Vehicle escape method, device and system

Abstract

The invention provides a vehicle escape method, device and system, and relates to the field of unmanned driving, automatic driving and unmanned vehicles. The vehicle escape method is applied to a decision-making end and comprises the steps that a takeover request of a target vehicle is received, and the takeover request comprises state information and / or environment information when the target vehicle is parked; based on the takeover request, a decision module built in the decision end is utilized to autonomously generate a target driving strategy; and sending the target driving strategy to the target vehicle, so that the target vehicle gets out of trouble according to the target driving strategy.

Description

Technical Field

[0001] This application relates to the fields of unmanned driving, autonomous driving, and unmanned vehicle technology, specifically to a method, device, and system for vehicle extrication from difficult situations. Background Technology

[0002] Currently, when unmanned mining vehicles encounter difficulties and stop, they mainly rely on remote manual takeover to resume operation. This results in low takeover efficiency, poor operational continuity, and is highly dependent on network conditions and labor costs, making it difficult to meet the needs of large-scale, efficient operation. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, apparatus and system for vehicle extrication.

[0004] Firstly, this application provides a vehicle extrication method applied to a decision-making end. The vehicle extrication method includes: receiving a takeover request from a target vehicle, wherein the takeover request includes the target vehicle's status information and / or environmental information when it is parked; based on the takeover request, autonomously generating a target driving strategy using a decision-making module built into the decision-making end; and sending the target driving strategy to the target vehicle so that the target vehicle can extricate itself from the predicament according to the target driving strategy.

[0005] In conjunction with the first aspect, in some implementations of the first aspect, the target driving strategy includes at least one of the following: traffic feasibility information, driving path information, drivable distance information, environmental obstacle information, and decision confidence information; optionally, traffic feasibility information includes whether it is possible to continue driving in the current driving direction; driving path information includes the type of adjustment of driving direction and / or the location of the target point; drivable distance information includes the distance that can be driven in the current driving direction; environmental obstacle information includes the obstacles targeted during detours and / or the types of obstacles that hinder the target vehicle's driving.

[0006] In conjunction with the first aspect, in some implementations of the first aspect, before the target vehicle attempts to escape from the predicament according to the target driving strategy, the vehicle escaping method further includes: determining the result of the validity judgment of the target driving strategy; and controlling the subsequent escaping process of the target vehicle based on the result of the validity judgment. Optionally, the validity judgment includes at least one of the following: determining whether the parking factors of the target vehicle fall within the preset autonomous escaping range; determining whether the position change of the target vehicle exceeds a preset range from sending the takeover request to receiving the target driving strategy; determining whether there is a target obstacle in the current perceived environment; and determining whether the quality of the image in the environmental information meets preset conditions. Optionally, the validity judgment is made by the target vehicle or the decision-making end.

[0007] In conjunction with the first aspect, in some implementations of the first aspect, the subsequent extrication process of the target vehicle is controlled based on the result of the validity judgment, including: if the result indicates that the validity judgment is passed, the target vehicle is controlled to extricate itself according to the target driving strategy; if the result indicates that the validity judgment is not passed, an auxiliary extrication response is provided to the target vehicle; optionally, the auxiliary extrication response is provided to the target vehicle, including: triggering a remote takeover of the target vehicle with human participation based on a takeover assistance request.

[0008] In conjunction with the first aspect, in some implementations of the first aspect, determining whether a target obstacle exists in the current perception environment includes: comparing the determination of the existence of the target obstacle based on environmental information with the determination of the existence of the target obstacle based on the current perception data of the target vehicle; if the two are consistent, the validity determination is passed; if the two are inconsistent, the validity determination is not passed.

[0009] In conjunction with the first aspect, in some implementations of the first aspect, based on the takeover request, the decision-making module built into the decision-making terminal autonomously generates the target driving strategy, including: based on the takeover request, using the decision-making module to make a takeover decision; if the decision result indicates that the decision-making terminal can autonomously extricate itself from the predicament, then using the decision-making module to autonomously generate the target driving strategy; optionally, if the decision result indicates that the decision-making terminal cannot autonomously extricate itself from the predicament, then providing an auxiliary extrication response to the target vehicle.

[0010] In conjunction with the first aspect, in some implementations of the first aspect, a takeover decision is made using a decision module based on the takeover request, including: if it is determined that the target vehicle meets preset triggering conditions, the takeover request is input to the decision module, and the takeover decision is made using the decision module; optionally, the triggering conditions include at least one of the following: the continuous parking time of the target vehicle exceeds a preset threshold; the parking factor of the target vehicle belongs to a preset decision-making factor; the current parking position of the target vehicle belongs to a preset triggering area; optionally, the decision module includes a visual language model, which is a model adjusted after target training data, and the target training data includes the original takeover data of the unmanned mining vehicle and the true values ​​of driving behavior calibrated based on the original takeover data.

[0011] In conjunction with the first aspect, in some implementations of the first aspect, the preset performance requirements of the visual language model during the training phase include at least one of the following: the dangerous misclassification rate is lower than a first threshold; the conservative misclassification rate is lower than a second threshold; and the inference time is lower than a third threshold. Wherein, the dangerous misclassification rate is the ratio of the visual language model misclassifying non-drivable scenes as drivable scenes, and the conservative misclassification rate is the ratio of the visual language model misclassifying drivable scenes as non-drivable scenes.

[0012] Secondly, this application provides a vehicle extrication method, applied to the vehicle end. The vehicle extrication method includes: upon detecting that the target vehicle is stopped and the stopping factors meet specified conditions, sending a takeover request to a decision-making end; wherein the takeover request includes the target vehicle's status information and / or environmental information; receiving a target driving strategy, and extricating the vehicle from the predicament according to the target driving strategy; wherein the target driving strategy is autonomously generated by the decision-making end based on a built-in decision module.

[0013] Thirdly, this application provides a vehicle escaping device for performing any of the above-mentioned vehicle escaping methods.

[0014] Fourthly, this application provides a vehicle extrication system, including a target vehicle and a decision-making terminal, the decision-making terminal having a built-in decision-making module. The target vehicle is configured to send a takeover request to the decision-making terminal when it detects that the target vehicle is stopped and the stopping factors meet specified conditions. The takeover request includes the vehicle's status information and / or environmental information. The decision-making terminal is configured to autonomously generate a target driving strategy based on the takeover request using the decision-making module and send it to the target vehicle. The target vehicle is further configured to extricate itself from the predicament according to the target driving strategy.

[0015] Fifthly, this application provides a computer-readable storage medium storing a computer program for performing the vehicle extrication methods described in the first and second aspects.

[0016] In a sixth aspect, this application provides an electronic device comprising: a processor; a memory for storing processor-executable instructions; the processor being configured to execute the method for vehicle extrication described in the first and second aspects.

[0017] In a seventh aspect, this application provides a computer program product including instructions that, when executed on an electronic device, cause the electronic device to perform the vehicle extrication described in the first and second aspects.

[0018] In this application, a decision-making module enables real-time analysis and decision-making of driving strategies, achieving intelligent and autonomous vehicle extrication processes. This avoids queuing issues caused by delays in manual response or limited resources, improving the smoothness of the target vehicle's movement and its reliability in harsh communication environments. Furthermore, the autonomous vehicle extrication method reduces the need for a remote-controlled cockpit and driver, thus lowering the human and hardware costs required for large-scale industrial expansion. Attached Figure Description

[0019] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 The diagram shown is a system architecture diagram of a vehicle extrication method provided in an embodiment of this application.

[0021] Figure 2 The diagram shown is a flowchart of a vehicle extrication method provided in an embodiment of this application.

[0022] Figure 3 The diagram shown is a flowchart of a vehicle extrication method provided in another embodiment of this application.

[0023] Figure 4 The diagram shown is a flowchart illustrating the steps of autonomously generating a target driving strategy based on a takeover request using a decision module built into the decision-making terminal, according to an embodiment of this application.

[0024] Figure 5 The diagram shown is a flowchart of a vehicle extrication method provided in another embodiment of this application.

[0025] Figure 6 The diagram shown is a schematic representation of the information interaction between the target vehicle and the decision-making end according to an embodiment of this application.

[0026] Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] With the development of autonomous driving technology in the mining transportation sector, driverless vehicles can autonomously drive and operate within pre-set operating areas and conditions, significantly improving the automation level and operational safety of mining transportation. However, due to the variability of the mining environment, vehicles inevitably encounter scenarios that current autonomous driving systems cannot handle, causing them to cease driving and operation. In such situations, a remote monitoring center is typically needed, where a remotely controlled driver can manually take over the vehicle through real-time video transmission and a remote control interface to help it escape the predicament and resume autonomous driving. However, this method has certain limitations.

[0029] First, in operational scenarios with high vehicle density and frequent takeover requests, relying solely on remote-controlled drivers for takeover is prone to queues, leading to low takeover efficiency and severely impacting operational productivity. Second, remote-controlled driving depends on factors such as network communication, the smoothness of video streaming, and command response latency. In complex environments like mines, network fluctuations or video transmission interruptions can prevent takeover commands from being issued reliably in real time, jeopardizing driving safety and operational timeliness. Third, a certain number of vehicles require a remote-controlled cockpit and a remote-controlled driver, and this cost increases with the scale of the industry. Fourth, for dispatchers in mining areas using automated driving transportation, frequent intervention in vehicle takeover coordination is necessary, making it difficult to manage vehicle scheduling from a holistic perspective, thus hindering further improvements in transportation efficiency.

[0030] To address the aforementioned technical problems, this application provides a vehicle extrication method applied to a decision-making end. The method includes: receiving a takeover request from a target vehicle, wherein the takeover request includes the target vehicle's status information and / or environmental information when parked; based on the takeover request, autonomously generating a target driving strategy using a decision-making module built into the decision-making end; and sending the target driving strategy to the target vehicle so that the target vehicle can extricate itself from the predicament according to the target driving strategy. This solution improves the unmanned extrication capability of the target vehicle and enables smooth driving of the target vehicle.

[0031] The following is a combination of... Figure 1 This application introduces a system architecture provided by one embodiment.

[0032] Figure 1 The diagram shown is a system architecture schematic of a vehicle extrication method provided in an embodiment of this application. Figure 1 As shown, the vehicle extrication system includes a target vehicle 110, a decision-making terminal 120, and a remote monitoring center 130. Communication between the target vehicle 110, the decision-making terminal 120, and the remote monitoring center 130 can be achieved, for example, through a wired or wireless network connection.

[0033] The target vehicle 110 is equipped with an autonomous driving system and a perception system. The perception system can acquire the target vehicle 110's own status information and the surrounding environment information in real time. When the target vehicle 110 is unable to move for an extended period of time, the autonomous driving system can identify the current parking factors, generate a takeover request accordingly, and upload the takeover request to the decision-making terminal 120 in real time via the communication link.

[0034] The decision-making terminal 120 has a built-in decision-making module. After receiving a takeover request, the decision-making terminal 120 autonomously generates a target driving strategy based on the takeover request and sends the target driving strategy to the target vehicle 110.

[0035] Optionally, the decision-making module can be deployed in the cloud or on the vehicle. Deploying the decision-making module on the vehicle can improve the real-time performance, independence, and data security of the decision-making process, and is suitable for scenarios with poor network conditions. Deploying the decision-making module in the cloud can make full use of its advantages of centralized computing, easy updates, and global collaboration.

[0036] The autonomous driving system of the target vehicle 110 can assess the reliability and safety of the target driving strategy and execute it if the target driving strategy is deemed safe and reliable. If the target driving strategy fails the reliability and safety assessment, the autonomous driving system sends a takeover request to the remote monitoring center 130 so that the remote driver can remotely control the target vehicle 110 to escape from the predicament.

[0037] Optionally, when the decision-making module is deployed on the vehicle: the vehicle is equipped with multiple decision-making algorithms, and the resource consumption of different decision-making algorithms is different; when the vehicle is driving normally (not trapped), the first decision-making algorithm is used for vehicle driving planning; when the vehicle is trapped and receives a takeover request, the second decision-making algorithm is used for vehicle driving planning; wherein, the resource consumption of the first decision-making algorithm is less than that of the second decision-making algorithm.

[0038] The following is combined Figures 2 to 6 This application describes in detail the vehicle extrication method provided in its embodiments.

[0039] Figure 2 The diagram shown is a flowchart illustrating a vehicle extrication method according to an embodiment of this application. Exemplarily, the vehicle extrication method provided in this embodiment is applied to... Figure 1 The decision-making endpoint 120 is shown. (As shown) Figure 2 As shown, the method for getting a vehicle out of trouble includes the following steps.

[0040] Step S210: Receive the takeover request from the target vehicle.

[0041] A takeover request refers to a command signal actively sent to the decision-making end when the autonomous driving system of the target vehicle determines that it can no longer drive safely.

[0042] Specifically, the takeover request includes the target vehicle's status information and / or environmental information when it is parked. Optionally, the status information includes the vehicle's desired driving direction, parking factors, etc., where parking factors are a structured description of the problems that prevent the target vehicle from continuing to drive, based on the fusion judgment of the perception, localization, and planning modules of the autonomous driving system. Environmental information includes environmental images, videos, point cloud information, etc., around the target vehicle, used to intuitively reflect the on-site conditions where the vehicle is located.

[0043] Step S220: Based on the takeover request, the target driving strategy is automatically generated using the decision-making module built into the decision-making terminal.

[0044] In this embodiment, the decision-making module can analyze and decide on the target vehicle's environment and parking factors based on the takeover request, and generate a target driving strategy. Specifically, the target driving strategy is a set of action instructions that the target vehicle can understand and operate.

[0045] Optionally, the decision-making module employs a Vision-Language Model (VLM). A Vision-Language Model is an artificial intelligence model capable of simultaneously processing and understanding multimodal information such as images and text, and performing reasoning and content generation based on this information. The Vision-Language Model can simultaneously process visual input (such as environmental information) and linguistic input (such as state information), mapping both to a shared semantic space to achieve cross-modal information alignment and reasoning.

[0046] For example, the image in the environmental information of the takeover request is input into the visual encoder of the decision module to extract rich visual features; the status information is integrated into text information that the decision module can understand through a preset prompt word template, and input into the text encoder of the decision module to obtain text features.

[0047] Optionally, the prompt word template includes the following parts: the vehicle's desired driving direction, parking factors, and problem instructions. For example, problem instructions can be categorized as follows: whether it is possible to continue driving in the current direction; the distance that can be traveled in the current direction; the reliability of the model's decision; whether continuing requires a left or right detour; whether it is necessary to switch driving directions before continuing towards the current target direction; if it is possible to continue driving, where are the key obstacle locations; if it is possible to continue driving, how much further can it travel along the current direction; if a detour is required, which obstacles need to be detoured and in what direction; whether the factors hindering the vehicle's driving as determined by the current autonomous driving system are correct, and what types of obstacles in the image might hinder the vehicle's driving; and what the ideal target point is if one wishes to escape the current driving predicament. It is understood that the specific form of the prompt word template or problem instructions can be set according to the specific application scenario, and this application does not impose specific restrictions on this.

[0048] The decision-making module maps visual and textual features to a shared semantic space, achieving cross-modal information alignment. For example, the decision-making module associates the "gravel" in the environmental information with the parking factor "the planned path is blocked by a stationary obstacle" described in the state information, providing a decision basis for the generation of subsequent target driving strategies.

[0049] Next, the decision-making module considers the surrounding environment and the problems encountered by the vehicle to determine the subsequent driving strategy. For example, when visual features indicate that there are scattered gravel in front of the target vehicle, and text features indicate that the parking factor is "the planned path is blocked by a stationary obstacle," the decision-making module can confirm the existence of the obstacle, identify its type, size, and position, determine the road surface passability, and further assess the obstacle's passability and detour space, ultimately outputting the corresponding target driving strategy. For example, the target driving strategy includes "detour around the gravel, and the detour direction is to the right."

[0050] Step S230: Send the target driving strategy to the target vehicle so that the target vehicle can get out of trouble according to the target driving strategy.

[0051] In this embodiment, after generating the target driving strategy, the decision-making end sends it to the target vehicle. Upon receiving the target driving strategy, the autonomous driving system converts it into specific throttle, brake, and steering control information and executes it. After execution, the target vehicle decides whether to resume autonomous driving based on the execution result. If the target driving strategy is successfully executed and the target vehicle has escaped the predicament, the target vehicle automatically resumes autonomous driving and continues to perform the predetermined driving task; if the target driving strategy fails to execute or the target vehicle has not escaped the predicament, a new takeover request is triggered, or a request is made for the remote driver to take over.

[0052] This embodiment achieves real-time analysis and decision-making of driving strategies through a decision-making module, realizing intelligent and autonomous vehicle extrication processes. This avoids queuing issues caused by delays in manual response or limited resources, improving the smoothness of the target vehicle's movement and its reliability in harsh communication environments. Furthermore, the autonomous vehicle extrication method reduces the need for a remote-controlled cockpit and driver, thus lowering the manpower and hardware costs required for large-scale industrial expansion.

[0053] In the above embodiments, the target driving strategy is the direct basis for enabling the vehicle to autonomously extricate itself from difficult situations. To ensure that the target driving strategy is clear, resolvable, and securely executable, the information content contained in the target driving strategy is defined in a specific and structured manner below.

[0054] In some embodiments, the target driving strategy includes at least one of the following: traffic feasibility information, driving path information, drivable distance information, environmental obstacle information, and decision confidence information.

[0055] Optionally, the feasibility information includes whether it is possible to continue traveling in the current direction, which can be expressed as "can continue traveling in the current direction" or "cannot continue traveling in the current direction".

[0056] The driving path information includes the type of driving direction adjustment and / or the location of the target point. For example, the type of driving direction adjustment includes going straight, detouring to the left, detouring to the right, reversing, etc. The target point includes at least one waypoint and / or destination. Ultimately, based on the location of the target point, the target vehicle can plan a local trajectory from its current location to the target point.

[0057] Driving distance information includes the distance that can be traveled in the current driving direction. The current driving direction can be the desired driving direction or an adjusted driving direction. The driving distance information in the current driving direction is the suggested distance that the target vehicle should travel in the current driving direction, which is a quantitative supplement to the driving path.

[0058] Environmental obstacle information refers to the description of obstacles directly related to the escape decision, specifically including obstacles to be addressed during detours and / or the types of obstacles hindering the target vehicle's movement. For example, the decision module first identifies the types of all obstacles surrounding the target vehicle, and then, combining obstacle type and location, vehicle parking factors, and feasibility assessment, makes an obstacle decision to identify obstacles requiring priority attention (e.g., a large rock that needs to be detoured), while also identifying obstacles that can be ignored or do not require intervention (e.g., loose rocks far from the driving path). In this case, only the obstacles to be addressed during detours and / or the types of obstacles hindering the target vehicle's movement need to be considered.

[0059] Decision confidence information is the reliability assessment of the target driving strategy by the decision-making module, and it is an important reference for evaluating the target driving strategy. When the reliability of the target driving strategy is too low, a new decision should be made, or manual intervention should be requested.

[0060] Optionally, the above information can independently constitute a minimal target driving strategy (e.g., the target driving strategy only includes traffic feasibility information), or multiple types of information can be combined to form structured information including traffic feasibility, driving path, driving distance, obstacle information and decision confidence, so as to obtain a more detailed and reliable target driving strategy.

[0061] In this embodiment, by providing multi-dimensional information such as feasibility information, driving route information, drivable distance information, environmental obstacle information, and decision confidence information, a complete and verifiable instruction set is formed, enhancing the operability and reliability of the target driving strategy. The aforementioned information can dynamically adjust the composition and granularity of the target driving strategy according to the complexity of different application scenarios, communication conditions, and vehicle-side computing power.

[0062] In the above embodiments, decision confidence information is an important reference for evaluating the reliability and effectiveness of the target driving strategy, but it is not the only factor in evaluating the target driving strategy. To ensure the safety of the target vehicle, this application further provides a pre-emptive validity verification mechanism.

[0063] Figure 3 The diagram shown is a flowchart illustrating a vehicle extrication method according to another embodiment of this application. Figure 3 As shown in the embodiment of this application, before the target vehicle gets out of trouble according to the target driving strategy, the following steps are also included.

[0064] Step S310: Determine the result of the effectiveness judgment for the target driving strategy.

[0065] Effectiveness assessment is a comprehensive, real-time evaluation of whether a target driving strategy is feasible in the current moment and environment. Its purpose is to verify whether the target driving strategy is still safe and feasible in the current real-time environment.

[0066] In some embodiments, the criteria for validity determination include the latest vehicle status, real-time environmental information, etc. Validity determination includes at least one of the following: determining whether the target vehicle's parking factors fall within a preset autonomous escape range; determining whether the target vehicle's position change during the period from sending a takeover request to receiving the target driving strategy exceeds a preset range; determining whether there are target obstacles in the current perceived environment; and determining whether the quality of the images in the environmental information meets preset conditions. The above criteria are described below.

[0067] Determining whether the target vehicle's parking factors fall within the predefined autonomous escape range involves re-evaluating the current parking factors before executing the target driving strategy and comparing them with the predefined autonomous escape range. The autonomous escape range includes typical scenarios authorized for autonomous handling (e.g., stationary obstacles, shallow water), while excluding dangerous or complex scenarios requiring manual intervention (e.g., scenarios involving human activity, serious vehicle malfunctions, extreme weather conditions). If the current parking factors do not fall within the autonomous escape range, the target driving strategy should be deemed invalid. This is because, considering the potential time lag between the target vehicle sending a takeover request and the final preparation for executing the target driving strategy, the actual parking factors of the target vehicle may have changed.

[0068] Determining whether the position change of the target vehicle between sending the takeover request and receiving the target driving strategy exceeds a preset range involves calculating the target vehicle's position change from the moment the takeover request is sent to the moment the target driving strategy is received, and comparing this position change with a preset range. If the position change exceeds the preset range, it means that the actual pose of the target vehicle no longer matches the pose information used by the decision-making end to make a judgment, causing the target driving strategy to be no longer applicable. In this case, the target driving strategy should be determined to have failed the validity check.

[0069] Among the criteria for determining whether a target obstacle exists in the current perception environment, the target obstacle can refer to high-risk obstacles with high dynamics, such as pedestrians and other vehicles. Once a target obstacle is perceived in real time in the current perception environment, it means that the current execution environment is a high-risk scenario.

[0070] Determining whether the quality of images in the environmental information meets preset conditions aims to assess the reliability of the decision-making basis. If the image quality used for the decision is found to be too low, it means that the basis for generating the target driving strategy is unreliable, and the target driving strategy should be deemed to have failed the validity assessment.

[0071] Optionally, the validity determination can be made by the target vehicle or the decision-making unit. For example, all of the above validity determinations can be made by the target vehicle. After receiving the target driving strategy, the target vehicle executes the above validity determinations and obtains the results. Alternatively, the above validity determinations can be made by the decision-making unit. If the image quality does not meet the preset conditions, the decision-making module is not used for decision-making, and an attempt is made to obtain a better quality image from the target vehicle, or manual intervention is requested directly. If the image quality meets the preset conditions, the target driving strategy is generated and sent to the target vehicle, which then completes the remaining validity determinations.

[0072] Next, based on the result of the validity assessment, the subsequent extrication process of the target vehicle is controlled. If the result indicates that the validity assessment has been passed, step S320 is executed; if the result indicates that the validity assessment has not been passed, step S330 is executed.

[0073] Step S320: Control the target vehicle to get out of trouble according to the target driving strategy.

[0074] Specifically, the autonomous driving system analyzes the target driving strategy, determines the local path from the current location to the escape target point, and controls the target vehicle to execute it. Furthermore, throughout the entire process, the target vehicle perceives its surrounding environment in real time, and, based on environmental obstacle information, assigns higher perception and monitoring weights to designated key obstacle areas near the path to ensure driving safety.

[0075] Step S330: Provide an auxiliary extrication response to the target vehicle.

[0076] For example, the assisted escape response includes automatically triggering new takeover requests to request the decision-making end to generate a new target driving strategy adapted to the current environment. If the target driving strategies generated by multiple takeover requests fail the validity judgment or time out, manual takeover is requested.

[0077] In some embodiments, providing an assisted extrication response to the target vehicle further includes: directly triggering a manned remote takeover of the target vehicle based on a takeover assistance request.

[0078] Specifically, as a last resort, when the aforementioned autonomous escape responses fail to resolve the issue, or when the autonomous driving system directly determines that the target vehicle is in a high-risk scenario, it will automatically send a takeover request to the remote monitoring center. In response to this takeover request, the remote driver will remotely take over the target vehicle to achieve manual assistance in escaping the predicament.

[0079] This application's embodiments construct a dynamic, closed-loop safety decision-making and execution system. By introducing real-time, multi-dimensional validity assessments, it ensures that only target driving strategies that have undergone validity assessments can be executed, thus improving the safety and reliability of target driving strategies. Simultaneously, the auxiliary escape response process provides reliable safety redundancy, enhancing robustness and overall safety in the face of dynamic and uncertain environments.

[0080] In some embodiments, determining whether a target obstacle exists in the current perception environment includes: comparing the determination of the existence of the target obstacle based on environmental information with the determination of the existence of the target obstacle based on the current perception data of the target vehicle.

[0081] Specifically, determining the existence of a target obstacle based on environmental information involves analyzing the environmental information collected when the target vehicle generates a takeover request to determine if the target obstacle exists. Determining the existence of a target obstacle based on the target vehicle's current perception data involves analyzing the real-time perception data acquired by the target vehicle's own sensors to determine if the target obstacle exists in the current perceived environment. Then, these two existence determinations are compared. If they match, it indicates that the environmental information upon which the decision was based matches the target vehicle's current real-time environment in terms of key safety elements, and the validity determination is considered successful. If they do not match, it indicates that the current real-time environment has changed, and there may be moving pedestrians or vehicles in the environment; in this case, the validity determination is considered unsuccessful.

[0082] In this embodiment, by introducing a cross-validation mechanism between environmental information and current perception data, the consistency between the decision-making basis and the real-time scenario in terms of core safety elements is ensured, the risk of environmental mismatch caused by target obstacles is avoided, the accuracy of effectiveness judgment is improved, and driving safety is enhanced.

[0083] The above embodiments illustrate how to determine the effectiveness of a target driving strategy to improve its reliability. To further improve the effective utilization of decision-making resources, before the decision module generates a detailed target driving strategy, the feasibility of the target vehicle autonomously escaping from trouble can be pre-verified, which will be described in detail below.

[0084] Figure 4 The diagram shown is a flowchart illustrating the steps of autonomously generating a target driving strategy based on a takeover request using a decision-making module built into the decision-making terminal, according to an embodiment of this application. Figure 4 As shown in the embodiments of this application, the step of autonomously generating a target driving strategy based on a takeover request using the decision-making module built into the decision-making terminal includes the following steps.

[0085] Step S410: Based on the takeover request, make a takeover decision using the decision module.

[0086] Specifically, upon receiving a takeover request, the decision-making module first performs a rapid analysis and evaluation of the request. For example, it can make a takeover decision based on preset decision rules or a decision model within the module, determining whether the target vehicle can extricate itself autonomously, and outputting a binary decision result, such as "can extricate itself autonomously" or "cannot extricate itself autonomously".

[0087] For example, if the status information carried in the takeover request indicates that there are moving obstacles around the target vehicle, or that the environmental complexity exceeds the understanding range of the decision module, or that the image carried in the takeover request is not clear enough, then the output is "Cannot escape autonomously".

[0088] Step S420: If the decision result indicates that the decision-making end can autonomously extricate itself from the predicament, then the decision-making module autonomously generates the target driving strategy.

[0089] Specifically, if the decision result indicates that the vehicle can autonomously extricate itself from the predicament, the decision module autonomously generates a target driving strategy. If the decision result indicates that the vehicle cannot autonomously extricate itself from the predicament, an auxiliary extrication response is provided to the target vehicle. The specific implementation of this step is similar to the above embodiment and will not be repeated here.

[0090] This application's embodiments introduce a pre-emptive feasibility check at the decision-making end to avoid complex calculations for invalid scenarios that exceed capability boundaries or are high-risk, thereby saving computing and communication resources at both the decision-making and vehicle ends. Simultaneously, this solution shortens the response path for manual takeover, optimizes overall system resource allocation, improves overall processing efficiency, reduces operating costs, and strengthens risk control through source diversion, ensuring that human resources can be more focused on complex situations that truly require intervention.

[0091] In some embodiments, based on a takeover request, a takeover decision is made using a decision module, including: if it is determined that the target vehicle meets preset triggering conditions, then the takeover request is input to the decision module, and the takeover decision is made using the decision module.

[0092] This step provides a pre-processing request filtering mechanism for takeover decision-making. Only when a takeover request meets the preset triggering conditions will it be input into the decision-making module to initiate the subsequent takeover decision-making process, thereby improving the overall response speed and processing efficiency of the decision-making module.

[0093] Optionally, the triggering conditions include at least one of the following: the target vehicle's continuous parking time exceeds a preset threshold; the target vehicle's parking factors belong to preset decision-making factors; the target vehicle's current parking location belongs to a preset triggering area. The triggering conditions mentioned above are described in detail below.

[0094] The timing for determining whether a target vehicle's continuous parking time exceeds a preset threshold begins when the vehicle enters an unexpected parking state. Only when the duration of the unexpected parking state exceeds the preset threshold is the vehicle considered to be in a predicament requiring external intervention, thus fulfilling the trigger condition. The preset threshold can be set based on scenario experience (e.g., 30 seconds). This condition aims to exclude brief, normal parking disturbances that may be caused by traffic flow or operational procedures, effectively preventing unnecessary takeover decisions due to short wait times.

[0095] The trigger condition, where the target vehicle's parking factors fall under the preset decision-making factors, includes typical parking factors that have been evaluated and are suitable for the decision-making module to attempt analysis and decision-making. Examples include the planned path being blocked by irregular obstacles or partial path loss. The trigger condition is only satisfied when the parking factors carried in the takeover request fall under the preset decision-making factors. Parking factors unsuitable for autonomous escape, such as sensor failure, are not preset decision-making factors, and in these cases, the trigger condition must be determined.

[0096] In the trigger condition that the target vehicle's current parking location falls within a preset trigger area, the preset trigger area is a pre-defined area within the target vehicle's operational scenario. For example, this trigger area includes major transport routes and areas near intersections, which typically have high requirements for traffic continuity and are prone to congestion. The trigger condition is considered met only when the target vehicle's parking location is within this trigger area. This condition prioritizes core operational areas and avoids invalid triggers in non-operational areas.

[0097] In this embodiment, by setting a composite triggering condition with multiple factors, the pre-management of the takeover decision-making process is realized, which effectively reduces the occupation of computing and communication resources by invalid takeover requests, enables decision-making resources to be called up in a priority and timely manner, and improves the overall response efficiency and processing pertinence of the decision-making end.

[0098] As described in the above embodiments, the decision-making module includes a visual language model. Since the general visual language model is not designed for unmanned driving scenarios in mines, it may have problems such as inaccurate identification of mine-specific obstacles (such as scattered gravel and special road conditions), misunderstanding of specialized terms and working conditions such as pitfalls and crushing stations, and inconsistencies between the reasoning logic and the actual dynamic constraints of the vehicle or operational safety regulations. These issues may lead to insufficient reliability of the generated target driving strategy or potential safety hazards.

[0099] Therefore, in this embodiment, the visual language model is a model adjusted using target training data. This target training data is takeover case data specifically collected for unmanned mining scenarios, including the original takeover data of the unmanned mining vehicle and the ground truth values ​​of driving behavior calibrated based on the original takeover data. The original takeover data of the unmanned mining vehicle includes the state information and environmental information of the unmanned mining vehicle when it is taken over; the ground truth values ​​of driving behavior calibrated based on the original takeover data refer to the driving behaviors taken by the remote-controlled driver or the driver inside the unmanned mining vehicle in response to the current driving difficulties.

[0100] The visual language model is then fine-tuned under supervised supervision using the training data from this target scenario, establishing a correlation between mine visual features and reasonable driving behavior. Furthermore, specific prompt word templates for mine scenarios can be designed. Through these methods, the general-purpose visual language model is transformed into a visual language model specifically for mine escape decision-making, thereby improving the accuracy, safety, and practicality of decision-making.

[0101] In some embodiments, to ensure the reliability and usability of the visual language model before actual deployment, its performance can be tested and evaluated after training is completed. Optionally, the preset performance requirements for the visual language model during the training phase include at least one of the following: a dangerous misclassification rate lower than a first threshold; a conservative misclassification rate lower than a second threshold; and inference time lower than a third threshold. These performance requirements are described below.

[0102] Within the performance requirement of keeping the hazard misjudgment rate below the first threshold, the hazard misjudgment rate is the percentage of non-drivable scenarios that the visual language model misclassifies as drivable. For example, misjudging a scene with a moving obstacle or a deep pit as one that can be traversed directly could lead to a collision with the obstacle, the vehicle falling into the pit, or even a rollover, directly threatening the safety of the vehicle itself, surrounding personnel, and equipment. Understandably, the hazard misjudgment rate directly relates to the baseline of driving safety and must therefore be controlled at a low first threshold, such as below 0.1%, to minimize driving accidents caused by misjudgments from the visual language model.

[0103] In the performance requirement of keeping the conservative misclassification rate below the second threshold, the conservative misclassification rate is the percentage of drivable scenarios that the visual language model misclassifies as drivable. For example, it might misclassify a passable shallow puddle or small gravel as impassable. Understandably, an excessively high conservative misclassification rate leads to unnecessary human intervention, reduces the smoothness of the target vehicle's movement, and consequently impacts the vehicle's operational efficiency. Therefore, it needs to be controlled within a reasonable second threshold, for example, below 5%.

[0104] In the performance requirement of inference time being below the third threshold, inference time refers to the average computation time required for the visual language model to process a single takeover request. To meet the real-time requirements of mine operations, inference time needs to be shorter than the third threshold determined based on the business scenario, for example, within 2 seconds.

[0105] It is understood that the specific values ​​of the first, second, and third thresholds mentioned above are merely exemplary and not restrictive. In practical applications, the above thresholds can be determined based on factors such as mining conditions, target vehicle performance parameters, the performance of the hardware computing platform used, and mine operation safety standards.

[0106] This application embodiment establishes quantitative testing standards for visual language models by setting multiple performance indicators, which not only ensures the reliability of visual language models, but also guides the direction of model training optimization. That is, under the premise of ensuring driving safety, it minimizes the conservative misjudgment rate and inference time, thereby improving the effectiveness and safety of visual language models.

[0107] The above text combined Figures 1 to 4 The following describes in detail an embodiment of a vehicle extrication method applied to the decision-making stage, in conjunction with... Figure 5 This application provides a detailed description of embodiments of the vehicle extrication method applied to the vehicle side. It should be understood that the descriptions of the vehicle-side method embodiments correspond to the descriptions of the embodiments applied to the decision-making side; therefore, any parts not described in detail can be referred to the above-described embodiments applied to the decision-making side.

[0108] Figure 5 The diagram shown is a flowchart illustrating a vehicle extrication method according to another embodiment of this application. Exemplarily, the vehicle extrication method provided in this embodiment is applied to... Figure 1 The target vehicle 110 shown, more specifically, is the autonomous driving system applied to the target vehicle 110. For example... Figure 5 As shown, the method for getting a vehicle out of trouble includes the following steps.

[0109] Step S510: If the target vehicle is detected to be parked and the parking factors meet the specified conditions, a takeover request is sent to the decision-making end.

[0110] Specifically, when the target vehicle stops, the autonomous driving system autonomously determines the factors contributing to the stop and judges whether the factors meet specified conditions. Optionally, these specified conditions can be the autonomous escape range described in the above embodiments, or they can be set separately.

[0111] If the parking factors meet the specified conditions, the target vehicle will proactively encapsulate a takeover request containing status information and / or environmental information, and send the takeover request to the decision-making end through the communication link.

[0112] Step S520: Receive the target driving strategy and extricate yourself from the predicament according to the target driving strategy.

[0113] Specifically, based on the takeover request, the decision-making unit autonomously generates a corresponding target driving strategy using its built-in decision-making module and sends it to the target vehicle via a communication link. The target vehicle receives the target driving strategy and attempts to extricate itself from the predicament according to it. The specific implementation of this step can be found in the above embodiment and will not be repeated here.

[0114] The above text combined Figures 1 to 5 Implementation examples of vehicle extrication methods applied to both the decision-making and vehicle ends are introduced below. Figure 6This application provides a detailed description of embodiments of the vehicle escaping system. It should be understood that the descriptions of the method embodiments correspond to the descriptions of the system embodiments; therefore, any parts not described in detail can also be referred to the above method embodiments.

[0115] In this embodiment, the vehicle extrication system includes a target vehicle and a decision-making terminal. The decision-making terminal has a built-in decision module, and the target vehicle and the decision-making terminal exchange information through a communication link. The following describes... Figure 6 This section introduces the specific applications of the target vehicle and the decision-making end, as well as the information exchange process between the two.

[0116] Figure 6 The diagram shown is a schematic representation of the information interaction between the target vehicle and the decision-making unit according to an embodiment of this application. Figure 6 As shown, the target vehicle is configured to generate a takeover request containing its own vehicle's status information and / or environmental information when it is detected that the target vehicle is parked and the parking factors meet specified conditions. The takeover request is then sent to the communication link and transmitted to the decision-making end via the communication link.

[0117] The decision-making unit is configured to autonomously generate a target driving strategy based on a takeover request using the decision module. Then, the target driving strategy is sent to the communication link, and subsequently transmitted to the target vehicle via the communication link.

[0118] The target vehicle is also configured to extricate itself from a difficult situation according to a target driving strategy. Before this, the target vehicle can first perform a validity check on the target driving strategy. If the check indicates that the validity check has been passed, the vehicle will proceed with the extrication according to the target driving strategy. If the check indicates that the validity check has not been passed, the vehicle will send an assistance request to the communication link. This assistance request can be transmitted to the decision-making end for a new extrication decision, or it can be transmitted to the remote monitoring center to request manual intervention.

[0119] Below, for reference Figure 7 This describes an electronic device according to embodiments of the present application. Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application.

[0120] like Figure 7 As shown, the electronic device 70 includes one or more processors 701 and memory 702.

[0121] The processor 701 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 70 to perform desired functions.

[0122] The memory 702 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 701 may execute the program instructions to implement the vehicle extrication methods of the various embodiments of this application described above and / or other desired functions. The computer-readable storage medium may also store various content such as target vehicle status information, environmental information, etc.

[0123] In one example, the electronic device 70 may also include an input device 703 and an output device 704, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0124] The input device 703 may include, for example, a keyboard, a mouse, etc.

[0125] The output device 704 can output various information to the outside, including the target vehicle's status information, environmental information, and the target driving strategy autonomously generated by the decision-making module. The output device 704 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0126] Of course, for the sake of simplicity, Figure 7 Only some of the components of the electronic device 70 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 70 may include any other suitable components depending on the specific application.

[0127] This application also provides a vehicle traction device for executing any of the above-described vehicle traction methods. The vehicle traction methods have been described in detail above and will not be repeated here.

[0128] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the vehicle extrication methods according to various embodiments of this application described above.

[0129] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0130] Furthermore, embodiments of this application can also be vehicles. Optionally, the vehicle includes an unmanned mining truck used in mining scenarios. When the vehicle stops and the stopping factors meet specified conditions, the vehicle can generate and send a takeover request to the decision-making end; in addition, the vehicle can also accept a target driving strategy autonomously generated by the decision-making end based on the built-in decision-making module, and extricate itself from the predicament according to the target driving strategy.

[0131] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the vehicle extrication methods according to various embodiments of this application described above.

[0132] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0133] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0134] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0135] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0136] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0137] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for getting a vehicle out of trouble, characterized in that, When applied to the decision-making process, the method includes: Receive a takeover request from the target vehicle, wherein the takeover request includes the target vehicle's status information and / or environmental information when it is parked; Based on the takeover request, the target driving strategy is autonomously generated using the decision-making module built into the decision-making terminal. The target driving strategy is sent to the target vehicle so that the target vehicle can get out of trouble according to the target driving strategy.

2. The vehicle extrication method according to claim 1, characterized in that, The target driving strategy includes at least one of the following: traffic feasibility information, driving route information, drivable distance information, environmental obstacle information, and decision confidence information; Preferably, the feasibility information includes whether it is possible to continue traveling in the current direction; The driving path information includes the type of driving direction adjustment and / or the location of the target point; The drivable distance information includes the drivable distance information in the current driving direction; The environmental obstacle information includes obstacles to be avoided during detours and / or types of obstacles that impede the target vehicle's movement.

3. The vehicle extrication method according to claim 1 or 2, characterized in that, Before the target vehicle attempts to escape according to the target driving strategy, the following is also included: Determine the result of the effectiveness assessment for the target driving strategy; Based on the results of the effectiveness assessment, the subsequent extrication process of the target vehicle is controlled; Preferably, the validity determination includes at least one of the following: determining whether the parking factors of the target vehicle fall within a preset autonomous escape range; determining whether the position change of the target vehicle during the period from sending the takeover request to receiving the target driving strategy exceeds a preset range; determining whether there is a target obstacle in the current perceived environment; and determining whether the quality of the image in the environmental information meets preset conditions. Preferably, the validity determination is made by the target vehicle or the decision-making terminal.

4. The vehicle extrication method according to claim 3, characterized in that, Based on the result of the validity judgment, the subsequent extrication process of the target vehicle is controlled, including: If the result indicates that the validity judgment is passed, then the target vehicle is controlled to extricate itself from the predicament according to the target driving strategy; If the result indicates that the validity judgment has not been passed, then an auxiliary extrication response is provided to the target vehicle; Preferably, providing the target vehicle with an assisted escape response includes: triggering a manned remote takeover of the target vehicle based on a takeover assistance request.

5. The vehicle extrication method according to claim 3, characterized in that, The determination of whether a target obstacle exists in the current perceived environment includes: The existence determination of the target obstacle based on the environmental information is compared with the existence determination of the target obstacle based on the current perception data of the target vehicle. If both are consistent, the validity check passes. If the two are inconsistent, the validity judgment fails.

6. The vehicle extrication method according to any one of claims 1 to 5, characterized in that, The step of autonomously generating a target driving strategy based on the takeover request using the decision-making module built into the decision-making terminal includes: Based on the takeover request, the takeover decision is made using the decision module; If the decision result indicates that the decision-making end can autonomously extricate itself from the predicament, then the decision-making module will autonomously generate the target driving strategy. Preferably, if the decision result indicates that the vehicle cannot extricate itself from the predicament on its own, then an auxiliary extrication response is provided to the target vehicle.

7. The vehicle extrication method according to claim 6, characterized in that, The step of making a takeover decision based on the takeover request using the decision module includes: If it is determined that the target vehicle meets the preset triggering conditions, the takeover request is input to the decision module, and the decision module is used to make a takeover decision. Preferably, the triggering condition includes at least one of the following: the continuous parking time of the target vehicle exceeds a preset threshold; the parking factor of the target vehicle belongs to a preset decision-making factor; the current parking position of the target vehicle belongs to a preset triggering area; Preferably, the decision-making module includes a visual language model, which is a model adjusted by target training data. The target training data includes the original takeover data of the unmanned mining vehicle and the true values ​​of driving behavior calibrated based on the original takeover data.

8. The vehicle extrication method according to claim 7, characterized in that, The preset performance requirements for the visual language model during the training phase include at least one of the following: The false alarm rate is below the first threshold; The conservative error rate is below the second threshold; The reasoning time is below the third threshold; The danger misjudgment rate is the percentage of non-drivable scenarios that the visual language model misjudges as drivable scenarios, and the conservative misjudgment rate is the percentage of drivable scenarios that the visual language model misjudges as non-drivable scenarios.

9. A method for getting a vehicle out of trouble, characterized in that, Applied to the vehicle end, the method includes: If a target vehicle is detected to be parked and the parking factors meet the specified conditions, a takeover request is sent to the decision-making end; wherein, the takeover request includes the status information and / or environmental information of the target vehicle; The system receives the target driving strategy and performs the extrication operation according to the target driving strategy; wherein the target driving strategy is generated autonomously by the decision-making terminal based on the built-in decision module.

10. A vehicle traction device, characterized in that, Used to perform the vehicle traction method according to any one of claims 1-8, or used to perform the vehicle traction method according to claim 9.

11. A vehicle traction system, characterized in that, This includes the target vehicle and the decision-making terminal, which has a built-in decision-making module. The target vehicle is configured to send a takeover request to the decision-making end when it is detected that the target vehicle is parked and the parking factors meet the specified conditions. The takeover request includes the vehicle's status information and / or environmental information. The decision-making terminal is configured to autonomously generate a target driving strategy based on the takeover request using the decision-making module, and send it to the target vehicle. The target vehicle is also configured to extricate itself from a difficult situation according to the target driving strategy.

Images