A vehicle automatic test method, device and medium based on an embodied robot
By adaptively adjusting the embodied robot and the vehicle testing process, the problems of dynamic adjustment of test paths and information convergence in existing technologies are solved, achieving high efficiency and reliability in automated vehicle testing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Filing Date
- 2026-03-13
- Publication Date
- 2026-07-10
AI Technical Summary
Existing vehicle testing technologies based on embodied robots and industrial control systems struggle to dynamically adjust test paths according to real-time changes in vehicle control response. They also lack a unified mechanism for adaptive screening and determining whether the test has reached information convergence, resulting in insufficient testing efficiency and instability of conclusions.
By obtaining the test task description, the robot pose calibration and vehicle operable parts identification are performed, a test initialization configuration set is established, an adaptive test candidate sequence is generated, the robot is driven to perform actions and vehicle control response data is collected to form a dynamic confidence state, the test candidate sequence is updated and a vehicle automation test report is generated.
It enables quantifiable state determination and dynamic management of vehicle control response, enhances the closed-loop consistency of the testing process and the reliability of test conclusions, and provides continuous, stable and traceable data support.
Smart Images
Figure CN122362784A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial control technology, and in particular to an automated vehicle testing method, equipment, and medium based on an embodied robot. Background Technology
[0002] As vehicle electronic and electrical architectures become increasingly complex and automated, vehicle testing is gradually evolving from manual operation to automation, and the application of embodied robots in vehicle testing is gaining attention. Through physical interaction between embodied robots and the vehicle's operable components, vehicle control behaviors can be triggered in real-world operating environments, and the vehicle's operating status can be observed and recorded by combining this with the acquisition of onboard signals. Simultaneously, the application of industrial control systems in vehicle testing is deepening, and the testing process is gradually showing a technological trend of parallel development of data-driven, closed-loop control, and state awareness.
[0003] In existing vehicle testing technologies based on embodied robots and industrial control systems, testing processes are mostly based on preset action sequences and fixed test scripts. This makes it difficult to dynamically adjust the test path according to real-time changes in vehicle control response, resulting in insufficient utilization of the test process's ability to distinguish control states. Especially when multiple control behavior states coexist, existing technologies lack a unified mechanism that can combine action execution results with onboard signal responses to adaptively filter test actions and determine whether the test has reached an information convergence state, thus limiting testing efficiency and the stability of test conclusions. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, this invention provides a vehicle automation testing method based on an embodied robot to solve the problem that test actions cannot be effectively distinguished and adaptively converged according to the control response during vehicle automation testing.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a vehicle automation testing method based on an embodied robot, comprising: acquiring a test task description of the target vehicle, and performing embodied robot pose calibration, vehicle operable component identification, and vehicle signal acquisition channel association to form a test initialization configuration set; evaluating the control state response differences corresponding to the test initialization configuration set under the current vehicle state to generate an adaptive test candidate sequence; driving the embodied robot to execute target action semantics based on the adaptive test candidate sequence and synchronously collecting the action execution process and vehicle control response to form an action response dataset; determining the consistency state of the vehicle control response under different action semantics based on the action response dataset, and performing confidence updates on predetermined control behavior states to form a dynamic confidence state; evaluating the discrimination ability of action semantics that have not yet been executed in the adaptive test candidate sequence based on the dynamic confidence state to form discrimination ability evaluation information, updating the adaptive test candidate sequence, generating a confidence convergence determination flag and an updated adaptive test candidate sequence; and performing convergence control on the execution process of the updated adaptive test candidate sequence based on the confidence convergence determination flag to generate a vehicle automation test report.
[0007] In a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps for forming the test initialization configuration set are as follows: Obtain the test task description of the target vehicle, and determine the vehicle test scope and action semantic type based on the test task description; Based on action semantic types, the embodied robot is driven to perform pose sampling and calibration within the vehicle test range to obtain a unified coordinate mapping relationship; Based on a unified coordinate mapping relationship and action semantic type, reachability determination is performed on candidate interactive objects within the vehicle test range to filter and obtain a set of operable parts; The robot is driven to perform controlled operations on the set of operable parts and collect corresponding changes in vehicle-mounted signals, thus establishing a correspondence between operable parts and vehicle-mounted signal acquisition channels. The test initialization configuration set is generated by summarizing the vehicle test range, action semantic type, unified coordinate mapping relationship, set of operable parts and operable behavior and their correspondence with the vehicle signal acquisition channel.
[0008] As a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps for generating the adaptive test candidate sequence are as follows: A joint traversal is performed on the action semantic types, operable component sets, and the correspondence between operable components and vehicle signal acquisition channels in the test initialization configuration set to construct the action semantic mapping relationship; The status of the vehicle signal acquisition channel corresponding to the action semantic mapping relationship is read to form a status snapshot; The control state response changes corresponding to the state snapshot under different action semantics are compared and calculated to obtain the distinguishability assessment identifier. Based on the distinguishability assessment identifier, each action semantic is sorted to generate an adaptive test candidate sequence.
[0009] As a preferred embodiment of the vehicle automation testing method based on embodied robots described in this invention, the steps for forming the motion response dataset are as follows: The action semantics in the adaptive test candidate sequence are scheduled and orchestrated to generate an action semantics execution scheduling sequence; Based on the action semantics execution scheduling sequence, the embodied robot is driven to perform target action semantics on the corresponding operable parts, and the start and end times of the action are recorded to form an action event trajectory; Under the constraints of the start and end time of the action event trajectory, vehicle control response data in the vehicle signal acquisition channel corresponding to the target action semantics are collected synchronously, and the vehicle control response data is associated and encapsulated with the action event trajectory to form an action response dataset.
[0010] In a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps for forming a dynamic confidence state are as follows: The vehicle control response data in the action response dataset is mapped to the predetermined control behavior state to form an action semantic response sequence. Consistency comparison is performed on the changes in vehicle control response data under different action semantics for the same control behavior state in the action semantic response sequence to obtain consistency judgment information; Based on the consistency determination information, the confidence level of the predetermined control behavior state is updated to generate a dynamic confidence state.
[0011] As a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps for forming the discrimination ability assessment information are as follows: Read the confidence distribution of each control behavior state in the dynamic confidence state, and filter the action semantics that have not yet been executed from the adaptive test candidate sequence to form a set of action semantics to be evaluated; By combining dynamic confidence states, the differences in vehicle control response data under different control behavior states are analyzed by state response trajectory comparison and deduction, and the discrimination ability metric value corresponding to each action semantic is calculated. The discrimination ability metrics corresponding to each action semantic are summarized and organized to form discrimination ability assessment information.
[0012] As a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps for generating the confidence convergence determination identifier and the updated adaptive test candidate sequence are as follows: The discrimination ability metric value from the discrimination ability assessment information is written into the action semantics of the adaptive test candidate sequence that have not yet been executed, forming the sorting basis to be updated; Based on the sorting criteria to be updated, the semantics of actions that have not yet been executed are reconstructed in order to generate an updated adaptive test candidate sequence. Based on the distribution of the discriminative power metric values corresponding to each action semantic in the updated adaptive test candidate sequence, it is determined whether the test process meets the preset confidence convergence condition, and a confidence convergence determination identifier is generated.
[0013] As a preferred embodiment of the vehicle automated testing method based on embodied robots described in this invention, the steps of performing convergence control on the execution process of the updated adaptive test candidate sequence according to the confidence convergence determination flag and generating a vehicle automated test report are as follows. Based on the confidence convergence determination flag, the execution direction of the updated adaptive test candidate sequence is determined, and an execution convergence control instruction is generated. According to the convergence control instruction, the execution process of the updated adaptive test candidate sequence is terminated to obtain the convergence execution trajectory; The execution order of action semantics, action response dataset, and dynamic confidence state corresponding to the converged execution trajectory are summarized and organized to generate a vehicle automation test report.
[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the vehicle automated testing method based on an embodied robot as described in the first aspect of the present invention.
[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the vehicle automated testing method based on an embodied robot as described in the first aspect of the present invention.
[0016] The beneficial effects of this invention are as follows: By associating the execution process of the embodied robot's actions with the corresponding vehicle-mounted signal acquisition data under unified action semantics and time constraints, the vehicle control response forms a quantifiable and continuously updatable state determination basis under different action triggering conditions. Based on the state determination, a confidence mechanism that is gradually updated with the testing process is adopted to dynamically manage the control behavior state in the industrial control system. This allows the confidence level of the control state to be continuously adjusted as evidence accumulates, providing continuous, stable, and traceable data support for the selection of test actions, the convergence control of the test process, and the generation of vehicle automation test reports. This enhances the closed-loop consistency of the test process and the reliability of the test conclusions. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the 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.
[0018] Figure 1 This is a flowchart of a vehicle automation testing method based on embodied robots.
[0019] Figure 2 A flowchart for generating the test initialization configuration set.
[0020] Figure 3 A flowchart for creating an action response dataset.
[0021] Figure 4 A flowchart for generating confidence convergence criteria. Detailed Implementation
[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides an automated vehicle testing method based on an embodied robot, comprising the following steps: S1. Obtain the test task description of the target vehicle, and perform robot pose calibration, vehicle operable component identification, and vehicle signal acquisition channel association to form a test initialization configuration set.
[0026] S1.1: Obtain the test task description of the target vehicle, and determine the vehicle test range and action semantic type based on the test task description; Specifically, before the target vehicle enters the vehicle automated testing process, the testers, based on the target vehicle's current testing phase, model information, and configuration, clarify the test objectives, involved vehicle functions, and permitted operational behaviors required for the current testing phase, thus forming a test task description. The testers then organize the vehicle parts corresponding to the vehicle functions mentioned in the test task description and limit the vehicle testing scope based on the distribution of these parts. Finally, the testers categorize the operational behaviors in the test task description according to their control objectives and operating methods, grouping behaviors with the same control objectives and operating methods into the same category, thus forming action semantic types that match the vehicle testing scope.
[0027] S1.2: Based on action semantic type, drive the embodied robot to perform pose sampling and calibration within the vehicle test range to obtain a unified coordinate mapping relationship; Specifically, within the vehicle test range, the embodied robot is assigned an operational posture that corresponds one-to-one with the action semantic type. According to the operational behavior defined by the action semantic type, the embodied robot is sequentially controlled to perform the corresponding posture positioning actions within the vehicle test range. During each operation posture positioning process, the spatial correspondence between the current posture of the embodied robot and the vehicle parts within the vehicle test range is recorded. The recorded spatial correspondence is then uniformly organized to establish a stable correspondence between the posture of the embodied robot and the position of the vehicle parts within the vehicle test range, resulting in a unified coordinate mapping relationship.
[0028] S1.3: Based on the unified coordinate mapping relationship and action semantic type, the reachability of candidate interactive objects within the vehicle test range is determined, and the set of operable parts is obtained by filtering. Specifically, based on a unified coordinate mapping relationship and action semantic type, vehicle parts that can physically interact with the embodied robot are identified one by one within the vehicle test range as candidate interaction objects. According to the unified coordinate mapping relationship, the positional relationship between the current pose of the embodied robot and the vehicle parts where the candidate interaction objects are located is mapped. Combined with the operation behavior limited by the action semantic type, it is determined whether the embodied robot can reach and act on the corresponding candidate interaction object in the current pose within the vehicle test range. Candidate interaction objects that can reach the vehicle parts where the candidate interaction objects are located and match the action semantic type are retained as operable parts, while candidate interaction objects that cannot reach the vehicle parts where the candidate interaction objects are located and cannot match the action semantic type are excluded, resulting in a set of operable parts.
[0029] S1.4: Drive the embodied robot to perform controlled operations on the set of operable parts and collect the corresponding changes in vehicle-mounted signals, and establish the correspondence between operable parts and vehicle-mounted signal acquisition channels; Specifically, according to the action semantic type corresponding to each operable component in the operable component set, within the vehicle test range, the embodied robot is sequentially driven to perform controlled operations on each operable component that are consistent with the action semantic type. During each controlled operation, the changes in the vehicle's onboard signals before and after the operation are recorded synchronously. The operable components corresponding to each controlled operation are matched one-to-one with the synchronously recorded changes in the onboard signals, so that the onboard signal acquisition channels associated with the same operable component remain consistent during repeated controlled operations, thus establishing a correspondence between operable components and onboard signal acquisition channels.
[0030] S1.5: Summarize the vehicle test range, action semantic type, unified coordinate mapping relationship, set of operable parts and operable behavior and the correspondence with the vehicle signal acquisition channel to generate a test initialization configuration set.
[0031] Specifically, the vehicle test range, action semantic type, unified coordinate mapping relationship, set of operable parts, and correspondence between operable behavior and vehicle signal acquisition channel are centrally organized according to the same target vehicle. The vehicle test range and action semantic type are recorded as test constraints. The unified coordinate mapping relationship is recorded as the spatial correspondence basis when the embodied robot performs actions. The set of operable parts is associated with the corresponding operation behavior and the correspondence between operable behavior and vehicle signal acquisition channel are also included in the record to form a test initialization configuration set.
[0032] S2. Evaluate the differences in control state response of the test initialization configuration set under the current vehicle state, and generate an adaptive test candidate sequence.
[0033] S2.1: Perform a joint traversal of the action semantic types, operable component sets, and the correspondence between operable components and vehicle signal acquisition channels in the test initialization configuration set to construct the action semantic mapping relationship; Specifically, the action semantic types, operable component sets, and the correspondence between operable components and vehicle signal acquisition channels in the test initialization configuration set are uniformly expanded. Each action semantic type is sequentially associated with the operable components in the operable component set, and the correspondence between operable components and vehicle signal acquisition channels is referenced synchronously during the association process. This ensures that each action semantic type can clearly correspond to the executable operable component and the associated vehicle signal acquisition channel, forming an action semantic mapping relationship.
[0034] S2.2: Perform status reading on the vehicle signal acquisition channel corresponding to the action semantic mapping relationship to form a status snapshot; Specifically, according to the action semantic mapping relationship, the vehicle status information presented by each vehicle signal acquisition channel at the same time point is read, and the vehicle status information is synchronously recorded with the corresponding action semantic type and operable parts, so that the status of the vehicle signal acquisition channels associated with the same action semantic type is kept consistent at the same time, thereby collecting a status snapshot.
[0035] S2.3: Compare and calculate the changes in control state response corresponding to the state snapshot under different action semantics to obtain the discrimination ability assessment identifier, and sort each action semantic according to the discrimination ability assessment identifier to generate an adaptive test candidate sequence.
[0036] Specifically, the vehicle signal acquisition channel status corresponding to the state snapshot under each action semantic is organized according to the action semantic type. The state snapshots corresponding to different action semantic types are compared item by item to clarify the degree of difference presented by each action semantic type when triggering changes in vehicle control status. The degree of difference is summarized into a discrimination ability assessment indicator. Based on the distribution of the discrimination ability assessment indicator among the action semantic types, the discrimination ability assessment indicators corresponding to each action semantic are arranged in order of the degree of difference from high to low to obtain the adaptive test candidate sequence.
[0037] S3. Based on the adaptive test candidate sequence, drive the embodied robot to execute the target action semantics and simultaneously collect the action execution process and vehicle control response to form an action response dataset.
[0038] S3.1: Schedule and orchestrate the action semantics in the adaptive test candidate sequence to generate an action semantic execution scheduling sequence; Specifically, according to the order of action semantics in the adaptive test candidate sequence, the action semantics are sorted one by one, and the action semantics at the top are given priority in the execution schedule. The remaining action semantics are then sequentially connected after the previous action semantics, so that the action semantics form a continuous and non-overlapping execution relationship in time order, thus forming the action semantic execution order. The action semantic execution order is recorded uniformly to form the action semantic execution scheduling sequence.
[0039] S3.2: Based on the action semantics, execute the scheduling sequence to drive the embodied robot to perform the target action semantics on the corresponding operable parts, and record the start and end times of the action to form the action event trajectory; Specifically, according to the execution order of action semantics in the action semantic execution scheduling sequence, within the vehicle test range, the embodied robot is driven to execute the target action semantics on the operable parts corresponding to each action semantics in turn, and the corresponding action start time and action end time are recorded when each target action semantic begins and ends. By continuously recording the action start time and action end time during the action semantic execution process, a clear sequential relationship is formed between each target action semantic in the time dimension, and the action event trajectory is formed by the aggregation.
[0040] It should be noted that the target action semantics are action semantics selected from the action semantics execution scheduling sequence according to the action semantics execution order and corresponding to the operable components.
[0041] S3.3: Under the constraints of the start and end time of the action event trajectory, synchronously collect vehicle control response data from the vehicle signal acquisition channel corresponding to the target action semantics, and associate and encapsulate the vehicle control response data with the action event trajectory to form an action response dataset.
[0042] Specifically, an execution time interval is formed based on the start and end times of the actions recorded in the action event trajectory. Within the execution time interval, the vehicle signal acquisition channel corresponding to the target action semantics is read, and the vehicle control response data presented in the corresponding vehicle signal acquisition channel is collected. The vehicle control response data is then organized according to the execution order and execution time interval of the corresponding action semantics in the action event trajectory, so that each segment of vehicle control response data is consistently associated with the corresponding action event trajectory. The action event trajectory and vehicle control response data are then combined to form an action response dataset.
[0043] S4. Based on the action response dataset, determine the consistency of the vehicle control response under different action semantics, and perform confidence updates on the predetermined control behavior state to form a dynamic confidence state.
[0044] S4.1: Organize the vehicle control response data in the action response dataset to correspond with the predetermined control behavior states to form an action semantic response sequence; Specifically, the vehicle control response data in the action response dataset is organized segment by segment according to the execution order of action semantics; based on the action semantic mapping relationship, the set of vehicle signal acquisition channels corresponding to the vehicle control response data is locked, and the changes in vehicle signal presented by the vehicle control response data within the time interval defined by the action event trajectory are organized into the state performance of vehicle signal acquisition channels; the state performance of vehicle signal acquisition channels of vehicle control response data is compared item by item with the state performance of vehicle signal acquisition channels corresponding to a predetermined control behavior state. When the state performance of vehicle control response data in the vehicle signal acquisition channel is consistent with the state performance of vehicle signal acquisition channel corresponding to a predetermined control behavior state, the vehicle control response data is assigned to the corresponding predetermined control behavior state, so that each segment of vehicle control response data clearly corresponds to a predetermined control behavior state; the matched vehicle control response data is arranged continuously according to the execution order of action semantics to form an action semantic response sequence.
[0045] It should be noted that the control behavior state is a summary and organization of the vehicle functions that need to be verified in the current test phase of the target vehicle before the start of vehicle automation testing, combined with the control state forms that the vehicle can present during normal operation. By sorting out the test objectives involved in the test task description item by item, each test objective that needs to be verified is corresponding to a specific control behavior state.
[0046] S4.2: Perform consistency comparison on the changes in vehicle control response data under different action semantics for the same control behavior state in the action semantic response sequence to obtain consistency judgment information; Specifically, the action semantic response sequence is grouped and organized according to the predetermined control behavior state, so that vehicle control response data triggered by different action semantics under the same control behavior state are collected in a centralized manner; the collected vehicle control response data is compared item by item in terms of time sequence and change trend to check whether the vehicle control response data under different action semantic triggers show consistent change characteristics, and the check results are recorded as consistency judgment information.
[0047] S4.3: Based on the consistency determination information, perform confidence updates on the predetermined control behavior state to generate a dynamic confidence state.
[0048] Specifically, consistency determination information is aggregated according to predetermined control behavior states. Based on the check results in the consistency determination information, the confidence level corresponding to each control behavior state is updated. This increases the confidence level of control behavior states that maintain a consistent response under different action semantics, while decreasing the confidence level of control behavior states that do not maintain a consistent response under different action semantics. The changes in confidence level of the control behavior states during the update process are recorded. All control behavior states and their corresponding changes in confidence level are then summarized to form a dynamic confidence level.
[0049] S5. Based on the dynamic confidence state, evaluate the discriminative ability of the semantics of actions that have not yet been executed in the adaptive test candidate sequence, form discriminative ability evaluation information, update the adaptive test candidate sequence, and generate a confidence convergence judgment flag and the updated adaptive test candidate sequence.
[0050] S5.1: Read the confidence distribution of each control behavior state in the dynamic confidence state, and filter the action semantics that have not yet been executed from the adaptive test candidate sequence to form a set of action semantics to be evaluated; Specifically, the changes in confidence levels corresponding to each control behavior state in the dynamic confidence state are read and organized to clarify the confidence level distribution of each control behavior state in the current testing phase; the confidence level distribution is used as a state reference to check each adaptive test candidate sequence, extract unexecuted action semantics from the adaptive test candidate sequence, and collect the unexecuted action semantics to form a set of action semantics to be evaluated.
[0051] S5.2: Combining dynamic confidence state, the differences in vehicle control response data under different control behavior states are analyzed by comparing state response trajectories, and the discrimination ability metric value corresponding to each action semantic is calculated. Specifically, by combining the confidence distribution of each control behavior state in the dynamic confidence state, each action semantic in the set of action semantics to be evaluated is associated with the vehicle control response data under different control behavior states. According to the action event trajectory in the action response dataset, the changes in the vehicle control response data in the time dimension are compared and expanded, so that the vehicle control response data associated with the same action semantic under different control behavior states form a comparable vehicle control response change trajectory. The response deviations between different control behavior states are compared and deduced item by item, and the discrimination ability metric value corresponding to each action semantic is calculated.
[0052] The expression for calculating the discriminative power metric for each action semantic is: ; in, Representing action semantics The corresponding discrimination ability metric; Indicates the number of control behavior states; Indicates the length of the execution time interval defined by the action event trajectory; In action semantics Under the triggering condition, the corresponding number When a control behavior state is in progress, the vehicle control response vector obtained synchronously by the vehicle signal acquisition channel is obtained by combining the vehicle control response data within the same time interval. Representing action semantics The average representation of the vehicle control response vector under all control behavior states; This represents a single action semantic that has not been executed within the set of action semantics to be evaluated; Indicates the index number of the control behavior status; This represents a continuous time variable within the execution time interval.
[0053] S5.3: Summarize and organize the discrimination ability measurement values corresponding to each action semantic to form discrimination ability assessment information.
[0054] Specifically, according to the order of the action semantics that have not been executed in the set of action semantics to be evaluated, the discrimination ability measures are centrally organized, and each discrimination ability measure is associated with the corresponding action semantic and recorded one by one, so that the discrimination ability level corresponding to different action semantics can be clearly distinguished, forming a correspondence between action semantics and discrimination ability measures; the correspondence between action semantics and discrimination ability measures is uniformly collected to form discrimination ability assessment information.
[0055] S5.4: Write the discrimination ability metric value from the discrimination ability assessment information into the action semantics of the adaptive test candidate sequence that has not yet been executed, forming the sorting basis to be updated; Specifically, based on the correspondence between action semantics and discrimination ability metrics recorded in the discrimination ability assessment information, each unexecuted action semantic in the adaptive test candidate sequence is searched for, and the discrimination ability metric corresponding to each unexecuted action semantic is added to the adaptive test candidate sequence, so that all unexecuted action semantics have clear discrimination ability labels. The unexecuted action semantics with completed discrimination ability labels are uniformly collected according to their original arrangement order in the adaptive test candidate sequence, so that the discrimination ability metric and action semantics remain consistent in the same sequence, forming the basis for updating the sorting.
[0056] S5.5: Based on the sorting criteria to be updated, reconstruct the semantic order of actions that have not yet been executed to generate an updated adaptive test candidate sequence; Specifically, the unexecuted action semantics in the sorting criteria to be updated are rearranged according to the distinguishing ability metric values from high to low. The unexecuted action semantics in the sorted criteria are then connected and arranged with the executed action semantics in the adaptive test candidate sequence to keep the original order of the executed action semantics unchanged. The reordered unexecuted action semantics are then arranged in sequence to form the updated adaptive test candidate sequence.
[0057] S5.6: Based on the distribution of the discriminative power metric values corresponding to each action semantic in the updated adaptive test candidate sequence, determine whether the test process meets the preset confidence convergence condition and generate a confidence convergence determination flag.
[0058] Specifically, based on the updated adaptive test candidate sequence, the discrimination ability metric values corresponding to each action semantic are collected sequentially, and the overall distribution of the discrimination ability metric values in the adaptive test candidate sequence is uniformly checked in the same test phase. Whether the discrimination ability metric values maintain the same ordering relationship before and after the sequence update, and whether the ordering changes due to the addition or subtraction of action semantics, are used as the judgment criteria. When the discrimination ability metric values maintain a consistent ordering state during the update process, the test process is judged to meet the preset confidence convergence condition, and a confidence convergence judgment mark is generated.
[0059] It should be noted that the confidence convergence condition is set based on the discrimination capability assessment information and dynamic confidence state during the vehicle automated testing process. Specifically, in the updated adaptive test candidate sequence, the discrimination capability metric value corresponding to each action semantic that has not yet been executed remains in a stable sorting state after multiple sequential reconstructions. Furthermore, in the dynamic confidence state, the change in confidence level corresponding to each control behavior state gradually decreases as the test progresses and enters a fixed change range. This indicates that it is difficult to introduce new action semantics to produce new discrimination effects on the determination of control behavior states, thus forming the basis for determining that the test process has reached the convergence stage.
[0060] S6. Based on the confidence convergence determination flag, perform convergence control on the execution process of the updated adaptive test candidate sequence and generate a vehicle automation test report.
[0061] S6.1: Based on the confidence convergence determination flag, determine the execution direction of the updated adaptive test candidate sequence and generate execution convergence control instructions; Specifically, based on the convergence determination status indicated by the confidence convergence determination flag, the updated adaptive test candidate sequence is checked as a whole. When the confidence convergence determination flag indicates that the test process has reached the convergence stage, it is determined that the action semantics that have not yet been executed in the updated adaptive test candidate sequence no longer need to be executed, and the execution direction of the action semantics after termination is determined. When the confidence convergence determination flag indicates that the test process has not yet reached the convergence stage, it is determined that the updated adaptive test candidate sequence still needs to be executed, and the execution direction of continuing to advance the action semantics according to the updated adaptive test candidate sequence is determined. The corresponding execution directions are uniformly organized to generate execution convergence control instructions.
[0062] S6.2: Based on the convergence control instruction, terminate the execution process of the updated adaptive test candidate sequence to obtain the converged execution trajectory; Specifically, based on the convergence control instruction, the action semantics included in the action semantic execution scheduling sequence in the updated adaptive test candidate sequence are uniformly constrained. When the convergence control instruction indicates to stop scheduling new action semantics, the action semantics that have been executed in the updated adaptive test candidate sequence are not expanded. The execution order, start and end time of the corresponding action semantics and the trajectory of the associated action events are centrally organized so that the updated adaptive test candidate sequence forms a clear termination boundary under the action of the convergence control instruction. All action semantic execution processes actually completed before the termination time of the updated adaptive test candidate sequence are completely preserved and the converged execution trajectory is obtained.
[0063] S6.3: Summarize and organize the action semantic execution order, action response dataset and dynamic confidence state corresponding to the converged execution trajectory to generate a vehicle automation test report.
[0064] Specifically, the execution sequence of action semantics in the convergent execution trajectory is used as the timeline. The action response dataset associated with each action semantic execution process is retrieved accordingly. The vehicle control response reflected in the action response dataset is synchronously compared and organized with the confidence changes of each control behavior state in the dynamic confidence state, so that the action semantic execution sequence, action response dataset, and dynamic confidence state form a complete correspondence within the same time frame. The action semantic execution sequence, action response dataset, and dynamic confidence state are uniformly summarized and arranged to form a vehicle automation test report that can fully reflect the vehicle's action execution process, response performance, and confidence state evolution during the automated testing process.
[0065] This embodiment also provides a computer device applicable to the vehicle automated testing method based on embodied robots, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the vehicle automated testing method based on embodied robots as proposed in the above embodiment.
[0066] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0067] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the vehicle automation testing method based on an embodied robot as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0068] In summary, this invention achieves this by: associating the embodied robot's action execution process with the corresponding vehicle-mounted signal acquisition data under unified action semantics and time constraints, thus forming a quantifiable and continuously updatable basis for vehicle control response under different action triggering conditions; and by employing a confidence mechanism that is gradually updated during the testing process to dynamically manage the control behavior state in the industrial control system, allowing the confidence level of the control state to be continuously adjusted as evidence accumulates. This provides continuous, stable, and traceable data support for the selection of test actions, the convergence control of the test process, and the generation of vehicle automation test reports, thereby enhancing the closed-loop consistency of the test process and the reliability of the test conclusions.
[0069] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A vehicle automated testing method based on an embodied robot, characterized in that: include, Obtain the test task description of the target vehicle, and perform robot pose calibration, vehicle operable component identification, and vehicle signal acquisition channel association to form a test initialization configuration set; The differences in control state response of the test initialization configuration set under the current vehicle state are evaluated to generate an adaptive test candidate sequence; Based on adaptive test candidate sequences, the embodied robot is driven to execute target action semantics and the action execution process and vehicle control response are collected simultaneously to form an action response dataset. Based on the action response dataset, the consistency status of vehicle control response under different action semantics is determined, and the confidence level of the predetermined control behavior state is updated to form a dynamic confidence state. Based on dynamic confidence states, the ability to distinguish actions in the adaptive test candidate sequence that have not yet been executed is evaluated to form the ability to distinguish evaluation information, update the adaptive test candidate sequence, and generate a confidence convergence judgment flag and the updated adaptive test candidate sequence. Based on the confidence convergence criterion, convergence control is applied to the execution process of the updated adaptive test candidate sequence to generate a vehicle automation test report.
2. The vehicle automated testing method based on a android as described in claim 1, characterized in that: The steps for forming the test initialization configuration set are as follows: Obtain the test task description of the target vehicle, and determine the vehicle test scope and action semantic type based on the test task description; Based on action semantic types, the embodied robot is driven to perform pose sampling and calibration within the vehicle test range to obtain a unified coordinate mapping relationship; Based on a unified coordinate mapping relationship and action semantic type, reachability determination is performed on candidate interactive objects within the vehicle test range to filter and obtain a set of operable parts; The robot is driven to perform controlled operations on the set of operable parts and collect corresponding changes in vehicle-mounted signals, thus establishing a correspondence between operable parts and vehicle-mounted signal acquisition channels. The test initialization configuration set is generated by summarizing the vehicle test range, action semantic type, unified coordinate mapping relationship, set of operable parts and operable behavior and their correspondence with the vehicle signal acquisition channel.
3. The vehicle automated testing method based on embodied robots as described in claim 2, characterized in that: The steps for generating adaptive test candidate sequences are as follows: A joint traversal is performed on the action semantic types, operable component sets, and the correspondence between operable components and vehicle signal acquisition channels in the test initialization configuration set to construct the action semantic mapping relationship; The status of the vehicle signal acquisition channel corresponding to the action semantic mapping relationship is read to form a status snapshot; The control state response changes corresponding to the state snapshot under different action semantics are compared and calculated to obtain the distinguishability assessment identifier. Based on the distinguishability assessment identifier, each action semantic is sorted to generate an adaptive test candidate sequence.
4. The automated vehicle testing method based on an embodied robot as described in claim 1, characterized in that: The steps for forming the action response dataset are as follows: The action semantics in the adaptive test candidate sequence are scheduled and orchestrated to generate an action semantics execution scheduling sequence; Based on the action semantics execution scheduling sequence, the embodied robot is driven to perform target action semantics on the corresponding operable parts, and the start and end times of the action are recorded to form an action event trajectory; Under the constraints of the start and end time of the action event trajectory, vehicle control response data in the vehicle signal acquisition channel corresponding to the target action semantics are collected synchronously, and the vehicle control response data is associated and encapsulated with the action event trajectory to form an action response dataset.
5. The vehicle automated testing method based on an embodied robot as described in claim 4, characterized in that: The steps for forming a dynamic confidence state are as follows: The vehicle control response data in the action response dataset is mapped to the predetermined control behavior state to form an action semantic response sequence. Consistency comparison is performed on the changes in vehicle control response data under different action semantics for the same control behavior state in the action semantic response sequence to obtain consistency judgment information; Based on the consistency determination information, the confidence level of the predetermined control behavior state is updated to generate a dynamic confidence state.
6. The automated vehicle testing method based on an embodied robot as described in claim 1, characterized in that: The steps for generating the differentiation ability assessment information are as follows: Read the confidence distribution of each control behavior state in the dynamic confidence state, and filter the action semantics that have not yet been executed from the adaptive test candidate sequence to form a set of action semantics to be evaluated; By combining dynamic confidence states, the differences in vehicle control response data under different control behavior states are analyzed by state response trajectory comparison and deduction, and the discrimination ability metric value corresponding to each action semantic is calculated. The discrimination ability metrics corresponding to each action semantic are summarized and organized to form discrimination ability assessment information.
7. The automated vehicle testing method based on an embodied robot as described in claim 6, characterized in that: The steps for generating the confidence convergence determination identifier and the updated adaptive test candidate sequence are as follows: The discrimination ability metric value from the discrimination ability assessment information is written into the action semantics of the adaptive test candidate sequence that have not yet been executed, forming the sorting basis to be updated; Based on the sorting criteria to be updated, the semantics of actions that have not yet been executed are reconstructed in order to generate an updated adaptive test candidate sequence. Based on the distribution of the discriminative power metric values corresponding to each action semantic in the updated adaptive test candidate sequence, it is determined whether the test process meets the preset confidence convergence condition, and a confidence convergence determination identifier is generated.
8. The vehicle automated testing method based on a android as described in claim 1, characterized in that: The steps for performing convergence control on the execution process of the updated adaptive test candidate sequence based on the confidence convergence determination flag and generating a vehicle automation test report are as follows. Based on the confidence convergence determination flag, the execution direction of the updated adaptive test candidate sequence is determined, and an execution convergence control instruction is generated. According to the convergence control instruction, the execution process of the updated adaptive test candidate sequence is terminated to obtain the convergence execution trajectory; The execution order of action semantics, action response dataset, and dynamic confidence state corresponding to the converged execution trajectory are summarized and organized to generate a vehicle automation test report.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the vehicle automated testing method based on a holographic robot as described in any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the vehicle automated testing method based on a holographic robot as described in any one of claims 1 to 8.