Method, apparatus, device and storage medium for trajectory generation

Abstract

Embodiments of the present disclosure relate to a method, device, equipment and storage medium for trajectory generation. The method comprises: in response to an abnormality occurring in a scene in which a target object is movable according to a first predicted trajectory, obtaining the first predicted trajectory and input information used to generate the first predicted trajectory, the first predicted trajectory being generated by a trajectory generation model based on the input information, the input information at least including visual information of the scene; determining a reference trajectory corresponding to the input information based on at least one candidate trajectory and the first predicted trajectory, the at least one candidate trajectory being obtained based on the input information; and updating the trajectory generation model based on the input information and the reference trajectory. In this way, the generalization capability of the trajectory generation model can be improved.

Description

Technical Field

[0001] The exemplary embodiments disclosed herein generally relate to the field of computer technology, and more specifically, to methods, apparatus, devices, and storage media for trajectory generation. Background Technology

[0002] Trajectory prediction technology primarily involves reasonably inferring the motion trends of surrounding dynamic objects in complex environments to assist in path planning decisions for movable target objects. Trajectory prediction typically relies on key technologies such as environmental perception, historical motion data analysis, and prediction of future scene information. With the improvement of perception and learning capabilities, especially the rise of end-to-end modeling paradigms, data-driven trajectory generation methods have gradually become mainstream. However, existing methods still face many challenges when dealing with complex dynamic environments. A training method for trajectory generation models that can effectively adapt to changes in complex environments is needed. Summary of the Invention

[0003] In a first aspect of this disclosure, a method for trajectory generation is provided. The method includes: in response to an anomaly occurring in a scene according to a first predicted trajectory of a movable target object, acquiring a first predicted trajectory and input information for generating the first predicted trajectory, the first predicted trajectory being generated by a trajectory generation model based on the input information, the input information including at least visual information specific to the scene; determining a reference trajectory corresponding to the input information based on at least one candidate trajectory and the first predicted trajectory, the at least one candidate trajectory being obtained based on the input information; and updating the trajectory generation model based on the input information and the reference trajectory.

[0004] In a second aspect of this disclosure, an apparatus for trajectory generation is provided. The apparatus includes: an acquisition module configured to acquire, in response to an anomaly occurring in a scene according to a first predicted trajectory of a movable target object, acquire a first predicted trajectory and input information for generating the first predicted trajectory, the first predicted trajectory being generated by a trajectory generation model based on the input information, the input information including at least visual information specific to the scene; a reference trajectory determination module configured to determine a reference trajectory corresponding to the input information based on at least one candidate trajectory and the first predicted trajectory, the at least one candidate trajectory being obtained based on the input information; and an update module configured to update the trajectory generation model based on the input information and the reference trajectory.

[0005] In a third aspect of this disclosure, an electronic device is provided. The device includes at least one processor; and at least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor. When executed by the at least one processor, the instructions cause the device to perform the method of the first aspect.

[0006] In a fourth aspect of this disclosure, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program that can be executed by a processor to implement the method of the first aspect.

[0007] In a fifth aspect of this disclosure, a computer program product is provided. The computer program product includes computer-executable instructions that, when executed by a processor, implement the method according to a first aspect of this disclosure.

[0008] It should be understood that the content described in this content section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0009] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0010] Figure 1 A schematic diagram of an example environment in which embodiments of the present disclosure can be implemented is shown;

[0011] Figure 2 A schematic diagram of the training architecture of a trajectory generation model according to some embodiments of the present disclosure is shown;

[0012] Figure 3A A schematic diagram of the architecture of a trajectory generation subsystem according to some embodiments of the present disclosure is shown;

[0013] Figure 3B A schematic diagram of an architecture for determining a reference trajectory according to some embodiments of the present disclosure is shown;

[0014] Figure 4 A flowchart of a process for trajectory generation according to some embodiments of the present disclosure is shown;

[0015] Figure 5 A schematic structural block diagram of an apparatus for trajectory generation according to some embodiments of the present disclosure is shown;

[0016] Figure 6 A block diagram of an electronic device in which one or more embodiments of the present disclosure may be implemented is shown. Detailed Implementation

[0017] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0018] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below.

[0019] As used in this paper, the term "model" refers to a system that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning is a machine learning algorithm that uses multiple layers of processing units to process inputs and provide corresponding outputs. In this paper, "model" may also be referred to as a "machine learning model," "machine learning network," "neural network," or simply "network," and these terms are used interchangeably.

[0020] In this document, unless explicitly stated otherwise, performing a step in response to A does not mean that the step is performed immediately after A, but may include one or more intermediate steps.

[0021] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0022] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure through appropriate means in accordance with relevant laws and regulations, and user authorization should be obtained.

[0023] For example, in response to receiving a user's active request, a prompt message is sent to the user to clearly inform the user that the requested operation will require the acquisition and use of the user's personal information, thereby enabling the user to choose whether to provide personal information to the software or hardware such as electronic devices, applications, servers or storage media that perform the operation of the technical solution disclosed herein, based on the prompt message.

[0024] As an optional but non-restrictive implementation, in response to a user's active request, a prompt message can be sent to the user, such as a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0025] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0026] Figure 1 A schematic diagram of an example environment in which embodiments of the present disclosure can be implemented is shown. References Figure 1 One or more movable objects, such as a first robot 110-1, a second robot 110-2, and a third robot 110-3, are deployed in environment 100. First robot 110-1, second robot 110-2, and third robot 110-3 can also be collectively or individually referred to as robot 110. Each robot 110 may be equipped with control devices to control its operation. For example, ... Figure 1 As shown, control devices 120-1, 120-2, and 120-3 are respectively deployed at the first robot 110-1, the second robot 110-2, and the third robot 110-3, which may also be collectively referred to as control device 120 or individually. In some embodiments, only one robot may be deployed in the environment 100, and only the movement path of this one robot needs to be planned. In some embodiments, multiple robots may be deployed in the environment 100. In this case, path planning can be performed simultaneously for multiple robots in the environment 100, or path planning can be performed only for any one robot in the environment 100.

[0027] In embodiments of this disclosure, robot 110 can be used for various suitable purposes, such as a transport robot for delivering goods, and control device 120 can be used to plan the movement path of robot 110. Robot 110 can, for example, perform parcel delivery services or warehouse management services. As an example, robot 110 can operate in indoor environments, such as warehouse environments.

[0028] exist Figure 1 In environment 100, control device 120 can be any type of microcontroller, programmable logic controller, industrial computer, single-board computer, field-programmable gate array, digital signal processor, multi-core processor, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, control device 120 can control robot 110 based on control commands it generates.

[0029] In some embodiments, the operation of robot 110 can be controlled via server 130. Server 130 can receive relevant data for various tasks and remotely control the operation of robot 110. Control device 120 collects data from robot 110 and communicates with server 130 via network 140 to send the data to server 130. Server 130 determines control commands for the robot based on the acquired data and sends the control commands to control device 120. Control device 120 controls robot 110 based on the received control commands. Server 130 can be various types of computing systems / servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, computing devices in cloud environments, etc.

[0030] In some embodiments, robot 110 can connect to and communicate with server 130 via network 140. Robot 110 can send relevant data on various tasks performed by robot 110 to server 130 in real time.

[0031] It should be understood that the structure and function of the various elements in environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of this disclosure. Furthermore, although in Figure 1 The illustration shows a robot, but embodiments of this disclosure can also be applied to other types of movable objects.

[0032] As briefly mentioned earlier, with the widespread application of autonomous mobile systems in fields such as industrial manufacturing, autonomous driving, and robotics, how to achieve safe and efficient path planning and trajectory prediction in complex environments has become one of the important research directions for mobile intelligent systems. The core objective of trajectory prediction is to generate a feasible and reasonable motion path for the moving body based on perceived environmental information and given task objectives.

[0033] Traditional trajectory planning methods typically employ a modular, phased approach. For example, a perception module first constructs an explicit environmental map, and then downstream modules perform path search and trajectory planning on this map. While this approach offers a clear structure, it inevitably introduces redundancy and loss during information transmission, impacting the overall efficiency and performance of trajectory generation.

[0034] With the development of deep learning technology, end-to-end trajectory generation architectures have gradually attracted attention. This approach attempts to output a complete, feasible trajectory by directly starting from raw perceptual inputs (such as camera images and laser data) and combining them with other information through a unified model. End-to-end methods have significant advantages in simplifying system structure and improving model synergy, and demonstrate strong performance in various scenarios. However, this approach still faces many challenges in practical applications, especially in terms of model generalization ability. Since most training relies on offline collected data, the model is prone to overfitting on the training set. Trajectory generation models lack the ability to cope with dynamic scene changes and scenarios requiring complex avoidance strategies.

[0035] In view of this, according to embodiments of the present disclosure, an improved scheme for trajectory generation is proposed to at least partially solve one or more of the above-mentioned problems. In this scheme, if a movable target object experiences an anomaly in its scene according to a first predicted trajectory, the first predicted trajectory and input information for generating the first predicted trajectory can be obtained. The first predicted trajectory is generated by a trajectory generation model based on the input information, which includes at least visual information specific to the scene. Further, a reference trajectory corresponding to the input information can be determined based on at least one candidate trajectory and the first predicted trajectory. The at least one candidate trajectory is obtained based on the input information. Then, the trajectory generation model is updated based on the input information and the reference trajectory.

[0036] In the embodiments of this disclosure, when an anomaly occurs during the execution of the target object according to the predicted trajectory, the corresponding input information is automatically acquired. A reference trajectory is determined using this input information and used as a training sample for trajectory generation. This approach implements a continuously evolving model optimization architecture, enabling the model to continuously adapt to new environments and tasks. In this way, targeted learning for abnormal scenarios can be achieved, effectively improving the adaptability of the trajectory generation model in complex or unseen environments and the reliability of trajectory generation.

[0037] Some exemplary embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0038] Figure 2 A schematic diagram of a training architecture 200 for a trajectory generation model according to some embodiments of the present disclosure is shown. The following will primarily refer to... Figure 1The example environment described herein is for illustrative purposes only and is not intended to limit the scope of the invention. Embodiments of this disclosure can also be applied to other types of environments. Furthermore, architecture 200 will be described relative to server 130, but this is merely exemplary. Architecture 200 may be implemented on a remote server of the target object (e.g., server 130), or on the target object's own control device (e.g., control device 120), or may be implemented collaboratively by the remote server and the control device.

[0039] In some embodiments, server 130 may, in response to an anomaly occurring in the scene of a movable target object according to a first predicted trajectory, acquire the first predicted trajectory and input information for generating the first predicted trajectory. The first predicted trajectory is generated by a trajectory generation model based on the input information.

[0040] like Figure 2 As shown, server 130 may include or deploy a trajectory generation subsystem 210. For example, trajectory generation subsystem 210 refers to an online environment that can run models and generate predicted trajectories in real time or periodically based on input information (or real-world perception information) of a movable target object.

[0041] In embodiments of this disclosure, the target object may include various suitable types of mobile objects, such as, but not limited to, autonomous driving systems (e.g., autonomous vehicles), drones, robots, etc. Such objects are capable of autonomously moving towards a predetermined target within their environment. For example, the target object may be one or more of the first robot 110-1, the second robot 110-2, and the third robot 110-3 described above (collectively or individually referred to as robot 110). Furthermore, for safe and efficient movement, the trajectory generation subsystem 210 needs to perceive the environment in which the target object is located. Based on the perception results, the next action of the target object can be determined (e.g., determining the direction of travel, planning a reasonable path, etc.).

[0042] Figure 3A A schematic diagram of the architecture 300 of a trajectory generation subsystem 210 according to some embodiments of the present disclosure is shown. In some embodiments, the input information includes at least visual information specific to the scene. Figure 3A As shown, the trajectory generation subsystem 210 can acquire multiple images 301 of the scene in which the movable target object is located. These images can typically cover the area in front of, to the sides of, or around the target object, and can reflect the spatial structure of the environment, passage space, and other content.

[0043] Furthermore, to support the trajectory generation model 312's perception of the scene, the trajectory generation subsystem 210 needs to extract image representations from the acquired images that can express the scene's semantics and spatial features. The trajectory generation subsystem 210 can determine the scene's image features 302 based on acquiring multiple images 301 from multiple angles of the scene where the movable target object is located. The image features 302 can be determined using a pre-trained image encoder 311. The encoder 311 can be a convolutional neural network, a transformer structure, a learning model that integrates multimodal perception, or any other image encoder structure that performs well in visual tasks. The task of the encoder 311 is to compress and represent the visual information (such as scene semantics, spatial geometry, etc.) contained in the image 301, extracting a set of intermediate image features 302 with high-level abstraction capabilities.

[0044] Continue to refer to Figure 3A Alternatively or additionally, the trajectory generation model 312 may generate a predicted trajectory 305 of the target object based on at least one of state information 303, task information 304, and image features 302. The state information indicates at least one of the target object's velocity, position, and orientation. The task information indicates constraint information related to generating the predicted trajectory.

[0045] The target object's state information 303 is used to describe its current motion state. The state information 303 may include, but is not limited to, the following: the target object's spatial position in the current coordinate system, used to determine its starting point or the context of its current environment; the target object's movement trend, which helps the trajectory prediction model infer its reasonable direction of motion and rate changes; and the target object's forward direction or orientation angle, etc.

[0046] The task information 304 of the target object is used to describe the external target or constraint conditions that the current trajectory prediction needs to meet, which may include, but are not limited to: target point location, used to guide the trajectory to be generated in a specific direction; task type, such as "driving on the side of the road", "avoiding pedestrians", "going to a designated area", etc., used to adjust the trajectory generation strategy; path preference or constraint, such as "avoiding congested areas as much as possible" "prioritizing the right side", etc., used to generate a trajectory that is more in line with actual needs.

[0047] In embodiments of this disclosure, the trajectory generation model 312 can adopt any structural form suitable for trajectory modeling or trajectory generation tasks. For example, it can be a recurrent neural network, an encoder-decoder structure, a Transformer-based sequence modeling network, a diffusion model, a flow matching model, or a combination of the above structures.

[0048] Continue to refer to Figure 2 The architecture 200 may further include a sample triggering subsystem 220 and a data acquisition subsystem 230. The trajectory generation subsystem 210 can generate a predicted trajectory based on the input perception information of the target object. The target object can move within its current scene according to the predicted trajectory. During this process, the sample triggering subsystem 220 can monitor the target object's movement status and behavioral outcomes, and determine whether the target object is experiencing any anomalies. Anomalies can include state anomalies, operational anomalies, and result anomalies. For example, state anomalies may include situations such as the target object braking suddenly or stopping at an unexpected location. Operational anomalies may refer to situations such as the target object failing to reach the target area, experiencing a collision, deviating from the trajectory, or failing to adjust the path. Result anomalies may include situations such as failing to reach the target area or endpoint, or the trajectory deviating from the preset route.

[0049] When an anomaly is determined to occur in the target object's environment, the sample triggering subsystem 220 can trigger a sample event signal, indicating that the current predicted trajectory and corresponding input information are "poorly performing samples." The data acquisition subsystem 230 can automatically send raw data 201, including the current input information and the predicted trajectory, to the server 130 when the sample event signal is triggered. The raw data 201 can form data samples associated with the abnormal behavior of the target object, which can then be used for subsequent policy optimization or model updates.

[0050] During the movement of the target object, the predicted trajectory generated by the trajectory generation model based on the current input information leads to the occurrence of anomalies. From the model's perspective, this input information can be understood as a "blind spot" or "unseen complex scene," which reflects the current capability boundary of the model. Therefore, the original data 201 can be used as sample data for model training.

[0051] In traditional trajectory generation model training methods, the model is typically fitted to offline training sets, including expert-labeled data. While the model may perform well on the offline training set, real-world deployment environments vary greatly, and the trajectory generation model may encounter situations never covered in the training set, resulting in poorly performing predicted trajectories. This highlights the limitation of the trajectory generation model's generalization ability. In such cases, it's possible to collect information on these poorly performing edge inputs from real-world deployment scenarios to dynamically correct the discrepancy between the training data distribution and the real-world environment distribution. This is a data-closed-loop adaptive mechanism. Training the trajectory generation model in this way can specifically improve its adaptability to complex scenarios.

[0052] Continue to refer to Figure 2Server 130 can determine a reference trajectory 203 corresponding to the input information based on at least one candidate trajectory and a first predicted trajectory. The at least one candidate trajectory can be obtained based on the input information using a strategy different from the trajectory generation model 312. Server 130 can determine an optimal reference trajectory 203 as a learning target by evaluating the performance differences between the first predicted trajectory and multiple candidate trajectories. Alternatively or additionally, the input information may also include at least one of the following: target object state information and task information.

[0053] Figure 3B A schematic diagram of an architecture 400 for determining a reference trajectory 203 according to some embodiments of the present disclosure is shown. Figure 3B As shown, the input information acquired by the data acquisition subsystem 230 may include multiple images 401 for generating the first predicted trajectory 403, and an obstacle map 402 of the target object.

[0054] Obstacle map 402 indicates the distribution of obstacles in the scene. Obstacle map 206, also known as an obstacle occupancy map, is a gridded representation of whether various spatial locations in the environment are occupied by obstacles. Each cell records a value indicating whether that location is occupied by an obstacle or the probability of it being occupied. An obstacle refers to an environmental entity or area that may obstruct, interfere with, or pose a collision risk to the normal path of a target object during its movement. Obstacles can be static or dynamic, and their types are not limited to walls, building structures, vehicles, equipment, pedestrians, shelves, boxes, etc., but can also be ground features such as elevation differences, potholes, or other obstacle areas that may affect passage.

[0055] In some embodiments, at least one candidate trajectory is obtained based on input information using a trajectory generation strategy different from the current trajectory generation model. The candidate trajectory can be used as a reference trajectory candidate and compared and evaluated with the first predicted trajectory 403 output by the current trajectory generation model to determine a better reference trajectory for updating the model.

[0056] Continue to refer to Figure 3B At least one candidate trajectory may include a candidate trajectory 404 generated by a first machine learning model 411, which is different from the trajectory generation model 312. The first machine learning model 411 may be another trajectory generation model, different from the trajectory generation model. For example, when the trajectory generation model 312 is based on a diffusion model, a model based on a variational autoencoder (VAE) structure can be used to generate the first candidate trajectory. The prediction results from the perspective of heterogeneous models are introduced through the first candidate trajectory.

[0057] Alternatively or additionally, the first machine learning model 411 can be a Vision Language Model (VLM). The server 130 can invoke the Vision Language Model to generate a second candidate trajectory based on the input information. The Vision Language Model can fuse visual perception information with semantic information such as language instructions, outputting structured trajectory or path candidate results. The Vision Language Model has strong general perception and understanding capabilities, providing a useful supplement for situations with complex scenes or semantic ambiguity.

[0058] Continue to refer to Figure 3B The candidate trajectory may also include the received manually annotated candidate trajectory 405. The candidate trajectory 405 may be an annotated trajectory obtained by the annotation module 412 based on the same input information and provided by a human annotator (e.g., a human expert or annotator). This trajectory can reflect human judgment on the reasonable behavior of the target object in the scene, and has high credibility and intuitive interpretability.

[0059] Continue to refer to Figure 3B The scoring module 413 can determine the reference trajectory 406 by comparing and sorting the multi-source candidate trajectories obtained through various suitable strategies and the first predicted trajectory 403. This helps to improve the accuracy and representativeness of the reference trajectory.

[0060] In some embodiments, the scoring module 413 can determine the corresponding trajectory scores of the first predicted trajectory and at least one candidate trajectory for the input information. The trajectory score can reflect the rationality, feasibility, or quality of each trajectory in the current input scenario. The trajectory score can be in the form of a specific score value, ranking, quality label, etc. The ranking module 413 can select the trajectory with the highest score from each trajectory based on its own trajectory score, as the reference trajectory 406 corresponding to the input information.

[0061] In some embodiments, the scoring module 413 may present the input information, the first predicted trajectory, and at least one candidate trajectory to a user (e.g., a human expert or annotator). The user can score or rank the trajectories based on intuitive judgment, thereby determining the corresponding trajectory scores for the first predicted trajectory and at least one candidate trajectory. The scoring module 413 may determine the corresponding trajectory score for each trajectory based on user input received for the first predicted trajectory and at least one candidate trajectory.

[0062] In some embodiments, the scoring module 413 may further utilize a second machine learning model to determine corresponding trajectory scores for the first predicted trajectory and at least one candidate trajectory, based on the input information. The second machine learning model may be a pre-trained reward model or ranking model, etc. For example, the second machine learning model may use the input information and the trajectory to be evaluated as input, and output a score for that trajectory in the current input scenario.

[0063] Through the process of determining candidate trajectory scores and selecting reference trajectories described above, server 130 can determine the optimal learning reference trajectory in a more objective and controllable manner. The reference trajectory can then be used to construct high-quality training samples, improving the performance of the trajectory generation model in complex environments.

[0064] In some embodiments, the candidate trajectories and reference trajectories determined by various means may indicate a continuous sequence of waypoints. Alternatively or additionally, candidate trajectories and reference trajectories may also refer to a class of intermediate control representations used to guide the movement of a target object, such as waypoints.

[0065] In some embodiments, server 130 can update the trajectory generation model based on input information and a reference trajectory. For example, reference... Figure 2 The server 130 can combine the input information with the corresponding reference trajectory to form a new training sample 204. The training subsystem 240 can use this training sample 204 to update the existing trajectory generation model. This training process can employ supervised learning, or it can use ranking learning or reinforcement learning, etc.

[0066] As an example, the training subsystem 240 can utilize the trajectory generation model to generate a second predicted trajectory for the target object based on the input information. Furthermore, the training subsystem 240 can update the trajectory generation model based on the difference between the second predicted trajectory and the reference trajectory. This difference can be measured in various ways, such as the Euclidean distance between trajectory points, path deviation, difference in motion direction, trajectory score difference, etc. The training subsystem 240 can construct a loss function to measure the degree of deviation of the second predicted trajectory from the reference trajectory and use this function to backpropagate and update the parameters of the trajectory generation model.

[0067] It should be noted that the trajectory generation model trained using the aforementioned data closed-loop mechanism can be deployed in actual operating systems for use by movable objects when performing operational tasks. For example, the trajectory generation model can generate predicted trajectories in real time to guide the movement of movable objects based on the perceived information of the current environment and the requirements of the target task, thereby guiding their path selection and behavior control within the scene.

[0068] As can be more clearly understood from the description of the various embodiments of this disclosure, in the embodiments of this disclosure, samples with poor inference performance are triggered and collected during the actual operation of the target object. These samples are then combined with candidate trajectories and reference trajectories generated through manual or algorithmic evaluation to form high-quality samples that can be used for training. In this way, the model's adaptability to abnormal scenarios and the robustness of behavioral decisions can be effectively improved. Compared with traditional methods that rely on static offline data training, continuous learning and performance iteration for complex dynamic environments can be achieved, thereby enhancing the generalization ability, security, and practicality of the trajectory generation model.

[0069] Figure 4 A flowchart of a process 400 for trajectory generation according to some embodiments of the present disclosure is shown. Process 400 may be executed by a remote server of the target object (e.g., server 130), or by a control device of the target object itself (e.g., control device 120), or may be performed collaboratively by the remote server and the control device. Reference is made below. Figure 1 Describe the process 400.

[0070] In box 410, server 130, in response to an anomaly occurring in the scene of a movable target object based on a first predicted trajectory, acquires the first predicted trajectory and input information used to generate the first predicted trajectory. The first predicted trajectory is generated by a trajectory generation model based on the input information, which includes at least visual information specific to the scene.

[0071] In box 420, server 130 determines a reference trajectory corresponding to the input information based on at least one candidate trajectory and a first predicted trajectory; and

[0072] In box 430, server 130 updates the trajectory generation model based on input information and reference trajectory.

[0073] In some embodiments, at least one candidate trajectory includes at least one of the following: a first candidate trajectory generated based on input information using another trajectory generation model different from the trajectory generation model, a second candidate trajectory generated based on input information using a visual language model, or a third candidate trajectory received with manual annotation.

[0074] In some embodiments, determining a reference trajectory corresponding to the input information includes: determining corresponding trajectory scores for a first predicted trajectory and at least one candidate trajectory for the input information; and selecting a reference trajectory from the first predicted trajectory and at least one candidate trajectory based on the corresponding trajectory scores.

[0075] In some embodiments, the corresponding trajectory score is determined by: presenting input information, a first predicted trajectory, and at least one candidate trajectory to a user; and receiving user input for the first predicted trajectory and at least one candidate trajectory, the user input including at least the corresponding trajectory scores for the first predicted trajectory and at least one candidate trajectory.

[0076] In some embodiments, the corresponding trajectory score is determined by using a second machine learning model to determine the corresponding trajectory scores of the first predicted trajectory and at least one candidate trajectory based on the input information.

[0077] In some embodiments, updating the trajectory generation model includes: using the trajectory generation model to generate a second predicted trajectory for the target object based on input information; and updating the trajectory generation model based on the difference between the second predicted trajectory and the reference trajectory.

[0078] In some embodiments, the input information further includes at least one of the following: state information of the target object, which indicates at least one of the target object's velocity, position, and orientation; or task information of the target object, which indicates constraint information related to generating the first predicted trajectory.

[0079] In some embodiments, the visual information includes multiple images acquired for the scene and an obstacle map indicating the distribution of obstacles in the scene.

[0080] In some embodiments, the multiple images include images of a scene captured from multiple angles.

[0081] Embodiments of this disclosure also provide corresponding apparatus for implementing the above methods or processes. Figure 5 A schematic structural block diagram of an apparatus 500 for trajectory generation according to some embodiments of the present disclosure is shown. The apparatus 500 may be implemented as or included in a server 130. The various modules / components in the apparatus 500 may be implemented by hardware, software, firmware, or any combination thereof.

[0082] refer to Figure 5 As shown, the device 500 includes an acquisition module 510 configured to acquire a first predicted trajectory and input information for generating the first predicted trajectory in response to an anomaly occurring in the scene of a movable target object according to a first predicted trajectory, the first predicted trajectory being generated by a trajectory generation model based on the input information, the input information including at least visual information for the scene; a reference trajectory determination module 520 configured to determine a reference trajectory corresponding to the input information based on at least one candidate trajectory and the first predicted trajectory, the at least one candidate trajectory being obtained based on the input information; and an update module 530 configured to update the trajectory generation model based on the input information and the reference trajectory.

[0083] In some embodiments, at least one candidate trajectory includes at least one of the following: a first candidate trajectory generated based on input information using another trajectory generation model different from the trajectory generation model, a second candidate trajectory generated based on input information using a visual language model, or a third candidate trajectory received with manual annotation.

[0084] In some embodiments, the reference trajectory determination module 520 is further configured to determine corresponding trajectory scores of a first predicted trajectory and at least one candidate trajectory for the input information; and to select a reference trajectory from the first predicted trajectory and at least one candidate trajectory based on the corresponding trajectory scores.

[0085] In some embodiments, the corresponding trajectory score is determined by: presenting input information, a first predicted trajectory, and at least one candidate trajectory to a user; and receiving user input for the first predicted trajectory and at least one candidate trajectory, the user input including at least the corresponding trajectory scores for the first predicted trajectory and at least one candidate trajectory.

[0086] In some embodiments, the corresponding trajectory score is determined by using a second machine learning model to determine the corresponding trajectory scores of the first predicted trajectory and at least one candidate trajectory based on the input information.

[0087] In some embodiments, the update module 530 is further configured to: generate a second predicted trajectory for the target object based on the input information using the trajectory generation model; and update the trajectory generation model based on the difference between the second predicted trajectory and the reference trajectory.

[0088] In some embodiments, the input information further includes at least one of the following: state information of the target object, which indicates at least one of the target object's velocity, position, and orientation; or task information of the target object, which indicates constraint information related to generating the first predicted trajectory.

[0089] In some embodiments, the visual information includes multiple images acquired for the scene and an obstacle map indicating the distribution of obstacles in the scene.

[0090] In some embodiments, the multiple images include images of a scene captured from multiple angles.

[0091] Figure 6 A block diagram is shown of an electronic device 600 in which one or more embodiments of the present disclosure may be implemented. The electronic device 600 may, for example, be used to implement... Figure 1 The control device 120 shown. It should be understood that, Figure 6 The electronic device 600 shown is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein.

[0092] refer to Figure 6 Electronic device 600 is in the form of a general-purpose electronic device. Components of electronic device 600 may include, but are not limited to, one or more processors or processing units 610, memory 620, storage device 630, one or more communication units 640, one or more input devices 650, and one or more output devices 660. Processing unit 610 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 620. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 600.

[0093] Electronic device 600 typically includes multiple computer storage media. Such media can be any available media accessible to electronic device 600, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 620 can be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 630 can be removable or non-removable media and can include machine-readable media, such as flash drives, disks, or any other media capable of storing information and / or data and accessible within electronic device 600.

[0094] Electronic device 600 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 6 As shown, disk drives for reading from or writing to removable, non-volatile disks (e.g., "floppy disks") and optical disk drives for reading from or writing to removable, non-volatile optical disks can be provided. In these cases, each drive can be connected to a bus (not shown) via one or more data media interfaces. Memory 620 may include computer program product 625 having one or more program modules configured to perform various methods or actions of various embodiments of this disclosure.

[0095] The communication unit 640 enables communication with other electronic devices via a communication medium. Additionally, the functionality of the components of the electronic device 600 can be implemented using a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, the electronic device 600 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.

[0096] Input device 660 can be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 660 can be one or more output devices, such as a monitor, speaker, printer, etc. Electronic device 600 can also communicate with one or more external devices (not shown) via communication unit 640 as needed. These external devices include storage devices, display devices, etc., and can communicate with one or more devices that enable user interaction with electronic device 600, or with any device that enables electronic device 600 to communicate with one or more other electronic devices (e.g., network card, modem, etc.). Such communication can be performed via input / output (I / O) interface (not shown).

[0097] According to an exemplary implementation of this disclosure, a computer-readable storage medium is provided that stores computer-executable instructions thereon, wherein the computer-executable instructions are executed by a processor to implement the methods described above. According to an exemplary implementation of this disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, which are executed by a processor to implement the methods described above.

[0098] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0099] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0100] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0101] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0102] Various implementations of this disclosure have been described above. The foregoing description is exemplary and not exhaustive, nor is it limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is determined to best explain the principles, practical applications, or improvements to technology in the market, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. A method for trajectory generation, comprising: In response to an anomaly occurring in the scene of a movable target object based on a first predicted trajectory, the first predicted trajectory and input information for generating the first predicted trajectory are obtained, wherein the first predicted trajectory is generated by the trajectory generation model based on the input information, and the input information includes at least visual information specific to the scene. as well as Based on at least one candidate trajectory and the first predicted trajectory, a reference trajectory corresponding to the input information is determined, wherein the at least one candidate trajectory is obtained based on the input information; as well as The trajectory generation model is updated based on the input information and the reference trajectory.

2. The method of claim 1, wherein the at least one candidate trajectory comprises at least one of the following: A first candidate trajectory is generated based on the input information using a different trajectory generation model. The second candidate trajectory generated based on the input information using a visual language model, or The third candidate trajectory received with manual annotation.

3. The method according to claim 1, wherein determining the reference trajectory corresponding to the input information comprises: Determine the corresponding trajectory scores of the first predicted trajectory and the at least one candidate trajectory for the input information; as well as Based on the corresponding trajectory score, the reference trajectory is selected from the first predicted trajectory and the at least one candidate trajectory.

4. The method according to claim 3, wherein the corresponding trajectory score is determined by: Presenting the input information, the first predicted trajectory, and the at least one candidate trajectory to the user; and Receive user input for the first predicted trajectory and the at least one candidate trajectory, the user input including at least the corresponding trajectory scores for the first predicted trajectory and the at least one candidate trajectory.

5. The method of claim 3, wherein the corresponding trajectory score is determined by: Using a second machine learning model, based on the input information, the corresponding trajectory scores for the first predicted trajectory and the at least one candidate trajectory are determined respectively.

6. The method of claim 1, wherein updating the trajectory generation model comprises: Using the trajectory generation model, a second predicted trajectory for the target object is generated based on the input information; as well as The trajectory generation model is updated based on the difference between the second predicted trajectory and the reference trajectory.

7. The method of claim 1, wherein the input information further comprises at least one of the following: The state information of the target object, wherein the state information indicates at least one of the target object's speed, position, and orientation, or The task information of the target object, wherein the task information indicates constraint information related to generating the first predicted trajectory.

8. The method of claim 1, wherein the visual information comprises a plurality of images acquired for the scene and an obstacle map indicating the distribution of obstacles in the scene.

9. The method of claim 8, wherein the plurality of images comprises images of the scene acquired from multiple angles.

10. An apparatus for trajectory generation, comprising: The acquisition module is configured to acquire the first predicted trajectory and input information for generating the first predicted trajectory in response to an anomaly occurring in the scene of a movable target object based on the first predicted trajectory, wherein the first predicted trajectory is generated by the trajectory generation model based on the input information, and the input information includes at least visual information specific to the scene. as well as The reference trajectory determination module is configured to determine a reference trajectory corresponding to the input information based on at least one candidate trajectory and the first predicted trajectory, wherein the at least one candidate trajectory is obtained based on the input information; as well as The update module is configured to update the trajectory generation model based on the input information and the reference trajectory.

11. An electronic device, comprising: At least one processing unit; as well as At least one memory, coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 9 when executed by the at least one processing unit.

12. A computer-readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the method according to any one of claims 1 to 9.

13. A computer program product comprising computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the method according to any one of claims 1 to 9.

Images