Vehicle control method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a vehicle control method and device, equipment and a storage medium. The method comprises the steps that in response to the situation that the autonomous vehicle is in a trapped state, at least one candidate trapped-out track is generated through a trapped-out model, and the trapped-out model is different from a track planning model used by the autonomous vehicle in a non-trapped state; sending an assistance request to the remote device, wherein the assistance request indicates at least one candidate de-trapping track; receiving an assistance message from the remote device, the assistance message indicating a confirmation of a target escape trajectory of the at least one candidate escape trajectory; and controlling the autonomous vehicle to run based on the target escape trajectory. In this way, according to the embodiment of the invention, the escape efficiency and the escape success rate of the autonomous vehicle can be effectively improved.

Description

TECHNICAL FIELD

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and in particular, to a vehicle control method, apparatus, device and computer readable storage medium. BACKGROUND

[0002] With the rapid development of autonomous driving technology, its performance in common driving scenarios has made important progress. However, in the actual driving environment, how to deal with when the autonomous vehicle is in a trapped state is also a key bottleneck for large-scale landing application of autonomous driving technology. How to accurately control the autonomous vehicle to complete the escape is the focus of attention. SUMMARY

[0003] In a first aspect of the present disclosure, a vehicle control method is provided. The method comprises: in response to an autonomous vehicle being in a trapped state, generating at least one candidate escape trajectory by using an escape model, the escape model being different from a trajectory planning model used by the autonomous vehicle in a non-trapped state; sending an assistance request to a remote device, the assistance request indicating the at least one candidate escape trajectory; receiving an assistance message from the remote device, the assistance message indicating confirmation of a target escape trajectory in the at least one candidate escape trajectory; and controlling the autonomous vehicle to travel based on the target escape trajectory.

[0004] In a second aspect of the present disclosure, an apparatus for vehicle control is provided. The apparatus comprises: a generation module configured to, in response to an autonomous vehicle being in a trapped state, generate at least one candidate escape trajectory by using an escape model, the escape model being different from a trajectory planning model used by the autonomous vehicle in a non-trapped state; a sending module configured to send an assistance request to a remote device, the assistance request indicating the at least one candidate escape trajectory; a receiving module configured to receive an assistance message from the remote device, the assistance message indicating confirmation of a target escape trajectory in the at least one candidate escape trajectory; and a control module configured to control the autonomous vehicle to travel based on the target escape trajectory.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device comprises 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. The instructions, when executed by the at least one processor, cause the device to perform the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer readable storage medium is provided. The computer readable storage medium has stored thereon computer executable instructions executable by a processor to implement the method of the first aspect.

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

[0008] It should be understood that nothing in the Summary is to be construed as a limitation on the scope of the embodiments of the present disclosure. Other features, aspects, and advantages of the present disclosure will become apparent from the following description, which is given by way of example only to illustrate the features and concepts underlying the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0009] The above and other features, aspects, and advantages of embodiments of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings. In the drawings, like reference numerals refer to like elements, wherein: Figure 1 A schematic diagram illustrating an example environment in which embodiments of the present disclosure can be implemented is shown; Figure 2 A flowchart of a vehicle control process according to some embodiments of the present disclosure is shown; Figure 3 A flowchart of a training process of an escape model according to some embodiments of the present disclosure is shown; Figure 4 A flowchart of a vehicle escape process according to some embodiments of the present disclosure is shown; Figure 5 A flowchart of a vehicle escape process according to some embodiments of the present disclosure is shown; Figure 6 A schematic block diagram of an apparatus for vehicle control according to certain embodiments of the present disclosure is shown; Figure 7 A block diagram of an electronic device capable of implementing various embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0010] Embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings. While several embodiments of the present disclosure are illustrated, it is to be understood that the present disclosure is not to be limited to the precise embodiments and that various changes and modifications can be effected therein by one skilled in the art. It should be understood that the drawings and detailed description thereto are not intended to limit the scope of the present disclosure to the precise embodiments illustrated. Rather, the embodiments are intended to illustrate by way of example the features and concepts underlying the present disclosure.

[0011] It should be noted that the headings provided in any section / subsection of the disclosure herein are not limitations. Various embodiments are described throughout this document and any type of embodiment can be included under any section / subsection. Furthermore, embodiments described in any section / subsection can be combined with any other embodiment described in the same section / subsection and / or a different section / subsection in any manner.

[0012] In the description of embodiments of the disclosure, the term "includes" and its derivatives, such as "including," should be understood in an open-ended way, that is, "including but not limited to." The term "based on" should be understood as "based at least in part 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 can also be included below. The terms "first," "second," and the like can refer to different or the same objects. Other explicit and implicit definitions can also be included below.

[0013] Data of users, acquisition and / or use of data, etc. can be involved in embodiments of the disclosure. These aspects all comply with corresponding laws and regulations and relevant provisions. In embodiments of the disclosure, all data collection, acquisition, processing, processing, forwarding, use, etc. are carried out on the premise that the user is aware of and confirms. Accordingly, when implementing embodiments of the disclosure, the type of data or information that can be involved, the scope of use, the use scenario, etc. should be notified to the user and authorized by the user in a proper manner according to relevant laws and regulations. The specific notification and / or authorization manner can vary according to the actual situation and application scenario, and the scope of the disclosure is not limited in this respect.

[0014] In the specification and embodiments of the disclosure, if personal information processing is involved, it will be processed on the premise of legality (for example, obtaining the consent of the subject of personal information, or being necessary for the performance of a contract, etc.), and only within the prescribed or agreed range. Users refuse to process personal information other than the necessary information required for basic functions, which will not affect the user's use of basic functions.

[0015] Conventionally, when the autonomous vehicle is in a trapped state, the remote device can also issue a trapped instruction to guide the autonomous vehicle to escape from the trapped state, but the trapped trajectory indicated in the trapped instruction is generally generated by manual point drawing operation. The accuracy of the trapped trajectory corresponding to the manual point drawing operation may not be high, which may affect the corresponding trapped success rate of the autonomous vehicle, and this generation method of the trapped trajectory will also bring problems of high labor cost and low trapped efficiency.

[0016] Embodiments of the present disclosure propose a vehicle control scheme. According to the scheme, at least one candidate escape trajectory can be generated by an escape model in response to the autonomous vehicle being in a trapped state, the escape model being different from a trajectory planning model used by the autonomous vehicle in a non-trapped state. Further, an assistance request can be sent to a remote device, the assistance request indicating the at least one candidate escape trajectory. Further, an assistance message can be received from the remote device, the assistance message indicating a confirmation of a target escape trajectory among the at least one candidate escape trajectory. Further, the autonomous vehicle can be controlled to travel based on the target escape trajectory.

[0017] Based on such a manner, embodiments of the present disclosure can automatically generate at least one candidate escape trajectory by an escape model for a remote device to select, to control the autonomous vehicle to complete escape based on a determined target escape trajectory. Compared with a method of generating an escape trajectory by manual dot operation, embodiments of the present disclosure can effectively improve the reliability of the escape trajectory and the generation efficiency of the escape trajectory, and can further improve the escape efficiency and the escape success rate of the autonomous vehicle.

[0018] Example Environment Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown.

[0019] An autonomous vehicle 110 is schematically shown in this example environment 100. The autonomous vehicle 110 can be any type of vehicle that can carry people and / or objects and move by a power system such as an engine. Examples of the autonomous vehicle 110 include, but are not limited to, a car, a truck, a bus, an electric vehicle, a motorcycle, a recreational vehicle, a train, and the like. As an example, the autonomous vehicle 110 of the present disclosure can be a vehicle with certain assisted driving capability or autonomous driving capability, which is also referred to as a smart driving vehicle.

[0020] The electronic device 120 can generate at least one candidate escape trajectory by an escape model in response to the autonomous vehicle 110 being in a trapped state, the escape model being different from a trajectory planning model used by the autonomous vehicle 110 in a non-trapped state. Further, the electronic device 120 can send an assistance request to a remote device, the assistance request indicating the at least one candidate escape trajectory. Further, the electronic device 120 can receive an assistance message from the remote device, the assistance message indicating a confirmation of a target escape trajectory among the at least one candidate escape trajectory. Further, the autonomous vehicle 110 can be controlled to travel based on the target escape trajectory.

[0021] In some embodiments, the electronic device 120 can be deployed on the autonomous vehicle 110, and can also be independent of other devices of the autonomous vehicle 110.

[0022] The electronic device 120 can be any type of mobile terminal, fixed terminal, or portable terminal including a mobile handset, a tablet computer, a laptop computer, a notebook computer, a netbook computer, a smartbook, a media computer, a multimedia tablet, a palmtop computer, a portable game console, a VR / AR device, a Personal Communication System (PCS) device, a personal navigation device, a Personal Digital Assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an electronic book device, a game device, or any combinations of the aforementioned and the like, including accessories and peripherals for these devices, or any combinations thereof. In some embodiments, the electronic device 120 can also support any type of interface to the user (such as “wearable” circuitry, etc.).

[0023] The electronic device 120 can also be a standalone physical server, a server cluster or distributed system of multiple physical servers, a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms, and the like basic cloud computing services. The electronic device 120 may, for example, include a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, and the like.

[0024] The remote device 130 can also be a standalone physical server, a server cluster or distributed system of multiple physical servers, a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms, and the like basic cloud computing services. The remote device 130 may, for example, include a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, and the like.

[0025] It should be understood that the structure and functionality of the various elements in the environment 100 are described for illustrative purposes only and are not intended to imply any limitations on the scope of the present disclosure.

[0026] Example Process Figure 2 A flowchart of a vehicle control process 200 is shown in accordance with some embodiments of the present disclosure. The process 200 can be implemented at the electronic device 120. The process 200 is described below with reference to the environment 100. Figure 1 The process 200 is described.

[0027] At block 210, the electronic device 120 generates at least one candidate escape trajectory for the autonomous vehicle 110 in response to the autonomous vehicle 110 being in the trapped state, the escape trajectory being generated by an escape model that is different from a trajectory planning model used by the autonomous vehicle 110 in a non-trapped state.

[0028] In some embodiments, the autonomous vehicle 110 can be a vehicle that navigates and travels independently of human operation, with some level of assisted driving or autonomous driving capability, such a vehicle is also referred to as a smart driving vehicle.

[0029] In some embodiments, the autonomous vehicle 110 being in the trapped state can represent that the autonomous vehicle 110 is in a state that it cannot move normally or continue to travel due to some reasons. In some embodiments, the trapped state can be related to any suitable factors, such as can be related to a perception obstacle, a system failure, a physical failure, and the like of the autonomous vehicle 110, which are not described herein. In some embodiments, the escape model can be deployed on the vehicle side, and can also be deployed on other devices.

[0030] In some embodiments, the escape model can be any suitable machine learning model, such as can be a multi-modal model that can be configured to support generating an escape trajectory based on any one form of data or a combination of at least two forms of data, such as text, speech, image, or video.

[0031] In some embodiments, the escape model can be configured to generate at least one reference escape trajectory for the autonomous vehicle 110 in the trapped state that can help the autonomous vehicle 110 escape, to ensure that the autonomous vehicle 110 can safely and successfully complete the escape.

[0032] It should be noted that the escape model is different from the trajectory planning model used by the autonomous vehicle 110 in the non-trapped state, the escape model corresponds to the trapped scenario of the autonomous vehicle 110, i.e., the autonomous vehicle 110 has certain difficulties in completing the escape without remote assistance from a remote device. The trajectory planning model can be configured to generate a travel trajectory for the autonomous vehicle 110 in the non-trapped state to assist the autonomous vehicle 110 in traveling. That is, the trajectory planning model can be configured to generate a travel plan under normal driving conditions, which can find a path for the autonomous vehicle 110 to successfully travel from a starting location to a destination based on factors such as travel time, energy consumption, comfort, safety, and the like, without the need for remote assistance from a remote device.

[0033] In some embodiments, the trajectory planning model can be any suitable machine learning model. The trajectory planning model can also be deployed on the vehicle side, and can also be deployed on other devices.

[0034] To enable the stuck-unstuck model to generate reliable unstuck trajectories, the electronic device 120 can obtain description information associated with the autonomous vehicle 110. The description information can be any suitable information that assists the stuck-unstuck model to accurately analyze the stuck state of the autonomous vehicle 110 to generate effective unstuck trajectories.

[0035] As an example, the description information can indicate at least one of: environment information associated with the autonomous vehicle 110, state information of the autonomous vehicle 110. As another example, the description information can also indicate predicted state information corresponding to other traffic participants. The other traffic participants can be other vehicles or pedestrians, etc. associated with the autonomous vehicle 110. The predicted state information can be state information corresponding to the other traffic participants that is predicted based on historical state information corresponding to the other traffic participants. The predicted state information can indicate a direction of movement of the other traffic participants, a speed of movement, etc. For example, the predicted state information can be a speed at which a vehicle on the left side of the autonomous vehicle 110 is predicted to travel based on a speed at which the vehicle has traveled in the past ten seconds.

[0036] In some embodiments, the state information of the autonomous vehicle 110 can include, but is not limited to, a speed of travel of the autonomous vehicle 110, a status of each perception module of the autonomous vehicle 110, a vehicle pose of the autonomous vehicle 110, etc.

[0037] In some embodiments, the environment information associated with the autonomous vehicle 110 can include, but is not limited to, map information and / or obstacle information determined based on perception information of the autonomous vehicle 110.

[0038] In some embodiments, the perception information can be information obtained by a perception module of the autonomous vehicle 110, and the perception module can include, but is not limited to, sensors of the autonomous vehicle 110, such as a camera, a lidar, etc. In some embodiments, the obstacle information can be other traffic participants, such as other vehicles, pedestrians, motorcycles, bicycles, construction zones, etc.

[0039] In some embodiments, the map information can be objects pre-labeled in an offline map or pre-set map data, and the objects can include, but are not limited to, fences, sidewalks, curbs, flower beds, road studs, etc.

[0040] In some embodiments, the environment information can include other any suitable information in addition to the map information and the obstacle information, such as road surface conditions (e.g., muddy, snowy) associated with the autonomous vehicle 110, weather conditions, state information of a traffic light, etc., which are not described herein.

[0041] Further, the electronic device 120 can provide the description information to the escape model to generate at least one candidate escape trajectory. For example, the electronic device 120 can input the description information to the escape model to obtain a set of escape trajectories output by the escape model. The set of escape trajectories can be specific guidance information indicating how to assist the autonomous vehicle 110 to recover from the trapped state to the non-trapped state, such as driving direction, obstacle avoidance, rotation angle, driving speed, and the like. Further, the electronic device 120 can determine the set of escape trajectories as the at least one candidate escape trajectory. To ensure the reliability of the escape trajectory, as an example, the electronic device 120 can output the set of escape trajectories and scores corresponding to the set of escape trajectories. Further, the electronic device 120 determines at least one escape trajectory in the set of escape trajectories that satisfies a condition as the at least one candidate escape trajectory.

[0042] In some embodiments, the electronic device 120 can generate vectorized data of the description information. As an example, the electronic device 120 can generate the vectorized data of the description information based on a predetermined feature extraction module. Since the content items indicated by the description information can be multiple, the vectorized data can be a tensor matrix mapped by embedding vectors corresponding to the multiple content items. For example, in response to the multiple content items indicated by the description text including information of other road users, driving speed of the autonomous vehicle 110, and information of lane lines associated with the autonomous vehicle 110, the electronic device 120 can generate a first feature vector corresponding to the information of other road users, a second feature vector corresponding to the driving speed of the autonomous vehicle 110, and a third feature vector corresponding to the information of lane lines. Further, the electronic device 120 can concatenate the first feature vector, the second feature vector, and the third feature vector into the tensor matrix.

[0043] Further, the electronic device 120 can process the vectorized data by using an encoder of the escape model to generate a sequence of fused features. As an example, the electronic device 120 can generate the sequence of fused features based on a self-attention mechanism and the vectorized data by using the encoder. Specifically, the attention mechanism can be used to indicate that the escape model dynamically focuses on information in different parts of the vectorized data when processing the vectorized data, thereby capturing the dependency between the parts. Specifically, since the vectorized data can indicate multiple content items, the electronic device 120 can use the encoder to interact and fuse the features corresponding to each content item with the features corresponding to other content items to output scene features rich in global interaction relationships for each content item.

[0044] Further, the electronic device 120 can generate at least one candidate escape trajectory based on the sequence of fused features by using a decoder of the escape model.

[0045] To improve the availability of the candidate escape trajectories and generate reliable escape trajectories that can support the autonomous vehicle 110 to escape, the electronic device 120 can obtain a plurality of query vectors. In some embodiments, the plurality of query vectors can correspond to different driving intents. For example, one query vector can correspond to a driving intent of turning left by the vehicle, and another query vector can correspond to a driving intent of following the vehicle.

[0046] In some embodiments, the query vector in the plurality of query vectors indicates at least one of a lateral offset sequence relative to a reference boundary, a longitudinal offset parameter relative to a start position. In some embodiments, the reference boundary can be a lane line, or any other appropriate boundary. In some embodiments, the lateral offset sequence can be a sequence of points determined by reference lines corresponding to different lanes, which can represent the horizontal position of the autonomous vehicle 110 on the road. In some embodiments, the lateral offset sequence relative to the reference boundary can represent a sequence of positions corresponding to a plurality of sampling points of the lateral reference path.

[0047] In some embodiments, the longitudinal offset parameter can be a displacement in the vertical direction, which can represent the vertical position change of the autonomous vehicle 110 on the road. In some embodiments, the longitudinal offset parameter can be a learnable parameter representing the degree of path advancement.

[0048] Specifically, the electronic device 120 can encode the lateral offset sequence and combine the obtained encoded data with the longitudinal offset parameter itself to generate the query vector.

[0049] Further, the electronic device 120 can provide the fusion feature sequence and the plurality of query vectors to the decoder to generate at least one escape trajectory. In some embodiments, the decoder can focus on feature elements corresponding to different driving intents based on the plurality of query vectors, thereby generating trajectory features of different modalities and reasonable.

[0050] In addition, by introducing the plurality of query vectors, the disclosure can also make the gradient calculated by the loss function corresponding to the escape model optimize the escape model in different ways, effectively improving the reliability of the escape trajectory generated by the escape model and the diversity of the modalities corresponding to the escape trajectory.

[0051] Taking the server 130 training the escape model as an example, the training process of the escape model is described below. Of course, the escape model can also be trained by other devices.

[0052] Server 130 can acquire scene data associated with a reference escape scenario. A reference escape scenario represents a situation where the reference vehicle is trapped, meaning it is currently unable to move or continue driving. Scene data may include, but is not limited to, descriptive data associated with the reference vehicle. The descriptive data may indicate at least one of the following: environmental information associated with the reference vehicle, and the reference vehicle's state information. As another example, the descriptive data may also indicate predicted state information for other traffic participants associated with the reference vehicle. Other traffic participants may be other vehicles or pedestrians associated with the reference vehicle, etc. Predicted state information may be the state information predicted for other traffic participants based on their historical state information. For example, predicted state information may be the speed of the reference vehicle to be driven in the future, determined based on the speed of the vehicle to its left over the past ten seconds.

[0053] In some embodiments, the state data of the reference vehicle may include, but is not limited to, the driving speed of the reference vehicle, the status of each sensing module of the reference vehicle, the vehicle attitude of the reference vehicle, etc.

[0054] In some embodiments, environmental information associated with the reference vehicle may include, but is not limited to, map information and / or obstacle information determined based on the reference vehicle's perception information.

[0055] In some embodiments, the perception information may be information obtained by a perception module of a reference vehicle, and this perception module may include, but is not limited to, sensors of the reference vehicle, such as cameras, lidar, etc. In some embodiments, obstacle information may be other traffic participants, such as other vehicles, pedestrians, motorcycles, bicycles, construction zones, etc.

[0056] In some embodiments, map information may be objects pre-marked in offline maps or pre-defined map data, and these objects may include, but are not limited to, fences, sidewalks, curbs, flower beds, kerb, etc.

[0057] In some embodiments, environmental information may include, in addition to map information and obstacle information, any other appropriate information, such as road conditions (e.g., mud, snow) associated with the reference vehicle, weather conditions, traffic light status information, etc., which will not be elaborated here.

[0058] Furthermore, server 130 can construct training samples based on scene data and reference escape trajectories associated with the reference escape scene.

[0059] In some embodiments, the reference escape trajectory can be a pre-labeled tag corresponding to the training sample, which can indicate a standard escape trajectory that can be referenced to control the reference vehicle to complete the escape in this scenario data. This standard escape trajectory is a confirmed reliable trajectory that can assist the reference vehicle in completing the escape. As an example, a remote device can send multiple points issued based on manual point drawing operations to the electronic device 120. Further, the electronic device 120 can generate a reference escape trajectory based on these multiple points and control the reference vehicle to drive based on this reference escape trajectory to complete the escape in this escape scenario. That is, this reference escape trajectory can be the trajectory traveled by the reference vehicle in the actual escape process in a scenario where the vehicle has successfully escaped the obstacle through manual point drawing operations. This trajectory can assist the reference vehicle in completing the escape.

[0060] In some embodiments, server 130 can train an escape model based on training samples. As an example, server 130 can provide training samples to the escape model to obtain a first escape trajectory. Further, server 130 can determine a first loss value based on a comparison of the first escape trajectory and a reference escape trajectory. Specifically, if the first escape trajectory includes multiple escape trajectories, server 130 can determine the first loss value based on a comparison of the trajectory features corresponding to the first escape trajectory and the trajectory features corresponding to the reference escape trajectory, as well as a comparison of confidence scores. For each of these multiple escape trajectories, the confidence score corresponding to that escape trajectory can indicate the degree of matching between that escape trajectory and the reference escape trajectory, where a higher degree of matching results in a higher confidence score. As an example, the confidence score corresponding to the escape trajectory with the highest degree of matching among the multiple escape trajectories can be a first value, and the confidence scores corresponding to the remaining escape trajectories can be second values, where the first value can be 1 and the second value can be 0.

[0061] Furthermore, server 130 can train an escape model based on the first loss value. It should be noted that the loss function required to determine the first loss value can be set according to requirements, which will not be elaborated here.

[0062] In other embodiments, server 130 may construct sample pairs based on training samples. These sample pairs may include positive and negative samples. In some embodiments, positive samples may be samples obtained by performing a first perturbation process on the training samples. The degree of perturbation corresponding to the first perturbation process may be less than a threshold, i.e., positive samples may be samples obtained by performing a small-amplitude perturbation on the training samples. As an example, server 130 may generate positive samples by adjusting the position of the reference vehicle, and / or adjusting the speed of the reference vehicle, and / or removing other vehicles that do not interact with the reference vehicle, to improve the robustness of the escape model.

[0063] In some embodiments, negative samples can be samples obtained by performing a second perturbation process on the training samples. The degree of perturbation corresponding to the second perturbation process can be greater than or equal to a threshold, that is, negative samples can be samples obtained by performing a significant perturbation on the training samples. As an example, server 130 can adjust the perturbation of training samples by disrupting causal relationships or introducing erroneous decision-making scenarios to generate negative samples. For example, server 130 can generate negative samples by deleting other vehicles that have key interactions with the reference vehicle, and / or changing the state of traffic lights, and / or adding obstacles, so that the escape model can learn the correct causal dependencies and enhance interaction understanding and generalization ability.

[0064] Furthermore, server 130 can acquire a first trajectory feature output by the escape model based on positive samples and a second trajectory feature output by the escape model based on negative samples. In some embodiments, the first and second trajectory features can be features output by a hidden layer of the escape model. In some embodiments, this hidden layer can be deployed at the output header of the escape model, and its output features are related to both the escape trajectory output by the escape model and the global features in the escape scenario.

[0065] Furthermore, server 130 can train an escape model based on the first trajectory features, the second trajectory features, and the reference trajectory features corresponding to the reference escape trajectory.

[0066] In some embodiments, the reference trajectory features can be the features output by the hidden layer of the escape model when training samples are input into the escape model. As an example, server 130 can determine a second loss value based on a comparison of the first trajectory features and the reference trajectory features, and a comparison of the second trajectory features and the reference trajectory features. Further, server 130 can train the escape model based on the second loss value. It should be noted that the goal of training the escape model based on the second loss value is to maximize the similarity between positive samples and training samples, and minimize the similarity between negative samples and training samples. That is, the smaller the difference between the first trajectory features and the reference trajectory features, and the larger the difference between the second trajectory features and the reference trajectory features, the smaller the second loss value. It should be noted that the loss function required to determine the second loss value can be set according to requirements, and will not be elaborated here.

[0067] By training the escape model in this way, the model can not only better fit the trajectory planning behavior in positive samples, but also avoid erroneous decision paths. In addition, this disclosure enhances the stability of decision-making through contrastive loss, enabling the escape model to maintain high decision-making performance in complex driving environments and effectively address the problems of data distribution bias and causal confusion.

[0068] Figure 3A flowchart of the training process of an escape model according to some embodiments of this disclosure has been drawn, and now, regarding... Figure 3 Please provide an explanation.

[0069] In box 301, server 130 can clean and label the training samples, specifically labeling the reference trajectory features corresponding to the training samples.

[0070] As an example, server 130 can perform at least one of the following cleaning operations on the training samples: noise removal, deduplication, etc.

[0071] As an example, the reference trajectory features can be the features output by the hidden layer of the escape model when training samples are input into the escape model. In some embodiments, this hidden layer can be deployed at the output head of the escape model, and its output features are related to both the escape trajectory output by the escape model and global features in the escape scenario.

[0072] In box 302, server 130 obtains positive samples.

[0073] As an example, server 130 can generate positive samples by adjusting the position of the reference vehicle, and / or adjusting the speed of the reference vehicle, and / or removing other vehicles that do not interact with the reference vehicle.

[0074] In box 303, server 130 acquires negative samples.

[0075] As an example, server 130 can manipulate training samples to generate negative samples by disrupting causal relationships or introducing erroneous decision-making scenarios. For instance, server 130 can interfere with training samples by removing other vehicles that have key interactions with the reference vehicle, and / or changing the state of traffic lights, and / or adding obstacles.

[0076] In box 304, server 130 inputs positive samples into the escape model to obtain first trajectory features.

[0077] As an example, server 130 inputs positive samples into the escape model to obtain the first trajectory features output by the hidden layer of the escape model.

[0078] In some embodiments, this hidden layer can be deployed at the output head of the escape model, and its output features are related to both the escape trajectory output by the escape model and the global features in the escape scenario.

[0079] In box 305, server 130 inputs negative samples into the escape model to obtain second trajectory features.

[0080] As an example, server 130 inputs negative samples into the escape model to obtain the second trajectory features output by the hidden layer of the escape model.

[0081] In some embodiments, this hidden layer can be deployed at the output head of the escape model, and its output features are related to both the escape trajectory output by the escape model and the global features in the escape scenario.

[0082] In box 306, server 130 trains an escape model based on the first trajectory features, the second trajectory features, and the reference trajectory features.

[0083] As an example, server 130 can determine a loss value based on a comparison of the first trajectory features and the reference trajectory features, and a comparison of the second trajectory features and the reference trajectory features. Furthermore, server 130 can train an escape model based on this loss value until a preset training completion condition is met.

[0084] As an example, training completion conditions may include, but are not limited to, training duration reaching a predetermined threshold, loss value reaching a minimum, etc.

[0085] To improve the availability and reliability of the escape trajectory, the electronic device 120 can filter at least a portion of the escape trajectories from a set of escape trajectories output by the model as at least one candidate escape trajectory, thereby filtering out escape trajectories with low reliability or low availability. Specifically, the electronic device 120 can obtain a set of reference trajectories output by the escape model. Further, the electronic device 120 can determine at least one reference trajectory that satisfies preset constraints from the set of reference trajectories.

[0086] In other embodiments, the electronic device 120 may acquire a set of reference routes output by the escape model. In this case, the set of reference routes only indicates the lateral reference information constituting the escape trajectory, and does not indicate the longitudinal reference information constituting the escape trajectory.

[0087] To improve the reliability and usability of the escape trajectory, the electronic device 120 can further determine at least one reference route from a set of reference routes that satisfies preset constraints. In some embodiments, the preset constraints may indicate at least one of the following: the reference route satisfies collision constraints with obstacles, the reference route satisfies kinematic constraints associated with the autonomous vehicle 110, and there exists a route for the autonomous vehicle 110 between the end point of the reference route and the end point of the journey reached by the autonomous vehicle 110.

[0088] In some embodiments, the fact that the reference route meets collision constraints with obstacles can indicate that the reference route will not collide with the obstacles. As an example, the obstacle can be a static obstacle. For example, the obstacle can be a curb, flower bed, or other obstacles determined based on map information, or it can be an obstacle that previously prevented the autonomous vehicle 110 from passing smoothly.

[0089] In some embodiments, kinematic constraints may indicate physical limitations on the autonomous vehicle 110, such as maximum acceleration, maximum deceleration, maximum steering angular velocity, curvature, etc. As an example, a preset constraint may indicate that the steering wheel angle of the autonomous vehicle 110 is less than or equal to 30 degrees.

[0090] In some embodiments, there is a route indication reference line between the end of the reference route and the end of the journey reached by the autonomous vehicle 110, which can guide the autonomous vehicle 110 to drive to the end of the global route of the autonomous vehicle 110, so as to ensure that the autonomous vehicle 110 can drive normally after the remote assistance ends.

[0091] Furthermore, the electronic device 120 can construct at least one candidate escape trajectory based on at least one reference route. In some embodiments, the at least one reference route only indicates reference information with lateral path and does not indicate reference information with longitudinal velocity, while the at least one candidate escape trajectory can indicate both reference information with lateral path and reference information with longitudinal velocity.

[0092] Specifically, the electronic device 120 can determine the boundary information of the driving area corresponding to at least one reference route. This boundary information indicates the area where the autonomous vehicle 110 can safely drive, preventing collisions with obstacles. This boundary information may include the geometric boundaries of the road corresponding to the driving area, such as lane lines, curbs, and guardrails.

[0093] Specifically, the electronic device 120 can generate this boundary information based on at least one reference route, with a certain offset set on the at least one reference route.

[0094] It should be noted that, in order to improve the smoothness of the trajectory and reduce sharp turns and abrupt changes, the electronic device 120 can first smooth the reference route. Furthermore, the electronic device 120 can generate this boundary information based on the reference route obtained after the smoothing process.

[0095] Furthermore, the electronic device 120 can search for at least one escape route from the driving area based on boundary information.

[0096] As an example, electronic device 120 can backtrack to generate a lane sequence based on at least one reference route (which may be a smoothed reference route). Further, electronic device 120 can determine a first safety constraint based on at least one reference route and the corresponding lane sequence. Specifically, electronic device 120 can infer the environmental information (such as information about other road users) and corresponding safety constraints that autonomous vehicle 110 needs to consider to escape a difficult situation based on at least one reference route and the corresponding lane sequence, to ensure that autonomous vehicle 110 can avoid collisions and remain within its lane during operation.

[0097] Furthermore, the electronic device 120 can search for at least one escape route from the driving area based on at least one reference route, a lane sequence, and boundary information of the driving area corresponding to at least one reference route. Specifically, the electronic device 120 can construct a search problem based on at least one reference route, a lane sequence, and boundary information of the driving area corresponding to at least one reference route. Further, the electronic device 120 can utilize a predetermined search algorithm to search for a set of escape routes from the driving area based on this search problem.

[0098] To ensure the reliability of the discovered escape routes, the electronic device 120 can determine at least one escape route from this set of escape routes that satisfies the first safety constraint. As an example, the electronic device 120 can determine at least one escape route that satisfies the first safety constraint based on this set of escape routes and the first safety constraint, using an Iterative Linear Quadratic Regulator (ILQR).

[0099] Furthermore, the electronic device 120 can determine driving speed information associated with at least one escape route. In some embodiments, the driving speed information can be a speed profile. The speed profile can be used to describe a graphical representation of the speed changes of the autonomous vehicle 110 on the escape route. In some embodiments, the driving speed information associated with at least one escape route is also the longitudinal reference information corresponding to the escape trajectory.

[0100] As an example, electronic device 120 can generate a second safety constraint based on at least one escape route and environmental information of autonomous vehicle 110. The environmental information can indicate other vehicles associated with autonomous vehicle 110. These other vehicles associated with autonomous vehicle 110 can be identified based on predicted state information corresponding to a group of vehicles associated with autonomous vehicle 110 (including other vehicles associated with autonomous vehicle 110), and can be vehicles that may have potential spatiotemporal interactions with autonomous vehicle 110. In some embodiments, the predicted state information can be the state information of a vehicle predicted from its historical state information.

[0101] To ensure that the autonomous vehicle 110 can travel at a reasonable speed and avoid collisions with other obstacles, the electronic device 120 can determine at least one candidate velocity profile based on a second safety constraint. As an example, the electronic device 120 can use a predetermined search algorithm based on the second safety constraint to search for at least one candidate velocity profile that meets the obstacle avoidance requirements. The predetermined search algorithm can be a greedy search algorithm, or other search algorithms, which will not be elaborated here.

[0102] Furthermore, to ensure both the autonomous vehicle 110 avoids collisions with other obstacles and maintains a comfortable trajectory, the electronic device 120 can determine a target speed profile from at least one candidate speed profile based on a second safety constraint and predetermined kinematic constraints. This target speed profile is essentially the driving speed information associated with at least one escape route. As an example, the electronic device 120 can determine the target speed profile from at least one candidate speed profile using ILQR based on the second safety constraint and predetermined kinematic constraints.

[0103] Furthermore, the electronic device 120 can construct at least one candidate escape trajectory based on the driving speed information and at least one escape route. That is, the driving speed information serves as the longitudinal reference information for the candidate escape trajectory, the at least one escape trajectory serves as the longitudinal reference information for the candidate escape trajectory, and the at least one candidate escape trajectory indicates global (lateral and longitudinal) reference information.

[0104] Back Figure 2 In box 220, electronic device 120 sends an assistance request to a remote device, the assistance request indicating at least one candidate escape trajectory.

[0105] For example, electronic devices can generate work orders to obtain remote assistance or rescue services.

[0106] In box 230, electronic device 120 receives an assistance message from a remote device, the assistance message indicating confirmation of a target escape trajectory among at least one candidate escape trajectory.

[0107] For example, in response to receiving an assistance request, the remote device may present an interactive interface (such as a control interface associated with the autonomous vehicle 110), and may display at least one selection control corresponding to at least one candidate escape trajectory in the interactive interface. Furthermore, in response to receiving a selection of the target selection control (corresponding to the target escape trajectory) among the at least one selection control, the remote device may send an instruction to the electronic device 120 to confirm the target escape trajectory.

[0108] In frame 240, electronic device 120 controls the autonomous vehicle 110 to drive based on the target escape trajectory.

[0109] In some embodiments, the target escape trajectory may indicate at least one of the following information: a sequence of points on the trajectory, speed changes, turning angles, etc.

[0110] As an example, electronic device 120 can convert the target escape trajectory into control commands that can be understood by the various execution units of autonomous vehicle 110. Furthermore, electronic device 120 can send control commands to the various execution units of autonomous vehicle 110 (such as steering system, drive system, braking system, etc.) to control autonomous vehicle 110 to drive according to the target escape trajectory, thereby achieving escape.

[0111] Figure 4 Flowcharts of vehicle extrication processes according to some embodiments of this disclosure are provided below. Figure 4 Please provide an explanation.

[0112] In frame 401, the vehicle collects and preprocesses surrounding map information, perception information, prediction information, and other information.

[0113] As an example, map information can be objects pre-marked in offline maps or pre-defined map data, which can include, but are not limited to, fences, sidewalks, curbs, flower beds, kerb, etc.

[0114] As an example, the perceived information may include obstacle information associated with the autonomous vehicle 110. The specific obstacle information may be other traffic participants, such as other vehicles, pedestrians, motorcycles, bicycles, construction zones, etc.

[0115] As an example, the predictive information can indicate the predicted state information of other traffic participants. Other traffic participants could be other vehicles or pedestrians associated with autonomous vehicle 110. The predicted state information can be the predicted state information of other traffic participants based on their historical state information. For example, the predicted state information could be the speed of autonomous vehicle 110 in the future, determined based on the speed of the vehicle to its left over the past ten seconds.

[0116] In some embodiments, preprocessing can be any appropriate data type conversion, coordinate transformation, or other operation.

[0117] In box 402, electronic device 120 can preprocess data and input it into the escape model so that the escape model can generate an escape trajectory.

[0118] As an example, the electronic device 120 can simultaneously output a set of escape trajectories and the corresponding scores for these trajectories. Further, the electronic device 120 identifies at least one escape trajectory in this set of trajectories whose score is greater than a threshold as the generated escape trajectory.

[0119] As an example, the escaping model is different from the trajectory planning model used by the autonomous vehicle 110 in a non-trapped state.

[0120] As an example, the obstacle avoidance model can be deployed on the vehicle side.

[0121] In box 403, the escape trajectory of electronic device 120 is fed back to remote device for confirmation.

[0122] Furthermore, taking multiple escape trajectories as an example, the remote device can present an interactive interface (such as a control interface associated with the autonomous vehicle 110), and can display multiple selection controls corresponding to these multiple escape trajectories in the interactive interface. Furthermore, in response to receiving a selection of the target selection control (corresponding to the target escape trajectory) among these multiple selection controls, the remote device can send an instruction to the electronic device 120 to confirm the target escape trajectory.

[0123] In frame 404, electronic device 120 controls autonomous vehicle 110 to perform and complete the extrication process.

[0124] In some embodiments, the electronic device 120 may, in response to receiving confirmation of the target escape trajectory, control the autonomous vehicle 110 to drive based on the target escape trajectory in order to complete the escape.

[0125] Figure 5 Flowcharts of vehicle extrication processes according to some embodiments of this disclosure are provided below. Figure 5 Please provide an explanation.

[0126] like Figure 5 As shown, the electronic device 120 can utilize the feature extraction module 501 to perform feature extraction operations on map information associated with the autonomous vehicle 110, obstacle information associated with the autonomous vehicle 110, and predicted state information corresponding to other traffic participants, to obtain a tensor matrix 502. This tensor matrix 502 can be determined based on the feature vectors corresponding to the map information associated with the autonomous vehicle 110, the feature vectors corresponding to the obstacle information associated with the autonomous vehicle 110, and the feature vectors corresponding to the predicted state information corresponding to other traffic participants. Further, the electronic device 120 can utilize the encoder 503 of the traction escape model to generate a fused feature sequence 504 based on the tensor matrix 502. Further, the electronic device 120 can utilize the decoder 505 of the traction escape model to generate a target feature sequence 507 based on the fused feature sequence 504 and multiple query vectors. These multiple query vectors can correspond to different driving intentions. Further, the electronic device 120 can utilize the planning head 508 to determine a set of reference routes and the corresponding scores for this set of reference routes based on the target feature sequence 507. Furthermore, the electronic device 120 can update the reference routes with scores greater than a threshold to this set of reference routes, so as to further perform post-processing operations 509 on this set of reference routes.

[0127] As an example, post-processing operation 509 may include, but is not limited to, the following: Electronic device 120 determines at least one reference route from a set of reference routes that satisfies preset constraints. Further, electronic device 120 may determine boundary information of the driving area corresponding to the at least one reference route. Further, electronic device 120 may search for at least one escape route from the driving area based on the boundary information. Further, electronic device 120 may determine driving speed information associated with the at least one escape route. Further, electronic device 120 may construct at least one candidate escape trajectory based on the driving speed information and the at least one escape route. Further, electronic device 120 may send this at least one candidate escape trajectory to a remote device. Further, in response to receiving confirmation of a target escape trajectory among the at least one candidate escape trajectory, electronic device 120 may control the autonomous vehicle 110 to drive based on the target escape trajectory to complete the escape.

[0128] In other embodiments, post-processing operation 509 may further include determining whether the candidate escape trajectory meets the requirements based on a predetermined model. For example, it may compare the candidate escape trajectory with a pre-defined set of escape trajectories that meet the availability requirements to determine whether the candidate escape trajectory meets the availability requirements. The predetermined model can be any suitable model, which will not be elaborated here.

[0129] Based on this approach, embodiments of this disclosure can automatically generate at least one candidate escape trajectory using an escape model for remote devices to select from, thereby controlling the autonomous vehicle to escape from trouble based on the determined target escape trajectory. Compared to the method of manually drawing points to generate escape trajectories, embodiments of this disclosure can effectively improve the reliability and generation efficiency of escape trajectories, and further improve the escape efficiency and success rate of autonomous vehicles.

[0130] Example Apparatus and Device Embodiments of this disclosure also provide corresponding apparatus for implementing the above methods or processes. Figure 6 A schematic structural block diagram of a vehicle control device 600 according to certain embodiments of the present disclosure is shown. The device 600 may be implemented as or included in the electronic device 120 discussed above. The various modules / components in the device 600 may be implemented by hardware, software, firmware, or any combination thereof.

[0131] like Figure 6 As shown, the device 600 includes a generation module 610 configured to generate at least one candidate escape trajectory using an escape model in response to the autonomous vehicle being in a trapped state, the escape model being different from the trajectory planning model used by the autonomous vehicle in an untrapped state; a sending module 620 configured to send an assistance request to a remote device, the assistance request indicating at least one candidate escape trajectory; a receiving module 630 configured to receive an assistance message from the remote device, the assistance message indicating confirmation of a target escape trajectory among the at least one candidate escape trajectory; and a control module 640 configured to control the autonomous vehicle to drive based on the target escape trajectory.

[0132] In some embodiments, the generation module 610 is further configured to: acquire descriptive information associated with the autonomous vehicle; and provide descriptive information to the escaping model to generate at least one candidate escaping trajectory.

[0133] In some embodiments, the description information indicates at least one of the following: environmental information associated with the autonomous vehicle; and state information of the autonomous vehicle.

[0134] In some embodiments, environmental information includes: map information; and / or obstacle information determined based on perception information from the autonomous vehicle.

[0135] In some embodiments, the generation module 610 is further configured to: generate vectorized data of descriptive information; process the vectorized data using the encoder of the escape model to generate a fused feature sequence; and generate at least one candidate escape trajectory based on the fused feature sequence using the decoder of the escape model.

[0136] In some embodiments, the generation module 610 is further configured to: acquire multiple query vectors, the multiple query vectors corresponding to different driving intentions; and provide a fused feature sequence and multiple query vectors to the decoder to generate at least one escape trajectory.

[0137] In some embodiments, the query vector among the plurality of query vectors indicates at least one of the following: a sequence of lateral offsets relative to a reference boundary; or a longitudinal offset parameter relative to a starting position.

[0138] In some embodiments, the escape model is trained based on the following process: acquiring scene data associated with a reference escape scenario; constructing training samples based on the scene data and a reference escape trajectory associated with the reference escape scenario; and training the escape model based on the training samples.

[0139] In some embodiments, training an escape model based on training samples includes: constructing sample pairs, which include positive samples and negative samples, based on the training samples; obtaining a first trajectory feature output by the escape model based on the positive samples and a second trajectory feature output by the escape model based on the negative samples; and training the escape model based on the first trajectory feature, the second trajectory feature, and a reference trajectory feature corresponding to a reference escape trajectory.

[0140] In some embodiments, the generation module 610 is further configured to: obtain a set of reference routes output by the escape model; determine at least one reference route from the set of reference routes that satisfies preset constraints; and construct at least one candidate escape trajectory based on the at least one reference route.

[0141] In some embodiments, the preset constraints indicate at least one of the following: the reference route satisfies collision constraints with obstacles; the reference route satisfies kinematic constraints associated with the autonomous vehicle; and a route for the autonomous vehicle exists between the end point of the reference route and the end point of the journey reached by the autonomous vehicle.

[0142] In some embodiments, the generation module 610 is further configured to: determine boundary information of a driving area corresponding to at least one reference route based on at least one reference route; search for at least one escape route from the driving area based on the boundary information; determine driving speed information associated with at least one escape route; and construct at least one candidate escape trajectory based on the driving speed information and at least one escape route.

[0143] The units included in device 600 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units may be implemented using software and / or firmware, such as machine-executable instructions stored on a storage medium. In addition to or as an alternative to machine-executable instructions, some or all of the units in device 600 may be implemented at least partially by one or more hardware logic components. By way of example and not limitation, exemplary types of hardware logic components that may be used include field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chips (SoCs), complex programmable logic devices (CPLDs), and so on.

[0144] Figure 7 A block diagram of an electronic device 700 in which one or more embodiments of the present disclosure may be implemented is shown. It should be understood that... Figure 7 The electronic device 700 shown is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. Figure 7 The electronic device 700 shown can be used to achieve Figure 1 The electronic device 120 shown.

[0145] like Figure 7 As shown, electronic device 700 is in the form of a general-purpose electronic device. Components of electronic device 700 may include, but are not limited to, one or more processors 710 or processing units, memory 720, storage device 730, one or more communication units 740, one or more input devices 750, and one or more output devices 760. Processor 710 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 720. In a multiprocessor system, multiple processors execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 700.

[0146] Electronic device 700 typically includes multiple computer storage media. Such media can be any accessible media that is accessible to electronic device 700, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 720 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 730 can be a removable or non-removable medium and can include machine-readable media, such as flash drives, disks, or any other media that can be used to store information and / or data (e.g., training data for training) and can be accessed within electronic device 700.

[0147] Electronic device 700 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 7 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 720 may include computer program product 725 having one or more program modules configured to perform various methods or actions of various embodiments of this disclosure.

[0148] The communication unit 740 enables communication with other electronic devices via a communication medium. Additionally, the functionality of the components of the electronic device 700 can be implemented using a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, the electronic device 700 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.

[0149] Input device 750 can be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 760 can be one or more output devices, such as a monitor, speaker, printer, etc. Electronic device 700 can also communicate with one or more external devices (not shown) via communication unit 740 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 700, or with any device that enables electronic device 700 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).

[0150] 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.

[0151] 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.

[0152] 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.

[0153] 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.

[0154] 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.

[0155] Various implementations of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they 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 chosen 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 vehicle control, comprising: In response to an autonomous vehicle being in a troubled state, at least one candidate escape trajectory is generated using an escape model, wherein the escape model is different from the trajectory planning model used by the autonomous vehicle in a non-troubled state. Send an assistance request to a remote device, the assistance request indicating the at least one candidate escape trajectory; Receive an assistance message from the remote device, the assistance message indicating confirmation of a target escape trajectory among the at least one candidate escape trajectory; as well as The autonomous vehicle is controlled to drive based on the target escape trajectory.

2. The method according to claim 1, wherein generating at least one candidate escape trajectory using the escape model comprises: Obtain descriptive information associated with the autonomous vehicle; as well as The descriptive information is provided to the escape model to generate the at least one candidate escape trajectory.

3. The method of claim 2, wherein the descriptive information indicates at least one of the following: First environmental information associated with the autonomous vehicle; The status information of the autonomous vehicle.

4. The method according to claim 3, wherein the first environmental information includes: Map information; and / or Obstacle information determined based on the perception information of the autonomous vehicle.

5. The method of claim 2, wherein providing the descriptive information to the escape model to generate the at least one candidate escape trajectory comprises: Generate vectorized data of the description information; The vectorized data is processed using the encoder of the escape model to generate a fused feature sequence; Using the decoder of the escape model, at least one candidate escape trajectory is generated based on the fused feature sequence.

6. The method of claim 5, wherein generating the at least one candidate escape trajectory based on the fused feature sequence using the decoder of the escape model comprises: Obtain multiple query vectors, which correspond to different driving intentions; as well as The decoder is provided with the fused feature sequence and the plurality of query vectors to generate the at least one escape trajectory.

7. The method of claim 6, wherein the query vector among the plurality of query vectors indicates at least one of the following: Lateral offset sequence relative to the reference boundary; The longitudinal offset parameter relative to the starting position.

8. The method of claim 1, wherein the escape model is trained based on the following process: Acquire scene data associated with the reference escape scenario; Based on the scenario data and the reference escape trajectory associated with the reference escape scenario, training samples are constructed; and The escape model is trained based on the training samples.

9. The method of claim 8, wherein training the escape model based on the training samples comprises: Based on the training samples, sample pairs are constructed, and the sample pairs include positive samples and negative samples; Obtain the first trajectory features output by the escape model based on the positive samples and the second trajectory features output by the escape model based on the negative samples; as well as The escape model is trained based on the first trajectory features, the second trajectory features, and the reference trajectory features corresponding to the reference escape trajectory.

10. The method of claim 1, wherein generating at least one candidate escape trajectory using an escape model comprises: Obtain a set of reference routes output by the escape model; Determine at least one reference route from the set of reference routes that satisfies the preset constraints; as well as Based on the at least one reference route, construct the at least one candidate escape trajectory.

11. The method of claim 10, wherein the preset constraint indicates at least one of the following: The reference route satisfies collision constraints with obstacles; The reference route satisfies the kinematic constraints associated with the autonomous vehicle; There exists a route for the autonomous vehicle between the end of the reference route and the end of the journey reached by the autonomous vehicle.

12. The method of claim 10, wherein constructing the at least one candidate escape trajectory based on the at least one reference route comprises: Based on the at least one reference route, determine the boundary information of the driving area corresponding to the at least one reference route; Based on the boundary information, at least one escape route is searched from the driving area; Determine the driving speed information associated with the at least one escape route; as well as Based on the driving speed information and the at least one escape route, construct the at least one candidate escape trajectory.

13. A device for vehicle control, comprising: The generation module is configured to generate at least one candidate escape trajectory using an escape model in response to an autonomous vehicle being in a trapped state, the escape model being different from the trajectory planning model used by the autonomous vehicle in an untrapped state. The sending module is configured to send an assistance request to a remote device, the assistance request indicating the at least one candidate escape trajectory; A receiving module is configured to receive an assistance message from the remote device, the assistance message indicating confirmation of a target escape trajectory among the at least one candidate escape trajectory; as well as The control module is configured to control the autonomous vehicle's movement based on the target escape trajectory.

14. An electronic device comprising: At least one processor; as well as At least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor, the instructions causing the electronic device to perform the method according to any one of claims 1 to 12 when executed by the at least one processor.

15. A computer-readable storage medium having stored thereon computer-executable instructions that, when executed by a processor, implement the method according to any one of claims 1 to 12.

16. 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 12.

Images