Trajectory planning method, device, equipment, vehicle, medium and product

Abstract

The application discloses a trajectory planning method and device, equipment, vehicle, medium and product. The method comprises the following steps: acquiring first environment information of an environment where a target vehicle is located, the first environment information being used for describing a plurality of first traffic participants in the environment where the target vehicle is located; predicting, by a pre-acquired task correlation predictor, based on the first environment information, to obtain a task correlation of each first traffic participant, the task correlation of the first traffic participant being used for representing an influence degree of the first traffic participant on a current driving behavior of the target vehicle; screening the plurality of first traffic participants according to the task correlation of each first traffic participant to obtain a target traffic participant; and performing trajectory planning on the target vehicle according to the target traffic participant to obtain a target trajectory. The first traffic participant with a larger task correlation can be screened out to participate in subsequent trajectory planning on the target vehicle, so that attention to redundant information is reduced, and the accuracy of trajectory planning is improved.

Description

Technical Field

[0001] This application belongs to the field of autonomous driving technology, and in particular relates to a trajectory planning method, device, equipment, vehicle, medium and product. Background Technology

[0002] End-to-end autonomous driving models can be broadly categorized into two types based on whether or not they utilize large language models. Autonomous driving models based on large language models can effectively understand the semantic information contained in images, thus achieving stronger generalization ability and interpretability. However, due to the large number of parameters in these models, deployment costs are high, making them unsuitable for real-time inference on low- to mid-range computing platforms. On the other hand, autonomous driving models that do not employ large language models, compared to traditional multi-segment models, can efficiently transmit potential information from training data, improving trajectory planning accuracy. Moreover, due to their smaller number of parameters, these models can perform real-time inference on low- to mid-range platforms.

[0003] Before performing trajectory planning, intelligent driving models need to understand the input perception data (detection results, feature quantities, etc. output by the perception module) and explore the relationship between the input data and the true trajectory to accurately predict the correct trajectory. However, the input perception results often contain a large amount of information irrelevant to trajectory prediction. This redundant information increases the difficulty for the model to understand the input data and explore the underlying driving behavior logic. If the influence of each object on the current driving behavior can be analyzed before the input perception data, most of the redundant information can be eliminated, allowing the intelligent driving system to focus on important objects and significantly improving the efficiency and performance of trajectory planning. This approach also aligns with human driving habits. For example, human drivers only need to focus on key information in the environment to make quick and accurate judgments. However, current intelligent driving models cannot achieve this, resulting in low accuracy in trajectory planning. Summary of the Invention

[0004] This application provides a trajectory planning method, apparatus, device, vehicle, medium, and product that can improve the accuracy of trajectory planning for target vehicles.

[0005] In a first aspect, embodiments of this application provide a trajectory planning method, the method comprising:

[0006] First environmental information of the environment in which the target vehicle is located is obtained, and the first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located.

[0007] Based on the first environmental information, the task relevance of each first traffic participant is obtained by predicting using a pre-acquired task relevance predictor. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle.

[0008] Multiple first traffic participants are filtered based on the task relevance of each first traffic participant to obtain the target traffic participants;

[0009] The target trajectory is obtained by planning the trajectory of the target vehicle based on the target traffic participants.

[0010] Secondly, embodiments of this application provide a trajectory planning device, the device comprising:

[0011] The acquisition module is used to acquire first environmental information of the environment in which the target vehicle is located. The first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located.

[0012] The prediction module is used to predict based on the first environmental information using a pre-acquired task relevance predictor to obtain the task relevance of each first traffic participant. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle.

[0013] The filtering module is used to filter multiple first traffic participants based on the task relevance of each first traffic participant to obtain the target traffic participants;

[0014] The planning module is used to plan the trajectory of the target vehicle based on the target traffic participants, and obtain the target trajectory.

[0015] Thirdly, embodiments of this application provide an electronic device, including: a processor and a memory storing computer program instructions; the processor executes the computer program instructions to implement the trajectory planning method as described in the first aspect.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the trajectory planning method as described in the first aspect.

[0017] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform the trajectory planning method as described in the first aspect.

[0018] By using the above method, the primary traffic participants with high task relevance can be selected to participate in the subsequent trajectory planning of the target vehicle, thereby reducing the focus on redundant information and improving the accuracy of the target vehicle trajectory planning. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the trajectory planning method provided in an embodiment of this application;

[0021] Figure 2 This is a schematic diagram of the data annotation process provided in an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of the data processing flow provided in the embodiments of this application;

[0023] Figure 4 This is a flowchart illustrating the trajectory planning method provided in an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the trajectory planning device provided in the embodiments of this application;

[0025] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0026] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0027] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0028] In all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. Additionally, when embodiments of this application require access to sensitive personal information, separate permission or consent from the user is obtained through pop-ups or redirects to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments obtained.

[0029] Explanation of the name:

[0030] Multimodal Large Language Models (MLLM);

[0031] Large Language Model (LLM).

[0032] In driving scenarios, the actions of surrounding vehicles or pedestrians are uncertain, and associating this uncertainty with optimizing the vehicle's trajectory requires considerable driving experience. However, this experience is difficult to learn through training AI models. Models can capture the relationship between trajectory changes and driving scenarios through training on large amounts of data. For example, when a vehicle ahead begins to decelerate, the trajectory predicted by the intelligent driving model will shorten. However, the intelligent driving model cannot understand the reason for the vehicle's deceleration and struggles to predict the timing of that deceleration. The root cause of this problem is that the intelligent driving model only learns the potential relationships between trajectory changes and input data in the training dataset, lacking a deep understanding of the logic behind driving behavior.

[0033] This problem leads to a lack of interpretability and efficiency in intelligent driving models. Since the model merely imitates the driving behavior of cars in the dataset, it lacks the ability to explain driving logic and understand the causal relationships between each action and the environment. Even if the intelligent driving model achieves relatively good results, its lack of interpretability makes risk assessment difficult. On the other hand, because the model cannot determine which objects in the environment have a significant impact on current driving behavior, it must analyze and process everything in the current environment to arrive at a conclusion. This is a highly inefficient process. Since only a portion of the objects in the environment actually affect the vehicle's driving, focusing too much on unimportant things not only wastes computing power but also distracts the intelligent driving model, making it difficult to learn human-like driving logic. This results in a lack of deep understanding of driving behavior logic, leading to a lack of interpretability in trajectory planning and low efficiency in the system's cognitive space.

[0034] Human drivers only need to focus on key information in the environment to make quick and accurate judgments. However, current intelligent driving systems cannot do this, resulting in low accuracy in trajectory planning.

[0035] To address the problems of the prior art, embodiments of this application provide a trajectory planning method, apparatus, device, vehicle, medium, and product. The trajectory planning method provided in this application embodiment will be described first below.

[0036] Figure 1 A flowchart illustrating a trajectory planning method provided in one embodiment of this application is shown. Figure 1 As shown, the trajectory planning method provided in this application embodiment is applied to an electronic device, which can be installed on the target vehicle or outside the target vehicle, and the electronic device is communicatively connected to the target vehicle; the trajectory planning method includes the following steps 101-104:

[0037] Step 101: Obtain first environmental information about the environment in which the target vehicle is located. The first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located.

[0038] The trajectory planning method in this application embodiment can be applied in the field of autonomous driving to perform trajectory planning for a target vehicle. The target vehicle can refer to an autonomous vehicle (hereinafter referred to as "autonomous vehicle"). In some application scenarios, the target vehicle can also refer to a non-autonomous vehicle. In this case, the purpose of trajectory planning for the target vehicle can be to assist the driver in driving the vehicle and avoid collision accidents.

[0039] The first traffic participant refers to the object in the environment in which the target vehicle is located. This can be a moving object such as a vehicle or pedestrian, or a fixed object such as a traffic sign or traffic light.

[0040] The initial environmental information can be obtained through sensors installed on the target vehicle, or it can be calculated based on information collected by the sensors on the target vehicle. The sensors collect data at regular intervals.

[0041] The first environmental information may include image information and / or text information. For example, multiple cameras installed on the target vehicle capture images of the environment in which the target vehicle is located. For instance, six cameras are installed on the target vehicle to capture images of the left front, right front, left side, right side, left rear, and right rear of the target vehicle, respectively, resulting in multiple first environmental images. The first environmental information includes these multiple first environmental images. For example, step 101, obtaining the first environmental information of the environment in which the target vehicle is located, includes:

[0042] Multiple first environmental images are obtained by capturing images of the environment in which the target vehicle is located using multiple cameras installed on the target vehicle.

[0043] Multiple first environmental images are processed to obtain multiple descriptive information entries. The first environmental information includes these multiple descriptive information entries, each used to describe a first traffic participant. As described above, by processing multiple first environmental images, multiple descriptive information entries are obtained. Each descriptive information entry may include the first traffic participant's identifier (i.e., ID), type (e.g., vehicle, pedestrian, traffic sign, traffic light), speed, acceleration, relative position, state, etc. Specifically, when the first traffic participant is a target object (e.g., vehicle or pedestrian), its descriptive information may include identifier, type, speed, acceleration, and relative position; when the first traffic participant is a traffic sign, its descriptive information may include identifier, type, and state (e.g., left turn sign, straight ahead sign, right turn sign, one-way street, etc.); when the first traffic participant is a traffic light, its descriptive information may include identifier, type, and state (e.g., whether the light is currently red or green, how many seconds remain before the light changes, etc.).

[0044] By acquiring images of the environment in which the target vehicle is located, multiple first environment images are obtained. These first environment images are then analyzed to obtain multiple descriptive information describing the first traffic participants. This facilitates the subsequent screening of first traffic participants who have a significant impact on the current driving behavior of the target vehicle based on these multiple descriptive information. As a result, when planning the trajectory of the target vehicle, these first traffic participants are given more attention, thereby reducing the focus on redundant information.

[0045] Step 102: Based on the first environmental information, the pre-acquired task relevance predictor makes a prediction to obtain the task relevance of each first traffic participant. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle.

[0046] The task relevance predictor is pre-trained and possesses the ability to predict the task relevance of traffic participants based on environmental information. In this step, the first environmental information is input into the task relevance predictor, which outputs the task relevance of each first traffic participant. The task relevance value can range from 0 to 1, where a higher task relevance indicates a greater impact on the current driving behavior of the target vehicle. If a first traffic participant has no direct impact on the current driving behavior of the target vehicle, its task relevance can be set to 0; if a first traffic participant has a decisive influence on the current driving behavior of the target vehicle, its task relevance can be set to 1. A higher task relevance indicates a greater influence of the first traffic participant on the current driving behavior of the target vehicle.

[0047] By applying a task relevance predictor to an end-to-end intelligent driving model (hereinafter referred to as the intelligent driving model), the task relevance of each first traffic participant can be predicted, thereby determining the influence of each first traffic participant on the current driving behavior of the target vehicle. The intelligent driving model can quickly understand the logical relationship between other traffic participants in the environment and the driving trajectory of the target vehicle, which can effectively improve the efficiency of the intelligent driving model in cognizing the environment, improve the intelligent driving model's understanding of driving behavior, and enable it to quickly plan an accurate driving trajectory.

[0048] Step 103: Filter multiple first traffic participants based on the task relevance of each first traffic participant to obtain the target traffic participant.

[0049] For example, the first traffic participant whose task relevance is greater than or equal to a first screening threshold is selected as the target traffic participant. The first screening threshold can be set according to actual conditions, for example, set to 0.8. This method can filter out the first traffic participants with high task relevance to participate in subsequent trajectory planning for the target vehicle, thereby reducing the focus on redundant information and improving the accuracy of trajectory planning.

[0050] Step 104: Perform trajectory planning for the target vehicle based on the target traffic participants to obtain the target trajectory.

[0051] For example, the trajectory of the target vehicle can be planned based on the description information of the target traffic participants.

[0052] In this embodiment, first environmental information about the environment in which the target vehicle is located is acquired. This first environmental information describes multiple first traffic participants in the environment. A pre-acquired task relevance predictor predicts the task relevance of each first traffic participant based on the first environmental information. The task relevance of each first traffic participant characterizes its influence on the current driving behavior of the target vehicle. Multiple first traffic participants are then filtered based on their task relevance to obtain target traffic participants. Finally, trajectory planning is performed on the target vehicle based on the target traffic participants to obtain the target trajectory. This method allows for the selection of first traffic participants with high task relevance to participate in subsequent trajectory planning for the target vehicle, thereby reducing the focus on redundant information and improving the accuracy of the target vehicle trajectory planning.

[0053] This application embodiment also provides a method for screening multiple first traffic participants. Specifically, step 103 involves screening multiple first traffic participants based on the task relevance of each first traffic participant to obtain target traffic participants, including steps 1031-1032, wherein:

[0054] Step 1031: If the first traffic participant is a target object, then calculate the first influence degree of the first traffic participant, which is used to characterize the influence degree of the first traffic participant on the current driving behavior of the target vehicle.

[0055] The first traffic participant belongs to the target object, for example, a vehicle or a pedestrian. The first influence degree of the first traffic participant is then calculated. Both the first influence degree and task relevance are used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle. However, they differ in that task relevance is based on a task relevance predictor, and its accuracy depends on the predictive power of the predictor. The first influence degree is obtained through a series of calculation strategies.

[0056] For example, the first influence degree is determined based on at least one of spatial influence degree, kinematic influence degree, and trajectory influence degree;

[0057] The spatial influence degree is determined according to at least one of the following:

[0058] Distance influence, which is determined based on the lateral and longitudinal distances between the first traffic participant and the target vehicle;

[0059] Directional influence, which is determined based on the orientation of the first traffic participant relative to the target vehicle;

[0060] And / or,

[0061] The kinematic influence is determined based on at least one of the following:

[0062] Speed ​​influence degree, which is determined based on the speed of the first traffic participant in a first direction and the speed of the target vehicle in a first direction, wherein the first direction is the direction of movement of the target vehicle;

[0063] Collision impact degree, which is determined based on the duration of the collision between the first traffic participant and the target vehicle;

[0064] Acceleration influence degree, which is determined based on the acceleration of the first traffic participant in the first direction and the acceleration of the target vehicle in the first direction;

[0065] And / or,

[0066] The trajectory influence is determined based on whether there is an intersection between the trajectory extension of the first traffic participant and the trajectory extension of the target vehicle, and the distance between the intersection and the target vehicle.

[0067] In the above, the first influence degree is determined based on at least one of spatial influence degree, kinematic influence degree, and trajectory influence degree. For example, the first influence degree can be directly adopted as spatial influence degree, kinematic influence degree, or trajectory influence degree; or, the first influence degree is the result of a weighted sum of any two of spatial influence degree, kinematic influence degree, and trajectory influence degree; or, the first influence degree is the result of a weighted sum of spatial influence degree, kinematic influence degree, and trajectory influence degree.

[0068] The spatial influence can be directly expressed as distance influence or direction influence, or the spatial influence can be expressed as a weighted sum of distance influence and direction influence.

[0069] The distance influence is determined based on the lateral and longitudinal distances between the first traffic participant and the target vehicle. For example, the lateral distance influence is calculated using a distance decay function based on the lateral distance; the larger the lateral distance, the smaller the lateral influence. Similarly, the longitudinal distance influence is calculated using a distance decay function based on the longitudinal distance; the larger the longitudinal distance, the smaller the longitudinal influence. The sum of the lateral and longitudinal influences is taken as the distance influence.

[0070] For example, the Euclidean distance between the first traffic participant and the target vehicle is calculated, and the distance influence degree D is determined according to the following expression (1):

[0071] (1)

[0072] Where d1, d2, and d3 are weighting coefficients, and the sum of d1, d2, and d3 is 1.

[0073] The directional influence is determined based on the direction of the first traffic participant relative to the target vehicle, wherein the directional influence of the first traffic participant decreases sequentially as it is located in front of, to the front side of, to the side of, behind, and to the rear side of the target vehicle.

[0074] The kinematic influence is determined based on at least one of velocity influence, collision influence, and acceleration influence. For example, the kinematic influence can be directly obtained by using any one of velocity influence, collision influence, and acceleration influence; or, the kinematic influence is the result of a weighted sum of any two of velocity influence, collision influence, and acceleration influence; or, the kinematic influence is the result of a weighted sum of velocity influence, collision influence, and acceleration influence.

[0075] The speed influence degree is determined based on the speed of the first traffic participant in a first direction and the speed of the target vehicle in the first direction, where the first direction is the direction of movement of the target vehicle. For example, the speed of the first traffic participant is mapped to the first direction; the speed vector difference between the first traffic participant and the target vehicle in the first direction is calculated. If the relative speed is positive, a larger influence degree is set; if the relative speed is negative, a smaller influence degree is set.

[0076] The collision impact is determined based on the duration of the collision between the first traffic participant and the target vehicle. The shorter the duration, the greater the collision impact.

[0077] The acceleration influence degree is determined based on the acceleration of the first traffic participant in the first direction and the acceleration of the target vehicle in the first direction; for example, the acceleration of the first traffic participant is mapped to the first direction; the acceleration vector difference between the first traffic participant and the target vehicle in the first direction is calculated, and if the relative acceleration is positive, a larger influence degree is set; if the relative acceleration is negative, a smaller influence degree is set.

[0078] The trajectory influence is determined based on whether the extended trajectory lines of the first traffic participant and the target vehicle intersect, and the distance between the intersection point and the target vehicle. For example, if the extended trajectory lines of the first traffic participant and the target vehicle do not intersect, a smaller trajectory influence is set; if they intersect, the trajectory influence is set according to the distance between the intersection point and the target vehicle, with a larger influence the closer the distance.

[0079] The above method can determine multiple degrees of influence affecting the current driving behavior of the target vehicle. The first degree of influence can be determined based on one or more of these multiple degrees of influence to improve the accuracy of the first degree of influence.

[0080] Step 1032: Determine the driving influence of the first traffic participant based on the first influence degree of the first traffic participant and the task relevance of the first traffic participant.

[0081] For example, the product of the first influence degree and the task relevance of the first traffic participant can be used as the driving influence degree of the first traffic participant. By comprehensively considering the first influence degree and task relevance to determine the driving influence degree, the accuracy of the driving influence degree can be improved.

[0082] For example, the weighted sum of the first influence degree and the task relevance of the first traffic participant can be used as the driving influence degree of the first traffic participant.

[0083] Step 1033: Filter the first traffic participants based on task relevance and driving impact to obtain the target traffic participants.

[0084] For example, the first traffic participant who meets the first preset condition is selected as the target traffic participant;

[0085] The first preset conditions include:

[0086] The first traffic participant is the target object, which includes vehicles or pedestrians;

[0087] The driving influence of the first traffic participant is greater than or equal to the average driving influence, where the average driving influence is the mean of the driving influence of the first traffic participants belonging to the target object.

[0088] The task relevance of the first traffic participant is greater than or equal to the second screening threshold.

[0089] For example, a first traffic participant with a driving influence greater than or equal to the average driving influence and a task relevance greater than or equal to 0.5 is selected, and this first traffic participant is used as the target traffic participant. In the above, the target traffic participants are selected through a first preset condition. Selecting only the target traffic participants for subsequent trajectory planning of the target vehicle reduces the focus on redundant information and improves the accuracy of trajectory planning.

[0090] By using steps 1031-1032 above, we can filter out the target traffic participants who have a significant impact on the current driving behavior of the target vehicle. By selecting only the target traffic participants to participate in the subsequent trajectory planning of the target vehicle, we can reduce the focus on redundant information and improve the accuracy of trajectory planning.

[0091] In one embodiment of this application, multiple first traffic participants are filtered based on the task relevance of each first traffic participant to obtain target traffic participants, including:

[0092] The first traffic participant who meets the second preset condition will be designated as the target traffic participant;

[0093] The second preset condition includes:

[0094] The first traffic participant is a traffic sign;

[0095] The task relevance of the first traffic participant is greater than or equal to the third screening threshold;

[0096] And / or,

[0097] The first traffic participant who meets the third preset condition will be designated as the target traffic participant;

[0098] The third preset condition includes:

[0099] The first traffic participant is a traffic signal light;

[0100] The task relevance of the first traffic participant is greater than or equal to the fourth screening threshold.

[0101] For example, the filtering threshold for traffic signs (i.e., the third filtering threshold, for example, 0.2) can be set lower than the filtering threshold for traffic lights (i.e., the fourth filtering threshold, for example, 0.5). In other words, different filtering thresholds are used for different types of first traffic participants. This allows for different filtering methods for different types of first traffic participants, improving the accuracy of the filtering. While effectively reducing redundant information, it avoids filtering out target traffic participants who have a significant impact on the current driving behavior of the target vehicle, thus improving the accuracy of trajectory planning for the target vehicle.

[0102] In another embodiment of this application, the task relevance predictor is trained according to the following steps:

[0103] Multiple target information is acquired, including multiple second environmental images, motion information of the sample vehicle, and second environmental information of the environment in which the sample vehicle is located. The second environmental information is used to describe multiple second traffic participants in the environment in which the sample vehicle is located. The multiple second environmental images are obtained by capturing images of the environment in which the sample vehicle is located through multiple cameras installed on the sample vehicle.

[0104] For each piece of target information, a prediction is made based on the target information and target cue words using a multimodal large language model to obtain the analysis results of the target information. These analysis results include the task relevance of each second traffic participant (i.e., the degree of influence of the second traffic participant on the current driving behavior of the sample vehicle); and...

[0105] Based on the analysis results of the target information, the labeled data of the target information is obtained; and,

[0106] Training samples are constructed based on the second environmental information in the target information and the labeled data of the target information;

[0107] The base model is trained using multiple training samples to obtain the task relevance predictor.

[0108] In the above, the task relevance of the second traffic participant can be automatically labeled using a multimodal large language model. Leveraging the powerful image and text understanding capabilities of the multimodal large language model, labeled data for target information can be generated quickly, significantly reducing the costs of manual labeling and time. This facilitates subsequent training of the base model using labeled training samples to obtain a task relevance predictor. In subsequent use, the task relevance predictor can be used to predict the task relevance of each traffic participant in the environment, thereby filtering out traffic participants with lower influence based on task relevance, reducing the impact of redundant information on trajectory planning, and improving the efficiency and accuracy of trajectory planning.

[0109] Specifically, the motion information of the sample vehicle may include one or more of the following: speed, angular velocity, acceleration, angular acceleration, direction of motion, straight, left turn, right turn, acceleration, deceleration, average speed, etc., without limitation.

[0110] Target prompts can include requirement prompts, scoring rule prompts, output template prompts, and so on.

[0111] For example, the prompt words could be:

[0112] 1. Identify all vehicles, pedestrians, traffic signs, and traffic lights that significantly affect the vehicle's trajectory;

[0113] 2. Assign an influence level of 0.0-1.0 to each influencing factor;

[0114] 3. Provide a detailed explanation of why each factor affects the vehicle's trajectory;

[0115] 4. Sort by degree of impact from highest to lowest.

[0116] For example, evaluation rule prompts could be:

[0117] 0.9-1.0: Decisive impact (immediate action required);

[0118] 0.7-0.8: Significant impact (requires proactive trajectory adjustment);

[0119] 0.4-0.6: Moderate impact (requires monitoring and may be subject to minor adjustments);

[0120] 0.1-0.3: Slight effect;

[0121] 0.0: No direct impact.

[0122] The output template prompts can be set according to actual output needs, and there are no restrictions here. For example, the types of traffic participants that can be selected in the output template prompts can be limited, such as vehicles, pedestrians, traffic signs, traffic lights, etc. The types of traffic participants output can only be selected from these categories provided by the template.

[0123] The analysis results also include the criteria for evaluating the impact of the task relevance of each second traffic participant. The criteria can be used to verify whether the multimodal large language model correctly scores the impact of the second traffic participant. Furthermore, if the impact score is incorrect, the multimodal large language model can be fine-tuned to improve the accuracy of the multimodal large language model in scoring the impact of the second traffic participant.

[0124] The output template prompts limit the impact assessment criteria available to the second traffic participant. In this case, the impact assessment criterion can only be selected from a variety of situations provided by the output template prompts. For example, impact assessment criteria such as "located to the left front of the vehicle and approaching the lane where the vehicle is located" or "relatively far away, with no intersection on the extended trajectory line of the vehicle" are used.

[0125] The target information and target cue words are predicted using a multimodal large language model to obtain the analysis results of the target information. The analysis results include the influence of each second traffic participant on the current driving behavior of the sample vehicle.

[0126] Task relevance predictors can understand (or simulate) human driving logic at the semantic level and explain driving logic by outputting task relevance (or impact assessment criteria).

[0127] Since the description of the analysis results may need to be adjusted for greater standardization, the analysis results are structured to obtain labeled data for the target information. Specifically, based on the analysis results of the target information, the labeled data for the target information is obtained, including:

[0128] The analysis results are format-converted to obtain the first text;

[0129] The first text is subjected to compliance testing and correction using a large language model to obtain the processing result;

[0130] If the processing result indicates compliance, then the first text will be used as the annotation data for the target information;

[0131] And / or,

[0132] If the processing result indicates non-compliance, and the processing result includes the corrected text obtained by the large language model correcting the first text, then the corrected text will be used as the annotation data for the target information;

[0133] And / or,

[0134] If the processing result indicates non-compliance, and the processing result does not include the corrected text obtained by the large language model correcting the first text, then correction information is obtained, and the correction information is used to correct the first text;

[0135] Based on the correction information, the corrected text is obtained;

[0136] The corrected text is used as the annotation data for the target information.

[0137] In the above, format conversion can be to convert the analysis results into JSON format. JSON format uses a key-value pair structure and has simple syntax. Compliance checks can be configured according to the actual situation. For example, they can check whether there are errors in the first text obtained after format conversion, such as a missing value for a key or a missing key for a value. Compliance checks can also check whether there are typos in the first text, etc.

[0138] Using a large language model for compliance detection and correction can yield one of three results. Regardless of the result, the final result will be the labeled data of the target information.

[0139] The labeled data can be in the following format:

[0140] Identifier 0001 (used to identify a second road user) has an impact score of 0.8. The impact score is determined by the fact that the second road user is located to the left front of the vehicle and is approaching the lane where the vehicle is located.

[0141] Identifier 0002 (used to identify a second traffic participant) has an impact of 0.1. The impact is judged based on the following criteria: it is far away and has no intersection with the extended trajectory of the vehicle.

[0142] Each target information will correspond to a set of labeled data.

[0143] Using the methods described above, and with the assistance of a large language model, labeled data with uniform target information can be obtained quickly, thus improving the efficiency of acquiring labeled data.

[0144] In the above, the correction information includes at least one of a second text and a correction rule, wherein the second text is the text obtained by manually correcting the first text, and the correction rule is a rule for correcting the first text.

[0145] Accordingly, based on the correction information, the corrected text is obtained, including:

[0146] If the correction information includes the second text but does not include the correction rule, then the second text is used as the correction text, or the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using a large language model to obtain the processing result.

[0147] If the correction information includes the correction rule but does not include the second text, then the large language model is learned and updated using the correction rule, and the first text is corrected using the updated large language model to obtain the corrected text;

[0148] If the correction information includes the correction rule and the second text, then the large language model is learned and updated using the correction rule, and the second text is used as the corrected text; or, the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using the large language model to obtain the processing result.

[0149] In the above, based on the correction information, three cases of the corrected text are obtained. Regardless of the case, the corrected text can be obtained, thereby achieving the annotation of the target information and improving the efficiency of data annotation.

[0150] The trajectory planning method provided in this application allows the intelligent driving model to focus on traffic participants in the environment that are highly relevant to driving behavior, and to reduce attention to irrelevant traffic participants, thereby planning driving trajectories efficiently and accurately.

[0151] Currently, the intelligent driving model includes a perception module and a trajectory planning module. The perception module is used to detect traffic participants around the vehicle, and the trajectory planning module is used to plan the trajectory based on the detected traffic participants.

[0152] The driving environment contains numerous traffic participants, each with its own behavioral logic. Furthermore, the scenario often includes multiple traffic lights, traffic signs, and other participants that may influence the driving route. However, not all traffic participants affect the vehicle's trajectory. Drivers only need to focus on nearby traffic participants and predict their trajectories to plan a path that avoids collisions. For a vehicle needing to proceed straight, the driver only needs to focus on the straight-ahead traffic light. In this straight-ahead task, the left-turn traffic light is irrelevant. Intelligent driving models should minimize their focus on it and concentrate more computational power on analyzing changes in the straight-ahead traffic light to efficiently understand key traffic participants in the environment. However, current intelligent driving models analyze all traffic participants detected by the perception module, consuming significant inference time and computational power.

[0153] Since most traffic participants have no impact on the driving behavior of the vehicle, enabling the model to identify which traffic participants have a significant impact on the current task and allowing the intelligent driving model to focus on key traffic participants while reducing attention to redundant information is crucial for achieving efficient intelligent driving solutions.

[0154] To achieve this goal, low-redundancy spatial cognition technology adjusts the attention of the intelligent driving model to each traffic participant based on task relevance, enabling the intelligent driving model to pay more attention to key traffic participants in the environment.

[0155] Task relevance refers to the degree of influence of traffic participants in the environment on current driving behavior. The stronger the influence of traffic participants on driving behavior, the greater the task relevance, and vice versa.

[0156] As shown in Figure 2, the task relevance of each traffic participant will be labeled using a multimodal large language model. This approach leverages the powerful image and text understanding capabilities of multimodal large language models to quickly generate a driving behavior logic dataset, thereby significantly reducing the costs of manual labeling and time.

[0157] Based on requirement prompts, scoring rule prompts, and output template prompts, the output data of a multimodal large language model can contain the required key information and have a standardized data structure, which facilitates subsequent processing.

[0158] The output data of the multimodal large language model is transformed into a labeled dataset that can be used for model training (training a task-related predictor) through data structuring. The specific process is as follows: Figure 3 As shown, specifically, the data output by the multimodal large language model is first preprocessed and converted into JSON format. Then, the large language model performs a compliance check on the converted data. If the data structure meets the requirements, it is added to the database as labeled data; otherwise, it is manually checked and modified. The data is then input back into the large language model for compliance checks until all data passes the checks. This process yields a labeled database containing the task relevance of each traffic participant and describing the driving logic.

[0159] As shown in Figure 4, the intelligent driving model includes a perception module, a task relevance predictor, and a trajectory planning module. The acquired environmental images surrounding the vehicle are input into the perception module, which analyzes these images to obtain perception results. These results include: the ID, type, speed, acceleration, direction, and relative position of other vehicles or pedestrians; the ID and type of traffic signs; and the ID, type, and status of traffic lights. The perception results are then input into the task relevance predictor to obtain the task relevance of each traffic participant.

[0160] The task relevance predictor is trained on a labeled database. After obtaining the task relevance of each traffic participant, it filters out those with less influence, inputting only those with a significant impact on the current driving behavior into the trajectory planning module. This assists the intelligent driving model in quickly understanding the influence of each traffic participant on the current driving behavior, reducing the impact of redundant information on the model's training, and improving the efficiency and accuracy of trajectory planning. This approach can reduce the difficulty of understanding changes in the cognitive environment, lower model training and deployment costs, reduce inference time, and improve trajectory planning accuracy without affecting the existing intelligent driving model architecture.

[0161] In this way, intelligent driving models can proactively identify traffic participants in the environment that have a significant impact on the vehicle, based on their understanding of human driving behavior logic, and quickly and accurately plan trajectories based on task relevance, thereby achieving low-redundancy spatial cognition capabilities.

[0162] In the above, the intelligent driving model quickly grasps the key elements in the driving environment by understanding the impact of various things in the driving environment on the vehicle's driving behavior, and only analyzes the key elements to achieve efficient driving capabilities similar to humans.

[0163] By using a multimodal large language model to semantically label the data, the impact of the trajectories of other traffic participants in various scenarios on the autonomous vehicle's driving behavior is analyzed. Then, a task relevance predictor learns from the sample data the degree and manner of the influence of other traffic participants on the autonomous vehicle, thereby improving the intelligent driving model's ability to understand driving behavior.

[0164] Figure 5 A structural diagram of the trajectory planning device provided in an embodiment of this application is shown. Figure 5 As shown, the trajectory planning device 500 includes:

[0165] The acquisition module 501 is used to acquire first environmental information of the environment in which the target vehicle is located. The first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located.

[0166] The prediction module 502 is used to make predictions based on the first environmental information through a pre-acquired task relevance predictor to obtain the task relevance of each first traffic participant. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle.

[0167] The filtering module 503 is used to filter multiple first traffic participants based on the task relevance of each first traffic participant to obtain the target traffic participant;

[0168] The planning module 504 is used to plan the trajectory of the target vehicle based on the target traffic participants to obtain the target trajectory.

[0169] In one embodiment of this application, the screening module 503 includes:

[0170] The calculation submodule is used to calculate the first influence degree of the first traffic participant if the first traffic participant belongs to the target object. The first influence degree is used to characterize the influence degree of the first traffic participant on the current driving behavior of the target vehicle.

[0171] The first determining submodule is used to determine the driving influence of the first traffic participant based on the first influence of the first traffic participant and the task relevance of the first traffic participant.

[0172] The first screening submodule is used to screen the first traffic participants based on task relevance and driving impact to obtain the target traffic participants.

[0173] In one embodiment of this application, the first influence degree is determined based on at least one of spatial influence degree, kinematic influence degree, and trajectory influence degree;

[0174] The spatial influence degree is determined according to at least one of the following:

[0175] Distance influence, which is determined based on the lateral and longitudinal distances between the first traffic participant and the target vehicle;

[0176] Directional influence, which is determined based on the orientation of the first traffic participant relative to the target vehicle;

[0177] And / or,

[0178] The kinematic influence is determined based on at least one of the following:

[0179] Speed ​​influence degree, which is determined based on the speed of the first traffic participant in a first direction and the speed of the target vehicle in a first direction, wherein the first direction is the direction of movement of the target vehicle;

[0180] Collision impact degree, which is determined based on the duration of the collision between the first traffic participant and the target vehicle;

[0181] Acceleration influence degree, which is determined based on the acceleration of the first traffic participant in the first direction and the acceleration of the target vehicle in the first direction;

[0182] And / or,

[0183] The trajectory influence is determined based on whether there is an intersection between the trajectory extension of the first traffic participant and the trajectory extension of the target vehicle, and the distance between the intersection and the target vehicle.

[0184] In one embodiment of this application, the first determining submodule is specifically used for:

[0185] The product of the first influence degree and the task relevance of the first traffic participant is taken as the driving influence degree of the first traffic participant.

[0186] In one embodiment of this application, the first screening submodule is specifically used for:

[0187] The first traffic participant who meets the first preset condition is designated as the target traffic participant;

[0188] The first preset conditions include:

[0189] The first traffic participant is the target object;

[0190] The driving influence of the first traffic participant is greater than or equal to the average driving influence, where the average driving influence is the mean of the driving influence of the first traffic participants belonging to the target object.

[0191] The task relevance of the first traffic participant is greater than or equal to the second screening threshold.

[0192] In one embodiment of this application, the screening module 503 includes a second screening submodule, which is used to select a first traffic participant who meets a second preset condition as a target traffic participant.

[0193] The second preset condition includes:

[0194] The first traffic participant is a traffic sign;

[0195] The task relevance of the first traffic participant is greater than or equal to the third screening threshold;

[0196] And / or,

[0197] The first traffic participant who meets the third preset condition will be designated as the target traffic participant;

[0198] The third preset condition includes:

[0199] The first traffic participant is a traffic signal light;

[0200] The task relevance of the first traffic participant is greater than or equal to the fourth screening threshold.

[0201] In one embodiment of this application, the screening module 503 includes a third screening submodule, which is used to select a first traffic participant whose task relevance is greater than or equal to a first screening threshold as the target traffic participant.

[0202] In one embodiment of this application, the task relevance predictor is trained according to the following steps:

[0203] Multiple target information is acquired, including multiple second environmental images, motion information of the sample vehicle, and second environmental information of the environment in which the sample vehicle is located. The second environmental information is used to describe multiple second traffic participants in the environment in which the sample vehicle is located. The multiple second environmental images are obtained by capturing images of the environment in which the sample vehicle is located through multiple cameras installed on the sample vehicle.

[0204] For each piece of target information, a prediction is made based on the target information and target cue words using a multimodal large language model to obtain the analysis result of the target information. The analysis result includes the task relevance of each of the second traffic participants; and...

[0205] Based on the analysis results of the target information, the labeled data of the target information is obtained; and,

[0206] Training samples are constructed based on the second environmental information in the target information and the labeled data of the target information;

[0207] The base model is trained using multiple training samples to obtain the task relevance predictor.

[0208] In one embodiment of this application, the annotation data of the target information is obtained based on the analysis results of the target information, including:

[0209] The analysis results are format-converted to obtain the first text;

[0210] The first text is subjected to compliance testing and correction using a large language model to obtain the processing result;

[0211] If the processing result indicates compliance, then the first text will be used as the annotation data for the target information;

[0212] And / or,

[0213] If the processing result indicates non-compliance, and the processing result includes the corrected text obtained by the large language model correcting the first text, then the corrected text will be used as the annotation data of the target information;

[0214] And / or,

[0215] If the processing result indicates non-compliance, and the processing result does not include the corrected text obtained by the large language model correcting the first text, then correction information is obtained, and the correction information is used to correct the first text;

[0216] Based on the correction information, the corrected text is obtained;

[0217] The corrected text is used as the annotation data for the target information.

[0218] In one embodiment of this application, the correction information includes at least one of a second text and a correction rule, wherein the second text is text obtained by manually correcting the first text, and the correction rule is a rule for correcting the first text.

[0219] In one embodiment of this application, obtaining the corrected text based on the corrected information includes:

[0220] If the correction information includes the second text but does not include the correction rule, then the second text is used as the correction text, or the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using a large language model to obtain the processing result.

[0221] And / or,

[0222] If the correction information includes the correction rule but does not include the second text, then the large language model is learned and updated using the correction rule, and the first text is corrected using the updated large language model to obtain the corrected text;

[0223] And / or,

[0224] If the correction information includes the correction rule and the second text, then the large language model is learned and updated using the correction rule, and the second text is used as the corrected text; or, the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using the large language model to obtain the processing result.

[0225] In one embodiment of this application, the acquisition module 501 includes:

[0226] The acquisition submodule is used to acquire images of the environment in which the target vehicle is located through multiple cameras installed on the target vehicle, and obtain multiple first environmental images;

[0227] The processing submodule is used to perform image processing on multiple first environmental images to obtain multiple descriptive information. The first environmental information includes multiple descriptive information, and each descriptive information is used to describe a first traffic participant.

[0228] The trajectory planning device 500 provided in this application embodiment can implement the various processes implemented in the aforementioned trajectory planning method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0229] Figure 6 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0230] The electronic device may include a processor 601 and a memory 602 storing computer program instructions.

[0231] Specifically, the processor 601 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0232] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 602 may include removable or non-removable (or fixed) media. Where appropriate, memory 602 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 602 is non-volatile solid-state memory.

[0233] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the first aspect of this disclosure.

[0234] In one example, the electronic device may also include a communication interface 603 and a bus 610. For example, Figure 6 As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 610 and complete communication with each other.

[0235] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0236] Bus 610 includes hardware, software, or both. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 610 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0237] Furthermore, in conjunction with the trajectory planning methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the trajectory planning methods in the above embodiments.

[0238] This application also provides a computer program product, wherein the instructions in the computer program product, when executed by the processor of an electronic device, cause the electronic device to perform any of the trajectory planning methods described in the above embodiments.

[0239] This application also provides a vehicle that includes the electronic devices described in the above embodiments.

[0240] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described as examples. However, the method process of this application is not limited to the specific steps described. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0241] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0242] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0243] The foregoing flowcharts and / or block diagrams of methods, apparatus (systems) according to embodiments of the present disclosure have described various aspects of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to create a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowcharts and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block of the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0244] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A trajectory planning method, characterized in that, The method includes: First environmental information of the environment in which the target vehicle is located is obtained, and the first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located. Based on the first environmental information, the task relevance of each first traffic participant is obtained by predicting using a pre-acquired task relevance predictor. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle. Multiple first traffic participants are filtered based on the task relevance of each first traffic participant to obtain the target traffic participants; Based on the target traffic participants, the target vehicle's trajectory is planned to obtain the target trajectory; Based on the task relevance of each first traffic participant, multiple first traffic participants are filtered to obtain target traffic participants, including: If the first traffic participant is a target object, then the first influence degree of the first traffic participant is calculated. The first influence degree is used to characterize the influence degree of the first traffic participant on the current driving behavior of the target vehicle. The target object is a vehicle or a pedestrian. The driving influence of the first traffic participant is determined based on the first influence of the first traffic participant and the task relevance of the first traffic participant. The primary traffic participants are selected based on task relevance and driving impact to obtain the target traffic participants.

2. The trajectory planning method according to claim 1, characterized in that, The first influence degree is determined based on at least one of spatial influence degree, kinematic influence degree, and trajectory influence degree; The spatial influence degree is determined according to at least one of the following: Distance influence, which is determined based on the lateral and longitudinal distances between the first traffic participant and the target vehicle; Directional influence, which is determined based on the orientation of the first traffic participant relative to the target vehicle; And / or, The kinematic influence is determined based on at least one of the following: Speed ​​influence degree, which is determined based on the speed of the first traffic participant in a first direction and the speed of the target vehicle in a first direction, wherein the first direction is the direction of movement of the target vehicle; Collision impact degree, which is determined based on the duration of the collision between the first traffic participant and the target vehicle; Acceleration influence degree, which is determined based on the acceleration of the first traffic participant in the first direction and the acceleration of the target vehicle in the first direction; And / or, The trajectory influence is determined based on whether there is an intersection between the trajectory extension of the first traffic participant and the trajectory extension of the target vehicle, and the distance between the intersection and the target vehicle.

3. The trajectory planning method according to claim 1, characterized in that, The driving influence of the first traffic participant is determined based on the first influence level and the task relevance of the first traffic participant, including: The product of the first influence degree and the task relevance of the first traffic participant is taken as the driving influence degree of the first traffic participant.

4. The trajectory planning method according to claim 1, characterized in that, Based on task relevance and driving impact, the primary traffic participants were screened to obtain the target traffic participants, including: The first traffic participant who meets the first preset condition is designated as the target traffic participant; The first preset conditions include: The first traffic participant is the target object; The driving influence of the first traffic participant is greater than or equal to the average driving influence, where the average driving influence is the mean of the driving influence of the first traffic participants belonging to the target object. The task relevance of the first traffic participant is greater than or equal to the second screening threshold.

5. The trajectory planning method according to claim 1, characterized in that, Based on the task relevance of each first traffic participant, multiple first traffic participants are filtered to obtain target traffic participants, including: The first traffic participant who meets the second preset condition will be designated as the target traffic participant; The second preset condition includes: The first traffic participant is a traffic sign; The task relevance of the first traffic participant is greater than or equal to the third screening threshold; And / or, The first traffic participant who meets the third preset condition will be designated as the target traffic participant; The third preset condition includes: The first traffic participant is a traffic signal light; The task relevance of the first traffic participant is greater than or equal to the fourth screening threshold.

6. The trajectory planning method according to claim 1, characterized in that, Based on the task relevance of each first traffic participant, multiple first traffic participants are filtered to obtain target traffic participants, including: The first traffic participant whose task relevance is greater than or equal to the first screening threshold is selected as the target traffic participant.

7. The trajectory planning method according to claim 1, characterized in that, The task relevance predictor is trained according to the following steps: Multiple target information is acquired, including multiple second environmental images, motion information of the sample vehicle, and second environmental information of the environment in which the sample vehicle is located. The second environmental information is used to describe multiple second traffic participants in the environment in which the sample vehicle is located. The multiple second environmental images are obtained by capturing images of the environment in which the sample vehicle is located through multiple cameras installed on the sample vehicle. For each of the target information, a prediction is made based on the target information and target cue words using a multimodal large language model to obtain the analysis result of the target information, and the analysis result includes the task relevance of each of the second traffic participants; as well as, Based on the analysis results of the target information, the labeled data of the target information is obtained; as well as, Training samples are constructed based on the second environmental information in the target information and the labeled data of the target information; The base model is trained using multiple training samples to obtain the task relevance predictor.

8. The trajectory planning method according to claim 7, characterized in that, Based on the analysis results of the target information, the labeled data of the target information is obtained, including: The analysis results are format-converted to obtain the first text; The first text is subjected to compliance testing and correction using a large language model to obtain the processing result; If the processing result indicates compliance, then the first text will be used as the annotation data for the target information; And / or, If the processing result indicates non-compliance, and the processing result includes the corrected text obtained by the large language model correcting the first text, then the corrected text will be used as the annotation data of the target information; And / or, If the processing result indicates non-compliance, and the processing result does not include the corrected text obtained by the large language model correcting the first text, then correction information is obtained, and the correction information is used to correct the first text; Based on the correction information, the corrected text is obtained; The corrected text is used as the annotation data for the target information.

9. The trajectory planning method according to claim 8, characterized in that, The correction information includes at least one of a second text and a correction rule, wherein the second text is text obtained by manually correcting the first text, and the correction rule is a rule for correcting the first text.

10. The trajectory planning method according to claim 9, characterized in that, Based on the correction information, the corrected text is obtained, including: If the correction information includes the second text but does not include the correction rule, then the second text is used as the correction text, or the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using a large language model to obtain the processing result. And / or, If the correction information includes the correction rule but does not include the second text, then the large language model is learned and updated using the correction rule, and the first text is corrected using the updated large language model to obtain the corrected text; And / or, If the correction information includes the correction rule and the second text, then the large language model is learned and updated using the correction rule, and the second text is used as the corrected text; or, the second text is used as the first text, and the process jumps to the step of performing compliance detection and correction on the first text using the large language model to obtain the processing result.

11. The trajectory planning method according to claim 1, characterized in that, Obtain primary environmental information about the target vehicle's location, including: Multiple first environmental images are obtained by capturing images of the environment in which the target vehicle is located using multiple cameras installed on the target vehicle. Multiple first environmental images are processed to obtain multiple descriptive information. The first environmental information includes multiple descriptive information, and each descriptive information is used to describe a first traffic participant.

12. A trajectory planning device, characterized in that, The device includes: The acquisition module is used to acquire first environmental information of the environment in which the target vehicle is located. The first environmental information is used to describe multiple first traffic participants in the environment in which the target vehicle is located. The prediction module is used to predict based on the first environmental information using a pre-acquired task relevance predictor to obtain the task relevance of each first traffic participant. The task relevance of the first traffic participant is used to characterize the influence of the first traffic participant on the current driving behavior of the target vehicle. The filtering module is used to filter multiple first traffic participants based on the task relevance of each first traffic participant to obtain the target traffic participants; The planning module is used to plan the trajectory of the target vehicle based on the target traffic participants to obtain the target trajectory; The filtering module includes: The calculation submodule is used to calculate the first influence degree of the first traffic participant if the first traffic participant belongs to the target object. The first influence degree is used to characterize the influence degree of the first traffic participant on the current driving behavior of the target vehicle. The target object is a vehicle or a pedestrian. The first determining submodule is used to determine the driving influence of the first traffic participant based on the first influence of the first traffic participant and the task relevance of the first traffic participant. The first screening submodule is used to screen the first traffic participants based on task relevance and driving impact to obtain the target traffic participants.

13. An electronic device, characterized in that, include: Processor and memory storing computer program instructions; When the processor executes the computer program instructions, it implements the trajectory planning method as described in any one of claims 1-11.

14. A vehicle, characterized in that, The vehicle includes the electronic equipment as described in claim 13.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the trajectory planning method as described in any one of claims 1-11.

16. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the trajectory planning method as described in any one of claims 1-11.

Images