Simulation evaluation method for automatic testing equipment of vehicle machine

Abstract

The invention, which relates to the technical field of equipment simulation testing, discloses a simulation evaluation method for automatic testing equipment of an in-vehicle infotainment system, and the method comprises the steps: collecting operation test data, carrying out the recognition processing through a deep convolutional neural network model, obtaining test task data, constructing a test action sequence, mapping the test action sequence into a control instruction, and carrying out the simulation evaluation of the in-vehicle infotainment system. The control instruction is optimized and adjusted according to a preset action rule base; driving the in-vehicle automatic test equipment to execute the control instruction to complete the test operation, recording the operation response time and interface state change data in the test process, calculating a movement track deviation value, calculating a test accuracy score, and when the test accuracy score is smaller than a preset threshold value, starting the in-vehicle automatic test equipment; and if not, supplementing the optimization parameters to the preset action rule base and triggering a new test. According to the invention, accurate control of the test equipment is realized by adopting the TCP / IP communication protocol, and the test strategy is continuously optimized by dynamically updating the preset action rule base, so that the automatic test efficiency of the vehicle machine is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of device simulation testing, and particularly to a simulation evaluation method for in-vehicle infotainment (IVI) system automation testing equipment. Background Art

[0002] Traditional IVI system testing mainly relies on manual operation, which has problems such as low testing efficiency, poor consistency, and high costs. In recent years, automation testing technology has gradually been applied to the field of IVI system testing, and the automation execution of the testing process has been achieved through technologies such as machine vision and robotic arms. However, existing IVI system automation testing solutions mostly adopt a fixed testing mode based on rules, lacking in-depth analysis of testing data and the ability of adaptive optimization, and it is difficult to cope with complex and changeable testing scenarios and the differentiated requirements of different vehicle models. At the same time, due to the complexity of IVI system interface interaction and the requirement for the accuracy of operation trajectories, existing testing equipment often has problems such as inaccurate operation and response delay during the execution of testing tasks, affecting the reliability of testing results.

[0003] Currently, the main technical difficulties in the field of IVI system automation testing are as follows: First, the ability of intelligent recognition and action planning for testing tasks is insufficient, and it is difficult to accurately understand the interface state and generate a reasonable test sequence; second, the execution accuracy and real-time control of testing actions are poor, and the stability and consistency of testing operations cannot be guaranteed; third, there is a lack of effective testing quality evaluation mechanisms and optimization feedback mechanisms, and it is difficult to continuously improve the testing effect. These problems seriously restrict the application and popularization of IVI system automation testing technology.

[0004] In view of the above problems, the present invention proposes a simulation evaluation method for in-vehicle infotainment (IVI) system automation testing equipment. This method realizes the intelligent recognition of testing tasks through a deep convolutional neural network, and combines a multi-dimensional scoring model and a dynamic optimization strategy to effectively improve the accuracy and reliability of test execution. Summary of the Invention

[0005] In view of the problems of inaccurate recognition of testing tasks, insufficient accuracy of action execution, and lack of adaptive optimization ability in existing IVI system automation testing equipment, the present invention is proposed.

[0006] Therefore, the problems to be solved by the present invention are how to achieve the intelligent recognition and adaptive execution of IVI system testing tasks, and how to establish a scientific testing evaluation and optimization mechanism, so as to improve the accuracy and reliability of IVI system automation testing.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a simulation evaluation method for a vehicle-mounted machine automatic test device, which includes collecting operation test data of an interface to be tested of the vehicle-mounted machine, and performing recognition processing on the operation test data through a deep convolutional neural network model to obtain test task data; constructing a test action sequence according to the test task data, mapping the test action sequence into a control instruction, and optimizing and adjusting the control instruction according to a preset action rule library; driving the vehicle-mounted machine automatic test device to execute the control instruction to complete the test operation, recording the operation response time and interface state change data during the test, and calculating the motion trajectory deviation value; calculating a test accuracy score based on the operation response time, interface state change data and motion trajectory deviation value, and when the test accuracy score is less than a preset threshold, supplementing optimization parameters to the preset action rule library and triggering a new test.

[0009] As a preferred solution of the simulation evaluation method for the vehicle-mounted machine automatic test device according to the present invention, wherein: the calculation method of the test accuracy score is to perform normalization processing on the operation response time by using a weighted summation method, and calculate the time deviation score based on the standard response time; use a deep learning image matching algorithm to compare the interface state images before and after the test, calculate the structural similarity index and local feature point matching degree, and generate an interface consistency score; based on the motion trajectory deviation value, use the root mean square error method to evaluate the fitting degree of the actual trajectory and the standard trajectory, and generate a trajectory accuracy score; fuse the time deviation score, the interface consistency score and the trajectory accuracy score through a multi-dimensional weighted model to obtain the test accuracy score; judge the test accuracy score and the preset threshold, analyze the specific dimension that causes the score to decrease, and extract the optimization parameters of the corresponding dimension; update the optimization parameters to the preset action rule library, automatically adjust the execution speed, waiting time and operation force of the test action, and trigger a new round of tests until the test accuracy score meets the requirements of the preset threshold.

[0010] As a preferred solution of the simulation evaluation method for the vehicle-mounted machine automatic test device according to the present invention, wherein: the specific formula of the test accuracy score is as follows:

[0011]

[0012] wherein, S is the test accuracy score, ω1 is the time score weight coefficient, ω1 is the interface score weight coefficient, ω3 is the trajectory score weight coefficient, t i is the actual response time of the i-th operation, t s is the standard response time, N is the total number of operations, SSIM(·) is the structural similarity between the interface states before and after the i-th operation, is the interface state image data before the test operation is performed at the i-th test point, The image data of the interface state after performing the test operation for the i-th test point, m is the number of interface state monitoring points, and ΔA is the motion trajectory deviation value.

[0013] When the test accuracy score is less than the preset threshold, if there is only a single dimension score anomaly, enter the fast optimization mode; if there are two or more dimension score anomalies at the same time, enter the collaborative optimization mode; when the test accuracy score is greater than or equal to the preset threshold, calculate the adjustment amplitude of the current optimal parameters and the initial parameters respectively, and update the current optimal parameter combination as the benchmark parameters for such test scenarios.

[0014] As a preferred solution of the simulation evaluation method for the in-vehicle machine automation test equipment according to the present invention, wherein: drive the in-vehicle machine automation test equipment to execute the control instruction to complete the test operation, record the operation response time and the interface state change data during the test, and calculate the motion trajectory deviation value, including: send the optimized control instruction sequence to the main controller of the in-vehicle machine automation test equipment according to the TCP / IP communication protocol, wherein the main controller drives the multi-degree-of-freedom robotic arm to execute the corresponding test operation after parsing the instruction parameters; set a microsecond-level timer to record the sending time and the actual response time of each control instruction, calculate the instruction execution delay time, and write the operation response time data into the test database in real time; use the camera module to collect the interface images during the test in real time, extract the interface state characteristic parameters, and record them in time series to form the interface state change data; use the laser displacement sensor to collect the spatial coordinate data of the end effector of the robotic arm in real time, compare the actual motion trajectory with the theoretical trajectory, eliminate the measurement noise through the Kalman filter algorithm, and calculate the motion trajectory deviation value.

[0015] As a preferred solution of the simulation evaluation method for the in-vehicle machine automation test equipment according to the present invention, wherein: the specific formula of the motion trajectory deviation value is as follows:

[0016]

[0017] Wherein, ΔA is the motion trajectory deviation value, P i (t) is the position function of the actual trajectory in the i-th dimension, is the position function of the theoretical trajectory in the i-th dimension, ε is the total number of dimensions, t is the test time variable, λ j is the weight coefficient of the j-th measurement point, ∈ j is the instantaneous error of the j-th measurement point, n is the total number of measurement points, H(ω) is the Hessian matrix used to evaluate the trajectory curvature, R(θ) is the rotation matrix used for attitude deviation compensation, det(·) is the matrix determinant operation, and tr(·) is the matrix trace operation.

[0018] As a preferred solution of the simulation evaluation method for the in-vehicle infotainment (IVI) automation test equipment according to the present invention, the following steps are included: constructing a test action sequence based on the test task data, mapping the test action sequence into control instructions, and optimizing and adjusting the control instructions according to a preset action rule library, including: decomposing the test task data based on the depth-first search algorithm into basic action units, constructing a directed acyclic graph according to the operation logic dependency relationship, and generating a test action sequence, where the basic action units include touch operations, swipe operations, long-press operations, and multi-touch operations; converting the test action sequence into corresponding control instructions according to an action mapping table, where the action mapping table includes action types, coordinate positions, execution durations, and trigger conditions; reading the optimization rules in the preset action rule library and verifying the instruction parameters of the control instructions, where the optimization rules include action interval thresholds, trigger condition constraints, and conflict detection rules; using a genetic algorithm to optimize the parameters of the control instruction sequence, and iteratively optimizing the instruction parameters that do not meet the optimization rules to generate an optimized control instruction sequence.

[0019] In a second aspect, the embodiments of the present invention provide a method for obtaining test task data, which is to collect the to-be-tested interface of the IVI from multiple angles using a high-resolution camera array to obtain the running test data of the interface display content, control layout, and text information; inputting the running test data into a deep convolutional neural network model, where the neural network model uses a VGG-16 network structure, and extracting hierarchical features of the interface image through several convolutional layers and pooling layers; setting an attention mechanism in the fully connected layer of the neural network model to perform object detection and semantic segmentation on the button area, text area, and input box area of the to-be-tested interface of the IVI; performing labeling processing on the results of object detection and semantic segmentation to generate test task data, where the test task data includes control types, position information, and operation attributes.

[0020] In a second aspect, the embodiments of the present invention provide a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program instructions are executed by the processor, the steps of the simulation evaluation method for the IVI automation test equipment as described in the first aspect of the present invention are implemented.

[0021] In a third aspect, the embodiments of the present invention provide a computer-readable storage medium, on which a computer program is stored, and: when the computer program instructions are executed by the processor, the steps of the simulation evaluation method for the IVI automation test equipment as described in the first aspect of the present invention are implemented.

[0022] The beneficial effects of the present invention are as follows: By using a deep convolutional neural network model to intelligently recognize and extract features from the in-vehicle infotainment (IVI) system interface, and combining with the attention mechanism to achieve precise positioning of test tasks, effectively solving the limitations of traditional manually defined features; By constructing a directed acyclic graph of test action sequences through a depth-first search algorithm and using a genetic algorithm for parameter optimization, significantly improving the execution efficiency and reliability of test instructions; Adopting multi-sensor fusion technology and a Kalman filtering algorithm to monitor and process data in real time during the test process, combined with microsecond-level response time recording and laser displacement measurement, to achieve high-precision test trajectory tracking and evaluation; At the same time, establishing a multi-dimensional scoring model based on time deviation, interface consistency, and trajectory accuracy, and through two modes of rapid optimization and collaborative optimization, realizing the adaptive adjustment of test parameters to ensure the accuracy and repeatability of test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. Among them:

[0024] Figure 1 FIG. is a flowchart of a simulation evaluation method for an in-vehicle infotainment system automation test device in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0026] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0027] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other from other embodiments.

[0028] Embodiment 1

[0029] Refer to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a simulation evaluation method for an in-vehicle infotainment system automation test device, including

[0030] S1: Collect the running test data of the in-vehicle unit to be tested, and identify and process the running test data through a deep convolutional neural network model to obtain test task data.

[0031] Specifically, the method for obtaining test task data is to use a high-resolution camera array to collect the in-vehicle unit to be tested from multiple angles, and obtain the running test data of the interface display content, control layout, and text information.

[0032] It should be noted that the camera array includes three fixed shooting positions: the front view, the left view, and the right view.

[0033] Furthermore, input the running test data into a deep convolutional neural network model. The neural network model uses the VGG-16 network structure, and extracts hierarchical features of the interface image through several convolutional layers and pooling layers; set an attention mechanism in the fully connected layer of the neural network model to perform object detection and semantic segmentation on the button area, text area, and input box area of the in-vehicle unit to be tested.

[0034] Even further, perform labeling processing on the results of object detection and semantic segmentation to generate test task data, where the test task data includes control type, position information, and operation attributes.

[0035] S2: Construct a test action sequence based on the test task data, map the test action sequence to a control instruction, and optimize and adjust the control instruction according to a preset action rule library.

[0036] Specifically, perform task decomposition on the test task data based on the depth-first search algorithm, split it into basic action units, and construct a directed acyclic graph according to the operation logic dependency relationship to generate a test action sequence, where the basic action units include touch operations, swipe operations, long-press operations, and multi-touch operations.

[0037] Furthermore, convert the test action sequence into corresponding control instructions according to the action mapping table, where the action mapping table includes action type, coordinate position, execution duration, and trigger condition; read the optimization rules in the preset action rule library, and verify the instruction parameters of the control instruction, where the optimization rules include action interval threshold, trigger condition constraint, and conflict detection rule.

[0038] It should be noted that for the action interval threshold rules: the minimum interval time between consecutive touch operations should be ≥ 200 ms to avoid accidental touches; the waiting time after a swipe operation should be ≥ 500 ms to ensure the completion of the interface animation; the minimum duration of a long - press operation should be ≥ 1000 ms to ensure the triggering of the long - press effect; the time difference between the contact press times of multi - touch operations should be ≤ 50 ms to ensure synchronization; the interval time for switching between different control - area operations should be ≥ 300 ms to ensure state updates. For the trigger - condition constraint rules: before a swipe operation, it is necessary to detect whether the target area supports the swipe property; before a long - press operation, it is necessary to verify whether the control has the long - press trigger function; for multi - touch operations, it is necessary to confirm whether the interface supports the multi - touch mode; for operations on pop - up controls, they should be executed after the pop - up is fully displayed; during the animation transition, the execution of subsequent operation instructions should be paused. For the conflict - detection rules: it is prohibited to perform multiple touch - type operations in the same area simultaneously; a swipe operation and a long - press operation cannot be executed in parallel in an overlapping area; multi - touch operations cannot be performed simultaneously with other types of operations; it is necessary to avoid triggering consecutive operations on mutually exclusive controls; ensure that there is no spatial overlap between the execution areas of adjacent instructions in the operation sequence.

[0039] Furthermore, the genetic algorithm is used to optimize the parameters of the control instruction sequence, and the instruction parameters that do not meet the optimization rules are iteratively optimized to generate an optimized control instruction sequence.

[0040] S3: Drive the in - vehicle infotainment (IVI) automated test equipment to execute the control instructions to complete the test operations, and record the operation response time and the interface state change data during the test, and calculate the motion - trajectory deviation value.

[0041] Specifically, according to the TCP / IP communication protocol, the optimized control instruction sequence is sent to the main controller of the IVI automated test equipment, and the main controller drives the multi - degree - of - freedom robotic arm to execute the corresponding test operations after parsing the instruction parameters.

[0042] Further, a microsecond - level timer is set to record the sending time and the actual response time of each control instruction, calculate the instruction execution delay time, and write the operation response time data into the test database in real - time; use the camera module to collect the interface images during the test in real - time, extract the interface state characteristic parameters, and record them in a time series to form the interface state change data.

[0043] Furthermore, a laser displacement sensor is used to collect the spatial coordinate data of the end - effector of the robotic arm in real - time, compare the actual motion trajectory with the theoretical trajectory, eliminate the measurement noise through the Kalman filter algorithm, and calculate the motion - trajectory deviation value. The specific formula is as follows:

[0044]

[0045] where ΔA is the motion - trajectory deviation value, P i(t) is the position function of the actual trajectory in the i-th dimension, is the position function of the theoretical trajectory in the i-th dimension, ε is the total number of dimensions, t is the test time variable, λ j is the weight coefficient of the j-th measurement point, ∈ j is the instantaneous error of the j-th measurement point, n is the total number of measurement points, H(ω) is the Hessian matrix used to evaluate the trajectory curvature, R(θ) is the rotation matrix used for attitude deviation compensation, det(·) is the matrix determinant operation, and tr(·) is the matrix trace operation.

[0046] S4: Calculate the test accuracy score based on the operation response time, interface state change data, and motion trajectory deviation value. When the test accuracy score is less than the preset threshold, supplement the optimization parameters to the preset action rule library and trigger a new test.

[0047] Specifically, use the weighted summation method to normalize the operation response time and calculate the time deviation score based on the standard response time; use the deep learning image matching algorithm to compare the interface state images before and after the test, calculate the structural similarity index and the local feature point matching degree, and generate the interface consistency score; based on the motion trajectory deviation value, use the root mean square error method to evaluate the fitting degree of the actual trajectory and the standard trajectory, and generate the trajectory accuracy score.

[0048] Furthermore, fuse the time deviation score, interface consistency score, and trajectory accuracy score through a multi-dimensional weighted model to obtain the test accuracy score. The specific formula is as follows:

[0049]

[0050] where S is the test accuracy score, ω1 is the time score weight coefficient, ω1 is the interface score weight coefficient, ω3 is the trajectory score weight coefficient, t i is the actual response time of the i-th operation, t s is the standard response time, N is the total number of operations, SSIM(·) is the structural similarity between the i-th interface states before and after, is the interface state image data before the i-th test point executes the test operation, is the interface state image data after the i-th test point executes the test operation, m is the number of interface state monitoring points, and ΔA is the motion trajectory deviation value.

[0051] It should be noted that the multi-dimensional weighted model is constructed based on three dimensions: the operation response time deviation score, the interface state consistency score, and the motion trajectory accuracy score.

[0052] Furthermore, the test accuracy score is compared with a preset threshold, the specific dimensions that lead to the score reduction are analyzed, and the optimization parameters for the corresponding dimensions are extracted. When the test accuracy score is less than the preset threshold, if there is only a single dimension with an abnormal score, the fast optimization mode is entered; if there are abnormal scores in two or more dimensions simultaneously, the collaborative optimization mode is entered; when the test accuracy score is greater than or equal to the preset threshold, the adjustment amplitudes of the current optimal parameters and the initial parameters are calculated respectively, and the current optimal parameter combination is updated as the benchmark parameters for such test scenarios.

[0053] It should be noted that the preset threshold is determined based on a comprehensive analysis of the probability density distribution characteristics of historical test data, the sensitivity thresholds of each dimension score, and the complexity of the test scenario.

[0054] Specifically, the fast optimization mode includes calculating the deviation rate between the abnormal dimension score and the historical mean. When the deviation rate is in the range of 20% to 50%, this abnormality is marked as a mild abnormality, and only the core parameters of this dimension are slightly adjusted. For example, if there is a mild abnormality in the time dimension, only the waiting time parameter is adjusted; if there is a mild abnormality in the interface consistency, only the image acquisition interval is adjusted; if there is a mild abnormality in the trajectory accuracy, only the execution speed is adjusted. If the deviation rate exceeds 50%, this abnormality is marked as a severe abnormality, and all relevant parameters of this dimension are adjusted simultaneously. For example, if there is a severe abnormality in the time dimension, both the waiting time and the number of retries are adjusted; if there is a severe abnormality in the interface consistency, both the image acquisition interval and the contrast sensitivity are adjusted; if there is a severe abnormality in the trajectory accuracy, both the execution speed is adjusted. The effect of each parameter adjustment is evaluated, and the optimal parameter combination is gradually approximated through the dichotomy method until the score of this dimension recovers above the preset threshold.

[0055] Furthermore, the collaborative optimization mode includes constructing a dimension correlation matrix to analyze the mutual influence relationship between each abnormal dimension; calculating the contribution weights of each dimension through the principal component analysis method to determine the optimization order; if the correlation coefficient between two dimensions is greater than 0.7, these two dimensions are classified as a strongly correlated group and a linkage optimization strategy is adopted. For example, when the time dimension and the interface consistency dimension are both abnormal and strongly correlated, the waiting time and the image acquisition interval are adjusted synchronously according to the correlation characteristics; if the correlation coefficient between dimensions is less than 0.3, it is classified as a weakly correlated group and an independent optimization strategy is adopted, and the optimization is carried out in descending order of the contribution weights. For the moderately correlated group with a correlation coefficient between 0.3 and 0.7, a mixed optimization strategy is adopted. Through the iterative optimization process, the parameter combination is continuously adjusted until the scores of all dimensions reach the preset threshold simultaneously.

[0056] Furthermore, update the optimization parameters to the preset action rule library, automatically adjust the execution speed, waiting time, and operation force of the test actions, and trigger a new round of tests until the test accuracy score reaches the preset threshold requirement.

[0057] This embodiment also provides a computer device applicable to the simulation evaluation method for in-vehicle infotainment (IVI) system automated test equipment, including 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 simulation evaluation method for in-vehicle infotainment (IVI) system automated test equipment as proposed in the above embodiment.

[0058] This computer device can be a terminal, and this computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the shell of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0059] In summary, the present invention uses a deep convolutional neural network model to intelligently identify and extract features of the in-vehicle infotainment (IVI) system interface, combines an attention mechanism to achieve precise positioning of test tasks, and effectively solves the limitations of traditional manually defined features; constructs a directed acyclic graph of test action sequences through a depth-first search algorithm and uses a genetic algorithm for parameter optimization, significantly improving the execution efficiency and reliability of test instructions; uses a multi-sensor fusion technology and a Kalman filtering algorithm to monitor and process data in real time during the test process, combines microsecond-level response time recording and laser displacement measurement to achieve high-precision test trajectory tracking and evaluation; at the same time, establishes a multi-dimensional scoring model based on time deviation, interface consistency, and trajectory accuracy, and realizes adaptive adjustment of test parameters through two modes of rapid optimization and collaborative optimization to ensure the accuracy and repeatability of test results.

[0060] Embodiment 2

[0061] Referring to Table 1, this is the second embodiment of the present invention. This embodiment provides a simulation evaluation method for in-vehicle infotainment (IVI) system automated test equipment. To verify the beneficial effects of the present invention, scientific demonstrations are carried out through economic benefit calculations and simulation experiments.

[0062] Specifically, a 2024 model intelligent in-vehicle system of a certain brand was selected as the test object. This in-vehicle system is equipped with a 12.3-inch high-definition touch display screen and uses the Android Automotive system. The test equipment includes: an automated test platform equipped with a 6-axis robotic arm (accuracy ±0.02 mm), a 4K high-speed industrial camera array (sampling rate of 200 fps), a 16-channel laser displacement sensor (sampling accuracy of 0.01 mm), and a high-performance industrial control computer (equipped with an RTX 4080 graphics card).

[0063] Furthermore, the in-vehicle interface data was collected through a high-resolution camera array. The camera array is distributed in a 120° fan shape to ensure complete coverage of the display screen surface. The collected image data was preprocessed and then input into an optimized VGG-16 network. This network consists of 5 convolutional blocks and 3 fully connected layers, combined with an improved attention mechanism, achieving a control recognition accuracy of 98.7%. Based on the recognition results, the system automatically generated a test task dataset containing 374 test points.

[0064] Even further, an improved depth-first search algorithm was used to decompose the test tasks, generating a total of 2183 basic action units, including operation types such as single-point touch, multi-point touch, swipe, and long press. These action units were converted into standard control instructions through an action mapping engine and optimized in combination with a preset action rule library. The rule library contains 186 optimization rules summarized from a large number of experiments, covering the optimal execution parameters for different operation types.

[0065] Specifically, during the test execution phase, the system sent the optimized control instructions to the test platform through Gigabit Ethernet. The main controller uses a real-time operating system with an interrupt response time <10 μs, ensuring the real-time execution of the instructions. At the same time, the laser displacement sensor array recorded the robotic arm movement trajectory at a sampling rate of 1 kHz, and the noise was processed through an improved Kalman filtering algorithm, with the trajectory reconstruction accuracy reaching ±0.05 mm.

[0066] Furthermore, as shown in Table 1, by comparing the data in the table, it can be clearly seen that the method of the present invention is significantly superior to the traditional manual testing method in various performance indicators. The specific analysis is as follows: In terms of the accuracy of control recognition, the method of the present invention reaches a high accuracy of 98.7%, which is 6.4 percentage points higher than 92.3% of the traditional manual testing method. This improvement is mainly due to the innovative application of the deep convolutional neural network model and the attention mechanism adopted by the present invention, realizing the accurate recognition and positioning of in-vehicle infotainment system interface controls. The average response time index shows that the method of the present invention reduces the response time to 98 ms, a reduction of 147 ms compared to 245 ms of the traditional method, and the response speed is increased by 60%. This significant improvement is attributed to the optimization of the TCP / IP communication protocol and the precise control of the microsecond-level timer adopted by the present invention, greatly improving the real-time performance of the testing system.

[0067] Table 1 Comparison Table of the Method of the Present Invention and the Traditional Method

[0068] Performance indicators Traditional manual testing method Method of the present invention Accuracy rate of control recognition 92.3％ 98.7％ Average response time 245ms 98ms Interface consistency score 0.882 0.978 Track precision deviation 0.43mm 0.12mm Test coverage 78.5％ 95.8％ Test efficiency 12 times per hour 96 times per hour Comprehensive score 0.823 0.947

[0069] Furthermore, in terms of the interface consistency score, the method of the present invention reaches a high score of 0.978, which is 10.9% higher than 0.882 of the traditional method. This reflects that the present invention can more accurately evaluate the changes in the interface state during the testing process through the deep learning image matching algorithm, ensuring the reliability of the test results. The trajectory accuracy deviation data shows that the method of the present invention controls the deviation within 0.12 mm, a reduction of 72.1% compared to 0.43 mm of the traditional method. This improvement in accuracy stems from the real-time monitoring of the laser displacement sensor and the noise processing technology of the Kalman filter algorithm adopted by the present invention, ensuring the high-precision execution of the testing operation.

[0070] Specifically, in terms of the test coverage rate, the method of the present invention achieves a coverage rate of 95.8%, which is 17.3 percentage points higher than 78.5% of the traditional method. This indicates that the present invention can cover the test scenarios more comprehensively, significantly improving the integrity and reliability of the test. In terms of the test efficiency index, the method of the present invention can complete 96 tests per hour, which is 8 times that of the traditional method at 12 times per hour. This significant improvement in efficiency benefits from the high degree of automation and the optimized test strategy of the present invention, greatly improving the time efficiency of the testing process. The comprehensive score data shows that the method of the present invention obtains a high score of 0.947, which is 15.1% higher than 0.823 of the traditional method. This comprehensive index fully proves the superiority of the present invention in overall performance.

[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A simulation evaluation method for vehicle computer automation test equipment, characterized in that: include, Collecting the running test data of the vehicle computer to be tested interface, and identifying and processing the running test data through a deep convolutional neural network model to obtain test task data; Constructing a test action sequence according to the test task data, mapping the test action sequence into a control instruction, and optimizing and adjusting the control instruction according to a preset action rule library; The vehicle computer automated test equipment is driven to execute the control instruction to complete the test operation, and the operation response time and interface state change data during the test are recorded, and the motion trajectory deviation value is calculated; The test accuracy score is calculated based on the operation response time, interface state change data and motion trajectory deviation value. When the test accuracy score is less than the preset threshold, the optimization parameters are added to the preset action rule library and a new test is triggered.

2. The simulation evaluation method for vehicle computer automation test equipment according to claim 1, characterized in that: The test accuracy score is calculated as follows: The operation response time is normalized using the weighted summation method, and the time deviation score is calculated based on the standard response time; Use deep learning image matching algorithm to compare interface status images before and after the test, calculate the structural similarity index and local feature point matching degree, and generate interface consistency score; Based on the motion trajectory deviation value, the root mean square error method is used to evaluate the fit between the actual trajectory and the standard trajectory, and the trajectory accuracy score is generated; The time deviation score, the interface consistency score and the trajectory accuracy score are integrated through a multi-dimensional weighted model to obtain a test accuracy score; The test accuracy score is judged against a preset threshold, the specific dimension causing the score to decrease is analyzed, and the optimization parameters of the corresponding dimension are extracted; The optimization parameters are updated to the preset action rule library, the execution speed, waiting time and operation intensity of the test action are automatically adjusted, and a new round of testing is triggered until the test accuracy score reaches the preset threshold requirement.

3. The simulation evaluation method for vehicle computer automation test equipment according to claim 2, characterized in that: The specific formula for the test accuracy score is as follows: Among them, S is the test accuracy score, ω1 is the time score weight coefficient, ω1 is the interface score weight coefficient, ω3 is the trajectory score weight coefficient, t i is the actual response time of the ith operation, t s is the standard response time, N is the total number of operations, SSIM(·) is the structural similarity before and after the i-th interface state, The interface status image data before the test operation is performed on the i-th test point. is the interface state image data after the test operation is performed on the i-th test point, m is the number of interface state detection points, and ΔA is the motion trajectory deviation value; When the test accuracy score is less than the preset threshold, if only a single dimension score is abnormal, it will enter the fast optimization mode; if two or more dimensions have score abnormalities at the same time, it will enter the collaborative optimization mode; When the test accuracy score is greater than or equal to the preset threshold, the adjustment range of the current optimal parameters and the initial parameters are calculated respectively, and the current optimal parameter combination is updated to the benchmark parameters for this type of test scenario.

4. The simulation evaluation method for vehicle computer automation test equipment according to claim 3, characterized in that: The vehicle computer automated test equipment is driven to execute the control instructions to complete the test operation, and the operation response time and interface state change data during the test are recorded, and the motion trajectory deviation value is calculated, including: Sending the optimized control instruction sequence to the main controller of the vehicle-mounted automated test equipment according to the TCP / IP communication protocol, wherein the main controller analyzes the instruction parameters and drives the multi-degree-of-freedom robotic arm to perform the corresponding test operation; Set a microsecond timer to record the sending time and actual response time of each control instruction, calculate the instruction execution delay time, and write the operation response time data into the test database in real time; The camera module is used to collect the interface images in real time during the test, extract the interface state characteristic parameters, and record the interface state change data in time series; A laser displacement sensor is used to collect the spatial coordinate data of the end effector of the robotic arm in real time, and the actual motion trajectory is compared with the theoretical trajectory. The Kalman filter algorithm is used to eliminate the measurement noise and calculate the motion trajectory deviation value.

5. The simulation evaluation method for vehicle computer automation test equipment according to claim 4, characterized in that: The specific formula of the motion trajectory deviation value is as follows: Among them, ΔA is the motion trajectory deviation value, P i (t) is the position function of the actual trajectory in the i-th dimension, is the position function of the theoretical trajectory in the i-th dimension, ε is the total number of dimensions, t is the test time variable, and λ j is the weight coefficient of the jth measurement point, ∈ j is the instantaneous error of the jth measurement point, n is the total number of measurement points, H(ω) is the Hessian matrix used to evaluate the trajectory curvature, R(θ) is the rotation matrix used for attitude deviation compensation, det(·) is the matrix determinant operation, and tr(·) is the matrix trace operation.

6. The simulation evaluation method for vehicle computer automation test equipment according to claim 5, characterized in that: Constructing a test action sequence according to the test task data, mapping the test action sequence into a control instruction, and optimizing and adjusting the control instruction according to a preset action rule library, including: Based on the depth-first search algorithm, the test task data is decomposed into basic action units, and a directed acyclic graph is constructed according to the operation logic dependency to generate a test action sequence, wherein the basic action units include touch operation, sliding operation, long press operation and multi-touch operation; Convert the test action sequence into corresponding control instructions according to the action mapping table, wherein the action mapping table includes action type, coordinate position, execution time and trigger condition; Reading optimization rules in a preset action rule library and verifying instruction parameters of the control instruction, wherein the optimization rules include action interval thresholds, trigger condition constraints, and conflict detection rules; A genetic algorithm is used to optimize the parameters of the control instruction sequence, and instruction parameters that do not meet the optimization rules are iteratively optimized to generate an optimized control instruction sequence.

7. The simulation evaluation method for vehicle computer automation test equipment according to claim 6, characterized in that: The method for obtaining the test task data is: Use a high-resolution camera array to collect data from multiple angles on the vehicle computer interface to be tested, and obtain the operating test data of the interface display content, control layout, and text information; Inputting the running test data into a deep convolutional neural network model, wherein the neural network model adopts a VGG-16 network structure and extracts hierarchical features of the interface image through a plurality of convolutional layers and pooling layers; An attention mechanism is set in the fully connected layer of the neural network model to perform target detection and semantic segmentation on the button area, text area and input box area of ​​the vehicle computer to be tested interface; The results of target detection and semantic segmentation are labeled to generate test task data, wherein the test task data includes control type, location information and operation attributes.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the simulation evaluation method for vehicle-mounted automated testing equipment described in any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the simulation evaluation method for vehicle-mounted automated testing equipment described in any one of claims 1 to 7 are implemented.

Images