Method and device for task planning of complex human-machine system, medium and computer equipment

Abstract

This invention discloses a method, apparatus, medium, and computer device for task planning in complex human-computer systems, comprising: responding to a planning instruction for a target human-computer interaction task, dividing the target human-computer interaction task into multiple task nodes, and obtaining task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters for each task node; determining the multi-dimensional workload of each task node based on the task attribute parameters, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload; generating multiple candidate task planning schemes using a preset scheme prediction model based on the multi-dimensional workload, task surrounding environment parameters, and human-computer function allocation core parameters; determining the scheme evaluation value of each candidate task planning scheme, and selecting the final task planning scheme for executing the target human-computer interaction task from each candidate task planning scheme based on the scheme evaluation value.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ergonomics and human factors, and in particular to a complex human-machine system task planning method, device, medium and computer equipment. BACKGROUND

[0002] With the increasing complexity of human-machine interaction and task planning of complex systems, the number and types of operations that need to be dealt with when performing diversified tasks have significantly increased. These tasks include not only daily maintenance and function testing of equipment, but also installation, repair and abnormal situation handling of equipment. Therefore, in order to ensure efficient completion of tasks and reduce operational risks, task planning is needed.

[0003] Currently, tasks are usually planned by humans to be completed in the shortest time. However, the shortest time for completing a task may increase the workload of personnel, thereby increasing the risk of task operation. In addition, the manual task planning method is time-consuming and labor-intensive, and may be affected by subjective factors, resulting in unreasonable task planning. SUMMARY

[0004] The present application provides a complex human-machine system task planning method, device, medium and computer equipment, which can improve the efficiency and rationality of task planning and reduce the risk of task operation.

[0005] According to a first aspect of the present application, a complex human-machine system task planning method is provided, comprising: In response to a planning instruction of a target human-machine interaction task, dividing the target human-machine interaction task into a plurality of task nodes, and obtaining a task attribute parameter, a task peripheral environment parameter and a human-machine function allocation core parameter of each task node; Based on the task attribute parameter, determining a multi-dimensional workload of each task node, wherein the multi-dimensional workload includes at least two of visual load, auditory load, cognitive load and psychomotor load; Based on the multi-dimensional workload, the task peripheral environment parameter and the human-machine function allocation core parameter, a plurality of candidate task planning schemes are generated by using a preset scheme prediction model; Determining a scheme evaluation value of each candidate task planning scheme, and selecting a final task planning scheme for executing the target human-machine interaction task in each candidate task planning scheme based on the scheme evaluation value.

[0006] Optionally, the determination of the scheme evaluation value of each candidate task planning scheme comprises: weighting sum of the multi-dimensional workloads, to obtain a node work load of each of the task nodes, and determine a comprehensive work load of the target candidate task planning scheme based on the node work load; determine a task execution time and a task execution risk of the target candidate task planning scheme, and determine a correlation among the comprehensive work load, the task execution time, and the task execution risk; determine a weight coefficient corresponding to each of the comprehensive work load, the task execution time, and the task execution risk, aggregate the comprehensive work load, the task execution time, and the task execution risk based on the correlation and the weight coefficient, and determine a scheme evaluation value of the target candidate task planning scheme based on an aggregation result.

[0007] Optionally, the generating of the multiple candidate task planning schemes based on the multi-dimensional workloads, the task peripheral environment parameters, and the human-machine function allocation core parameters utilizes a preset scheme prediction model, including: determining a load feature vector corresponding to each of the multi-dimensional workloads, an environment feature vector corresponding to each of the task peripheral environment parameters, and a function allocation feature vector corresponding to each of the human-machine function allocation core parameters; mining latent features from the load feature vector, the environment feature vector, and the function allocation feature vector to obtain a latent mining feature vector; inputting the latent mining feature vector into the preset scheme prediction model for planning scheme prediction to obtain the multiple candidate task planning schemes for the target human-machine interaction task.

[0008] Optionally, the mining of the latent features from the load feature vector, the environment feature vector, and the function allocation feature vector to obtain the latent mining feature vector includes: determining a vector length and a vector direction of each of the load feature vector, the environment feature vector, and the function allocation feature vector as a target feature vector, and determining a task scene complexity of the target human-machine interaction task; determining a vector transformation parameter of the target feature vector based on the vector length, determining a length-direction coupling coefficient of the target feature vector based on the vector length and the vector direction, determining an environment perception factor of the target feature vector based on the task scene complexity, and determining a vector weight of the target feature vector based on the length-direction coupling coefficient and the environment perception factor; performing vector transformation on the corresponding load feature vector, environment feature vector and function allocation feature vector based on the vector transformation parameters corresponding to the load feature vector, the environment feature vector and the function allocation feature vector respectively, and performing weighted fusion on the vector-transformed load feature vector, the vector-transformed environment feature vector and the vector-transformed function allocation feature vector based on the vector weights corresponding to the load feature vector, the environment feature vector and the function allocation feature vector respectively, and taking the weighted fusion result as the potential mining feature vector.

[0009] Optionally, after selecting the final task planning scheme for performing the target human-computer interaction task in each candidate task planning scheme based on the scheme evaluation value, the method further comprises: monitoring whether the task surrounding environment parameters and the task time window change in real time, and if yes, re-planning the target human-computer interaction task.

[0010] Optionally, after determining the scheme evaluation value of each candidate task planning scheme, the method further comprises: determining the required tool, the predicted task completion time and the predicted workload of performing each candidate task planning scheme, and sorting each candidate task planning scheme based on the scheme evaluation value; displaying each sorted candidate task planning scheme in a human-computer interaction interface based on the required tool, the predicted task completion time and the predicted workload.

[0011] Optionally, before generating multiple candidate task planning schemes by using a preset scheme prediction model based on the multi-dimensional workload, the task surrounding environment parameters and the human-computer function allocation core parameters, the method further comprises: constructing a preset initial scheme generation model; obtaining a sample data set, wherein the sample data set includes the multi-dimensional workload, the task surrounding environment parameters and the human-computer function allocation core parameters of a sample task with a task planning scheme label; dividing the sample data set into a training set and a test set, training the preset initial scheme generation model by using the training set, testing the trained preset initial scheme generation model by using the test set, and finally taking the trained preset initial scheme generation model meeting the test condition as the preset scheme prediction model.

[0012] According to a second aspect of the present application, a complex human-computer system task planning device is provided, comprising: The acquisition unit is used to respond to the planning instructions of the target human-computer interaction task, divide the target human-computer interaction task into multiple task nodes, and acquire the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node. The determining unit is configured to determine the multi-dimensional workload of each task node based on the task attribute parameters, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload. The scheme generation unit is used to generate multiple candidate task planning schemes based on the multi-dimensional workload, the task surrounding environment parameters, and the human-machine function allocation core parameters, using a preset scheme prediction model. The scheme selection unit is used to determine the scheme evaluation value of each candidate task planning scheme, and based on the scheme evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

[0013] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described complex human-machine system task planning method.

[0014] According to a fourth aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described complex human-machine system task planning method.

[0015] The present invention provides a method, apparatus, medium, and computer equipment for complex human-machine system task planning. Compared with the current method of manually planning tasks based on minimizing task time, the present invention plans tasks by comprehensively analyzing parameters of the task's surrounding environment, core parameters of human-machine function allocation, visual load, auditory load, cognitive load, and psychomotor load. This ensures the rationality of task planning, avoids excessive workload on workers, and reduces the risk of fatigue operation. By generating multiple solutions through models, the limitations of a single solution are avoided, improving adaptability to complex scenarios. The entire process of complex human-machine system task planning in the present invention does not require human intervention, thereby improving the efficiency and rationality of task planning. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart of a task planning method for a complex human-machine system provided by an embodiment of the present invention is shown; Figure 2 This invention provides a flowchart of another task planning method for complex human-machine systems. Figure 3 This diagram illustrates the structure of a task planning device for a complex human-machine system according to an embodiment of the present invention. Figure 4 This invention provides a schematic diagram of the structure of another complex human-machine system task planning device according to an embodiment of the present invention. Figure 5 A schematic diagram of the physical structure of a computer device provided in an embodiment of the present invention is shown. Detailed Implementation

[0017] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0018] Currently, manually planning tasks based on the shortest possible time increases the workload on staff, thereby increasing the risk of task operation. In addition, manual task planning is time-consuming and labor-intensive, and due to subjective factors, there may be unreasonable task planning.

[0019] To address the aforementioned problems, embodiments of the present invention provide a task planning method for complex human-machine systems, such as... Figure 1 As shown, the method includes: 101. In response to the planning instructions of the target human-computer interaction task, the target human-computer interaction task is divided into multiple task nodes, and the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node are obtained.

[0020] The target human-computer interaction task can be a complex human-computer system task in any scenario such as aircraft, space system, aircraft carrier, nuclear energy, etc.; the task attribute parameters include, but are not limited to, task location coordinates, tools required to perform the task, pre-set priority of the task, task execution time window, resource consumption, dependencies, and task risk coefficient; the task surrounding environment parameters include, but are not limited to, task surrounding lighting and visual conditions, gravity environment, terrain, temperature, and communication conditions; the core parameters of human-computer function allocation include, but are not limited to, the task complexity of the task node, the skill requirements and equipment resources required to perform the task node.

[0021] In this embodiment of the invention, if the target human-computer interaction task is a complex human-computer system task, after receiving the task planning instruction, the complex human-computer system task is broken down into task nodes such as satellite launch, orbit adjustment, scientific experiment, and data transmission. Then, the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node are determined by means of data collection by sensors or other devices or by means of data acquisition from a database.

[0022] 102. Based on task attribute parameters, determine the multi-dimensional workload of each task node, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload.

[0023] In this embodiment of the invention, a preset workload prediction model can be pre-constructed, and then the preset workload prediction model can be used to predict the multi-dimensional workload of each task node. Based on this, the method includes: constructing a preset initial workload prediction model; obtaining a sample dataset, wherein the sample dataset includes task attribute parameters of sample task nodes with multi-dimensional workload labels, the multi-dimensional workload labels including visual workload labels, auditory workload labels, cognitive workload labels, and psychomotor workload labels, and the task attribute parameters including but not limited to the task location coordinates of the sample task node, the tools required to perform the task, the pre-set priority of the task, the task execution time window, resource consumption, dependencies, and task risk coefficient; dividing the sample dataset into a training set and a test set, training the preset initial workload prediction model using the training set, and testing the trained preset initial workload prediction model using the test set, and finally using the trained preset initial workload prediction model that meets the test conditions as the preset workload prediction model. Specifically, in the model training process, the preset initial workload prediction model is first constructed, and then the sample dataset is obtained. It is ensured that the dataset contains all necessary files. Transform the data into a format that the initial workload prediction model can understand, and then train and test the model. Specifically, first, divide the dataset: use randomness or a specific strategy (such as stratified sampling) to divide the sample dataset into training and test sets. Then, use the training set to train the model, and use the test set to test the trained model and evaluate its performance on unseen data. Calculate and record metrics such as precision and recall on the test set. If the model performance does not meet requirements, return to the training phase for further iterations or adjustments. This process yields a satisfactory initial workload prediction model.

[0024] Furthermore, the model structure of the preset workload prediction model includes a feature extraction layer, a feature enhancement layer, and a workload prediction layer. After training and constructing the preset workload prediction model, for each task node, its corresponding task attribute parameters are input into the preset workload prediction model. The feature extraction layer extracts the parameter features of the task attribute parameters, the feature enhancement layer enhances the parameter features output by the feature extraction layer, and the workload prediction layer uses the enhanced features output by the feature enhancement layer to predict the workload, thus obtaining the visual workload, auditory workload, cognitive workload, and psychomotor workload of the corresponding task node. This embodiment of the invention predicts multi-dimensional workload through a model without human intervention, thereby improving the prediction accuracy and efficiency of workload.

[0025] In another embodiment of the present invention, the preset workload prediction model can also be a VCP (Visual, Auditory, Cognitive, Psychomotor) model. Multi-dimensional workload prediction is performed using the VCP model.

[0026] 103. Based on multi-dimensional workload, task surrounding environment parameters, and human-machine function allocation core parameters, generate multiple candidate task planning schemes using a preset scheme prediction model.

[0027] In this embodiment of the invention, to improve the prediction accuracy of the preset scheme prediction model, it is first necessary to train and construct the preset scheme prediction model. Based on this, the method includes: constructing a preset initial scheme generation model; obtaining a sample dataset, wherein the sample dataset includes multi-dimensional workload of sample tasks with task planning scheme labels, task surrounding environment parameters, and human-machine function allocation core parameters; dividing the sample dataset into a training set and a test set, using the training set to train the preset initial scheme generation model, and using the test set to test the trained preset initial scheme generation model, and finally using the trained preset initial scheme generation model that meets the test conditions as the preset scheme prediction model.

[0028] Specifically, during model training, a pre-defined initial scheme generation model is first constructed, followed by the acquisition of a sample dataset. The dataset is ensured to contain all necessary files. The data is then converted to a format understandable by the pre-defined initial scheme generation model. Finally, the model is trained and tested. Specifically, the dataset can be divided first: using randomness or a specific strategy (such as stratified sampling), the sample dataset is divided into training and test sets. The training set is then used to train the model, and the test set is used to test the trained model, evaluating its performance on unseen data. Precision, recall, and other metrics on the test set are calculated and recorded. If the model performance does not meet requirements, it can return to the training phase for further iterations or adjustments. This process yields a pre-defined scheme generation model that meets the requirements. The pre-defined scheme generation model includes an input layer, an embedding layer, a feature extraction layer, and a scheme prediction layer. In the process of predicting task planning schemes using a preset scheme prediction model, firstly, multi-dimensional workload, the parameters of the task's surrounding environment, and the core parameters of human-machine function allocation are input into the preset scheme prediction model through an input layer. An embedding layer performs low-dimensional densification processing on the input parameter pairs while preserving semantic relevance. A feature extraction layer extracts the parameter features of the output parameters from the embedding layer. Finally, a scheme prediction layer predicts the scheme based on the output features of the feature extraction layer, resulting in multiple candidate task planning schemes. This embodiment of the invention plans tasks by comprehensively analyzing the parameters of the task's surrounding environment, the core parameters of human-machine function allocation, visual load, auditory load, cognitive load, and psychomotor load. This ensures the rationality of task planning, avoids excessive workload for staff, and reduces the risk of fatigue operation. The model generates multiple schemes, avoiding the limitations of a single scheme and improving adaptability to complex scenarios. The entire task planning process of this invention's complex human-machine system requires no human intervention, thereby improving the efficiency and rationality of task planning.

[0029] 104. Determine the evaluation value of each candidate task planning scheme, and based on the evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

[0030] In this embodiment of the invention, after predicting multiple candidate task planning schemes, a comprehensive analysis is performed on indicators such as task completion time, workload, and task execution risk. Based on this comprehensive analysis result, the final task planning scheme for executing the target human-computer interaction task is selected from among the multiple candidate task plans. By comprehensively analyzing indicators such as task completion time, workload, and task execution risk when determining the task execution scheme, this embodiment of the invention ensures the rationality of the task planning, thereby guaranteeing the stable and safe execution of the task.

[0031] Furthermore, in order to achieve reasonable task planning, when the environment or task execution time changes, the task needs to be replanned. Based on this, the method includes: real-time monitoring of whether the environmental parameters around the task and the task time window have changed; if so, the target human-computer interaction task is replanned.

[0032] Specifically, when environmental factors such as lighting and visual conditions, gravity, terrain, temperature, and communication status change, or when the task execution time window changes, it is necessary to re-plan the tasks that have not yet been executed. This involves re-planning the task based on the current environmental parameters, task attribute parameters, and core human-machine function allocation parameters. This enables dynamic task planning, allowing the planned tasks to adapt to environmental and time constraints, thus ensuring the rationality and accuracy of the task planning.

[0033] The present invention provides a task planning method for complex human-machine systems. Compared with the current method of manually planning tasks based on minimizing task time, the present invention plans tasks by comprehensively analyzing parameters of the task's surrounding environment, core parameters of human-machine function allocation, visual load, auditory load, cognitive load, and psychomotor load. This ensures the rationality of task planning, avoids excessive workload on staff, and reduces the risk of fatigue operation. By generating multiple solutions through models, the limitations of a single solution are avoided, improving adaptability to complex scenarios. The entire task planning process for complex human-machine systems in the present invention does not require human intervention, thereby improving the efficiency and rationality of task planning.

[0034] Furthermore, to better illustrate the process of planning complex human-machine system tasks described above, and as a refinement and extension of the above embodiments, this invention provides another method for planning complex human-machine system tasks, such as... Figure 2 As shown, the method includes: 201. In response to the planning instructions of the target human-computer interaction task, the target human-computer interaction task is divided into multiple task nodes, and the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node are obtained.

[0035] Specifically, upon receiving a task planning instruction, the target human-computer interaction task is divided into multiple task nodes based on requirements such as function, operational phase, and the collaboration method between the operator and the intelligent system. For example, by function: the overall task is broken down into functional modules, such as navigation, communication, and scientific experiments; by operational phase: the task execution flow is divided into phases, such as preparation, execution, and completion; by the collaboration method between the operator and the intelligent system: it is determined which tasks are executed by the operator, which by the intelligent system, and the interaction points between them. Then, the task attribute parameters, surrounding environment parameters, and core parameters for human-computer function allocation for each task node are retrieved from the database.

[0036] 202. Based on task attribute parameters, determine the multi-dimensional workload of each task node, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload.

[0037] Specifically, based on task attribute parameters, the visual load, auditory load, cognitive load, and psychomotor load required to execute each task node are determined respectively.

[0038] 203. Determine the load feature vector corresponding to the multi-dimensional workload, the environmental feature vector corresponding to the task's surrounding environment parameters, and the function allocation feature vector corresponding to the core parameters of human-machine function allocation.

[0039] In the embodiments of the present invention, feature extraction models, such as CNN models, can be used to extract the load feature vector corresponding to the multi-dimensional workload, the environmental feature vector corresponding to the task's surrounding environmental parameters, and the function allocation feature vector corresponding to the core parameters of human-machine function allocation.

[0040] 204. Perform potential feature mining on the load feature vector, environmental feature vector, and functional allocation feature vector to obtain potential feature vectors.

[0041] For embodiments of the present invention, to make fuller use of data, it is necessary to extract more latent features. Therefore, step 204 specifically includes: taking any one of the load feature vector, the environment feature vector, and the function allocation feature vector as a target feature vector; determining the vector length and vector direction of the target feature vector; and determining the task scenario complexity of the target human-computer interaction task; determining the vector transformation parameters of the target feature vector based on the vector length; determining the length-direction coupling coefficient of the target feature vector based on the vector length and the vector direction; and determining the environment of the target feature vector based on the task scenario complexity. The perception factor, based on the length-direction coupling coefficient and the environmental perception factor, determines the vector weight of the target feature vector; based on the vector transformation parameters corresponding to the load feature vector, the environmental feature vector, and the function allocation feature vector, respectively, vector transformation is performed on the corresponding load feature vector, environmental feature vector, and function allocation feature vector; and based on the vector weights corresponding to the load feature vector, environmental feature vector, and function allocation feature vector, the transformed load feature vector, environmental feature vector, and function allocation feature vector are weighted and fused, and the weighted fusion result is used as the potential mining feature vector.

[0042] Specifically, the process of determining potential mining feature vectors is illustrated by taking any one of the following feature vectors—load feature vector, environment feature vector, and function allocation feature vector—as a target feature vector. First, the vector length and direction of the target feature vector are determined, along with the task scenario complexity of the target human-computer interaction task. In determining task scenario complexity, the diversity of environmental factors, the clarity and diversity of task objectives, the differences in user characteristics, the richness of interaction methods, and the constraints of system resources within the task scenario are first determined. The diversity of environmental factors is determined by factors such as lighting conditions, background noise, and spatial layout of the task scenario. The clarity and diversity of task objectives are determined by the number of task objectives, their hierarchical structure and interdependencies, whether the objectives are clear and quantifiable, and whether there are conflicts or differences in priority. The differences in user characteristics are determined by the user's skill level, experience background, cognitive ability, and preferences. The richness of interaction methods is determined by the supported interaction modes, such as voice, gestures, touch, and their combinations. The constraints of system resources are determined by hardware conditions such as computing resources, storage capacity, and network bandwidth. Next, scores were assigned to the diversity of environmental factors, the clarity and diversity of task objectives, the differences in user characteristics, the richness of interaction methods, and the constraints of system resources. Corresponding scores were obtained for each of these factors: diversity of environmental factors, clarity and diversity of task objectives, differences in user characteristics, richness of interaction methods, and constraints of system resources. Weighting coefficients were then determined for each of these scores based on actual needs. These weighting coefficients were then used to calculate a weighted sum of the scores to obtain a comprehensive scenario score. Finally, the complexity of the target human-computer interaction task was determined based on the comprehensive scenario score; for example, a higher comprehensive scenario score indicates greater task scenario complexity.

[0043] Furthermore, in determining the target feature vector vector length and vector direction Then, the length-direction coupling coefficient is determined according to the following formula. :

[0044] in, The balancing weight coefficients are set according to actual needs. To control the length-direction coupling coefficient Parameters for determining the steepness of a logic function This is the offset parameter for the angle cosine value. Therefore, the length-direction coupling coefficients of the load characteristic vector, environmental characteristic vector, and functional allocation characteristic vector can be determined using the above method. Simultaneously, the environmental perception factor is determined using the following formula. :

[0045] in, For the complexity of the task scenario, An adjustment factor set according to actual needs, used to control the overall adjustment range. The steepness parameter, set according to actual needs, affects the rate of change of the function. This represents the baseline complexity or threshold parameter for the scene. Further, the target feature vector is determined according to the following formula. Vector weights :

[0046] in, Vector length The attenuation coefficient. Therefore, following the above method, the vector weights of the load characteristic vector, environmental characteristic vector, and functional allocation characteristic vector can be determined respectively.

[0047] Further, the sum of squares of the vector lengths corresponding to the load feature vector, environmental feature vector, and function allocation feature vector are determined. The ratio of the square of the vector length of the load feature vector to the sum of squares is used as the vector transformation parameter for the load feature vector, the ratio of the square of the vector length of the environmental feature vector to the sum of squares is used as the vector transformation parameter for the environmental feature vector, and the ratio of the square of the vector length of the function allocation feature vector to the sum of squares is used as the vector transformation parameter for the function allocation feature vector. The vector transformation parameter is multiplied by the corresponding feature vector, and the product is used as the transformed feature vector of the corresponding feature vector. Finally, based on the vector weights of the load feature vector, environmental feature vector, and function allocation feature vector, the corresponding transformed feature vectors are weighted and fused to obtain the potential mining feature vector.

[0048] This invention generates new feature combinations by mining latent features from load feature vectors, environmental feature vectors, and functional allocation feature vectors. These combined features may contain complex nonlinear relationships between the original features, enabling the model to capture more refined and richer information in the data. In other words, it can make full use of the relationships between various data, extract more latent features, make fuller use of data, and obtain more accurate prediction results for the subsequent schemes, thus meeting the needs of practical application scenarios.

[0049] 205. Input the potential mining feature vector into the preset scheme prediction model to predict the planning scheme and obtain multiple candidate task planning schemes for the target human-computer interaction task.

[0050] Specifically, the potential mining feature vectors are directly input into the preset scheme prediction model, and the preset scheme prediction model outputs multiple candidate task planning schemes.

[0051] In another embodiment of the present invention, the variant modeling mechanism of SysML (Systems Modeling Language) can be used to pass the parameter characteristics of the parent module to the child module through generalization, and generate multiple candidate task planning schemes by adjusting the parameter values ​​of the child module. Task characteristics include time windows and resource constraints, while child module parameter values ​​include task order and function allocation methods.

[0052] 206. Determine the evaluation value of each candidate task planning scheme, and based on the evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

[0053] In this embodiment of the invention, to select the optimal task planning scheme from multiple candidate task planning schemes, it is first necessary to determine the scheme evaluation value of each candidate task planning scheme. Based on this, step 206 specifically includes: taking any candidate task planning scheme in each of the candidate task planning schemes as a target candidate task planning scheme, performing a weighted summation of the multi-dimensional workload to obtain the node workload of each task node, and determining the comprehensive workload of the target candidate task planning scheme based on the node workload; determining the task execution time and task execution risk of the target candidate task planning scheme, and determining the correlation between the comprehensive workload, the task execution time, and the task execution risk; determining the weight coefficients corresponding to the comprehensive workload, the task execution time, and the task execution risk respectively, and aggregating the comprehensive workload, the task execution time, and the task execution risk based on the correlation and the weight coefficients, and determining the scheme evaluation value of the target candidate task planning scheme based on the aggregation result.

[0054] Specifically, for each task node, weighting coefficients for visual load, auditory load, cognitive load, and psychomotor load are determined according to actual needs. Based on these weighting coefficients, the visual load, auditory load, cognitive load, and psychomotor load are weighted and summed to obtain the node workload of each task node. The sum of the node workloads of multiple task nodes constituting a candidate task planning scheme is taken as the comprehensive workload of the corresponding candidate task planning scheme. Furthermore, it is necessary to determine the correlation coefficients between the comprehensive workload, task execution time, and task execution risk. Taking the calculation of the correlation coefficient between the comprehensive workload W and the task execution time T as an example, the covariance of the comprehensive workload W and the task execution time T in all selected task planning schemes is first determined. And determine the standard deviation of the comprehensive workload W respectively. and the standard deviation of task execution time T Then, the correlation coefficient between the overall workload and task execution time is calculated according to the following formula. :

[0055] Therefore, the correlation coefficients between overall workload and task execution risk, and between task execution time and task execution risk, can be calculated using the above method. For each candidate task planning scheme, taking any one of the candidate task planning schemes as a target candidate task planning scheme as an example, the overall workload, task execution time, and task execution risk corresponding to the target candidate task planning scheme are scored, resulting in the overall workload score, task execution time score, and task execution risk score for that candidate task planning scheme. Furthermore, the scheme evaluation value of target candidate task planning scheme i is calculated according to the following formula. :

[0056] Where m and j are the identifiers of the overall workload score, task execution time score, and task execution risk score, and n is the number of the overall workload score, task execution time score, and task execution risk score, which is 3. To determine the correlation coefficient between any two of the comprehensive workload, task execution time, and task execution risk, and The comprehensive workload score, task execution time score, and task execution risk score are any two of the following: The formula above means that, based on the weighted coefficients corresponding to the comprehensive workload, task execution time, and task execution risk, the comprehensive workload score, task execution time score, and task execution risk score are weighted and summed to obtain a weighted sum. The comprehensive workload score is multiplied by the task execution time score, and the product is multiplied by the corresponding correlation coefficient to obtain the first product. The comprehensive workload score is multiplied by the task execution risk score, and the product is multiplied by the corresponding correlation coefficient to obtain the second product. The task execution time score is multiplied by the task execution risk score, and the product is multiplied by the corresponding correlation coefficient to obtain the third product. The first, second, and third products are added together to obtain the correlation sum. Finally, the correlation sum and the weighted sum are added together to obtain the evaluation value of the corresponding candidate task planning scheme. Therefore, the evaluation value of each candidate task planning scheme can be determined according to the above method.

[0057] Furthermore, schemes with evaluation values ​​exceeding a preset evaluation threshold are selected as the final task planning schemes for performing the target human-computer interaction task. The preset evaluation threshold is set based on actual needs.

[0058] Furthermore, each candidate task planning scheme can be visualized for staff to select and understand. Based on this, the method includes: determining the tools required to execute each candidate task planning scheme, the estimated task completion time, and the estimated workload; ranking each candidate task planning scheme based on the scheme evaluation value; and displaying each ranked candidate task planning scheme on a human-computer interaction interface based on the required tools, the estimated task completion time, and the estimated workload.

[0059] Specifically, the multi-dimensional workload of each task node is weighted and summed to obtain the node workload of each task node. For each candidate task planning scheme, the node workloads of each task node constituting that candidate task planning scheme are summed to obtain the expected workload of that candidate task planning scheme. Then, based on the scheme evaluation values, each candidate task planning scheme is sorted, for example, in descending order of evaluation values. Finally, the required tools, expected task completion time, and expected workload are used as labels to display each candidate task planning scheme in the order listed. This embodiment of the invention, through the labeling of required tools, expected task completion time, and expected workload, and the sorting mechanism of scheme evaluation values, can quickly filter out the optimal scheme, providing decision-makers with data-driven information references and reducing subjective judgment errors.

[0060] According to another complex human-machine system task planning method provided by the present invention, compared with the current method of manually planning tasks based on minimizing task time, the present invention plans tasks by comprehensively analyzing the surrounding environmental parameters, core parameters of human-machine function allocation, visual load, auditory load, cognitive load, and psychomotor load. This ensures the rationality of task planning, avoids excessive workload on staff, and reduces the risk of fatigue operation. By generating multiple solutions through models, the limitations of a single solution are avoided, and the adaptability to complex scenarios is improved. The entire complex human-machine system task planning process of the present invention does not require human intervention, thereby improving the efficiency and rationality of task planning.

[0061] Furthermore, as Figure 1 In specific implementation, embodiments of the present invention provide a task planning device for complex human-machine systems, such as... Figure 3 As shown, the device includes: an acquisition unit 31, a determination unit 32, a scheme generation unit 33, and a scheme selection unit 34.

[0062] The acquisition unit 31 can be used to respond to the planning instructions of the target human-computer interaction task, divide the target human-computer interaction task into multiple task nodes, and acquire the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node.

[0063] The determining unit 32 can be used to determine the multi-dimensional workload of each task node based on the task attribute parameters, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload.

[0064] The scheme generation unit 33 can be used to generate multiple candidate task planning schemes based on the multi-dimensional workload, the task surrounding environment parameters, and the human-machine function allocation core parameters, using a preset scheme prediction model.

[0065] The scheme selection unit 34 can be used to determine the scheme evaluation value of each candidate task planning scheme, and based on the scheme evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

[0066] In specific application scenarios, in order to determine the evaluation value of each candidate task planning scheme, such as Figure 4 As shown, the scheme selection unit 34 includes a weighting module 341, a first determination module 342, and an aggregation module 343.

[0067] The weighting module 341 can be used to take any candidate task planning scheme in each of the candidate task planning schemes as a target candidate task planning scheme, perform weighted summation on the multi-dimensional workload to obtain the node workload of each task node, and determine the comprehensive workload of the target candidate task planning scheme based on the node workload.

[0068] The first determining module 342 can be used to determine the task execution time and task execution risk of the target candidate task planning scheme, and to determine the correlation between the comprehensive workload, the task execution time and the task execution risk.

[0069] The aggregation module 343 can be used to determine the weight coefficients corresponding to the comprehensive workload, the task execution time, and the task execution risk, respectively. Based on the correlation and the weight coefficients, the comprehensive workload, the task execution time, and the task execution risk are aggregated, and the scheme evaluation value of the target candidate task planning scheme is determined based on the aggregation result.

[0070] In specific application scenarios, in order to generate multiple candidate task planning schemes using a preset scheme prediction model, the scheme generation unit 33 includes a second determination module 331, a feature mining module 332, and a scheme prediction module 333.

[0071] The second determining module 331 can be used to determine the load feature vector corresponding to the multi-dimensional workload, the environmental feature vector corresponding to the task surrounding environment parameters, and the function allocation feature vector corresponding to the human-machine function allocation core parameters.

[0072] The feature mining module 332 can be used to perform potential feature mining on the load feature vector, the environment feature vector, and the function allocation feature vector to obtain potential mined feature vectors.

[0073] The scheme prediction module 333 can be used to input the potential mining feature vector into the preset scheme prediction model to predict the planning scheme, and obtain multiple candidate task planning schemes for the target human-computer interaction task.

[0074] In specific application scenarios, to determine potential mining feature vectors, the feature mining module 332 can be specifically used to take any one of the load feature vector, the environment feature vector, and the function allocation feature vector as a target feature vector, determine the vector length and vector direction of the target feature vector, and determine the task scenario complexity of the target human-computer interaction task; determine the vector transformation parameters of the target feature vector based on the vector length, determine the length-direction coupling coefficient of the target feature vector based on the vector length and the vector direction, determine the environmental perception factor of the target feature vector based on the task scenario complexity, and determine the vector weight of the target feature vector based on the length-direction coupling coefficient and the environmental perception factor; perform vector transformation on the corresponding load feature vector, environment feature vector, and function allocation feature vector based on the vector transformation parameters corresponding to the load feature vector, environment feature vector, and function allocation feature vector, and perform weighted fusion on the vector-transformed load feature vector, environment feature vector, and function allocation feature vector based on the vector weights corresponding to the load feature vector, environment feature vector, and function allocation feature vector, and use the weighted fusion result as the potential mining feature vector.

[0075] In specific application scenarios, in order to replan the tasks of complex human-machine systems, the device also includes a task replanning unit 35.

[0076] The task replanning unit 35 can be used to monitor in real time whether the environmental parameters of the task's surrounding environment and the task's time window have changed. If so, the target human-computer interaction task will be replanned.

[0077] In specific application scenarios, the device also includes a task display unit 36 ​​for displaying task plans.

[0078] The task display unit 36 ​​can be used to determine the tools required to execute each candidate task planning scheme, the estimated task completion time, and the estimated workload, and to rank each candidate task planning scheme based on the scheme evaluation value; and to display each ranked candidate task planning scheme in the human-computer interaction interface based on the required tools, the estimated task completion time, and the estimated workload.

[0079] In specific application scenarios, in order to construct a prediction model of a preset scheme, the device also includes a model building unit 37.

[0080] The model building unit 37 can be used to build a preset initial scheme generation model; obtain a sample dataset, wherein the sample dataset includes multi-dimensional workload of sample tasks with task planning scheme labels, task surrounding environment parameters, and human-machine function allocation core parameters; divide the sample dataset into a training set and a test set, use the training set to train the preset initial scheme generation model, and use the test set to test the trained preset initial scheme generation model, and finally use the trained preset initial scheme generation model that meets the test conditions as the preset scheme prediction model.

[0081] It should be noted that other corresponding descriptions of the functional modules involved in the complex human-machine system task planning device provided in this embodiment of the invention can be found in [reference]. Figure 1 The corresponding descriptions of the methods shown will not be repeated here.

[0082] Based on the above, Figure 1 Accordingly, this embodiment of the invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the following steps: responding to a planning instruction for a target human-computer interaction task, dividing the target human-computer interaction task into multiple task nodes, and obtaining task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters for each task node; based on the task attribute parameters, determining the multi-dimensional workload of each task node, wherein the multi-dimensional workload includes at least two of visual workload, auditory workload, cognitive workload, and psychomotor workload; based on the multi-dimensional workload, the task surrounding environment parameters, and the human-computer function allocation core parameters, generating multiple candidate task planning schemes using a preset scheme prediction model; determining the scheme evaluation value of each candidate task planning scheme, and based on the scheme evaluation value, selecting the final task planning scheme for executing the target human-computer interaction task from each candidate task planning scheme.

[0083] Based on the above, Figure 1 The method shown and as Figure 3 The embodiment of the device shown in the invention also provides a physical structure diagram of a computer device, such as... Figure 5As shown, the computer device includes: a processor 41, a memory 42, and a computer program stored in the memory 42 and executable on the processor. Both the memory 42 and the processor 41 are mounted on a bus 43. When the processor 41 executes the program, it performs the following steps: In response to a planning instruction for a target human-computer interaction task, it divides the target human-computer interaction task into multiple task nodes and obtains task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters for each task node; Based on the task attribute parameters, it determines the multi-dimensional workload of each task node, wherein the multi-dimensional workload includes at least two of visual workload, auditory workload, cognitive workload, and psychomotor workload; Based on the multi-dimensional workload, the task surrounding environment parameters, and the human-computer function allocation core parameters, it generates multiple candidate task planning schemes using a preset scheme prediction model; It determines the scheme evaluation value of each candidate task planning scheme and, based on the scheme evaluation value, selects the final task planning scheme for executing the target human-computer interaction task from each candidate task planning scheme.

[0084] Through the technical solution of this invention, the present invention plans tasks by comprehensively analyzing the parameters of the task's surrounding environment, the core parameters of human-machine function allocation, visual load, auditory load, cognitive load, and psychomotor load. This ensures the rationality of task planning, avoids the workload of staff, and reduces the risk of fatigue operation. By generating multiple solutions through the model, the limitations of a single solution are avoided, and the adaptability to complex scenarios is improved. The entire task planning process of the complex human-machine system of this invention does not require human intervention, thereby improving the efficiency and rationality of task planning.

[0085] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. 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 task planning method for complex human-machine systems, characterized in that, include: In response to the planning instructions of the target human-computer interaction task, the target human-computer interaction task is divided into multiple task nodes, and the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node are obtained. Based on the task attribute parameters, the multi-dimensional workload of each task node is determined, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload. Based on the multi-dimensional workload, the task's surrounding environment parameters, and the core parameters of human-machine function allocation, a variety of candidate task planning schemes are generated using a preset scheme prediction model. Determine the evaluation value of each candidate task planning scheme, and based on the evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

2. The method according to claim 1, characterized in that, The process of determining the evaluation value of each candidate task planning scheme includes: Each candidate task planning scheme is taken as a target candidate task planning scheme, and the multi-dimensional workload is weighted and summed to obtain the node workload of each task node. Based on the node workload, the comprehensive workload of the target candidate task planning scheme is determined. Determine the task execution time and task execution risk of the target candidate task planning scheme, and determine the correlation between the overall workload, the task execution time, and the task execution risk; The weight coefficients corresponding to the overall workload, the task execution time, and the task execution risk are determined respectively. Based on the correlation and the weight coefficients, the overall workload, the task execution time, and the task execution risk are aggregated. Based on the aggregation result, the scheme evaluation value of the target candidate task planning scheme is determined.

3. The method according to claim 1, characterized in that, Based on the multi-dimensional workload, the task's surrounding environment parameters, and the core parameters of human-machine function allocation, a preset prediction model is used to generate multiple candidate task planning schemes, including: Determine the load feature vector corresponding to the multi-dimensional workload, the environmental feature vector corresponding to the task's surrounding environment parameters, and the function allocation feature vector corresponding to the human-machine function allocation core parameters, respectively. Potential feature mining is performed on the load feature vector, the environment feature vector, and the function allocation feature vector to obtain potential mined feature vectors; The potential mining feature vector is input into the preset scheme prediction model to predict the planning scheme, thereby obtaining multiple candidate task planning schemes for the target human-computer interaction task.

4. The method according to claim 3, characterized in that, The process of performing latent feature mining on the load feature vector, the environment feature vector, and the function allocation feature vector to obtain latent feature vectors includes: Take any one of the load feature vector, the environment feature vector, and the function allocation feature vector as a target feature vector, determine the vector length and vector direction of the target feature vector, and determine the task scenario complexity of the target human-computer interaction task. The vector transformation parameters of the target feature vector are determined based on the vector length, the length-direction coupling coefficient of the target feature vector is determined based on the vector length and the vector direction, the environmental perception factor of the target feature vector is determined based on the task scenario complexity, and the vector weight of the target feature vector is determined based on the length-direction coupling coefficient and the environmental perception factor. Based on the vector transformation parameters corresponding to the load feature vector, the environment feature vector, and the function allocation feature vector, vector transformation is performed on the corresponding load feature vector, environment feature vector, and function allocation feature vector. Based on the vector weights corresponding to the load feature vector, the environment feature vector, and the function allocation feature vector, the transformed load feature vector, environment feature vector, and function allocation feature vector are weighted and fused, and the weighted fusion result is used as the potential mining feature vector.

5. The method according to claim 1, characterized in that, After selecting the final task planning scheme for performing the target human-computer interaction task from each of the candidate task planning schemes based on the scheme evaluation value, the method further includes: The system monitors in real time whether the environmental parameters surrounding the task and the task time window change. If so, the target human-computer interaction task is replanned.

6. The method according to claim 1, characterized in that, After determining the evaluation value of each candidate task planning scheme, the method further includes: Determine the tools required to execute each candidate task planning scheme, the estimated task completion time, and the estimated workload, and rank each candidate task planning scheme based on the scheme evaluation value; Based on the required tools, the estimated task completion time, and the estimated workload, the planning schemes for each of the ranked candidate tasks are displayed in the human-computer interaction interface.

7. The method according to claim 1, characterized in that, Before generating multiple candidate task planning schemes using a preset prediction model based on the multi-dimensional workload, the task's surrounding environment parameters, and the core parameters of human-machine function allocation, the method further includes: Construct a model based on a pre-defined initial scheme; Obtain a sample dataset, wherein the sample dataset includes multi-dimensional workload, task surrounding environment parameters, and human-machine function allocation core parameters of sample tasks labeled with task planning schemes; The sample dataset is divided into a training set and a test set. The training set is used to train the preset initial scheme generation model, and the test set is used to test the trained preset initial scheme generation model. Finally, the trained preset initial scheme generation model that meets the test conditions is used as the preset scheme prediction model.

8. A task planning device for complex human-machine systems, characterized in that, include: The acquisition unit is used to respond to the planning instructions of the target human-computer interaction task, divide the target human-computer interaction task into multiple task nodes, and acquire the task attribute parameters, task surrounding environment parameters, and human-computer function allocation core parameters of each task node. The determining unit is configured to determine the multi-dimensional workload of each task node based on the task attribute parameters, wherein the multi-dimensional workload includes at least two of the following: visual workload, auditory workload, cognitive workload, and psychomotor workload. The scheme generation unit is used to generate multiple candidate task planning schemes based on the multi-dimensional workload, the task surrounding environment parameters, and the human-machine function allocation core parameters, using a preset scheme prediction model. The scheme selection unit is used to determine the scheme evaluation value of each candidate task planning scheme, and based on the scheme evaluation value, select the final task planning scheme for performing the target human-computer interaction task from each candidate task planning scheme.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images