Method and device for intelligent recommendation of human-robot interaction test scenarios for space robots

Abstract

The patent discloses a kind of human-computer interaction test scene intelligent recommendation method and device for space robot, constructs the correlation factor layer of space robot human-computer interaction test scene and analysis and correlation classification rule, and correlation factor is decomposed into tree structure according to analysis rule, the correlation coefficient between each factor is calculated, the correlation coefficient size is used as correlation classification standard, constructs tree space robot meta scene set;Quantify the evaluation of expert to the meta scene set in data set as reward function, and construct the intelligent recommendation total module composed of robot capability decomposition module and Tree DDPG module, train intelligent recommendation model with the aid of space robot meta scene set and corresponding reward function, obtain the well-trained human-computer interaction test scene intelligent recommendation model;According to the ability characteristics of space robot and the preset interaction task, call the well-trained human-computer interaction test scene intelligent recommendation model, generate test scene sequence and corresponding test index set sequence.

Description

Technical Field

[0001] This invention belongs to the field of human factors evaluation technology for space robots, and particularly relates to an intelligent recommendation method and device for human-computer interaction testing scenarios for space robots. Background Technology

[0002] With the space station becoming operational and manned lunar exploration missions progressing, more and more robots will participate in manned spaceflight missions. Before space robots perform space exploration missions, designing reasonable test scenarios and selecting appropriate test index sequences based on robot performance and mission characteristics, and conducting human-robot interaction performance tests, is of great significance for ensuring the implementation of space exploration missions. However, the ergonomic evaluation standards for space robots are still under development. Currently, the method of designing test scenarios and selecting test indicators using expert knowledge involves iterative iterations to continuously optimize the test plan. Obviously, this model results in a heavy workload for experts and makes it difficult to respond quickly to new products. Furthermore, there is currently no systematic summary or corresponding dataset for human-robot interaction task evaluation indicators and testing methods. Moreover, there is a lack of quantitative evaluation standards for test scenarios designed for specific robots performing specific tasks. Therefore, it is difficult to apply relevant intelligent recommendation algorithms in the field of artificial intelligence to recommend human-robot interaction test scenarios for space robots without data and reward functions.

[0003] Therefore, it is necessary to construct an intelligent recommendation system for human-computer interaction test scenarios for space robots. This system should include hierarchical analysis and association classification rules of human factors ergonomics indicators, a set of space robot test scenarios containing various evaluation indicators, and an intelligent recommendation algorithm for test scenario sets that integrates convolutional neural networks and reinforcement learning, so as to recommend test scenario sequences based on robot capability characteristics and preset tasks.

[0004] Already, some methods have been developed to select robot capabilities and corresponding collaborative capability evaluation indicators using correlation coefficients, thereby constructing an evaluation indicator system. For example, patent application CN115781764A discloses a collaborative robot collaborative capability testing task and its comprehensive evaluation method. However, the dimensions of human-robot interaction evaluation indicators and the diversity of testing methods for space robots are more complex, and it is difficult to construct a reasonable comprehensive evaluation indicator system using only correlation coefficients.

[0005] To evaluate the performance of human-computer interaction in space, a performance evaluation method for a space human-computer interaction system, disclosed in application publication number CN112016786 A, constructs an event tree model and analyzes human-computer interaction characteristics to analyze the impact of human error, environment, equipment, and various coupling factors on performance. This method focuses more on human-computer interaction performance evaluation. However, human-computer interaction ergonomics evaluation is not merely an ergonomics evaluation; it involves evaluation indicators across various dimensions such as user experience, usability, and human-computer ergonomics. Furthermore, based on these evaluation indicators and considering the task capabilities and testing tasks of various space robots, reasonable test scenarios need to be designed. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing an intelligent recommendation method and apparatus for human-computer interaction testing scenarios for space robots.

[0007] The objective of this invention is achieved through the following technical solution: an intelligent recommendation method for human-computer interaction testing scenarios for space robots, comprising the following steps:

[0008] (1) Construct a layer of related factors for human-computer interaction test scenarios of space robots, and set rules for analysis and classification. Decompose the related factors into a tree structure according to the analysis rules, calculate the correlation coefficient between each factor, and use the correlation coefficient as the classification standard to construct a tree-shaped meta-scenario set for space robots.

[0009] (2) The evaluation of experts on the meta-scene set in the dataset is quantified into a reward function, and an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module is constructed. The intelligent recommendation model is trained with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios.

[0010] (3) Based on the capabilities and characteristics of the space robot and the preset interactive tasks, call the trained intelligent recommendation model for human-computer interaction test scenarios to generate a sequence of test scenarios and a corresponding sequence of test index sets.

[0011] Furthermore, step (1) specifically includes the following sub-steps:

[0012] (1.1) Based on the meta-scene set containing robot capability set, interaction task set, evaluation index set and test method set, with robot capability, interaction task, evaluation index and test method as the main root node, the hierarchical analysis and association classification rules of the factors related to the space robot test scenario are constructed, and the decomposition is carried out layer by layer until the smallest unit. The decomposition results of each layer are stored in the corresponding leaf node. Each leaf node constructs a corresponding test meta-scene set, thus constructing a tree-like space robot meta-scene set.

[0013] (1.2) Use a scale to assess each leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and evaluation metrics to obtain the first correlation score set X. i ; and use a scale to evaluate the same leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and testing methods to obtain a second correlation score set Y. i ;

[0014] According to formula P i =Cov(X) i ,Y i ) / (σX i *σY i Solve for H of each leaf node i The correlation coefficients among various factors such as robot capabilities, interactive tasks, evaluation metrics, and testing methods are used to obtain the correlation coefficient set {P}, where Cov(X) = ... i ,Y i ) represents the first associated score set X i Second associated rating set Y i The covariance between them, σX i For the first set of related scores X i Standard deviation, σY i For the second associated score set Y i The standard deviation.

[0015] Furthermore, step (2) specifically includes the following sub-steps:

[0016] (2.1) The evaluations of experts on each test meta-scene set within the space robot meta-scene set are quantified into a reward function. First, according to the formula... The reward value is normalized, where r i ′ represents the normalized score of the i-th test meta-scene set, r i For the expert's score on the i-th test meta-scene set, r max r is the highest score given by experts across all test scenarios. min The lowest score given by experts across all test scenarios;

[0017] Then construct the reward function according to the following formula:

[0018]

[0019] Among them, R i Let be the reward function for the i-th test meta-scene set, and Δ be the threshold for selecting the reinforcement learning reward function;

[0020] (2.2) Construct an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module. Train the intelligent recommendation model with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios.

[0021] Furthermore, step (2.2) specifically includes the following sub-steps:

[0022] (2.2.1) Construct a general intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module;

[0023] (2.2.2) In the robot capability decomposition module composed of a one-dimensional convolutional neural network, the hierarchical parsing and association classification rules are used to decompose the interactive task set into a meta-task set {S}: {S} = {s1, s2, ..., s a ,…,s M}, where S a Let {S} be any meta-task; and decompose the robot capability set into a test function sequence set {I} based on the robot capabilities required by the meta-task set {S}: {I} = {i1, i2, ..., i k ,…,i N}, where i k For any test function sequence, the set of test function sequences {I} is then combined with the corresponding meta-tasks in order to transform it into a functional test selection matrix {IS}: {IS} = {i1-s1,i2-s2,…,i k -s k ,…,i N -s N}, where i k -s k Select for any functional test; when i k -s k =0 indicates that in metatask s k The following does not require testing the function sequence i k Perform the test, when i k -s k =1 indicates that in metatask s k The following requires testing the functional sequence i k Conduct testing;

[0024] (2.2.3) In the Tree DDPG module, which consists of the Tree 1D-CNN module and the DDPG module, the functional test selection matrix {IS} is input into the Tree 1D-CNN module for hierarchical classification to obtain the evaluation index and test method set {E} matched by the functional test selection matrix {IS}.

[0025] The Tree 1D-CNN module consists of a root node module, an evaluation metric branch, and a test method branch;

[0026] The root node module selects i based on the input functional test. k -s k The test function sequence i was searched out from the tree-like spatial robot meta-scene set. k The corresponding test meta-scene set;

[0027] Subsequently, the evaluation metric branch and the testing method branch are based on the searched corresponding test meta-scenario set and test function sequence i. k Further subdivision yields the test function sequence i. k Matching evaluation metrics and testing methods E k ;

[0028] (2.2.4) Subsequently, the evaluation metrics and test method set {E} are input into the DDPG module. The reward value of the DDPG module adopts the test function sequence i. k The maximum corresponding reward value

[0029] Let a be the action performed by the agent at each time step. t The corresponding state is q t =[i k -s k [,{E},{P}];Reward value G t The solution can be obtained using the following formula:

[0030] G t =r t +γ*r t-1 +…+γ N-t *r N

[0031] Where γ is the attenuation factor, γ∈[0,1]; r t The reward for step t;

[0032]

[0033] Where s represents the current state; R j |s t =s represents the test function sequence i in the current state. t The reward function for the corresponding test meta-scene set; (P st＝s ) 2 m represents the sum of squares of the correlation coefficients in the set {P} of correlation coefficients under the current state; e This indicates the proportion of the squared value of the correlation coefficient in the reward function;

[0034] Within a preset training step, the DDPG module gradually converges towards the maximum value of the evaluation metric and the reward value of the test method set {E} output by the Tree 1D-CNN module, thus obtaining a trained intelligent recommendation model for human-computer interaction test scenarios.

[0035] The present invention also includes an intelligent recommendation device for human-computer interaction test scenarios for space robots, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, it is used for the aforementioned intelligent recommendation method for human-computer interaction test scenarios for space robots.

[0036] The present invention also includes a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the above-described intelligent recommendation method for human-computer interaction test scenarios for space robots.

[0037] The beneficial effects of this invention are: compared with traditional machine learning-based intelligent recommendation algorithms, the network structure designed in this invention has the advantage of rapid expansion. That is, when new indicators or new robot-related testing methods need to be added to the model, only a portion of new data needs to be selected for training according to the established rules, without the need to retrain the entire system, and the weights can be updated. Attached Figure Description

[0038] Figure 1 A flowchart of an intelligent recommendation method for human-computer interaction testing scenarios for space robots;

[0039] Figure 2 This is a schematic diagram of the meta-scene set;

[0040] Figure 3 A schematic diagram of the overall structure of the intelligent recommendation module;

[0041] Figure 4 This is a structural diagram of an intelligent recommendation device for human-computer interaction testing scenarios for space robots. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0043] In this invention, the terms "first," "second," etc. (if present) in the invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0044] Example 1

[0045] like Figure 1 As shown, this invention provides an intelligent recommendation method for human-computer interaction testing scenarios for space robots, comprising the following steps:

[0046] (1) Construct a layer of related factors for human-computer interaction test scenarios of space robots, and set rules for analysis and classification. Decompose the related factors into a tree structure according to the analysis rules, calculate the correlation coefficient between each factor, and use the correlation coefficient as the classification standard to construct a tree-shaped meta-scenario set for space robots.

[0047] Step (1) specifically includes the following sub-steps:

[0048] (1.1) Based on the meta-scene set containing the robot capability set, interaction task set, evaluation index set and test method set, with robot capability, interaction task, evaluation index and test method as the main root nodes, the hierarchical analysis and association classification rules of the factors related to the space robot test scenario are constructed, and the decomposition is carried out layer by layer until the smallest unit. The decomposition results of each layer are stored in the corresponding leaf nodes. Each leaf node constructs a corresponding test meta-scene set, thus constructing a tree-like space robot meta-scene set.

[0049] In this embodiment, the meta-scenario set includes a robot capability set, an interaction task set, an evaluation index set, and a test method set. Using robot capabilities, interaction tasks, evaluation indices, and test methods as the main root nodes, it decomposes layer by layer into hierarchical analysis and association classification rules for factors related to the space robot test scenario, down to the smallest unit. The decomposition results at each layer are stored as corresponding leaf nodes. Figure 2 As shown, "Robot Capabilities" is the first layer of the robot capability set, "Interactive Tasks" is the first layer of the interactive task set, "Evaluation Metrics" is the first layer of the evaluation metric set, and "Testing Methods" is the first layer of the testing method set. "Robot Capabilities," "Interactive Tasks," "Evaluation Metrics," and "Testing Methods" serve as the primary root nodes. The four related factors—"Perception," "Cognition," "Decision-Making," and "Execution"—are decomposed at the same level and stored under the same leaf node, H. i The robot's capabilities, interaction tasks, evaluation metrics, and testing methods are used to construct a corresponding test meta-scenario set D. i Furthermore, the related factors such as "touch", "vision", "voice", ..., "other", "natural language understanding", "scene understanding", ..., "task understanding", "human-controlled robot", "human-machine collaboration", ..., "autonomous operation", "robotic arm", ..., "other" are also decomposition results of the same layer and stored under the same leaf node;

[0050] By decomposing the meta-scene set layer by layer, a tree-like spatial robot meta-scene set {D1,…,D} is constructed. i ,…,D Q}, where D1 is the test meta-scenario set constructed by the main root node, D Q The minimum test meta-scenario set constructed for the terminal leaf nodes.

[0051] (1.2) Use a scale to assess each leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and evaluation metrics to obtain the first correlation score set X. i ; and use a scale to evaluate the same leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and testing methods to obtain a second correlation score set Y. i ;

[0052] According to formula P i =Cov(X) i ,Y i ) / (σX i *σY i Solve for H of each leaf node i The correlation coefficients among various factors such as robot capabilities, interactive tasks, evaluation metrics, and testing methods are used to obtain the correlation coefficient set {P}, where Cov(X) = ... i ,Y i ) represents the first associated score set X i Second associated rating set Y i The covariance between them, σX i For the first set of related scores X i Standard deviation, σY i For the second associated score set Y i The standard deviation.

[0053] (2) The evaluation of experts on the meta-scene set in the dataset is quantified into a reward function, and an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module is constructed. The intelligent recommendation model is trained with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios.

[0054] Step (2) specifically includes the following sub-steps:

[0055] (2.1) The evaluations of experts on each test meta-scene set within the space robot meta-scene set are quantified into a reward function. First, according to the formula... The reward value is normalized, where r i ′ represents the normalized score of the i-th test meta-scene set, r iFor the expert's score on the i-th test meta-scene set, r max r is the highest score given by experts across all test scenarios. min The lowest score given by experts across all test scenarios;

[0056] Then construct the reward function according to the following formula:

[0057]

[0058] Among them, R i Let be the reward function for the i-th test meta-scene set, and Δ be the threshold for selecting the reinforcement learning reward function. By reducing the reward value of the scheme with lower score, the model is prompted to converge to the combination of schemes with higher score.

[0059] (2.2) Construct an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module. Train the intelligent recommendation model with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios.

[0060] Step (2.2) specifically includes the following sub-steps:

[0061] (2.2.1) Construct a general intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module;

[0062] (2.2.2) In the robot capability decomposition module composed of a one-dimensional convolutional neural network, the hierarchical parsing and association classification rules are used to decompose the interactive task set into a meta-task set {S}: {S} = {s1, s2, ..., s a ,…,s M}, where S a Let {S} be any meta-task; and decompose the robot capability set into a test function sequence set {I} based on the robot capabilities required by the meta-task set {S}: {I} = {i1, i2, ..., i k ,…,i N}, where i k For any test function sequence, the set of test function sequences {I} is then combined with the corresponding meta-tasks in order to transform it into a functional test selection matrix {IS}: {IS} = {i1-s1,i2-s2,…,i k -s k ,…,i N -s N}, where i k -s k Select for any functional test; when i k -s k =0 indicates that in metatask sk The following does not require testing the function sequence i k Perform the test, when i k -s k =1 indicates that in metatask s k The following requires testing the functional sequence i k Conduct testing;

[0063] (2.2.3) In the Tree DDPG module, which consists of the Tree 1D-CNN module and the DDPG module, the functional test selection matrix {IS} is input into the Tree 1D-CNN module for hierarchical classification to obtain the evaluation index and test method set {E} matched by the functional test selection matrix {IS}.

[0064] The Tree 1D-CNN module consists of a root node module, an evaluation metric branch, and a test method branch;

[0065] The root node module selects i based on the input functional test. k -s k The test function sequence i was searched out from the tree-like spatial robot meta-scene set. k The corresponding test meta-scene set;

[0066] Subsequently, the evaluation metric branch and the testing method branch are based on the searched corresponding test meta-scenario set and test function sequence i. k Further subdivision yields the test function sequence i. k Matching evaluation metrics and testing methods E k ;

[0067] The set of matching evaluation metrics and test methods {E} consists of each test function sequence i k Matching evaluation metrics and testing methods E k composition.

[0068] (2.2.4) Subsequently, the evaluation metrics and test method set {E} are input into the DDPG module. The reward value of the DDPG module adopts the test function sequence i. k The maximum corresponding reward value

[0069] Let a be the action performed by the agent at each time step. t The corresponding state is q t =[i k -s k [,{E},{P}];Reward value G t The solution can be obtained using the following formula:

[0070] G t =r t +γ*rt-1 +…+γ N-t *r N

[0071] Where γ is the attenuation factor, γ∈[0,1]; r t The reward for step t;

[0072]

[0073] Where s represents the current state; R j |s t =s represents the test function sequence i in the current state. t The reward function for the corresponding test meta-scene set; (P st＝s ) 2 m represents the sum of squares of the correlation coefficients in the set {P} of correlation coefficients under the current state; e This indicates the proportion of the squared value of the correlation coefficient in the reward function;

[0074] Within a preset training step, the DDPG module gradually converges towards the maximum value of the evaluation metric and the reward value of the test method set {E} output by the Tree 1D-CNN module, thus obtaining a trained intelligent recommendation model for human-computer interaction test scenarios.

[0075] (3) Based on the capabilities and characteristics of the space robot and the preset interaction tasks, the trained intelligent recommendation model for human-computer interaction test scenarios is invoked to generate a sequence of test scenarios and a corresponding sequence of test index sets, specifically:

[0076] By leveraging the hierarchical classification rules of factors associated with space robot test scenarios, the robot's capabilities are decomposed into sequential test functions. Combined with the meta-tasks obtained from the predetermined interaction task decomposition, a sequential test function selection matrix is ​​generated. A pre-trained intelligent recommendation model for human-computer interaction test scenarios is then invoked to output test scenarios. Since the output test scenarios are sequential numerical results, they must be decoded according to predetermined rules to output readable test scenario solutions.

[0077] Example 2

[0078] This embodiment relates to an intelligent recommendation device for human-computer interaction test scenarios for space robots, including a memory and one or more processors. The memory stores executable code. When the one or more processors execute the executable code, it is used for the intelligent recommendation method for human-computer interaction test scenarios for space robots in Embodiment 1 above. The device embodiment can be applied to any device with data processing capabilities, such as a computer or other similar devices.

[0079] like Figure 4At the hardware level, this knowledge distillation device includes a processor, internal bus, network interface, memory, and non-volatile storage, and may also include other hardware required for the business logic. The processor reads the corresponding computer program from the non-volatile storage into memory and then executes it to achieve the above-mentioned functionality. Figure 1 The method is illustrated. Of course, in addition to software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0080] Improvements in a technology can be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology can now be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement in methodology cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog are the most commonly used. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0081] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0082] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0083] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0086] Example 3

[0087] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements the intelligent recommendation method for human-computer interaction test scenarios for space robots described in Embodiment 1 above.

[0088] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent recommendation of human-computer interaction testing scenarios for space robots, characterized in that, Includes the following steps: (1) Construct a layer of related factors for human-computer interaction test scenarios of space robots, and set rules for analysis and classification. Decompose the related factors into a tree structure according to the analysis rules, calculate the correlation coefficient between each factor, and use the correlation coefficient as the classification standard to construct a tree-shaped meta-scenario set for space robots. (2) The evaluation of experts on the meta-scene set in the dataset is quantified into a reward function, and an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module is constructed. The intelligent recommendation model is trained with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios. (3) Based on the capabilities and characteristics of the space robot and the preset interactive tasks, call the trained intelligent recommendation model for human-computer interaction test scenarios to generate a sequence of test scenarios and a corresponding sequence of test index sets.

2. The intelligent recommendation method for human-computer interaction testing scenarios for space robots according to claim 1, characterized in that, Step (1) specifically includes the following sub-steps: (1.1) Based on the meta-scene set containing robot capability set, interaction task set, evaluation index set and test method set, with robot capability, interaction task, evaluation index and test method as the main root node, the hierarchical analysis and association classification rules of the factors related to the space robot test scenario are constructed, and the decomposition is carried out layer by layer until the smallest unit. The decomposition results of each layer are stored in the corresponding leaf node. Each leaf node constructs a corresponding test meta-scene set, thus constructing a tree-like space robot meta-scene set. (1.2) Use a scale to assess each leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and evaluation metrics to obtain the first correlation score set X. i ; and use a scale to evaluate the same leaf node H i Subjectively evaluate the correlation between the robot's capabilities, interactive tasks, and testing methods to obtain a second correlation score set Y. i ; According to formula P i =Cov(X) i ,Y i ) / (σX i *σY i Solve for H of each leaf node i The correlation coefficients among various factors such as robot capabilities, interactive tasks, evaluation metrics, and testing methods are used to obtain the correlation coefficient set {P}, where Cov(X) i ,Y i ) represents the first associated score set X i Second associated rating set Y i The covariance between them, σX i For the first set of related scores X i Standard deviation, σY i For the second associated score set Y i The standard deviation.

3. The intelligent recommendation method for human-computer interaction testing scenarios for space robots according to claim 1, characterized in that, Step (2) specifically includes the following sub-steps: (2.1) The evaluations of experts on each test meta-scene set within the space robot meta-scene set are quantified into a reward function. First, according to the formula... The reward value is normalized, where r i ′ represents the normalized score of the i-th test meta-scene set, r i For the expert's score on the i-th test meta-scene set, r max r is the highest score given by experts across all test scenarios. min The lowest score given by experts across all test scenarios; Then construct the reward function according to the following formula: Among them, R i Let be the reward function for the i-th test meta-scene set, and Δ be the threshold for selecting the reinforcement learning reward function; (2.2) Construct an intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module. Train the intelligent recommendation model with the help of the spatial robot meta-scene set and the corresponding reward function to obtain a trained intelligent recommendation model for human-computer interaction test scenarios.

4. The intelligent recommendation method for human-computer interaction testing scenarios for space robots according to claim 3, characterized in that, Step (2.2) specifically includes the following sub-steps: (2.2.1) Construct a general intelligent recommendation module consisting of a robot capability decomposition module and a Tree DDPG module; (2.2.2) In the robot capability decomposition module composed of a one-dimensional convolutional neural network, the hierarchical parsing and association classification rules are used to decompose the interactive task set into a meta-task set {S}: {S} = {s1, s2, ..., s a ,…,s M }, where S a Let {S} be any meta-task; and decompose the robot capability set into a test function sequence set {I} based on the robot capabilities required by the meta-task set {S}: {I} = {i1, i2, ..., i k ,…,i N }, where i k For any test function sequence, the set of test function sequences {I} is then combined with the corresponding meta-tasks in order to transform it into a functional test selection matrix {IS}: {IS} = {i1-s1,i2-s2,…,i k -s k ,…,i N -s N }, where i k -s k Select for any functional test; when i k -s k =0 indicates that in metatask s k The following does not require testing the function sequence i k Perform the test, when i k -s k =1 indicates that in metatask s k The following requires testing the functional sequence i k Conduct testing; (2.2.3) In the Tree DDPG module, which consists of the Tree 1D-CNN module and the DDPG module, the functional test selection matrix {IS} is input into the Tree 1D-CNN module for hierarchical classification to obtain the evaluation index and test method set {E} matched by the functional test selection matrix {IS}. The Tree 1D-CNN module consists of a root node module, an evaluation metric branch, and a test method branch; The root node module selects i based on the input functional test. k -s k The test function sequence i was searched out from the tree-like spatial robot meta-scene set. k The corresponding test meta-scene set; Subsequently, the evaluation metric branch and the testing method branch are based on the searched corresponding test meta-scenario set and test function sequence i. k Further subdivision yields the test function sequence i. k Matching evaluation metrics and testing methods E k ; (2.2.4) Subsequently, the evaluation metrics and test method set {E} are input into the DDPG module. The reward value of the DDPG module adopts the test function sequence i. k The maximum corresponding reward value Let a be the action performed by the agent at each time step. t The corresponding state is qt = [ik-sk, {E}, {P}]; the reward value is G. t The solution can be obtained using the following formula: G t =r t +γ*r t-1 +…+c N-t *r N Where γ is the attenuation factor, γ∈[0,1]; r t The reward for step t; Where s represents the current state; R j |s t =s represents the test function sequence i in the current state. t The reward function for the corresponding test meta-scene set; (P st＝s ) 2 m represents the sum of squares of the correlation coefficients in the set {P} of correlation coefficients under the current state; e This indicates the proportion of the squared value of the correlation coefficient in the reward function; Within a preset training step, the DDPG module gradually converges towards the maximum value of the evaluation metric and the reward value of the test method set {E} output by the Tree 1D-CNN module, thus obtaining a trained intelligent recommendation model for human-computer interaction test scenarios.

5. An intelligent recommendation device for human-computer interaction testing scenarios for space robots, characterized in that, The device includes a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the intelligent recommendation method for human-computer interaction test scenarios for space robots as described in any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, It stores a program that, when executed by a processor, implements the intelligent recommendation method for human-computer interaction test scenarios for space robots as described in any one of claims 1-4.

Images