Methods, equipment, media and programs for scheduling and training models of autonomous vehicles

By acquiring multi-dimensional perception data in real time and calling the trained vehicle control model for human-like scheduling, the problem of control stability and safety of unmanned vehicles in complex dynamic scenarios is solved, achieving driving behavior similar to that of a human driver and improving the stability and safety of driving control.

CN123067145APending Publication Date: 2026-11-10LINGONG GROUP (JINAN) HEAVY MACHINERY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611318355.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-28
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing autonomous vehicle control methods are difficult to adapt to complex, dynamic, and time-varying real-world conditions. They cannot dynamically adjust control strategies in conjunction with the real-time driving environment, resulting in lag in vehicle speed regulation, deviation in braking response, poor driving smoothness, and insufficient environmental adaptability and safety.

Method used

By acquiring multi-dimensional operational perception data in real time, the vehicle control model trained with human driving behavior sample data is invoked to generate vehicle motion control commands. Combined with the driving decision model and trajectory optimization control model, human-like scheduling is performed to improve the stability of vehicle driving control in complex dynamic scenarios.

Benefits of technology

It achieves human-like alignment of driving behavior in autonomous vehicles, improves the stability and safety of driving control in complex dynamic scenarios, narrows the gap with human drivers in terms of scenario adaptability and interaction rationality, and enhances environmental adaptability and operational robustness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067145A_ABST
    Figure CN123067145A_ABST
Patent Text Reader

Abstract

This invention discloses a method, device, medium, and program for scheduling and training unmanned vehicles. The method includes: acquiring multi-dimensional operational perception data of a target unmanned vehicle in a target driving scenario in real time; invoking a vehicle control model to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein the vehicle control model is trained using sample data of target human driving behavior; and scheduling the target unmanned vehicle in the target driving scenario using the vehicle motion control commands. The technical solution of this invention can achieve human-like alignment of unmanned vehicle driving behavior, thereby improving the driving control stability of unmanned vehicles in complex dynamic scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of autonomous driving technology, and in particular to an autonomous vehicle scheduling and model training method, device, medium and program. Background Technology

[0002] Autonomous vehicle control is a core technology that ensures stable vehicle speed, precise braking, and safe operation. It directly determines the smoothness and reliability of vehicle driving and is an important foundation for autonomous driving systems to achieve autonomous driving.

[0003] Currently, most autonomous vehicle control systems employ traditional PID (Proportional-Integral-Derivative) control or model predictive control based on vehicle dynamics models. Existing solutions lack adaptive optimization designs for complex, dynamic, and time-varying real-world autonomous driving conditions. They cannot dynamically adjust control strategies in conjunction with the real-time driving environment, making it difficult to adapt to the control requirements of unconventional driving scenarios such as complex road surfaces and sudden situations. This easily leads to problems such as lag in speed regulation, deviations in braking response, and decreased ride smoothness, reducing the environmental adaptability and driving safety of autonomous vehicles and failing to meet the stable driving control requirements of high-level autonomous driving systems in complex dynamic scenarios. Summary of the Invention

[0004] This invention provides a method, device, medium, and program for scheduling and training unmanned vehicles, which can achieve human-like alignment of the driving behavior of unmanned vehicles, thereby improving the driving control stability of unmanned vehicles in complex dynamic scenarios.

[0005] According to one aspect of the present invention, an autonomous vehicle scheduling method is provided, applied to autonomous vehicles, comprising: Real-time acquisition of multi-dimensional operational perception data of the target autonomous vehicle in the target driving scenario; The vehicle control model is invoked to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein, the vehicle control model is trained using sample data of the target human driving behavior. The target unmanned vehicle in the target driving scenario is scheduled using the vehicle motion control commands.

[0006] According to another aspect of the present invention, a method for training an autonomous vehicle scheduling model is provided, comprising: Acquire full-condition human driving behavior data; The driving ability level of each human driver is obtained by scoring the full-condition human driving behavior data according to the preset test index system. When the driving ability level of the target human driver is determined to be the first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as the target human driving behavior sample data. The vehicle control model is trained based on the target human driving behavior sample data; The trained vehicle control model is distributed to each autonomous vehicle in the target driving scenario, so that the autonomous vehicles can execute the autonomous vehicle scheduling method described in any embodiment of the present invention through the vehicle control model.

[0007] According to another aspect of the present invention, an unmanned vehicle dispatching device is provided, configured in an unmanned vehicle, comprising: The multi-dimensional operational perception data acquisition module is used to acquire multi-dimensional operational perception data of the target unmanned vehicle in the target driving scenario in real time. The vehicle motion control command generation module is used to call the vehicle control model to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein, the vehicle control model is trained using sample data of target human driving behavior; The target autonomous vehicle scheduling module is used to schedule target autonomous vehicles in the target driving scenario through the vehicle motion control commands.

[0008] According to another aspect of the present invention, an unmanned vehicle scheduling model training device is provided, comprising: The full-condition manual driving behavior data acquisition module is used to acquire full-condition manual driving behavior data; The driving ability level determination module is used to score the full-condition human driving behavior data according to a preset test index system to obtain the driving ability level of each human driver. The target manual driving behavior sample data filtering module is used to select the full-condition manual driving behavior data of the target manual driver as the target manual driving behavior sample data when the driving ability level of the target manual driver is determined to be the first preset level. The vehicle control model training module is used to train the vehicle control model based on the target human driving behavior sample data; The vehicle control model distribution module is used to distribute the trained vehicle control model to each unmanned vehicle in the target driving scenario, so that the unmanned vehicles can execute the unmanned vehicle scheduling method described in any embodiment of the present invention through the vehicle control model.

[0009] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the autonomous vehicle scheduling method or the autonomous vehicle scheduling model training method according to any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the unmanned vehicle scheduling method or the unmanned vehicle scheduling model training method according to any embodiment of the present invention.

[0011] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the autonomous vehicle scheduling method or the autonomous vehicle scheduling model training method according to any embodiment of the present invention.

[0012] This invention acquires multi-dimensional operational perception data of a target autonomous vehicle in a target driving scenario in real time, and calls a vehicle control model trained using target human driving behavior sample data to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data. Furthermore, the vehicle motion control commands are used to schedule the target autonomous vehicle in the target driving scenario. The vehicle control model is trained in the cloud. Specifically, by acquiring full-condition human driving behavior data, scoring the full-condition human driving behavior data according to a preset test index system, the driving ability level of each human driver is obtained. If the driving ability level of the target human driver is determined to be a first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as target human driving behavior sample data. Further, the vehicle control model is trained based on the target human driving behavior sample data, and the trained vehicle control model is distributed to each autonomous vehicle in the target driving scenario. This solution solves the defects of poor scenario adaptability and insufficient dynamic control performance in existing autonomous vehicle scheduling methods, enabling human-like alignment of autonomous vehicle driving behavior, thereby improving the driving control stability of autonomous vehicles in complex dynamic scenarios.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of an unmanned vehicle scheduling method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a method for training an unmanned vehicle scheduling model according to Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of an unmanned vehicle dispatching device provided in Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of an unmanned vehicle scheduling model training device provided in Embodiment 4 of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in Embodiment 5 of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," and "third," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] Example 1 Figure 1This is a flowchart of an autonomous vehicle scheduling method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where a vehicle control model trained using human driving behavior sample data is used to achieve human-like control of autonomous vehicles. This method can be executed by an autonomous vehicle scheduling device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device, which can be an autonomous vehicle. Correspondingly, as... Figure 1 As shown, the method includes the following operations: S110: Real-time acquisition of multi-dimensional operational perception data of the target unmanned vehicle in the target driving scenario.

[0019] The target driving scenario can be an application scenario corresponding to the autonomous vehicle performing a preset task. For example, the target driving scenario may include, but is not limited to, open-pit mine transportation scenarios, port collection and distribution scenarios, and industrial park in-plant transfer scenarios. This embodiment of the invention does not limit the specific type of the target driving scenario. The target autonomous vehicle can be any autonomous vehicle within the target driving scenario. Multi-dimensional operational perception data can be a data set collected by the target autonomous vehicle during operation in the target driving scenario, covering multiple dimensions such as the vehicle's own state, surrounding environment, interactive objects, and operational conditions.

[0020] In this embodiment of the invention, the target unmanned vehicle operating in the target driving scenario can collect and acquire its own multi-dimensional operational perception data in real time using sensing and onboard hardware units such as onboard LiDAR, millimeter-wave radar, onboard cameras, positioning modules, and vehicle controllers. This multi-dimensional operational perception data integrates the vehicle's own operating parameters, environmental perception results, surrounding interactive target motion information, spatiotemporal positioning data, and operational status data, providing data support for vehicle decision-making and control.

[0021] It should be noted that the target driving scenario may include only driverless vehicles, or it may include both manually driven vehicles and driverless vehicles. This embodiment of the invention does not limit this.

[0022] S120. The vehicle control model is invoked to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein, the vehicle control model is trained using sample data of the target human driving behavior.

[0023] The vehicle control model can be trained using sample data of target human driving behavior and is capable of outputting motion control commands for the autonomous vehicle. The sample data of target human driving behavior can be a collection of driving behavior data from excellent drivers. The vehicle motion control commands can be instructions used to regulate the motion state of the target autonomous vehicle.

[0024] Accordingly, after acquiring the multi-dimensional operational perception data of the target autonomous vehicle in the target driving scenario, this data can be input into the vehicle control model trained using sample data of the target human driving behavior. Furthermore, the vehicle control model can output vehicle motion control commands adapted to the current operating scenario based on the multi-dimensional operational perception data of the target autonomous vehicle in the target driving scenario.

[0025] In this embodiment of the invention, the vehicle control model includes a driving decision model and a trajectory optimization control model. The step of calling the vehicle control model to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data may include: inputting the multi-dimensional operational perception data into the driving decision model to obtain a sequence of time-series control parameters for the target autonomous vehicle; wherein the sequence of time-series control parameters includes drive control parameters, braking control parameters, and gear shifting control parameters; and calling the trajectory optimization control model to perform rolling time-domain optimization based on the multi-dimensional boundary constraints of vehicle operation and the sequence of time-series control parameters to determine the vehicle motion control commands for the target autonomous vehicle.

[0026] The vehicle control model can include a driving decision model and a trajectory optimization control model. The driving decision model can be a model for outputting a driving decision sequence. The trajectory optimization control model can be a model for optimizing the driving decision sequence output by the driving decision model. The temporal control parameter sequence can be a driving decision sequence output by the driving decision model, arranged in chronological order. For example, the temporal control parameter sequence can include, but is not limited to, drive control parameters, braking control parameters, and gear shifting control parameters. This embodiment of the invention does not limit the specific content included in the temporal control parameter sequence. Drive control parameters can be parameters used to control the vehicle's power output and achieve driving actions. Braking control parameters can be control parameters used to control the magnitude and timing of braking force applied by the vehicle's braking system. Gear shifting control parameters can be control parameters used to control the vehicle's gearbox to complete gear shifting actions. The multi-dimensional boundary constraints of vehicle operation can be a set of multi-dimensional limit thresholds and constraint rules pre-set for the vehicle's motion state, actuator output, and operating conditions during the autonomous vehicle's operation in the target driving scenario. In this embodiment of the invention, the vehicle control model can be a control model with a two-layer architecture. The upper layer can be a human-like driving decision model based on bidirectional LSTM+Transformer (Long Short-Term Memory-Transformer) temporal deep learning, and the lower layer can be an MPC (Model Predictive Control) trajectory tracking controller based on a dynamic model. When the vehicle control model generates vehicle motion control commands for the target autonomous vehicle based on multi-dimensional operational perception data, the multi-dimensional operational perception data can first be input into the upper-layer driving decision model. The driving decision model then outputs a sequence of temporal control parameters, including drive control parameters, braking control parameters, and gear shifting control parameters, based on the multi-dimensional operational perception data. On this basis, a trajectory optimization control model can be invoked to perform rolling time-domain optimization based on the multi-dimensional boundary constraints of vehicle operation and the temporal control parameter sequence to obtain the vehicle motion control commands for the target autonomous vehicle. This allows for precise control of the target autonomous vehicle based on the vehicle motion control commands. In a specific example, the multidimensional boundary constraints for vehicle operation may include, but are not limited to, vehicle dynamics constraints, safety distance constraints, and road adhesion coefficient constraints. This embodiment of the invention does not limit the specific content included in the multidimensional boundary constraints for vehicle operation. Vehicle dynamics constraints may be a set of constraint limits characterizing the physical limits of a vehicle's motion capabilities. Safety distance constraints may be pre-set, dynamically adjustable distance threshold rules used to limit the minimum distance that the target autonomous vehicle should maintain between itself and surrounding obstacles, other operating vehicles, pedestrians, or road boundaries to prevent collision risks. Road adhesion coefficient constraints may be a boundary characterizing the maximum tire grip capability that the road surface can provide.

[0027] S130. The target unmanned vehicle in the target driving scenario is scheduled through the vehicle motion control command.

[0028] Specifically, after generating the vehicle motion control commands for the target autonomous vehicle, the target autonomous vehicle can execute the vehicle motion control commands sequentially according to the time sequence, thereby achieving precise scheduling of its own driving status.

[0029] Therefore, this invention, through training a human-like vehicle control model using sample data of target human driving behavior, enables the target autonomous vehicle to align its driving behavior with the human-like vehicle motion control commands generated by the vehicle control model. This human-like driving behavior improves the smoothness and stability of autonomous vehicles in complex dynamic scenarios such as traffic flow, non-motorized vehicle weaving, and the appearance of temporary obstacles. It narrows the gap between autonomous vehicles and human drivers in terms of scenario adaptability and interaction rationality, enhances the environmental adaptability and operational robustness of the autonomous driving system in the face of uncertain dynamic traffic elements, and further ensures the driving safety of autonomous vehicles throughout the entire traffic flow.

[0030] In an optional embodiment of the present invention, the unmanned vehicle scheduling method may further include: acquiring environmental visual perception data collected by the onboard camera of the target unmanned vehicle; calling a hand target localization and detection model to perform hand detection on the environmental visual perception data; if it is determined that the environmental visual perception data includes hand movements, calling a target feature extraction model to extract hand movement feature information based on the environmental visual perception data; calling a target classification model to classify the hand movements based on the hand movement feature information to obtain a gesture classification result; determining a first vehicle control command based on the gesture classification result, and adjusting the vehicle's driving motion state based on the first vehicle control command.

[0031] The environmental visual perception data can be image data collected in real time by the target autonomous vehicle, used to characterize the vehicle's surrounding environment. The hand target localization and detection model can be a model used to detect the presence of hand movements in the environmental visual perception data. The target feature extraction model can be a model used to extract features of hand movements included in the environmental visual perception data. The hand movement feature information can be the feature information of hand movements included in the environmental visual perception data extracted by the target feature extraction model. The target classification model can be a model used to classify hand movements. The gesture classification result can be used to characterize the category to which the hand movements included in the environmental visual perception data belong. The first vehicle control command can be a vehicle control command determined based on the gesture classification result, used to control the target autonomous vehicle.

[0032] In this embodiment of the invention, a standardized gesture recognition subsystem based on machine vision can be deployed on the target unmanned vehicle. This subsystem reuses the existing front-view, side-view, and rear-view explosion-proof starlight-level cameras of the target unmanned vehicle in terms of hardware, eliminating the need for extensive additional hardware installation and effectively reducing vehicle modification and scenario adaptation costs. The subsystem completes model training based on pre-defined standardized interactive gestures. For open-pit mine transportation scenarios, the standardized interactive gestures can directly adopt the conventional operating gestures of existing mine drivers and excavator operators. These gestures are broadly categorized into two types: loading and unloading gestures and road traffic interaction gestures, fully covering mainstream human-vehicle interaction scenarios in mines. On-site personnel do not need to relearn and adapt to new gestures, demonstrating strong adaptability and practicality.

[0033] Furthermore, the target autonomous vehicle can collect environmental visual perception data through its onboard camera and call a hand target localization and detection model, such as the RT-DETR (Real-Time Detection Transformer) model, to perform hand detection on the environmental visual perception data, thereby determining whether the environmental visual perception data includes valid hand movements. For harsh working conditions such as strong light, dust, and low light at night in open-pit mines, this system can effectively improve image quality in complex scenes through optimized algorithms such as image dehazing enhancement, infrared illumination, and multi-frame image fusion recognition, significantly improving the accuracy and stability of gesture recognition.

[0034] When hand movements are detected in the environmental visual perception data, a target feature extraction model, such as HRNet (High-Resolution Network), can be further invoked to extract 21 3D keypoints of the hand movements included in the environmental visual perception data, accurately obtaining hand movement feature information. After obtaining the hand movement feature information, a target classification model can be invoked to classify the hand movements based on the hand movement feature information, obtaining the gesture classification result. For example, the target classification model can be a CNN (Convolutional Neural Network) model.

[0035] This system sets a 95% confidence threshold for gesture classification. When the confidence level of the gesture classification result is greater than or equal to 95%, the system can accurately parse the gesture semantics, convert it into a first vehicle control command, and adjust the vehicle's driving motion state according to the first vehicle control command. When the confidence level of the gesture classification result is lower than the threshold, that is, when there is uncertainty in the recognition result, the target unmanned vehicle immediately remains stationary, and the back-end dispatch center's takeover confirmation mechanism is triggered, completely avoiding vehicle misoperation caused by recognition errors and ensuring the safety of mining operations.

[0036] In an optional embodiment of the present invention, the unmanned vehicle scheduling method may further include: acquiring environmental acoustic perception data collected by the onboard digital intercom of the target unmanned vehicle; wherein the environmental acoustic perception data includes voice interaction commands output by each manually driven vehicle in the target driving scenario; if it is determined that the environmental acoustic perception data includes the vehicle number of the target unmanned vehicle, determining a second vehicle control command and interactive response information of the target unmanned vehicle based on the voice interaction commands included in the environmental acoustic perception data; adjusting the vehicle's driving motion state according to the second vehicle control command, and broadcasting the interactive response information through the onboard digital intercom of the target unmanned vehicle.

[0037] The environmental acoustic perception data can be collected in real time by the target autonomous vehicle to characterize the acoustic features of the vehicle's surrounding environment. The voice interaction commands can be the interaction commands output by each manually driven vehicle in the target driving scenario. The second vehicle control command can be the vehicle motion state adjustment command determined based on the voice interaction commands included in the environmental acoustic perception data. The interactive response information can be the response information generated based on the voice interaction commands.

[0038] In this embodiment of the invention, a voice interaction subsystem can be deployed on the target unmanned vehicle. To enable interaction between the various driving vehicles in the target driving scenario, digital walkie-talkies of the same frequency band and model can be installed on each unmanned and manually driven vehicle in the target driving scenario. These digital walkie-talkies are compatible with existing communication channels and equipment in the target operation scenario and establish a connection with the vehicle domain controller via a serial port to receive and transmit voice signals.

[0039] Furthermore, the target autonomous vehicle can collect environmental acoustic perception data, including voice interaction commands output by various manually driven vehicles in the target driving scenario, via an onboard digital intercom. After obtaining the environmental acoustic perception data, the target driving vehicle can filter background noise such as engine and mechanical noise using an AI (Artificial Intelligence) noise reduction model. Then, it uses a customized trained ASR (Automatic Speech Recognition) model and NLU (Natural Language Understanding) model to recognize the environmental acoustic perception data. The constructed corpus covers existing dispatching language, operational interaction statements, industry professional terms, and local regional accents in the target driving scenario. When the vehicle's vehicle number is detected in the environmental acoustic perception data, the voice interaction commands included in the environmental acoustic perception data can be parsed and converted into second vehicle control commands. At the same time, interactive response information can be generated through TTS (Text To Speech) speech synthesis and broadcast externally via the onboard digital intercom. In the above scheme, the target autonomous vehicle can accurately recognize voice interaction commands from the driver via digital walkie-talkie, such as requests for yielding, overtaking, and route conflict negotiation, automatically execute the corresponding actions, and respond to the driver via walkie-talkie through TTS (Text-to-Speech) voice synthesis. Simultaneously, if an emergency command is recognized, a safe stopping operation will be immediately performed, and an alarm message will be pushed to the upstream dispatch center platform.

[0040] Optionally, a V2X (Vehicle to Everything) vehicle-to-vehicle communication module can be deployed on each vehicle in the target driving scenario. This module enables point-to-point information interaction between various vehicle terminals within the target driving scenario, serving as a redundant communication path under extreme and harsh conditions.

[0041] In an optional embodiment of the present invention, the target unmanned vehicle communicates with a dispatch center, which can be used to: acquire the operating status information of the target unmanned vehicle in real time; dispatch the target unmanned vehicle according to a preset emergency response procedure when it is determined that the operating status of the target unmanned vehicle is abnormal; and determine a third vehicle control command according to vehicle traffic constraints and send the third vehicle control command to the target unmanned vehicle when it is determined that the interaction between the target unmanned vehicle and the human-driven vehicle in the target driving scenario fails.

[0042] Among them, the preset emergency response procedures can be pre-configured response procedures corresponding to various emergency scenarios. Vehicle traffic constraints can be pre-set restrictive rules that vehicles must follow to pass. The third vehicle control command can be a control command based on the vehicle operating status determined by the vehicle traffic constraints.

[0043] In this embodiment of the invention, the dispatch center, acting as the control center for each vehicle in the target driving scenario, is capable of establishing communication connections with each vehicle. The dispatch center adopts a microservice architecture and is configured with a dual-machine hot standby redundancy mechanism to ensure long-term stable system operation. The overall architecture can be divided into a data layer, a service layer, and an application layer, enabling unified access to real-time operational data from all unmanned and manually driven vehicles, and seamless integration with the existing dispatching, billing, and maintenance systems of the target operation scenario. During operation, the dispatch center utilizes a high-precision map to visualize the location, operating status, tasks, and alarm information of all vehicles, supports high-frequency data refresh, and features functions such as trajectory playback, real-time video retrieval, and electronic fence control.

[0044] Specifically, the dispatch center can obtain real-time operational status information of the target autonomous vehicle. When the operational status information indicates that an abnormal situation has occurred in the target autonomous vehicle, the dispatch center can dispatch the target autonomous vehicle according to the preset emergency response procedure, automatically completing alarm push, vehicle braking, and issuing rescue instructions. When the operational status information indicates that the interaction between the target autonomous vehicle and the human-driven vehicle in the target driving scenario has failed, the dispatch center can allocate passage permissions to both vehicles according to vehicle passage constraints, generate a third vehicle control command for the target autonomous vehicle based on the allocated passage permissions, and issue the third vehicle control command to the target autonomous vehicle. This serves as a fallback to resolve passage conflicts and prevent operational congestion and passage deadlock problems.

[0045] Optionally, the dispatch center can also monitor the operational compliance of all vehicles in real time and synchronize violation data to the billing system. Simultaneously, it automatically compiles comprehensive operational data within the target driving scenario, generating benchmarking analysis reports on operational efficiency, operating costs, and safety indicators, providing data support for iterative optimization of control algorithms and on-site operational management.

[0046] Optionally, vehicle traffic constraints can be formulated in conjunction with the safety procedures and existing operating habits of the target driving scenario, adopting a unified traffic constraint standard for both driverless and manually driven vehicles. In a specific example, vehicle traffic constraints may include, but are not limited to, rules such as yielding from branch lines to main lines, empty vehicles yielding to loaded vehicles, and downhill vehicles yielding to uphill vehicles; prohibitions against overtaking on curves, overtaking on slopes, changing lanes over solid lines, and cutting in line; and priority rules such as emergency vehicles having the highest priority and normal operating vehicles having a first-come, first-served basis. This embodiment of the invention does not limit the specific content included in the vehicle traffic constraints.

[0047] Optionally, a strong binding mechanism can be established between rule enforcement and driver trip billing. All violations are recorded in real time, and violations are not included in performance evaluation. Multiple violations trigger penalties, while compliant operations are rewarded, thus solving the problem of existing rules lacking binding force.

[0048] In an optional embodiment of the present invention, the unmanned vehicle scheduling method may further include: obtaining a queue number and a target access berth issued by a scheduling center; entering the work area of ​​the target driving scenario to perform a work task based on the queue number, the target access berth, the spatial map of the target driving scenario, and vehicle traffic constraints; wherein the scheduling center is used to determine the transportation task and transportation route of each driving vehicle in the target driving scenario based on the production condition information of the target driving scenario; when it is determined that the target unmanned vehicle enters within a first preset distance range of the work area of ​​the target driving scenario, determining the queue number of the target unmanned vehicle based on the transportation task and transportation route of each driving vehicle in the target driving scenario; determining the target access berth of the target unmanned vehicle based on the queue number of the target unmanned vehicle; and issuing the queue number and the target access berth to the target unmanned vehicle.

[0049] The queue number can be assigned by the dispatch center to each vehicle in the target driving scenario to identify the order in which vehicles enter the work area to perform tasks. The target parking space can be a queuing lane and waiting area determined based on the queue number of the target autonomous vehicle. Production status information can be information reflecting the production operation status in the target driving scenario. The first preset distance range can be a pre-set queuing distance.

[0050] In this embodiment of the invention, a closed-loop queuing management mechanism can be established for the three high-conflict areas in the target driving scenario: loading area, unloading area, and refueling area. This mechanism involves "unified allocation of serial numbers by the platform, strict execution by individual vehicles, and real-time monitoring of violations." The core operating principle of queuing according to arrival order and prohibiting queue jumping is strictly enforced, thus standardizing vehicle operation order from the source of the process.

[0051] Specifically, the dispatch center can match transportation tasks to each working vehicle in the field based on the real-time production conditions of the target driving scenario, and dynamically plan the optimal transportation route to achieve a balanced distribution of the overall workload. When a target unmanned vehicle enters the first preset distance range of the work area of ​​the target driving scenario, the dispatch center can assign a queue number to the target unmanned vehicle according to the transportation tasks and routes of each driving vehicle in the target driving scenario, and determine the target parking space for the target unmanned vehicle based on the queue number. Furthermore, the dispatch center can issue the queue number and target parking space to the target unmanned vehicle. On this basis, after receiving the dispatch instruction, the target unmanned vehicle can drive in an orderly manner according to regulations, based on its own queue number, target parking space, scene space map, and vehicle traffic constraints, and is prohibited from changing lanes arbitrarily throughout the process. It must strictly follow the queue order to enter the work area to complete loading, unloading, recharging, and other operations. If an abnormal situation occurs, such as a vehicle malfunctioning and stopping, or voluntarily abandoning the work task, the dispatch center will automatically fill the queue number to ensure the continuous and smooth operation process. Understandably, this solution also applies to manually driven vehicles within the target driving scenario. The dispatch center can simultaneously assign queue numbers and parking spaces to manually driven vehicles, ensuring that they enter the site and operate in accordance with regulations.

[0052] To strengthen order management across all scenarios, this solution establishes differentiated regulatory and constraint mechanisms for different operating entities: Unmanned vehicles strictly adhere to platform dispatch instructions throughout the entire process, and cutting in line, changing lanes, or overtaking without authorization is strictly prohibited; manually driven vehicles receive queuing information in real time via onboard terminals and simultaneously receive voice order reminders. Vehicle violations are synchronized to the billing system in real time, enabling traceability of violations and enforcement of rights; the cabs of excavators on-site display the overall queuing order, requiring operators to strictly follow the order, with violations directly linked to individual performance evaluations. This solution, through comprehensive order management of both unmanned and manually driven vehicles, completely resolves the problems of vehicles cutting in line and chaotic mixed operations in traditional operating models, effectively avoiding pain points such as reduced operational efficiency and decreased earnings for human drivers, and significantly improving the standardization and efficiency of the overall operating scenario.

[0053] This invention acquires multi-dimensional operational perception data of a target autonomous vehicle in a target driving scenario in real time, and then uses a vehicle control model trained with sample data of target human driving behavior to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data. Furthermore, the vehicle motion control commands are used to schedule the target autonomous vehicle in the target driving scenario. This solution addresses the shortcomings of existing autonomous vehicle scheduling methods, such as poor scenario adaptability and insufficient dynamic control performance. It enables human-like alignment of autonomous vehicle driving behavior, thereby improving the stability of autonomous vehicle driving control in complex dynamic scenarios.

[0054] Example 2 Figure 2 This is a flowchart of a training method for an autonomous vehicle scheduling model provided in Embodiment 2 of the present invention. This embodiment is applicable to situations where a vehicle control model is trained using high-confidence human driving behavior sample data. The method can be executed by an autonomous vehicle scheduling model training device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. This electronic device can be a terminal device or a server device, as long as it can execute the autonomous vehicle scheduling model training method. The embodiments of the present invention do not limit the specific type of electronic device. Accordingly, as... Figure 1 As shown, the method includes the following operations: S210, Obtain full-condition manual driving behavior data.

[0055] Among them, full-condition manual driving behavior data can be a collection of driving behavior data of multiple drivers in different driving scenarios.

[0056] In this embodiment of the invention, when training the vehicle control model, full-condition manual driving behavior data can first be acquired. In a specific example, the data collection target can be an S-level excellent driver with more than 3 years of driving experience, ranking in the top 20% of single-trip transportation efficiency, zero safety accidents, and fuel consumption and vehicle wear below average. The collection scope can cover all open-pit mine operation scenarios, including straight driving on flat roads, uphill and downhill driving, curves, following other vehicles, loading / unloading alignment, queuing start and stop, meeting other vehicles, and obstacle avoidance. The collection period can cover more than 30 operating shifts and adapt to all operating conditions, such as empty / full load, sunny / rainy / dusty conditions, and day and night. Exemplarily, the full-condition manual driving behavior data can include vehicle operating status data, scene environment data, and driver operation timing data, etc. This embodiment of the invention does not limit the specific content included in the full-condition manual driving behavior data. Among them, vehicle operation status data may include, but is not limited to, location, speed, acceleration and deceleration, throttle opening, braking pressure, gear, load, etc.; scene environment data may include, but is not limited to, road slope curvature, following distance, work area location, excavator status, etc.; driver operation timing data may include, but is not limited to, throttle, braking and gear shifting operation timing, predictive operation lead time and scenario-based operation strategies, etc.

[0057] S220. The full-condition manual driving behavior data is scored according to the preset test index system to obtain the driving ability level of each manual driver.

[0058] The preset test index system can be a pre-established index system for quantitatively evaluating human driving behavior data under all working conditions. The driving ability level can be a comprehensive driving level classification obtained by quantitatively evaluating the driver's human driving behavior data based on the preset test index system.

[0059] Accordingly, after acquiring full-condition human driving behavior data, a pre-established four-dimensional quantitative evaluation system encompassing efficiency, economy, safety, and ride comfort can be used to score the data, thereby determining the driving ability level of each human driver. Specifically, the efficiency dimension assesses indicators such as single-trip time, average speed over a section, and alignment time; the economy dimension assesses indicators such as fuel consumption, braking frequency, and the number of rapid accelerations and decelerations related to tire wear; the safety dimension assesses indicators such as following distance maintenance rate and the number of risk triggers; and the ride comfort dimension assesses indicators such as acceleration / deceleration fluctuation rate and braking pressure change rate.

[0060] In a specific example, if the human driver's driving ability score is greater than or equal to 90 points, the human driver's driving ability level can be S level; if the human driver's driving ability score is in the range of 80-89 points, the human driver's driving ability level can be A level; if the human driver's driving ability score is less than 80 points, the human driver's driving ability level can be B level.

[0061] S230. When the driving ability level of the target human driver is determined to be the first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as the target human driving behavior sample data.

[0062] The first preset level can be a pre-set threshold for driving ability level.

[0063] Accordingly, after determining the driving ability level of each human driver, the full-condition human driving behavior data corresponding to drivers with driving ability levels of S and A can be used as the target human driving behavior sample data. The above scheme only retains the optimal driving behavior data that is suitable for the target driving scenario, which can effectively solve the problems of lack of filtering and scenario mismatch in existing anthropomorphic driving data.

[0064] Optionally, after determining the target human driving behavior sample data, data cleaning and scene labeling processing can be performed on the target human driving behavior sample data, and data augmentation operations can be carried out for different load states, road slopes, weather conditions, and lighting environments, so as to improve the model's scene generalization ability and reduce the risk of model overfitting.

[0065] S240. Train the vehicle control model based on the target human driving behavior sample data.

[0066] Correspondingly, after obtaining the target human driving behavior sample data, the target human driving behavior sample data can be input into the vehicle control model to train the vehicle control model.

[0067] S250. The trained vehicle control model is then distributed to each unmanned vehicle in the target driving scenario.

[0068] In this embodiment of the invention, the vehicle control model can be trained in the cloud, and then the trained vehicle control model can be distributed to each autonomous vehicle in the target driving scenario. This allows each autonomous vehicle in the target driving scenario to execute the autonomous vehicle scheduling method described in any embodiment of the invention through the vehicle control model. The above solution abandons the traditional, conservative model control logic and completely replicates the driving behavior of a dedicated, highly skilled driver in the target driving scenario. This aligns the behavioral rhythm of autonomous vehicles with that of human drivers, addressing efficiency gaps and allowing the driving behavior of autonomous vehicles to be predicted by human drivers, reducing the risk of mixed-route conflicts.

[0069] This invention acquires full-condition human driving behavior data, scores the data according to a preset test index system, and obtains the driving ability level of each human driver. When the driving ability level of a target human driver is determined to be a first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as target human driving behavior sample data. Furthermore, a vehicle control model is trained based on the target human driving behavior sample data, and the trained vehicle control model is distributed to each autonomous vehicle in the target driving scenario. This solution addresses the shortcomings of existing autonomous vehicle scheduling methods, such as poor scenario adaptability and insufficient dynamic control performance. It enables human-like alignment of autonomous vehicle driving behavior, thereby improving the driving control stability of autonomous vehicles in complex dynamic scenarios.

[0070] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information (such as gesture information, voice information, etc.) involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0071] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant regions.

[0072] It should be noted that any arrangement or combination of the technical features in the above embodiments also falls within the protection scope of this invention.

[0073] Example 3 Figure 3 This is a schematic diagram of an unmanned vehicle dispatching device provided in Embodiment 3 of the present invention, as shown below. Figure 3 As shown, the device is configured in an autonomous vehicle and includes: a multi-dimensional operational perception data acquisition module 310, a vehicle motion control command generation module 320, and a target autonomous vehicle scheduling module 330, wherein: The multi-dimensional operational perception data acquisition module 310 is used to acquire multi-dimensional operational perception data of the target unmanned vehicle in the target driving scenario in real time.

[0074] The vehicle motion control command generation module 320 is used to call the vehicle control model to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein, the vehicle control model is trained using target human driving behavior sample data.

[0075] The target unmanned vehicle scheduling module 330 is used to schedule the target unmanned vehicles in the target driving scenario through the vehicle motion control commands.

[0076] This invention acquires multi-dimensional operational perception data of a target autonomous vehicle in a target driving scenario in real time, and then uses a vehicle control model trained with sample data of target human driving behavior to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data. Furthermore, the vehicle motion control commands are used to schedule the target autonomous vehicle in the target driving scenario. This solution addresses the shortcomings of existing autonomous vehicle scheduling methods, such as poor scenario adaptability and insufficient dynamic control performance. It enables human-like alignment of autonomous vehicle driving behavior, thereby improving the stability of autonomous vehicle driving control in complex dynamic scenarios.

[0077] Optionally, the vehicle control model includes a driving decision model and a trajectory optimization control model; the vehicle motion control command generation module 320 is specifically used to: input the multi-dimensional operation perception data into the driving decision model to obtain the time-series control parameter sequence of the target unmanned vehicle; wherein, the time-series control parameter sequence includes drive control parameters, braking control parameters, and gear shifting control parameters; and call the trajectory optimization control model to perform rolling time-domain optimization solution based on the multi-dimensional boundary constraints of vehicle operation and the time-series control parameter sequence to determine the vehicle motion control command of the target unmanned vehicle.

[0078] Optionally, the above-mentioned device may further include a gesture interaction module, used to acquire environmental visual perception data collected by the on-board camera of the target autonomous vehicle; call a hand target localization and detection model to perform hand detection on the environmental visual perception data; if it is determined that the environmental visual perception data includes hand movements, call a target feature extraction model to extract hand movement feature information based on the environmental visual perception data; call a target classification model to classify the hand movements based on the hand movement feature information to obtain a gesture classification result; determine a first vehicle control command based on the gesture classification result, and adjust the vehicle's driving motion state according to the first vehicle control command.

[0079] Optionally, the above-mentioned device may further include a voice interaction module for acquiring environmental acoustic perception data collected by the vehicle-mounted digital intercom of the target unmanned vehicle; wherein the environmental acoustic perception data includes voice interaction commands output by each manually driven vehicle in the target driving scenario; if it is determined that the environmental acoustic perception data includes the vehicle number of the target unmanned vehicle, a second vehicle control command and interactive response information of the target unmanned vehicle are determined according to the voice interaction commands included in the environmental acoustic perception data; the vehicle driving motion state is adjusted according to the second vehicle control command, and the interactive response information is broadcast through the vehicle-mounted digital intercom of the target unmanned vehicle.

[0080] Optionally, the target autonomous vehicle communicates with the dispatch center, which is used to: acquire the operating status information of the target autonomous vehicle in real time; dispatch the target autonomous vehicle according to a preset emergency response procedure when it is determined that the operating status of the target autonomous vehicle is abnormal; and determine a third vehicle control command according to vehicle traffic constraints and send the third vehicle control command to the target autonomous vehicle when it is determined that the interaction between the target autonomous vehicle and the human-driven vehicle in the target driving scenario fails.

[0081] Optionally, the above-mentioned device may further include a queue management module, used to obtain a queue number and target access berth issued by the dispatch center; enter the work area of ​​the target driving scenario to perform work tasks according to the queue number, the target access berth, the spatial map of the target driving scenario, and vehicle traffic constraints; wherein, the dispatch center is used to determine the transportation tasks and transportation routes of each driving vehicle in the target driving scenario according to the production condition information of the target driving scenario; when it is determined that the target unmanned vehicle enters within a first preset distance range of the work area of ​​the target driving scenario, determine the queue number of the target unmanned vehicle according to the transportation tasks and transportation routes of each driving vehicle in the target driving scenario; determine the target access berth of the target unmanned vehicle according to the queue number of the target unmanned vehicle; and issue the queue number and the target access berth to the target unmanned vehicle.

[0082] The aforementioned unmanned vehicle scheduling device can execute the unmanned vehicle scheduling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the unmanned vehicle scheduling method provided in any embodiment of the present invention.

[0083] Since the unmanned vehicle scheduling device described above is capable of executing the unmanned vehicle scheduling method in the embodiments of the present invention, those skilled in the art can understand the specific implementation methods and various variations of the unmanned vehicle scheduling device in this embodiment based on the unmanned vehicle scheduling method described in the embodiments of the present invention. Therefore, how the unmanned vehicle scheduling device implements the unmanned vehicle scheduling method in the embodiments of the present invention will not be described in detail here. Any device used by those skilled in the art to implement the unmanned vehicle scheduling method in the embodiments of the present invention falls within the scope of protection of this application.

[0084] Example 4 Figure 4 This is a schematic diagram of an unmanned vehicle scheduling model training device provided in Embodiment 4 of the present invention, as shown below. Figure 4 As shown, the device includes: a full-condition manual driving behavior data acquisition module 410, a driving ability level determination module 420, a target manual driving behavior sample data filtering module 430, a vehicle control model training module 440, and a vehicle control model distribution module 450, wherein: The full-condition manual driving behavior data acquisition module 410 is used to acquire full-condition manual driving behavior data.

[0085] The driving ability level determination module 420 is used to score the full-condition manual driving behavior data according to a preset test index system to obtain the driving ability level of each manual driver.

[0086] The target manual driving behavior sample data filtering module 430 is used to select the full-condition manual driving behavior data corresponding to the target manual driver as the target manual driving behavior sample data when the driving ability level of the target manual driver is determined to be a first preset level.

[0087] The vehicle control model training module 440 is used to train the vehicle control model based on the target human driving behavior sample data.

[0088] The vehicle control model distribution module 450 is used to distribute the trained vehicle control model to each unmanned vehicle in the target driving scenario, so that the unmanned vehicles can execute the unmanned vehicle scheduling method described in any embodiment of the present invention through the vehicle control model.

[0089] This invention acquires full-condition human driving behavior data, scores the data according to a preset test index system, and obtains the driving ability level of each human driver. When the driving ability level of a target human driver is determined to be a first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as target human driving behavior sample data. Furthermore, a vehicle control model is trained based on the target human driving behavior sample data, and the trained vehicle control model is distributed to each autonomous vehicle in the target driving scenario. This solution addresses the shortcomings of existing autonomous vehicle scheduling methods, such as poor scenario adaptability and insufficient dynamic control performance. It enables human-like alignment of autonomous vehicle driving behavior, thereby improving the driving control stability of autonomous vehicles in complex dynamic scenarios.

[0090] The aforementioned autonomous vehicle scheduling model training device can execute the autonomous vehicle scheduling model training method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment can be found in the autonomous vehicle scheduling model training method provided in any embodiment of the present invention.

[0091] Since the autonomous vehicle scheduling model training device described above is capable of executing the autonomous vehicle scheduling model training method in the embodiments of the present invention, those skilled in the art can understand the specific implementation and various variations of the autonomous vehicle scheduling model training device in this embodiment based on the autonomous vehicle scheduling model training method described in the embodiments of the present invention. Therefore, how the autonomous vehicle scheduling model training device implements the autonomous vehicle scheduling model training method in the embodiments of the present invention will not be described in detail here. Any device used by those skilled in the art to implement the autonomous vehicle scheduling model training method in the embodiments of the present invention falls within the scope of protection of this application.

[0092] Example 5 Figure 5 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0093] like Figure 5 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0094] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0095] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as autonomous vehicle scheduling methods or autonomous vehicle scheduling model training methods.

[0096] In some embodiments, the autonomous vehicle scheduling method or autonomous vehicle scheduling model training method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the autonomous vehicle scheduling method or autonomous vehicle scheduling model training method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the autonomous vehicle scheduling method or autonomous vehicle scheduling model training method by any other suitable means (e.g., by means of firmware).

[0097] Optionally, the autonomous vehicle scheduling method, applied to autonomous vehicles, may include: acquiring multi-dimensional operational perception data of a target autonomous vehicle in a target driving scenario in real time; invoking a vehicle control model to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data; wherein the vehicle control model is trained using target human driving behavior sample data; and scheduling the target autonomous vehicle in the target driving scenario using the vehicle motion control commands.

[0098] Optionally, the autonomous vehicle scheduling model training method may include: acquiring full-condition human driving behavior data; scoring the full-condition human driving behavior data according to a preset test index system to obtain the driving ability level of each human driver; when the driving ability level of the target human driver is determined to be a first preset level, using the full-condition human driving behavior data corresponding to the target human driver as target human driving behavior sample data; training the vehicle control model based on the target human driving behavior sample data; and distributing the trained vehicle control model to each autonomous vehicle in the target driving scenario, so that the autonomous vehicles can execute the autonomous vehicle scheduling method described in any embodiment of the present invention through the vehicle control model.

[0099] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0100] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0101] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0102] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0103] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0104] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0105] This application also discloses a computer program product, which includes a computer program that, when executed by a processor, implements the autonomous vehicle scheduling method or autonomous vehicle scheduling model training method provided in any embodiment of this application. This program product belongs to the same inventive concept as the autonomous vehicle scheduling method or autonomous vehicle scheduling model training method disclosed in the embodiments of this application, and therefore will not be described in detail here.

[0106] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0107] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for scheduling unmanned vehicles, characterized in that, Applications in autonomous vehicles include: Real-time acquisition of multi-dimensional operational perception data of the target autonomous vehicle in the target driving scenario; The vehicle control model is invoked to generate vehicle motion control commands for the target unmanned vehicle based on the multi-dimensional operational perception data; wherein, the vehicle control model is trained using sample data of the target human driving behavior. The target unmanned vehicle in the target driving scenario is scheduled using the vehicle motion control commands.

2. The method according to claim 1, characterized in that, The vehicle control model includes a driving decision model and a trajectory optimization control model; the step of invoking the vehicle control model to generate vehicle motion control commands for the target autonomous vehicle based on the multi-dimensional operational perception data includes: The multi-dimensional operational perception data is input into the driving decision model to obtain the time-series control parameter sequence of the target unmanned vehicle; wherein, the time-series control parameter sequence includes drive control parameters, braking control parameters and gear shifting control parameters; The trajectory optimization control model is invoked to perform rolling time-domain optimization based on the multi-dimensional boundary constraints of vehicle operation and the sequence of timing control parameters to determine the vehicle motion control command for the target unmanned vehicle.

3. The method according to claim 1, characterized in that, The method further includes: Acquire environmental visual perception data collected by the onboard camera of the target unmanned vehicle; The hand target localization and detection model is invoked to detect hands in the environmental visual perception data; If it is determined that the environmental visual perception data includes hand movements, the target feature extraction model is invoked to extract hand movement feature information based on the environmental visual perception data; The target classification model is invoked to classify the hand movements based on the hand movement feature information, thereby obtaining the gesture classification result; The first vehicle control command is determined based on the gesture classification result, and the vehicle's driving motion state is adjusted according to the first vehicle control command.

4. The method according to claim 1, characterized in that, The method further includes: The environmental acoustic perception data collected by the vehicle-mounted digital intercom of the target unmanned vehicle is acquired; wherein, the environmental acoustic perception data includes voice interaction commands output by each manually driven vehicle in the target driving scenario; If the environmental acoustic perception data includes the vehicle number of the target unmanned vehicle, the second vehicle control command and interactive response information of the target unmanned vehicle are determined based on the voice interaction commands included in the environmental acoustic perception data. Adjust the vehicle's driving motion state according to the second vehicle control command, and broadcast the interactive response information through the onboard digital intercom of the target unmanned vehicle.

5. The method according to any one of claims 1, 3, or 4, characterized in that, The target unmanned vehicle communicates with the dispatch center, which is used for: Real-time acquisition of the operating status information of the target unmanned vehicle; If it is determined that the target unmanned vehicle is in an abnormal operating state, the target unmanned vehicle shall be dispatched according to the preset emergency response procedure. If it is determined that the interaction between the target autonomous vehicle and the human-driven vehicle in the target driving scenario fails, a third vehicle control command is determined based on the vehicle traffic constraints, and the third vehicle control command is sent to the target autonomous vehicle.

6. The method according to claim 1, characterized in that, The method further includes: Obtain the queue number and target berth issued by the dispatch center; Enter the work area of ​​the target driving scenario to perform the work task according to the queue number, the target berth, the spatial map of the target driving scenario, and the vehicle traffic constraints; The dispatch center is used to determine the transportation tasks and routes of each driving vehicle in the target driving scenario based on the production condition information of the target driving scenario; when it is determined that the target unmanned vehicle enters the work area of ​​the target driving scenario within a first preset distance range, it determines the queue number of the target unmanned vehicle based on the transportation tasks and routes of each driving vehicle in the target driving scenario; it determines the target access parking space of the target unmanned vehicle based on the queue number of the target unmanned vehicle; and it sends the queue number and the target access parking space to the target unmanned vehicle.

7. A method for training an autonomous vehicle scheduling model, characterized in that, include: Acquire full-condition human driving behavior data; The driving ability level of each human driver is obtained by scoring the full-condition human driving behavior data according to the preset test index system. When the driving ability level of the target human driver is determined to be the first preset level, the full-condition human driving behavior data corresponding to the target human driver is used as the target human driving behavior sample data. The vehicle control model is trained based on the target human driving behavior sample data; The trained vehicle control model is distributed to each autonomous vehicle in the target driving scenario, so that the autonomous vehicles can execute the autonomous vehicle scheduling method according to any one of claims 1-6 through the vehicle control model.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the autonomous vehicle scheduling method according to any one of claims 1-6, or to execute the autonomous vehicle scheduling model training method according to claim 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the autonomous vehicle scheduling method according to any one of claims 1-6, or to execute the autonomous vehicle scheduling model training method according to claim 7.

10. A computer program product, characterized in that, It includes a computer program / instruction, wherein when the computer program / instruction is executed by a processor, it implements the autonomous vehicle scheduling method according to any one of claims 1-6, or performs the autonomous vehicle scheduling model training method according to claim 7.