met at any time step, the candidate trajectory is removed, and the trajectory that meets the safety threshold requirement at all prediction times is retained as the safe trajectory output.

[0259] This safety verification method first acquires the future motion information of dynamic obstacles, then performs time-by-time collision discrimination between candidate trajectories and prediction results. This allows the planning results to be directly based on the future motion states of traffic participants, thereby outputting multiple safe trajectories for subsequent intent clustering and trajectory optimization. Since collision risks are screened out before entering subsequent evaluation, subsequent processing is only carried out on safe trajectories, making the trajectory selection process clearer and improving real-time performance and decision stability in complex traffic environments.

[0260] In some alternative implementations, S302 above includes:

[0261] S3021. Obtain the collision time between the candidate trajectory and the predicted motion trajectory.

[0262] Specifically, collision time refers to the time required for a candidate trajectory to collide with the predicted motion trajectory. Obtaining the collision time between the candidate trajectory and the predicted motion trajectory provides a quantitative basis for subsequent screening, allowing candidate trajectories to be distinguished based on the length of the collision time.

[0263] S3022. Extract candidate trajectories whose collision time is greater than a preset time threshold to obtain multiple safe trajectories.

[0264] Specifically, the preset time threshold refers to a pre-set threshold used to determine whether the collision time meets the requirements. Candidate trajectories with collision times greater than the preset time threshold are extracted, candidate trajectories with collision times that do not meet the requirements are excluded, and candidate trajectories with collision times that meet the requirements are retained, so that the obtained safe trajectory has sufficient avoidance margin in the time dimension.

[0265] The collision time characterizes the earliest contact moment between the candidate trajectory and the predicted trajectory in the time domain. Its value is determined by comparing the position points of the candidate trajectory at discrete sampling moments with the position points of the predicted trajectory at the corresponding moments. The preset time threshold limits the minimum safety margin allowed to proceed to subsequent processing. The threshold can be set according to the vehicle's current speed, braking capacity, road adhesion coefficient, and obstacle motion state, and can remain a fixed value under different road scenarios or be updated in real time by the controller.

[0266] In its implementation, the system first performs time parameterization on each candidate trajectory based on the vehicle's current pose, velocity, and acceleration, and then predicts the vehicle's motion trajectory in the future time domain based on the historical position, velocity, and heading of the dynamic obstacles around the vehicle.

[0267] The controller aligns the candidate trajectory with the predicted motion trajectory of the corresponding obstacle on a unified time axis and calculates the spatial distance between the two at each sampling moment. When the spatial distance is less than the set collision radius for the first time, the moment is recorded as the collision time.

[0268] If a candidate trajectory does not come into contact with any predicted motion trajectory within the prediction time domain, its collision time can be recorded as a value greater than the upper limit of the prediction time domain, thus being determined to meet the safety conditions.

[0269] The controller further compares the collision time corresponding to each candidate trajectory with a preset time threshold, retaining only candidate trajectories with collision times greater than the threshold as safe trajectories. In this way, trajectories where collisions occur too close are eliminated in advance during the collision verification process, while trajectories that meet the minimum time safety margin are output to subsequent compliance verification and intent filtering stages. In practical applications, the collision radius, sampling period, and time threshold can all be configured according to vehicle size, control frequency, and scenario risk level; this application does not impose any limitations on these settings.

[0270] Based on the above processing, collision verification can quickly separate near-collision trajectories from safe trajectories in the candidate trajectories, ensuring that the set of trajectories entering the subsequent decision-making stage maintains a high temporal safety margin. Since the collision time directly reflects the proximity between the candidate trajectory and the dynamic obstacle, this verification method can complete safety screening in the initial stage of trajectory planning, thereby making the output of safe trajectories more in line with the real-time motion control requirements of the vehicle.

[0271] This invention obtains the predicted motion trajectory of a dynamic obstacle, performs collision verification between the candidate trajectory and the predicted motion trajectory, and extracts a safe trajectory based on the collision time and a preset time threshold. This ensures that the candidate trajectory has been excluded from subsequent processing because the collision time does not meet the requirements, thereby improving the reliability of subsequent processing.

[0272] Figure 4 Flowchart of the vehicle motion control method provided in this application Figure 4 ,like Figure 4 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0273] S401. Obtain the lane markings of the lane where the vehicle is located and the status of the traffic lights in front of the vehicle.

[0274] Specifically, lane markings refer to the lines on the lane in which the vehicle is located that indicate the lane boundaries. Traffic light status refers to the current indication status of the traffic lights ahead of the vehicle. Obtaining the lane markings of the vehicle's lane and the traffic light status ahead of the vehicle provides a basis for subsequent compliance verification, allowing the safety trajectory to be judged by combining lane boundaries and traffic light indications.

[0275] S402. Based on lane markings and traffic light status, the safety trajectory is verified to obtain multiple valid trajectories.

[0276] Specifically, the safety trajectory is verified based on lane markings and traffic light status. Safety trajectories that do not meet the requirements indicated by lane markings and traffic light status are excluded, while compliant safety trajectories are retained, ensuring that the obtained valid trajectory is compliant with regulations.

[0277] Lane markings are used to characterize the boundary rules of the vehicle's current lane. The system can capture road images through an onboard camera and extract solid lines, dashed lines, and lane boundary positions using a visual recognition model. Traffic light status is used to characterize the traffic constraints at the intersection ahead. The system can identify traffic light colors, arrow directions, and countdown status based on a forward-facing camera and map the identification results to signs indicating whether to proceed, wait, or restricted passage. Lane markings and traffic light status together constitute the road rule input for trajectory compliance judgment. Lane markings are used to determine whether the trajectory crosses solid lane lines, while traffic light status is used to determine whether the trajectory still meets the passage requirements under traffic light restrictions.

[0278] During the actual verification process, the control module discretizes the safe trajectory into multiple trajectory points within the planning time domain and matches them with the spatial location of the current lane markings and the traffic constraints corresponding to the upcoming traffic light status. If the lateral projection of a safe trajectory crosses a solid lane line, or if it still shows a tendency to cross the stop line while the light is red, the trajectory is deemed non-compliant and discarded. If a safe trajectory remains within the lane boundary and meets the stop, waiting, or permitted passage status under traffic light constraints, the trajectory is retained as a valid trajectory. By jointly verifying lane markings and traffic light status, the output trajectory can simultaneously satisfy lane rules and traffic light rules, providing a stable set of valid trajectories for subsequent trajectory selection.

[0279] This method introduces road rule constraints after safety verification, ensuring that trajectories that only meet collision safety requirements but not traffic regulations are promptly filtered out. This results in multiple valid trajectories being output for subsequent intent clustering and target trajectory selection. Since compliance judgments are directly based on lane markings and traffic light status, the reasons for invalid trajectories have a clear correspondence, thus improving the interpretability of trajectory decisions and reducing the number of trajectories that do not comply with road traffic rules entering subsequent calculation stages.

[0280] In some alternative implementations, S402 above includes:

[0281] S4021. Determine whether the safe trajectory involves crossing solid lane lines or violating traffic light status.

[0282] Specifically, by comparing the safe trajectory with lane markings and traffic light status, it can be determined whether the safe trajectory crosses solid lane lines or violates traffic light status, providing a basis for subsequent deletion of unqualified safe trajectories.

[0283] S4022. If a safe track exists, delete the corresponding track to obtain multiple valid tracks.

[0284] Specifically, safe tracks that involve crossing solid lane lines or violating traffic light status will be deleted, while safe tracks that do not involve such behavior will be retained, so that the resulting valid tracks meet the requirements indicated by lane markings and traffic light status.

[0285] Among them, solid lane lines are road markings used to define the boundaries of vehicle passage. The safety trajectory will be compared with the boundary relationship of the vehicle's current lane and adjacent lanes after planning. Traffic light status refers to the current traffic signal light indication in front of the vehicle, which is used to constrain whether the vehicle can pass through the stop line or enter the intersection.

[0286] Crossing a solid lane line refers to the sequence of trajectory points crossing the lane boundary where lane changing or crossing is prohibited after the safe trajectory unfolds in time; violating traffic light status refers to the vehicle corresponding to the safe trajectory crossing the stop line or entering the controlled area when the traffic light is red or yellow and prohibits passage.

[0287] In practical implementation, after obtaining the safe trajectory output by the planning module, the system combines lane line vector data from the high-precision map with traffic light status data obtained through visual perception or roadside perception to determine the compliance of each safe trajectory. The lane line vector data includes at least the spatial coordinates, direction attributes, and effective range of the solid lane lines, while the traffic light status data includes at least the signal color, corresponding lane group, and status duration.

[0288] The safe trajectory is discretized into multiple time-point trajectory points, and the positional relationship of each trajectory point relative to the solid lane line and the stop line is calculated. When any trajectory point crosses the solid lane line, or when the end of the trajectory crosses the stop line and enters the conflict area while the light is red, it is determined that the safe trajectory has a violation.

[0289] The system deletes safety tracks deemed to be in violation, retaining only those tracks that do not cross solid lane lines and do not violate traffic light status as valid tracks. This deletion process can either directly clear the corresponding index item in the track cache or mark them as invalid in the track set and remove them uniformly. In practical applications, other models of this judgment module can also be selected, and this application does not limit this choice.

[0290] This implementation method verifies road marking constraints and traffic light constraints in parallel, so that trajectories that clearly do not meet traffic rules are eliminated before entering subsequent intent clustering. This ensures that the output set of valid trajectories remains compliant with road regulations and provides a more stable candidate basis for subsequent trajectory optimization.

[0291] With this specific implementation method, the safety trajectory has completed the compliant screening of solid lane lines and traffic light status before entering the subsequent processing, which can make invalid trajectories leave the trajectory set in a timely manner, reduce the subsequent calculation burden, and make the final retained valid trajectory more in line with the current road traffic rules.

[0292] This invention verifies the compliance of safe trajectories by acquiring lane markings and traffic light status, and deletes corresponding trajectories based on whether there are behaviors such as crossing solid lane lines or violating traffic light status, so that valid trajectories are compliant at the regulatory level, thereby improving the compliance of target trajectory determination.

[0293] Figure 5 Flowchart of the vehicle motion control method provided in this application Figure 5 ,like Figure 5 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0294] S501: Obtain driver operation signals and navigation expectations based on operating parameters and navigation data.

[0295] Specifically, driver operation signals refer to signals that reflect the driver's operational behavior. Navigation expectations refer to the expectations given by the navigation system related to the vehicle's driving actions.

[0296] The driver's operating signals and navigation expectations are obtained based on operating parameters and navigation data, providing a data source for subsequent matching, so that the intent label can be compared with the driver's operating signals and navigation expectations.

[0297] S502. Match the intent label with the driver's operation signal and navigation expectation respectively to obtain the matching result.

[0298] Specifically, the matching result refers to the result obtained after matching the intent label with the driver's operation signal or navigation expectation. Matching the intent label with the driver's operation signal and navigation expectation respectively yields the matching result, providing a basis for subsequent queries of the preset mapping table, allowing factor values ​​to be determined based on the matching result.

[0299] S503. Based on the matching results, query the preset mapping table to obtain the values ​​of the driver intention factor and navigation factor.

[0300] Specifically, the preset mapping table refers to a pre-defined mapping table used to determine factor values ​​based on the matching results. The driver intent factor is a factor used to evaluate the degree of match between the candidate intent set and the driver's operational intent. The navigation factor is a factor used to evaluate the degree of match between the candidate intent set and navigation expectations.

[0301] The preset mapping table is queried based on the matching results to obtain the values ​​of the driver intention factor and navigation factor. The values ​​of discrete factors are determined according to the matching relationship, providing data basis for subsequent weighted summation.

[0302] Among them, driver operation signals are used to represent the driver's real-time operating intentions. They can be generated by combining operational parameters such as the vehicle's current accelerator pedal opening, brake pedal opening, steering angle, gear status, and steering wheel fine-tuning, along with road type, speed limit information, and expected driving direction from navigation data. Navigation expectations are used to represent the navigation system's expectations for driving behavior. They can be constructed based on the current navigation path, the next intersection turn instruction, target lane information, and remaining path distance.

[0303] Intent labels are used to identify the driving behavior category corresponding to each candidate intent set, such as going straight, turning left, turning right, changing lanes, or slowing down. The matching results are used to characterize the consistency between intent labels and driver operation signals and navigation expectations.

[0304] In practical implementation, the system can pre-establish a preset mapping table indexed by intent tags, which records the values ​​of driver intent factors and navigation factors corresponding to different matching combinations. After acquiring the driver's operation signals and navigation expectations, the system calculates the degree of matching between them and each intent tag, and uses the matching results as the search key to query the preset mapping table to output the corresponding factor values.

[0305] The driver intent factor can be represented by a discrete hierarchy or a normalized numerical value to quantify the consistency between the current operation and the intent label; the navigation factor can also be represented by a discrete hierarchy or a normalized numerical value to quantify the consistency between the candidate intent set and the navigation expectation. The mapping table can be stored in the storage unit of the vehicle controller or generated from calibration data. In practical applications, other storage structures and organization methods can also be selected for the mapping table, which is not limited in this application.

[0306] This factor value method can convert operating parameters and navigation data into quantitative results on the driver's side and the navigation side, respectively, and complete the lookup output through matching results. This allows different candidate intent sets to have a unified numerical expression form when they are subsequently weighted and summed, thus providing a stable input for determining the calculated value of the target set.

[0307] The driver-side factors and navigation-side factors of the candidate intent set can be determined separately and quantified uniformly. The correspondence between factor values ​​and intent labels is clear, which facilitates the direct use of preset results in subsequent weighted calculations, reduces the amount of real-time calculations on site, and improves the interpretability and consistency of the candidate intent set evaluation results.

[0308] In some alternative implementations, it also includes:

[0309] S504. Predict the total deceleration of surrounding vehicles when executing the effective trajectory, and take the negative value of the predicted value as the value of the interaction factor.

[0310] Specifically, the interaction factor refers to a factor used to evaluate the degree of influence of the candidate intent set on the motion of surrounding vehicles. The total deceleration magnitude refers to the total deceleration of surrounding vehicles in order to avoid the vehicle.

[0311] The total deceleration magnitude of surrounding vehicles when executing an effective trajectory is predicted to obtain a predicted value, and the negative value of the predicted value is taken as the value of the interaction factor, so that the value of the interaction factor is determined according to the degree of influence of the trajectory on surrounding vehicles.

[0312] S505. Calculate the ratio of the average speed of the effective trajectory to the road speed limit, and use the ratio as the value of the efficiency factor.

[0313] Specifically, the efficiency factor is a factor used to evaluate the traffic efficiency of the trajectory corresponding to the candidate intent set. The road speed limit refers to the maximum speed allowed on the road where the vehicle is located. The ratio of the average speed of the effective trajectory to the road speed limit is calculated, and this ratio is used as the value of the efficiency factor, ensuring that the value of the efficiency factor is determined based on the traffic efficiency of the trajectory.

[0314] The interaction factor is used to quantify the interaction between a vehicle and surrounding vehicles when the vehicle is traveling along the effective trajectory. Surrounding vehicles include target vehicles that are in the same lane or adjacent lane as the current vehicle and may generate speed responses during the trajectory execution period.

[0315] Based on the relative position, relative speed, lane relationship, and trajectory time domain length of the current vehicle and surrounding vehicles, a pre-established longitudinal response prediction model is invoked to cumulatively predict the possible deceleration of surrounding vehicles within a given time period, obtaining a predicted value for the total deceleration magnitude. This predicted value is then negatively represented as the interaction factor value. Using a negative value indicates a more significant deceleration of surrounding vehicles, resulting in a smaller value for the corresponding interaction factor, thus reflecting the degree of influence of the trajectory on the surrounding traffic flow in the subsequent weighted summation.

[0316] The efficiency factor is used to quantify the driving efficiency of the effective trajectory. The average speed of the effective trajectory is calculated from the total trajectory length and the estimated travel time. The road speed limit is determined by navigation data, road attribute information or lane-level map information.

[0317] After obtaining the effective trajectory, the average speed is calculated based on the trajectory's geometric length, speed constraints for each trajectory segment, and the time percentage of each segment. This average speed is then compared to the road speed limit to obtain the efficiency factor. This value characterizes the travel speed level of the effective trajectory under road constraints; a higher value indicates higher trajectory efficiency.

[0318] In practical applications, both the interaction factor and the efficiency factor are used as factor items in the candidate intent set for weighted calculation, and together with the driver intent factor and navigation factor, they constitute the trajectory evaluation input. After completing the effective trajectory screening, the system writes the values ​​of the above two factors into the feature vector of the corresponding candidate intent set, then performs weighted summation according to preset weights, and finally outputs the calculated value.

[0319] By mapping the effects of surrounding vehicle deceleration to the average speed level of the trajectory as quantitative factors, trajectory evaluation can simultaneously reflect the environmental interaction costs and traffic efficiency, thus providing a calculable basis for determining the target set.

[0320] The trajectory evaluation process can separate and quantify the impact on surrounding vehicles and the trajectory traffic efficiency, and participate in subsequent decision-making in a unified factor form. Therefore, the calculated value of the candidate intent set is closer to the actual traffic interaction state and road operation state, and ultimately improves the stability and scenario adaptability of the target trajectory selection results.

[0321] In this embodiment of the invention, the intent tag is matched with the driver's operation signal and navigation expectation respectively. The values ​​of the driver's intent factor and navigation factor are obtained by querying a preset mapping table based on the matching result. The values ​​of the interaction factor and efficiency factor are calculated based on the effective trajectory, so that the discrete factor and the continuous factor are determined in their own way, providing a complete data basis for subsequent weighted summation.

[0322] Figure 6 Flowchart of the vehicle motion control method provided in this application Figure 6 ,like Figure 6 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0323] S601. Extract the feature vector of the effective trajectory.

[0324] Specifically, a feature vector is a vector composed of features extracted from valid trajectories and used for intent clustering. Extracting feature vectors from valid trajectories provides data for subsequent matching with preset intent labels.

[0325] S602. Match the feature vector with the preset intent label to obtain the intent label corresponding to each valid trajectory.

[0326] Specifically, preset intent labels refer to pre-defined intent labels used for matching with feature vectors. Intent labels are labels used to identify the driving intent corresponding to the candidate intent set. Matching feature vectors with preset intent labels yields the intent labels corresponding to each valid trajectory, allowing valid trajectories to be distinguished according to driving intent.

[0327] S603. Cluster the valid trajectories based on the intent labels to obtain multiple candidate intent sets.

[0328] Specifically, the candidate intent set refers to the set of trajectories corresponding to the same intent label, obtained after clustering valid trajectories by intent. Valid trajectories are clustered based on intent labels, grouping valid trajectories with the same intent label into the same set, while trajectories with different driving intentions are assigned to their respective candidate intent sets.

[0329] Among them, the feature vector is used to characterize the set of motion attributes of the effective trajectory. The set of motion attributes includes at least the starting and ending positions of the trajectory, path curvature, heading change, longitudinal displacement, lateral displacement, and deviation from the center line of the lane.

[0330] The preset intent labels are a pre-established set of labels that correspond to the typical driving intentions of a vehicle on urban roads, intersections, or highways. Typical driving intentions include going straight, turning left, turning right, changing lanes, making a U-turn, and following another vehicle.

[0331] When matching feature vectors with preset intent labels, the feature vectors can be input into a classification model, the matching scores of each label can be output, and the label with the highest score can be taken as the intent label of the corresponding valid trajectory.

[0332] In practical implementation, the classification model can adopt an intention recognition model based on neural networks. The input layer of the neural network receives feature vectors, the hidden layer is used to extract nonlinear combination features of the trajectory, and the output layer corresponds to a preset set of intention labels in order to distinguish the motion patterns of different trajectories.

[0333] To enhance the stability of label matching, the trajectory duration, number of sampling points, rate of change of acceleration, and cumulative change of heading angle can be input into the model as auxiliary features, so that the same type of driving intention can still maintain consistent label output under different road scenarios.

[0334] This model can be deployed as a lightweight inference module in an onboard processor to reduce the real-time computing burden. In practical applications, other models of this module can also be selected, and this application does not limit them.

[0335] When clustering valid trajectories based on intent tags, valid trajectories with the same intent tag are grouped into the same candidate intent set, while trajectories corresponding to different intent tags are formed into multiple candidate intent sets respectively.

[0336] If multiple valid trajectories exist under a certain intent label, they can be further subdivided and aggregated based on the similarity between feature vectors to ensure that trajectories in the same candidate intent set maintain a consistent motion trend.

[0337] Similarity can be calculated using cosine similarity, Euclidean distance, or Mahalanobis distance. Cluster centers can be determined by the feature mean of the trajectory corresponding to the label, thus obtaining a set of candidate intentions with a clear structure.

[0338] The system first converts valid trajectories into computable feature vectors, then uses preset intent labels to assign intents to the trajectories, and finally forms multiple candidate intent sets based on the intent labels. This allows subsequent factor values ​​and weighted calculations to be performed directly on intents of the same type. This processing method separates the intent attributes of valid trajectories from comfort evaluation, reduces the entry of irrelevant trajectories into subsequent stages, and improves the interpretability and scene adaptability of the candidate intent sets.

[0339] In some alternative implementations, S601 above includes:

[0340] S6011. Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs.

[0341] Specifically, the trajectory curvature distribution characteristics refer to the features that reflect how the trajectory curvature changes with the trajectory. The correlation characteristics between the trajectory and traffic signs refer to the features that reflect the relationship between the trajectory and traffic signs.

[0342] Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs, so that the feature vector contains information about the trajectory shape and its relationship with traffic signs.

[0343] S6012. Add the trajectory curvature distribution features and associated features to the feature vector.

[0344] Specifically, the trajectory curvature distribution features and related features are added to the feature vector, so that the feature vector adds information about trajectory shape and relationship with traffic signs on the basis of the original features, providing a more complete basis for subsequent matching.

[0345] Among them, the trajectory curvature distribution feature is used to characterize the curvature change of the effective trajectory in different arc length segments. Specifically, it can be expressed by calculating the curvature of discrete sampling points of the trajectory and statistically analyzing the frequency, peak value, mean or variance of each curvature interval.

[0346] The correlation features between trajectories and traffic signs are used to describe the spatial and rule-based relationships between valid trajectories and traffic signs. Specifically, based on the vehicle's current positioning results, traffic sign detection results, and trajectory point sets, it can be calculated whether the trajectory covers the area affected by traffic signs and whether it has a constraint relationship with signs related to speed limits, no-entry, yielding, or turn instructions. These features can be jointly generated from trajectory point sequences, road topology information, and perception results, and participate in subsequent intent label matching in the form of numerical vectors.

[0347] In practical implementation, the effective trajectory can be resampled at equal intervals to obtain the trajectory point coordinate sequence. Then, the local curvature can be calculated based on the heading changes and arc length intervals of adjacent points, and the curvature values ​​of each sampled segment can be encoded into curvature distribution vectors according to preset intervals. At the same time, the spatial coordinates, scope of action and category information of traffic signs can be combined to determine the correlation between the trajectory and various types of traffic signs, and the judgment result can be encoded into an association feature vector.

[0348] Subsequently, the curvature distribution features and correlation features are concatenated or expanded with the original trajectory features along the same dimension to form a complete feature vector, which is then input into the intent clustering module. Using this method, the feature vector not only contains information about the trajectory's geometric changes but also information about the trajectory's response to traffic sign constraints, making the intent label matching results closer to road rules and driving intentions.

[0349] In some alternative implementations, after S602 above, the following is also included:

[0350] S6021. If the associated feature indicates a trajectory that conflicts with a traffic sign, the intent label of the corresponding valid trajectory will be determined as an invalid intent.

[0351] Specifically, invalid intent refers to an intent label that is determined to be invalid due to a conflict between the trajectory and traffic signs. If the associated feature indicates that the trajectory conflicts with a traffic sign, the intent label of the corresponding valid trajectory will be determined as invalid intent, and the trajectory that conflicts with the traffic sign will not be processed further.

[0352] Among them, the feature vector is used to characterize the motion pattern of the effective trajectory and its relationship with road semantic information, the preset intent label is used to represent the candidate driving intent corresponding to the effective trajectory, and the invalid intent is used to mark the intent label corresponding to the trajectory that conflicts with traffic signs, specifically by excluding it from the normal candidate intent set.

[0353] The associated characteristics can be composed of the spatial relationship between the trajectory and traffic signs, the correspondence between the trajectory's direction of travel and the meaning of the signs, and whether there are any behaviors such as crossing boundaries, running away, or violating instructions in the trajectory's time sequence.

[0354] In practical processing, feature vectors can be generated first based on the trajectory point sequence of the effective trajectory, and traffic sign categories, locations and effective ranges can be extracted from map elements, perception results or vehicle recognition results. Then, the correlation features between the trajectory and traffic signs can be calculated.

[0355] When the associated features indicate that the valid trajectory crosses the area under the influence of a no-entry sign, travels in the opposite direction to the meaning of the sign, or enters a lane that is not allowed to pass under the constraints of a directional sign, the trajectory is determined to conflict with the traffic sign, and the intent label originally matched by the trajectory is set to invalid intent.

[0356] This process can be completed by the processor according to a preset rule table. The rule table stores the conflict determination conditions corresponding to different traffic sign types in advance. When the processor reads that the conflict conditions are met, it outputs an invalid intent flag.

[0357] This invention extracts feature vectors of valid trajectories and matches them with preset intent labels. Based on the matching results, the valid trajectories are clustered to obtain multiple candidate intent sets. When the associated feature indicates that the trajectory conflicts with a traffic sign, the intent label is determined to be an invalid intent. This allows trajectories with different driving intentions to be distinguished first, and trajectories that conflict with traffic signs to be excluded, thereby improving the accuracy of subsequent processing.

[0358] Figure 7 Flowchart of the vehicle motion control method provided in this application Figure 7 ,like Figure 7 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0359] S701. Calculate the smoothing value and style matching value of each valid trajectory in the target set.

[0360] Specifically, the smoothness value is a numerical value used to evaluate the smoothness of the trajectory. The style matching value is a numerical value used to evaluate the degree of matching between the trajectory and the preset driving style. The smoothness value and style matching value of each valid trajectory in the target set are calculated to provide data for subsequent weighted summation.

[0361] S702. The smoothness value is obtained by weighted summation of the smoothness value and style matching value according to the preset harmonic coefficient.

[0362] Specifically, the preset harmonic coefficient refers to a pre-set coefficient used to weight and sum the smoothing value and the style matching value. The smoothing value is the value obtained by weighting and summing the smoothing value and the style matching value according to the preset harmonic coefficient.

[0363] The smoothness value is obtained by weighted summation of the smoothness value and style matching value according to the preset harmonic coefficient, so that the smoothness of the trajectory and the style matching degree participate in the calculation according to their respective coefficients.

[0364] S703. The effective trajectory with the smallest smoothness value is determined as the target trajectory.

[0365] Specifically, the target trajectory refers to the trajectory determined after smoothness screening and used for output to the vehicle control module. The effective trajectory with the smallest smoothness value is determined as the target trajectory, so that the target trajectory meets the requirements in the smoothness dimension.

[0366] Among them, the smoothness value is used to characterize the smoothness of the trajectory motion. Its calculation basis can be the comprehensive smoothness cost of the trajectory curvature change, acceleration change and jerk change. The style matching value is used to characterize the consistency between the trajectory and the preset driving style. Its calculation basis can be the square of the difference between the average acceleration of the effective trajectory and the style reference acceleration.

[0367] The harmonic coefficient is used to balance the proportion of the two types of indicators in the comprehensive evaluation, and can be preset as a fixed weight according to the current vehicle control mode, road condition type or comfort constraints.

[0368] In practical implementation, the discrete trajectory point sequence can be obtained by sampling each valid trajectory in the target set first. Then, the smoothness cost is calculated based on the changes in heading angle, velocity and curvature of adjacent trajectory points, and the normalized cost is used as the smoothing value. At the same time, the longitudinal acceleration sequence corresponding to the trajectory is extracted, the average acceleration is calculated and the difference is calculated with the preset style reference acceleration. The squared difference is used as the style matching value.

[0369] Then, the smoothing value and style matching value are weighted and summed according to the harmonic coefficient to obtain the smoothness value corresponding to each valid trajectory, and the smallest smoothness value is selected as the target trajectory.

[0370] The smoothness value calculation can be completed by the trajectory scoring module in the vehicle controller. The scoring module can be run in real time using an on-board processor combined with a parameter table in memory. In practical applications, other models of this component can also be selected, and this application does not limit this.

[0371] The above processing allows effective trajectories in the target set to undergo a secondary screening based on motion smoothness and driving style consistency after satisfying the predetermined intent constraints, thereby outputting target trajectories that better meet the current control requirements of the vehicle. Since smoothness values ​​and style matching values ​​are uniformly mapped to smoothness values, and only the magnitude of the smoothness values ​​is compared in the end, the trajectory optimization results have clear evaluation criteria, and the output trajectory maintains consistency between control response and ride comfort.

[0372] In some alternative implementations, S701 above includes:

[0373] S7011. Obtain the longitudinal and lateral jerk of each valid trajectory.

[0374] Specifically, longitudinal jerk refers to the derivative of the vehicle's longitudinal acceleration with respect to time. Lateral jerk refers to the derivative of the vehicle's lateral acceleration with respect to time.

[0375] The longitudinal and lateral jerk of each valid trajectory are obtained to provide data for subsequent calculation of the smoothing value.

[0376] S7012. Discretely sum the squares of the longitudinal jerk and the lateral jerk to obtain a smooth value.

[0377] Specifically, discrete summation refers to the process of accumulating and summing the values ​​at discrete trajectory points. Discrete summation is performed on the sum of the squares of the longitudinal and lateral jerk to obtain a smoothed value, which is then determined based on the accumulated result of the acceleration variations along the trajectory.

[0378] Among them, longitudinal jerk is used to characterize the rate of change of acceleration of the effective trajectory in the driving direction, and lateral jerk is used to characterize the rate of change of acceleration of the effective trajectory in the lateral offset direction. Both can be calculated based on the velocity and acceleration sequences corresponding to discrete sampling points of the trajectory. The sampling points can correspond to equally spaced moments within the planning period, and the sampling interval is consistent with the vehicle control cycle to match the execution rhythm of the control module.

[0379] In actual calculations, the longitudinal jerk sequence and the lateral jerk sequence of each discrete point can be extracted according to the sampling time of the trajectory. Then, the longitudinal jerk and the lateral jerk at the same sampling point are squared and added together to form the instantaneous smoothing cost of the sampling point. Subsequently, the instantaneous smoothing costs of all sampling points are discretely accumulated to obtain the smoothing value of the corresponding effective trajectory.

[0380] Discrete summation can be performed using fixed sampling weights or by normalizing according to the sampling interval to ensure that the smooth values ​​between trajectories of different lengths are comparable.

[0381] In one feasible approach, the trajectory smoothing value is calculated by the trajectory planning module after resampling the candidate trajectories in the target set. The resampled trajectory points are stored in array form in chronological order. The calculation unit sequentially reads the jerk data in the array and performs a sum of squares, with the output smoothing value serving as the input parameter for subsequent smoothness filtering. Since the smoothing value directly reflects the intensity of the trajectory jerk change, the smaller the value, the more continuous the acceleration change in the longitudinal and lateral directions, and the more suitable the trajectory is for inclusion in the target trajectory filtering results.

[0382] In some alternative implementations, S7012 above includes:

[0383] S70121. Obtain the spatiotemporal weighting factor, which includes the weight of the starting segment and the weight of the middle and later segments.

[0384] Specifically, the spatiotemporal weighting factor refers to the factor used to assign different weights to the sum of squared jerk at different positions of the trajectory, including the weight of the starting segment and the weight of the middle and later segments.

[0385] The starting segment weight refers to the weight assigned to the sum of squares of the accelerations of the starting segment of the trajectory.

[0386] The weighting of the middle and later segments refers to the weight assigned to the sum of squared jerkes in the middle and later segments of the trajectory. The spatiotemporal weighting factor is obtained to provide a basis for subsequent position-weighted summation.

[0387] S70122. Divide the trajectory points in the effective trajectory into the starting segment and the middle and rear segments according to the preset ratio.

[0388] Specifically, the preset ratio refers to the pre-defined proportion used to divide the trajectory points into the starting segment and the middle and later segments. Dividing the trajectory points in the valid trajectory into the starting segment and the middle and later segments according to the preset ratio provides a basis for subsequent segmented weighted summation.

[0389] S70123. The sum of the squares of the starting segment and the middle and later segments is weighted according to the weight of the starting segment and the weight of the middle and later segments to obtain the smoothed value.

[0390] Specifically, the sum of the squares of the starting segment and the middle and later segments is weighted according to the weights of the starting segment and the middle and later segments to obtain the smoothed value, so that the starting segment and the middle and later segments participate in the smoothed value calculation according to their respective weights.

[0391] Among them, discrete summation is used to accumulate the smoothness cost on the trajectory point sequence, spatiotemporal weighting factor is used to assign different weights to trajectory points at different times and spatial locations, starting segment weight is used to increase the smoothness cost weight near the starting point of the trajectory, and middle and later segment weight is used to adjust the smoothness cost weight in the middle and later segments of the trajectory.

[0392] By dividing the effective trajectory into a starting segment and a middle and later segment according to a preset ratio, and then accumulating the sum of squares of the two segments by applying corresponding weights, the abrupt changes in the initial stage and the overall smoothness of the subsequent stage can be included in the same smoothing value calculation.

[0393] In the specific implementation, a sequence of trajectory points is first generated based on discrete sampling points of the effective trajectory. Then, the coverage area of ​​the starting segment is determined according to the first trajectory point corresponding to the vehicle's current position. The preset ratio can be set as the proportion of the front segment and the proportion of the rear segment according to the control sensitive interval within the planning period. The weights of the starting segment and the middle and rear segments can be given by a fixed coefficient table, or obtained by mapping vehicle speed, road curvature, or traffic constraints.

[0394] After squaring the longitudinal and lateral accelerometer velocities of the trajectory points, the values ​​are accumulated in the initial and middle / rear segments respectively, multiplied by their corresponding weights, and then summed to obtain a smoothing value. A smaller smoothing value indicates a smoother acceleration change in the segmented trajectory, making it suitable for subsequent smoothness screening. In practical applications, other combinations of weighting coefficients can be selected, and this application does not impose any limitations on this.

[0395] By adopting the above method, the calculation of the smoothing value retains the differentiated constraints between the initial stage and the middle and later stages of the trajectory, enabling the segmented weighted results to more accurately reflect the smoothness characteristics during vehicle start-up and continuous driving, thereby providing a stable evaluation basis for target trajectory selection. Since the smoothing value is obtained through discrete summation, it directly corresponds to the trajectory sampling points, facilitating real-time calculation and output in vehicle motion control.

[0396] In some optional implementations, the above-mentioned S701 further includes:

[0397] S7013, Obtain the average acceleration of the preset driving style and each effective trajectory.

[0398] Specifically, the preset driving style refers to the pre-set driving style used to determine the style reference acceleration. The average acceleration refers to the average value of the acceleration at each point on the trajectory.

[0399] The average acceleration of the preset driving style and each effective trajectory is obtained to provide data for subsequent determination of style reference acceleration and calculation of style matching value.

[0400] S7014. Match the preset driving style with the preset style reference acceleration mapping table to obtain the style reference acceleration.

[0401] Specifically, the preset style reference acceleration mapping table refers to a pre-set mapping table used to determine the style reference acceleration based on the preset driving style.

[0402] Style reference acceleration refers to the reference acceleration determined based on the preset driving style and used to calculate the style matching value.

[0403] The preset driving style is matched with the preset style reference acceleration mapping table to obtain the style reference acceleration, so that the style reference acceleration is determined according to the preset driving style.

[0404] S7015. Calculate the square of the difference between the average acceleration and the style reference acceleration to obtain the style matching value.

[0405] Specifically, the square of the difference between the average acceleration and the style reference acceleration is calculated to obtain the style matching value, which is then determined based on the deviation between the trajectory's average acceleration and the style reference acceleration.

[0406] The preset driving style defines the vehicle's driving preference mode, which may include styles such as smooth, normal, or aggressive. A preset style reference acceleration mapping table establishes the correspondence between driving styles and reference accelerations; this mapping table can be obtained from historical driving data and pre-stored in the vehicle controller or trajectory planning module. The average acceleration of each valid trajectory characterizes the overall longitudinal acceleration of the trajectory, typically obtained by averaging the longitudinal acceleration at each sampling time along the trajectory. The style reference acceleration serves as a benchmark value corresponding to the driving style and is used for comparison with the trajectory's average acceleration.

[0407] In its implementation, after obtaining each valid trajectory from the target set, the trajectory planning module first reads the corresponding style reference acceleration based on the currently selected preset driving style. Then, it extracts the longitudinal acceleration sequence from the temporal sampling points of the valid trajectories and performs an arithmetic mean on the sequence to obtain the average acceleration of the valid trajectory. Next, it performs a difference calculation between the average acceleration and the style reference acceleration, and squares the difference to obtain the style matching value.

[0408] If the average acceleration of a valid trajectory is closer to the reference acceleration corresponding to the driving style, its style matching value is smaller; if the deviation is large, the style matching value is increased accordingly. In practical applications, the mapping table can also be stored in the form of a lookup table index, a hash index, or database records; this application does not limit this.

[0409] In this way, the style matching value can reflect the degree of fit between the effective trajectory and the preset driving style, and is used in conjunction with the smoothness value for smoothness screening, so that the target trajectory can maintain the acceleration characteristics consistent with the predetermined driving preferences while meeting the vehicle control requirements, thereby improving the consistency of trajectory results and scene adaptability.

[0410] This invention calculates the smoothness value and style matching value of each effective trajectory in the target set, and obtains the smoothness value by weighting and summing them according to a preset harmonic coefficient. The effective trajectory with the smallest smoothness value is determined as the target trajectory, and the trajectory is filtered under the same driving intention. This makes the determination of the target trajectory take into account both the smoothness of the trajectory and the degree of matching with the driving style, thereby improving the accuracy of the target trajectory determination.

[0411] Figure 8 Flowchart of the vehicle motion control method provided in this application Figure 8 ,like Figure 8 As shown, this embodiment, based on the above embodiments, provides a detailed description of the vehicle motion control method, which includes:

[0412] S801. Obtain current scene information, including traffic light status and remaining distance of navigation path.

[0413] Specifically, current scene information refers to information reflecting the vehicle's current location, including traffic light status and remaining distance on the navigation path. Obtaining current scene information provides a basis for subsequent adjustments to preset weights.

[0414] S802. Adjust the preset weights of each factor according to the current scene information.

[0415] Specifically, preset weights refer to the pre-defined weights used to perform a weighted summation of the values ​​of each factor. Adjusting the preset weights of each factor based on the current scenario information allows the participation of each factor in the weighted summation to vary with the current scenario, thereby making the calculated value more closely match the evaluation requirements of the current scenario.

[0416] The current scenario information characterizes the combined state of the road environment and navigation constraints in which the vehicle is located. The traffic light status indicates the current traffic control conditions and reflects different signal constraints such as red, green, or yellow lights. The remaining distance of the navigation path characterizes the remaining driving distance of the vehicle from the navigation endpoint or critical path point. The preset weights of each factor characterize the influence of driver intention factors, navigation factors, interaction factors, efficiency factors, and smoothness-related factors in the comprehensive decision-making process. The weights can be pre-stored by the controller and updated according to changes in the scenario during operation.

[0417] In the actual implementation, after receiving the traffic light status and the remaining distance of the navigation path, the controller first encodes the two for the scene, and then determines the weight adjustment rules according to the preset mapping relationship.

[0418] When the traffic light status indicates that there is a red light constraint ahead and the remaining distance of the navigation path is short, the system increases the weight of factors related to compliance and navigation objectives, and decreases the weight of factors related to efficiency.

[0419] When the traffic light indicates green and the remaining distance of the navigation path is relatively long, the system increases the weight of the efficiency factor and the interaction factor, making the trajectory selection more focused on the continuity of traffic.

[0420] When the remaining distance approaches the preset threshold and the traffic light status is in the switching phase, the system simultaneously increases the weight of the navigation factor and the smoothness-related factor to ensure that the target trajectory maintains a more stable control output as it approaches the destination or intersection.

[0421] The aforementioned weight adjustment results serve as input parameters for subsequent weighted summation, directly impacting the calculation process of each candidate intent set.

[0422] In terms of working principle, the scene perception weighting mechanism introduces traffic control constraints and navigation process into the factor weighting process, so that the evaluation focus of different road conditions changes synchronously with the scene. Based on this, the controller performs differentiated evaluation of the candidate intent set and outputs a target trajectory that is more in line with the current road traffic requirements.

[0423] By adopting this specific implementation method, the system can dynamically adjust the influence ratio of each factor when traffic lights change and navigation progress changes, so that the trajectory decision is consistent with the current road constraints, thereby improving the scenario adaptability and control stability of the weighted evaluation results.

[0424] This invention improves the adaptability of the calculated value to the evaluation of the candidate intent set by acquiring current scene information and adjusting the preset weights of each factor accordingly, so that the participation of each factor in the weighted summation changes with the current scene.

[0425] In some optional implementations, a driving mode signal is acquired, the driving mode signal is matched with multiple preset driving styles to obtain a preset driving style, and a preset style reference acceleration mapping table is queried according to the preset driving style to obtain the style reference acceleration.

[0426] Specifically, the driving mode signal refers to the signal reflecting the driving mode selected by the driver. The preset driving style refers to a pre-set driving style used to determine the style reference acceleration. The preset style reference acceleration mapping table refers to a pre-set mapping table used to determine the style reference acceleration based on the preset driving style. Style reference acceleration refers to the reference acceleration determined based on the preset driving style and used to calculate the style matching value. By acquiring the driving mode signal and matching it with multiple preset driving styles, the currently used preset driving style is obtained. Based on this, the preset style reference acceleration mapping table is consulted to obtain the style reference acceleration. This ensures that the calculation of the style matching value is based on the driving mode selected by the driver, thereby ensuring that the determination of the target trajectory is consistent with the driving mode.

[0427] This invention obtains a preset driving style by acquiring a driving mode signal and matching it with multiple preset driving styles. Based on this, a style reference acceleration is obtained by querying a mapping table. This allows the calculation of the style matching value to vary with the driving mode selected by the driver, thereby improving the consistency between the target trajectory determination and the driving mode.

[0428] In some optional implementations, current environmental perception information is acquired, the driving scenario type is determined based on the current environmental perception information, and the style reference acceleration is adjusted based on the driving scenario type.

[0429] Specifically, current environmental perception information refers to information collected by the perception module that reflects the current environmental conditions of the vehicle. Driving scenario type refers to information determined based on the current environmental perception information that reflects the scenario category in which the vehicle is located. Style reference acceleration refers to reference acceleration determined based on a preset driving style and used to calculate the style matching value. By acquiring current environmental perception information and determining the driving scenario type accordingly, and then adjusting the style reference acceleration, the calculation of the style matching value changes with the driving scenario type, thereby matching the determination of the target trajectory with the current driving scenario.

[0430] This invention improves the adaptability of target trajectory determination to driving scenarios by acquiring current environmental perception information and determining the driving scenario type accordingly, and then adjusting the style reference acceleration so that the calculation of style matching value can change with the driving scenario type.

[0431] Figure 9 A schematic diagram of the vehicle motion control device provided in this application is shown below. Figure 9 As shown, the vehicle motion control device 90 provided in this embodiment includes:

[0432] The acquisition module 901 is used to acquire multiple candidate trajectories, operating parameters, and navigation data of the vehicle.

[0433] The verification module 902 is used to verify the candidate trajectory and obtain multiple valid trajectories;

[0434] Clustering module 903 is used to cluster valid trajectories to obtain multiple candidate intent sets, each candidate intent set corresponding to an intent label;

[0435] Analysis module 904 is used to determine the values ​​of each candidate intent set on multiple factors based on intent labels, operation parameters, navigation data, and valid trajectories in each candidate intent set;

[0436] The analysis module 904 is also used to perform weighted summation of the values ​​according to the preset weights corresponding to each factor to obtain the calculated value of each candidate intent set;

[0437] Module 905 is used to determine the set of candidate intents with the highest calculated values ​​as the target set;

[0438] The filtering module 906 is used to perform smoothness filtering on the valid trajectories in the target set to determine the target trajectory;

[0439] The control module 907 is used to output the target trajectory to the vehicle control module to control the vehicle's movement.

[0440] In one possible embodiment, the verification module 902 is specifically used for:

[0441] The candidate trajectories are subjected to security checks to obtain multiple safe trajectories;

[0442] The safety trajectory was verified for compliance, resulting in multiple valid trajectories.

[0443] In one possible embodiment, the verification module 902 is further specifically used for:

[0444] Obtain the predicted motion trajectories of dynamic obstacles around the vehicle;

[0445] Collision verification is performed between the candidate trajectory and the motion prediction trajectory to obtain multiple safe trajectories.

[0446] In one possible embodiment, the verification module 902 is further specifically used for:

[0447] Obtain the collision time between the candidate trajectory and the predicted motion trajectory;

[0448] Candidate trajectories with collision times greater than a preset time threshold are extracted to obtain multiple safe trajectories.

[0449] In one possible embodiment, the verification module 902 is further specifically used for:

[0450] Get the lane markings of the lane where the vehicle is located and the status of the traffic lights in front of the vehicle;

[0451] The safety trajectory is verified based on lane markings and traffic light status, resulting in multiple valid trajectories.

[0452] In one possible embodiment, the verification module 902 is further specifically used for:

[0453] Determine whether the safe trajectory involves crossing solid lane lines or violating traffic light status;

[0454] If a valid track exists, delete the corresponding safe track to obtain multiple valid tracks.

[0455] In one possible embodiment, the analysis module 904 is specifically used for:

[0456] Driver operation signals and navigation expectations are obtained based on operating parameters and navigation data;

[0457] The intent labels are matched with the driver's operation signals and navigation expectations respectively to obtain the matching results;

[0458] The values ​​of the driver intent factor and navigation factor are obtained by querying the preset mapping table based on the matching results.

[0459] In one possible embodiment, the analysis module 904 is further specifically used for:

[0460] The total deceleration of surrounding vehicles during the execution of the effective trajectory is predicted to obtain a predicted value, and the negative value of the predicted value is taken as the value of the interaction factor.

[0461] Calculate the ratio of the average speed of the effective trajectory to the road speed limit, and use this ratio as the value of the efficiency factor.

[0462] In one possible embodiment, clustering module 903 is specifically used for:

[0463] Extract the feature vector of the effective trajectory;

[0464] The feature vectors are matched with preset intent labels to obtain the intent labels corresponding to each valid trajectory.

[0465] Clustering of valid trajectories based on intent labels yields multiple candidate intent sets.

[0466] In one possible embodiment, the clustering module 903 is further specifically used for:

[0467] Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs;

[0468] Add the trajectory curvature distribution features and associated features to the feature vector.

[0469] In one possible embodiment, the clustering module 903 is further specifically used for:

[0470] If the associated feature indicates a trajectory that conflicts with a traffic sign, the intent label of the corresponding valid trajectory will be determined as an invalid intent.

[0471] In one possible embodiment, the filtering module 906 is specifically used for:

[0472] Calculate the smoothing value and style matching value of each valid trajectory in the target set;

[0473] The smoothness value is obtained by weighted summation of the smoothness value and style matching value based on the preset harmonic coefficient;

[0474] The effective trajectory with the smallest smoothness value is determined as the target trajectory.

[0475] In one possible embodiment, the filtering module 906 is further specifically used for:

[0476] Obtain the longitudinal and lateral jerk of each valid trajectory;

[0477] Discretely sum the squares of the longitudinal and lateral jerks to obtain a smooth value.

[0478] In one possible embodiment, the filtering module 906 is further specifically used for:

[0479] Obtain the spatiotemporal weighting factor, which includes the weight of the starting segment and the weight of the middle and later segments;

[0480] The trajectory points in the effective trajectory are divided into the starting segment and the middle and rear segments according to a preset ratio;

[0481] The smoothed value is obtained by weighting the sum of the squares of the starting segment and the middle and later segments according to the weights of the starting segment and the middle and later segments.

[0482] In one possible embodiment, the filtering module 906 is further specifically used for:

[0483] Obtain the average acceleration of the preset driving style and each effective trajectory;

[0484] Match the preset driving style with the preset style reference acceleration mapping table to obtain the style reference acceleration;

[0485] The style matching value is obtained by calculating the square of the difference between the average acceleration and the style reference acceleration.

[0486] In one possible embodiment, the analysis module 904 is further configured to:

[0487] Obtain current scene information, including traffic light status and remaining distance of the navigation path;

[0488] Adjust the preset weights of each factor based on the current scenario information.

[0489] The vehicle motion control device 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.

[0490] Furthermore, the vehicle motion control device provided in this application, as a core module of the intelligent driving domain controller planning stack, is integrated into the vehicle's electronic and electrical architecture. The system interacts with the upstream perception module and the downstream tracking controller via shared memory or inter-process communication (IPC) interfaces. The overall architecture supports modular configuration, and the processing thresholds and weight coefficients at each level can be dynamically adjusted according to vehicle parameters, road types, and user preferences.

[0491] In an urban road intersection scenario, a vehicle approaches the intersection at 60 km / h, with pedestrians crossing, oncoming right-turning vehicles cutting in, and navigation instructions indicating a left turn. After the trajectory generation module outputs multiple candidate trajectories, the system performs a trajectory continuity check. Based on the current vehicle speed of 60 km / h, it uses piecewise linear interpolation to set a lateral position continuity threshold of 2.5 meters. If the lateral offset of the candidate trajectory's starting point exceeds this threshold, it is discarded. Subsequently, a vehicle dynamics limit check is performed, setting a lateral acceleration threshold of 4 m / s², and trajectories exceeding this value are discarded. Static obstacle collision checks are then performed on the pre-screened trajectories, employing a two-stage detection strategy: first, a rapid intersection test is performed based on axis-aligned bounding boxes to filter out potentially overlapping trajectory-obstacle pairs; then, the separated axis theorem (SAT) is used to perform precise geometric calculations on the retained pairs, and a preset safety margin is introduced to output a safe trajectory. In the compliance verification stage, the system deletes trajectories that cross solid lane lines or violate traffic light instructions based on lane markings and traffic light status, resulting in a valid trajectory set.

[0492] The system extracts feature vectors from valid trajectories for intent clustering. To improve the completeness of feature representation, curvature continuity and acceleration continuity features are further introduced as supplementary indicators in the continuity check. The curvature difference and acceleration difference between the trajectory start point and the vehicle's current position are calculated. If the difference exceeds the corresponding dynamic threshold, the trajectory is determined to be discontinuous. After clustering, each candidate intent set enters the factor evaluation stage. The safety factor weight is configured as 3.0, and the navigation factor weight is configured as 1.5. The system assesses the interaction risk by calculating the minimum time-to-collision (TTC) between each trajectory and dynamic obstacles. For example, the TTC for a straight-ahead intent is 1.8 seconds, for a left-turn intent it is 1.2 seconds, and for a right-turn avoidance intent it is 2.0 seconds. At the intent decision layer, the system prioritizes the intents with the highest safety factor scores that meet navigation constraints as the target intent set.

[0493] During the trajectory comfort selection phase, the system calculates the smoothness value and style matching value of the effective trajectory. The smoothness value is calculated using a spatiotemporal weighted summation method, with a weighting coefficient of 1.5 for the initial segment and 0.8 for the middle and later segments to enhance the suppression of acceleration changes during the initial acceleration phase. The style matching value is calculated based on a preset driving style matching reference acceleration: 0.2g when the vehicle switches to Comfort mode and 0.3g when switching to Sport mode. The system weights and sums the smoothness value and style matching value according to preset harmonic coefficients to obtain the smoothness value, and then selects the trajectory with the minimum smoothness value as the target trajectory and outputs it to the control module.

[0494] To enhance decision-making robustness in complex interaction scenarios, the system introduces dynamic obstacle pre-screening logic during the pre-screening stage. Trajectories with a minimum time-to-travel (TTC) of less than 2.0 seconds compared to the predicted path of the dynamic obstacle are directly eliminated. Simultaneously, a dynamic obstacle trajectory prediction and avoidance mechanism is introduced, assigning higher priority to trajectories with a minimum TTC greater than 3.0 seconds compared to the predicted path of the obstacle. The system further deploys a real-time feedback module, collecting actual interaction results and riding experience after the target trajectory is executed, and inputting the feedback data back into the intent decision layer to dynamically adjust the weights of various factors. The context awareness module reconstructs intent priorities in real time based on traffic light status and the remaining distance of the navigation path. When the remaining distance is short and the system is approaching a traffic light switching phase, it automatically increases the influence ratio of navigation and compliance factors. Furthermore, the system can access current environmental perception information to determine the driving scenario type. When rain or slippery road surfaces are identified, it automatically reduces style reference acceleration to ensure the stability and scenario adaptability of vehicle motion control.

[0495] Figure 10 A schematic diagram of the structure of the electronic device provided in this application. Figure 10 As shown, the electronic device 100 provided in this embodiment includes at least one processor 1001 and a memory 1002. Optionally, the electronic device 100 further includes a communication component 1003. The processor 1001, memory 1002, and communication component 1003 are connected via a bus 1004.

[0496] In the specific implementation process, at least one processor 1001 executes computer execution instructions stored in memory 1002, causing at least one processor 1001 to perform the above method.

[0497] The specific implementation process of processor 1001 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0498] This electronic device can be deployed in the planning and control unit of an autonomous vehicle. The memory 1002 is pre-written with computer execution instructions such as trajectory verification, intent clustering, factor weighting, and ride comfort filtering. After receiving multi-source perception data, positioning information, and navigation information, the processor 1001 first verifies the validity of candidate trajectories, then assigns the verified trajectories to the target intent set, and calculates the weights of each trajectory factor in conjunction with vehicle operating parameters and navigation constraints, thereby completing weighted evaluation and ride comfort filtering, and finally outputting the target trajectory and corresponding control instructions. Because invalid trajectories are eliminated in advance, and intent judgment and comfort evaluation are performed in stages, redundant calculations are reduced and real-time performance is improved. Furthermore, the clear boundaries of each processing stage make the trajectory selection process easier to interpret, thereby improving the stability of vehicle motion control and adaptability to complex scenarios.

[0499] 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.

[0500] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0501] 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.

[0502] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0503] This computer program product encapsulates the processing flow of trajectory verification, intent clustering, factor value and weighting, and smoothness screening into an executable program. Deployed in the vehicle planning and control system, the processor completes the layer-by-layer processing of candidate trajectories according to preset steps. Because the program first eliminates obviously invalid trajectories during execution, then performs intent set judgment on the valid trajectories, and combines operational parameters and navigation information for weighted screening, the trajectory decision-making chain is transformed from a hybrid evaluation to a phased processing. This reduces invalid calculations entering subsequent stages, improving real-time performance and stability; at the same time, it makes the boundaries of intent judgment, comfort evaluation, and control output clearer, thus obtaining more interpretable and adaptable trajectory control results in urban roads, intersections, or highway scenarios.

[0504] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0505] 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.

[0506] 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.

[0507] 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.

[0508] 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.

[0509] 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.

[0510] 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.

[0511] 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.

[0512] 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.

Claims

1. A vehicle motion control method, characterized in that, include: Acquire multiple candidate trajectories, operating parameters, and navigation data for the vehicle; The candidate trajectories are validated to obtain multiple valid trajectories; The valid trajectories are clustered to obtain multiple candidate intent sets, each of which corresponds to an intent label. Based on the intent tags, operation parameters, and navigation data, as well as the valid trajectories in each candidate intent set, determine the values ​​of each candidate intent set on multiple factors; The values ​​are weighted and summed according to the preset weights corresponding to each factor to obtain the calculated value of each candidate intent set; The set of candidate intents with the highest calculated values ​​is determined as the target set; Smoothness screening is performed on the valid trajectories in the target set to determine the target trajectory; The target trajectory is output to the vehicle control module to control the vehicle's movement.

2. The method according to claim 1, characterized in that, Validating the candidate trajectory includes: The candidate trajectories are subjected to security verification to obtain multiple secure trajectories; The safety trajectory is verified for compliance, resulting in multiple valid trajectories.

3. The method according to claim 2, characterized in that, The candidate trajectory is subjected to security verification, including: Obtain the predicted motion trajectories of dynamic obstacles around the vehicle; The candidate trajectory is compared with the motion prediction trajectory to obtain multiple safe trajectories.

4. The method according to claim 3, characterized in that, The collision verification between the candidate trajectory and the predicted motion trajectory includes: Obtain the collision time between the candidate trajectory and the predicted motion trajectory; Candidate trajectories with collision times greater than a preset time threshold are extracted to obtain multiple safe trajectories.

5. The method according to claim 2, characterized in that, The compliance verification of the security trajectory includes: Obtain the lane markings of the lane where the vehicle is located and the status of the traffic lights in front of the vehicle; The safety trajectory is verified based on the lane markings and traffic light status to obtain multiple valid trajectories.

6. The method according to claim 5, characterized in that, The safety trajectory is verified based on the lane markings and traffic light status, including: Determine whether the safety trajectory involves crossing solid lane lines or violating traffic light status. If a valid track exists, delete the corresponding safe track to obtain multiple valid tracks.

7. The method according to claim 1, characterized in that, Determine the values ​​of each candidate intent set on multiple factors, including: Based on the operating parameters and navigation data, obtain the driver's operating signals and navigation expectations; The intent labels are matched with the driver's operation signals and navigation expectations respectively to obtain the matching results; The values ​​of the driver intent factor and navigation factor are obtained by querying the preset mapping table based on the matching results.

8. The method according to claim 7, characterized in that, Also includes: The total deceleration magnitude of surrounding vehicles during the execution of the effective trajectory is predicted to obtain a predicted value, and the negative value of the predicted value is taken as the value of the interaction factor. Calculate the ratio of the average speed of the effective trajectory to the road speed limit, and use the ratio as the value of the efficiency factor.

9. The method according to claim 1, characterized in that, Intent clustering is performed on the valid trajectories, including: Extract the feature vector of the effective trajectory; The feature vectors are matched with preset intent labels to obtain the intent labels corresponding to each valid trajectory; Based on the intent tags, the valid trajectories are clustered to obtain multiple candidate intent sets.

10. The method according to claim 9, characterized in that, The step of extracting the feature vector of the effective trajectory further includes: Extract the trajectory curvature distribution features and the correlation features between the trajectory and traffic signs; The trajectory curvature distribution features and the associated features are added to the feature vector.

11. The method according to claim 10, characterized in that, The step of matching the feature vector with a preset intent label further includes: If the associated feature indicates a trajectory that conflicts with a traffic sign, then the intent label of the corresponding valid trajectory will be determined as an invalid intent.

12. The method according to claim 1, characterized in that, Smoothness filtering of valid trajectories in the target set includes: Calculate the smoothing value and style matching value of each valid trajectory in the target set; The smoothness value is obtained by weighted summation of the smoothness value and style matching value based on the preset harmonic coefficient; The effective trajectory with the smallest smoothness value is determined as the target trajectory.

13. The method according to claim 12, characterized in that, Calculating the smoothing value of each valid trajectory in the target set includes: Obtain the longitudinal and lateral jerk of each valid trajectory; Discretely sum the squares of the longitudinal jerk and the lateral jerk to obtain a smooth value.

14. The method according to claim 13, characterized in that, The discrete summation includes: Obtain the spatiotemporal weighting factor, which includes the starting segment weight and the middle and later segment weight; The trajectory points in the effective trajectory are divided into a starting segment and a middle and rear segment according to a preset ratio; The smoothed value is obtained by weighting the sum of the squares of the starting segment and the middle and later segments according to the weights of the starting segment and the middle and later segments.

15. The method according to claim 12, characterized in that, Calculate style match values, including: Obtain the average acceleration of the preset driving style and each effective trajectory; Match the preset driving style with the preset style reference acceleration mapping table to obtain the style reference acceleration; The style matching value is obtained by calculating the square of the difference between the average acceleration and the style reference acceleration.

16. The method according to claim 1, characterized in that, Also includes: Obtain current scene information, including traffic light status and remaining distance of the navigation path; Adjust the preset weights of each factor based on the current scene information.

17. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1 to 16.

18. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 16.

Images