Behavior plan generation method and device based on man-machine cooperation

Abstract

The embodiment of the invention provides a behavior plan generation method and device based on man-machine cooperation, and the method comprises the steps: receiving project information, carrying out the semantic reasoning of the received project information based on an ontology library, and decomposing a project into a plurality of executable sub-projects; evaluating a current project environment, preconditions and constraints based on historical data in a case library, and recommending appropriate strategies and behavior arrangements for each sub-project; allocating resources required by each sub-project in the project execution process based on the resource library, and outputting a resource allocation scheme; generating an execution sequence of each sub-project according to the front-back dependency relationship of the project, and dynamically adjusting the execution sequence according to the state change; and based on the dynamic manual intervention threshold, adjusting the priority, resource allocation and execution strategy of the project through a feedback mechanism, and generating a behavior plan. According to the scheme, the efficiency and accuracy of project formulation and execution can be effectively improved through a machine autonomous planning and human-assisted behavior plan formulation mode.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of project planning, and in particular, to a human-machine collaboration-based behavior plan generation method and device and a computer readable storage medium storing a computer program. BACKGROUND

[0002] In the execution process of modern complex projects, the formulation of behavior plans plays a crucial role. Traditional behavior plan formulation methods rely mainly on human guidance, with artificial experts filtering, judging and reasoning information to produce decisions. Machines are usually used as data sources and computing tools to support human decision-making. However, this "human-led, machine-assisted" mode has several obvious technical challenges. For example, machines mainly provide data processing and computing support in traditional processes, but they cannot understand the basis and reasons for human decision-making, making it difficult to fully exploit the potential of data in complex project planning. Especially in project environments that require efficient and accurate decision-making, artificial intelligence lacks sufficient intelligent reasoning and decision support capabilities, and the efficiency and quality of plan formulation cannot be maximized.

[0003] In the existing behavior plan formulation process, project decomposition and behavior sequence arrangement mainly rely on human understanding of complex situations. Although machines can help analyze data, the understanding of project details and their reasonable decomposition still rely on human experience, which leads to inaccurate project division or failure to fully consider the mutual relationship between projects and the logic of execution. As the complexity of projects increases, traditional solutions cannot effectively handle the fusion and processing of multi-source heterogeneous data. The collaborative use of multiple data sources such as historical case libraries and knowledge bases still cannot achieve autonomous and efficient behavior plan formulation, especially when decisions need to be made under multiple conditions. The comprehensive analysis and reasoning capabilities of machines for different information are relatively weak.

[0004] Although machines can handle large amounts of data, human intervention is still required when dealing with high-risk or high-uncertainty projects. In traditional solutions, human intervention is often limited by processes, technology and experience, and it is difficult to provide real-time feedback and adjustments when machines make decisions. How to find a balance between machines and humans in the behavior plan formulation process to ensure efficient collaboration between the two is a bottleneck of traditional solutions. SUMMARY

[0005] The embodiments described herein provide a human-machine collaboration-based behavior plan generation method and device and a computer readable storage medium storing a computer program, which maximizes the advantages of artificial intelligence and machine intelligence while ensuring project execution efficiency and accuracy, improving overall efficiency and reliability.

[0006] According to a first aspect of this disclosure, a method for generating behavior plans based on human-computer collaboration is provided, comprising: receiving project information; performing semantic reasoning on the received project information based on an ontology library to decompose the project into multiple executable sub-projects; evaluating the current project environment, preconditions, and constraints based on historical data in a case library, and recommending appropriate strategies and behavioral arrangements for each sub-project; allocating resources required by each sub-project during project execution based on a resource library, and outputting a resource allocation scheme; generating an execution sequence for each sub-project based on the project's dependencies, and dynamically adjusting the execution sequence according to state changes; and adjusting the project's priority, resource allocation, and execution strategy through a feedback mechanism based on a dynamic human intervention threshold to generate a behavior plan.

[0007] In some embodiments of this disclosure, receiving project information and performing semantic reasoning on the received project information based on an ontology library to decompose the project into multiple executable sub-projects includes: extracting project objectives, time constraints, resource requirements, and status requirements from the received project information using natural language processing technology; standardizing the extracted information through ontology mapping based on a unified ontology representation in the ontology library; and analyzing the inherent relationships of the project through semantic reasoning, and decomposing the project into multiple executable sub-projects according to the project's complexity and specific objectives.

[0008] In some embodiments of this disclosure, evaluating the current project environment, preconditions, and constraints based on historical data in the case library, and recommending appropriate strategies and behavioral arrangements for each sub-project includes: automatically recommending multiple feasible behavioral arrangements for each sub-project based on similar cases in the historical case library; checking whether the current project environment meets the preconditions and subsequent states required for the recommended behavioral arrangements, and adjusting and optimizing the behavioral arrangements if not; arranging a sequence of behavioral executions according to the priority, dependencies, and resource requirements of the sub-projects, and verifying whether the sequence of behavioral executions meets the time, resource, and project requirements; and adjusting the order of behaviors according to the verification results, and re-verifying the constraints after each adjustment to ensure that all constraints are met.

[0009] In some embodiments of this disclosure, the resource allocation scheme for each sub-project during project execution is based on a resource library. This includes: during resource allocation, calculating the impact factor of each sub-project on the overall completion time to determine which projects need to be prioritized; using fuzzy logic to convert resource usage into fuzzy values; adjusting project priorities based on these fuzzy values; and adjusting the execution order and priority of projects based on load conditions. Furthermore, using virtualization technology, the resource requirements of multiple projects are over-configured to reserve resources for high-priority projects. When a high-priority project is executed, low-priority projects are suspended or delayed until the high-priority project is completed, at which point resources are returned to the low-priority projects. Finally, by parsing logs during project execution, abnormal resource utilization is detected in real time. When abnormal resource utilization occurs, the dependencies between projects are reorganized and adjusted based on resource bottlenecks. A knowledge graph between projects, resources, and constraints is constructed, and the resource allocation scheme is adjusted through semantic analysis within the knowledge graph.

[0010] In some embodiments of this disclosure, generating the execution sequence of each sub-project based on the project's dependencies and dynamically adjusting the execution sequence according to state changes includes: constructing a directed acyclic graph to describe the dependencies, resource allocation, and dynamic changes between projects, performing topological sorting on the project graph to obtain the project execution order; or, transforming the objectives of minimizing completion time, reducing resource usage, and reducing energy consumption into constraints, and solving the optimal scheduling scheme based on mixed-integer linear programming; or, encoding project dependencies as chromosomes, performing crossover and mutation operations in a genetic algorithm, and optimizing the project execution order and resource allocation by selecting chromosomes with high fitness.

[0011] In some embodiments of this disclosure, the generation of a behavior plan, based on a dynamic human intervention threshold, involves adjusting the project's priority, resource allocation, and execution strategy through a feedback mechanism. This includes: dynamically adjusting the human intervention threshold based on the threat context, the confidence level of the machine's decision, and the matching degree of past experience; when the intervention threshold is triggered, feeding back the human modifications and adjustments to the behavior plan to the machine, and using reinforcement learning and meta-learning algorithms to optimize the dynamic intervention threshold and decision-making mechanism; and forming a dynamic autonomous arbitration closed loop through situational awareness and data fusion, decision confidence assessment, autonomous boundary arbitration, feedback, and knowledge updates.

[0012] In some embodiments of this disclosure, dynamically adjusting the human intervention threshold based on the threat context, the confidence level of the machine's decision, and the matching degree of past experience includes: in low-risk operations, the machine autonomously executes the decision; for high-risk operations, experts adjust the machine's strategy or intervene directly through gesture recognition or natural language commands; the machine's decision confidence level is assessed in real time, and when the confidence level is high, the machine executes autonomously, while when the confidence level is low, human intervention is initiated; based on past experience, the similarity between the current situation and past projects is analyzed, and if the current project is similar to historical projects, the confidence level of the decision is increased and the human intervention threshold is reduced, otherwise the human intervention threshold is increased.

[0013] In some embodiments of this disclosure, a dynamic autonomous arbitration closed loop is formed through situational awareness and data fusion, decision confidence assessment, autonomy boundary arbitration, feedback, and knowledge updating. This includes: the machine collects and fuses data in real time through sensor networks, log analysis, and threat intelligence platform tools to form a multi-dimensional situational awareness map; humans review key nodes through a visual interface and supplement business context information; the machine uses Bayesian neural networks and Monte Carlo methods to assess decision confidence and quantify the uncertainty of predictions; humans generate human confidence levels by comparing historical case libraries and verifying business logic; and an arbitration function is applied to determine the attribution of decision-making power. The arbitration function is: , in, , For dynamic threshold parameters, For human confidence level, To ensure the credibility of the decision, Let be the objective function. , , These represent machine decisions, human decisions, and historical decisions, respectively. The machine records each decision and its execution result, uses reinforcement learning to update the policy network, and manually labels misjudged cases to trigger incremental training of the case library and rule library.

[0014] According to a second aspect of this disclosure, a behavior plan generation device based on human-computer collaboration is provided. The device includes at least one processor and at least one memory storing a computer program. When the computer program is executed by the at least one processor, the device causes the following actions: receiving project information; performing semantic reasoning on the received project information based on an ontology library to decompose the project into multiple executable sub-projects; evaluating the current project environment, preconditions, and constraints based on historical data in a case library, and recommending appropriate strategies and behavioral arrangements for each sub-project; allocating resources required by each sub-project during project execution based on a resource library, and outputting a resource allocation scheme; generating an execution sequence for each sub-project based on the project's dependencies, and dynamically adjusting the execution sequence according to state changes; and adjusting the project's priority, resource allocation, and execution strategy through a feedback mechanism based on a dynamic human intervention threshold to generate a behavior plan.

[0015] According to a third aspect of this disclosure, a computer-readable storage medium storing a computer program is provided, wherein the computer program, when executed by a processor, implements the steps of the human-computer collaboration-based behavior plan generation method according to a first aspect of this disclosure.

[0016] The behavior plan generation method and apparatus based on human-machine collaboration provided in this disclosure can effectively improve the efficiency and accuracy of project formulation and execution through a behavior plan formulation mode that combines machine autonomous planning with human assistance. By fully utilizing knowledge bases and intelligent reasoning, the machine plays a leading role in project analysis and decision-making, while humans perform evaluation and optimization in the final stage, jointly completing the efficient behavior plan formulation. By setting dynamic thresholds, the option of human intervention can be introduced during project execution, adjusting the plan according to external changes or unforeseen circumstances. This retains the efficiency of automatic plan generation by the machine while combining the flexibility of human intervention, ensuring that the generated behavior plan not only has high execution efficiency but also better meets the expected goals. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments will be briefly described below. It should be understood that the drawings described below only relate to some embodiments of this disclosure and are not intended to limit this disclosure, wherein: Figure 1 This is an exemplary flowchart of a human-computer collaboration-based behavior plan generation method according to embodiments of the present disclosure; Figure 2 This is a schematic block diagram of a human-computer collaboration-based behavior plan generation apparatus according to embodiments of the present disclosure.

[0018] It should be noted that the elements in the attached diagram are schematic and not drawn to scale. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the described embodiments of this disclosure without creative effort are also within the scope of protection of this disclosure.

[0020] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this subject matter pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having meanings consistent with their meanings in the context of the specification and in the relevant art, and shall not be interpreted in an idealized or overly formal form unless otherwise explicitly defined herein. Furthermore, terms such as “first” and “second” are used only to distinguish one component (or part of a component) from another component (or another part of a component).

[0021] In cybersecurity behavior planning, the collaborative process between machines and humans requires efficient and flexible mechanisms to ensure the simultaneous utilization of automated systems and expert experience in complex and high-risk environments. This disclosure provides an adaptive behavior plan generation method based on a closed-loop mechanism of dynamic autonomous arbitration, utilizing knowledge bases, semantic reasoning, and deep analysis to achieve a seamless integration of machine autonomy and human intervention.

[0022] Figure 1 This is an exemplary flowchart of a human-computer collaboration-based behavior plan generation method according to embodiments of the present disclosure. Figure 1 At block S102 of method 100 shown, project information is received, semantic reasoning is performed on the received project information based on the ontology library, and the project is decomposed into multiple executable sub-projects.

[0023] According to one embodiment of this disclosure, core elements such as project objectives, time constraints, resource requirements, and status requirements can be extracted from received project information using natural language processing technology. Then, based on a unified ontology representation in an ontology library, the extracted information is standardized through ontology mapping. This unified representation using the ontology library allows the project to interface with and be associated with other projects and elements. Semantic reasoning is used to analyze the project's internal relationships, and the project is decomposed into multiple executable sub-projects based on its complexity and specific objectives. Each sub-project has a clearly defined starting state, target state, and potential resource requirements.

[0024] Subsequently, in step S104, the current project environment, preconditions, and constraints are evaluated based on historical data in the case library, and appropriate strategies and behavioral arrangements are recommended for each sub-project.

[0025] During project execution, the machine proactively searches for similar project execution cases in a historical case library and automatically recommends behavioral arrangements for each sub-project based on these similar cases. Case similarity can be measured by factors such as feature matching, project type, environmental conditions, and execution results. These behavioral arrangements consider the strategies, methods, tools used in historical cases, and their performance under specific environmental conditions. The machine checks whether the current project environment meets the prerequisites and subsequent states required for the recommended behavioral arrangements; if not, it adjusts and optimizes the behavioral arrangements.

[0026] Next, in box S106, the resources required by each sub-project during the project execution process are allocated based on the resource library, and a resource allocation plan is output.

[0027] Specifically, during resource allocation, calculating the impact factor of each sub-project on the overall completion time effectively determines which projects need to be prioritized. In the case of multiple projects running concurrently, some projects may be delayed due to resource constraints. In such situations, scheduling project execution based on Slack Time ensures that projects on the critical path are not delayed, thus avoiding impact on the overall project schedule.

[0028] By using fuzzy logic, resource usage is transformed into fuzzy values ​​(e.g., low, medium, high load). Project priorities are adjusted based on these fuzzy values, and the execution order and priority of projects are adjusted according to load conditions. Virtualization technology can be used to over-configure the resource requirements of multiple projects, thereby temporarily resolving resource shortages under high load conditions. For example, virtual memory and container technologies can "virtualize" physical resources, allowing multiple projects to share a resource pool. When certain resources are not fully utilized in the short term, they can be temporarily provided to other projects, optimizing resource utilization. Furthermore, resource reservation and preemption is an efficient scheduling mechanism, especially suitable for scenarios where high-priority projects need to preempt limited resources. Resources can be reserved for high-priority projects; when a high-priority project executes, low-priority projects are suspended or delayed until the high-priority project completes, at which point the resources are returned to the low-priority projects. This mechanism ensures that urgent projects are not delayed.

[0029] By analyzing logs during project execution, abnormal resource utilization can be detected in real time. When abnormal resource utilization occurs, the dependencies between projects are reorganized and adjusted based on resource bottlenecks. A knowledge graph is constructed between projects, resources, and constraints, and the resource allocation scheme is adjusted through semantic analysis in the knowledge graph.

[0030] By combining the aforementioned priority-driven conflict resolution, elastic resource scheduling, and data-driven adaptive mechanisms, resource utilization efficiency can be significantly improved, conflicts can be reduced, and the timely execution of high-priority projects can be ensured.

[0031] Subsequently, in step S108, the execution sequence of each sub-project is generated according to the project's dependencies, and the execution sequence is dynamically adjusted according to the state changes.

[0032] Graph-based scheduling optimization algorithms combine traditional graph algorithms, operations research methods, and modern intelligent algorithms to efficiently solve complex project scheduling problems. Different algorithms have different applicable scenarios and advantages. For example, classic graph algorithms are suitable for scheduling projects with clear dependencies and can effectively handle issues related to project execution order and parallelism. Mixed-integer programming and constraint solving are suitable for handling complex scheduling problems with multiple constraints and objectives, and can obtain optimal solutions. Intelligent optimization algorithms, by simulating natural evolution or learning processes, can perform intelligent scheduling and optimization in complex environments.

[0033] Specifically, a directed acyclic graph (DAG) can be constructed to describe the dependencies, resource allocation, and dynamic changes between projects. A topological sort of the project graph can then be performed to obtain the execution order of the projects. This topological sorting ensures that all dependencies are satisfied.

[0034] The objectives of minimizing completion time, reducing resource usage, and lowering energy consumption are transformed into constraints. The optimal scheduling scheme is then solved using mixed-integer linear programming (ILP). ILP is a special case of multi-objective linear programming that uses integer decision variables to represent discrete choices in project scheduling, such as whether a project should be executed or its order. ILP obtains the optimal scheduling decision through an optimized solution process.

[0035] Project dependencies are encoded as chromosomes, and crossover and mutation operations are performed in a genetic algorithm. By selecting chromosomes with high fitness, the execution order and resource allocation of projects are optimized. Through continuous iteration, the genetic algorithm explores the optimal project resource allocation method, maximizing the system efficiency.

[0036] These algorithms can be flexibly combined to address a variety of scheduling needs and challenges.

[0037] Finally, in step S110, based on the dynamic human intervention threshold, the project's priority, resource allocation, and execution strategy are adjusted through a feedback mechanism to generate a behavior plan.

[0038] Although machines can autonomously generate project plans, human intervention can still filter, adjust, and optimize the final action plan. Humans can modify the generated plan based on actual circumstances, providing suggestions or feedback to ensure its operability and effectiveness in actual execution. This allows for continuous improvement of the system.

[0039] According to one embodiment of this disclosure, the human intervention threshold is dynamically adjusted based on the threat context, the confidence level of the machine's decision, and the degree of matching with past experience. The dynamic intervention threshold defines a boundary where human intervention becomes necessary when factors such as project complexity, risk level, and system uncertainty exceed this threshold. Specifically, in low-risk operations, the machine autonomously executes decisions; for high-risk operations, experts adjust the machine's strategy or intervene directly through gesture recognition or natural language commands. For example, for low-risk operations (such as blocking known malicious IPs, applying patch templates, etc.), the machine can execute autonomously. For high-risk operations (such as blocking critical business ports, data erasure commands, etc.), human experts intervene through gesture recognition or natural language commands, adjusting strategy parameters and deciding whether to continue execution.

[0040] The machine's decision confidence level is assessed in real time, i.e., how confident the system is in executing the current action plan. When confidence is high, the machine can execute autonomously; when confidence is low, human intervention is initiated. Furthermore, the similarity between the current situation and past projects can be analyzed based on past experience. If the current project is similar to historical projects, the decision confidence level is increased, and human intervention is reduced; otherwise, the threshold for human intervention is raised.

[0041] When the intervention threshold is triggered, human modifications and adjustments to the behavior plan are fed back to the machine, and reinforcement learning and meta-learning algorithms are used to optimize the dynamic intervention threshold and decision-making mechanism. Reinforcement learning enables the system to learn the optimal decision path through continuous trial and error and strategy adjustment. For example, after executing a project multiple times, the system learns the optimal sequence of behaviors and decision-making strategies based on reward and punishment mechanisms. When the project or environment changes, meta-learning algorithms help the system adjust its learning strategy, enabling it to make effective decisions with less data and experience.

[0042] Through situational awareness and data fusion, decision confidence assessment, autonomy boundary arbitration, feedback, and knowledge updates, a dynamic autonomous arbitration closed loop is formed. For example, machines collect and fuse various types of data in real time using tools such as sensor networks, log analysis, and threat intelligence platforms (TIP). This data constructs a multi-dimensional situational awareness map, covering information such as asset topology, vulnerability distribution, and attack chain modeling. Experts use visual interfaces to review key nodes, such as zero-day exploit paths or lateral movement anomaly traffic, supplementing the machine-collected data with business context (such as core asset priorities and compliance requirements) to ensure that the information is more accurate and meets actual needs.

[0043] The machine uses a Bayesian neural network (BNN) and Monte Carlo Dropout to assess the credibility of decisions and quantify the uncertainty of predictions. Experts compare the results with a historical case library and verify them in conjunction with business logic to generate human confidence scores, further validating the machine's decisions cognitively.

[0044] The decision-making power is determined using an arbitration function, which is:

[0045] in, , For dynamic threshold parameters, For human confidence level, To ensure the credibility of the decision, Let be the objective function. , , These represent machine decisions, human decisions, and historical decisions, respectively. The arbitration function balances machine efficiency with human judgment to ensure optimal decision-making under different circumstances.

[0046] The machine records every decision and execution result, uses reinforcement learning to update the policy network, and manually labels misjudged cases to trigger incremental training of the case and rule bases, thereby further optimizing the machine's decision-making logic. After multiple rounds of reasoning, optimization, and manual adjustments, a behavior plan that meets the expected goals is finally generated, including specific project arrangements, resource allocation, and timelines. After the plan is generated, the system provides support for actual execution, including monitoring and evaluation mechanisms to ensure the project proceeds smoothly as planned.

[0047] Figure 2 This is a schematic block diagram of a human-computer collaboration-based behavior plan generation apparatus according to embodiments of the present disclosure. Figure 2 As shown, the device 200 may include a processor 210 and a memory 220 storing a computer program. When the computer program is executed by the processor 210, the device 200 is made capable of performing actions such as... Figure 1The steps of the method are shown below. In one example, device 200 can receive project information, perform semantic reasoning on the received project information based on an ontology library, decompose the project into multiple executable sub-projects; evaluate the current project environment, preconditions, and constraints based on historical data in a case library, and recommend appropriate strategies and behavioral arrangements for each sub-project; allocate resources required by each sub-project during project execution based on a resource library, and output a resource allocation scheme; generate an execution sequence for each sub-project based on the project's dependencies, and dynamically adjust the execution sequence according to state changes; and, based on a dynamic human intervention threshold, adjust the project's priority, resource allocation, and execution strategy through a feedback mechanism to generate a behavior plan.

[0048] In some embodiments of this disclosure, the device 200 can extract project objectives, time constraints, resource requirements, and status requirements from received project information using natural language processing technology; standardize the extracted information through ontology mapping based on a unified ontology representation in an ontology library; and decompose the project into multiple executable sub-projects according to the project's complexity and specific objectives by analyzing the project's internal relationships through semantic reasoning.

[0049] In some embodiments of this disclosure, the device 200 can automatically recommend multiple feasible behavioral arrangements for each sub-project based on similar cases in a historical case library; check whether the current project environment meets the prerequisite and subsequent states required for the recommended behavioral arrangements, and if not, adjust and optimize the behavioral arrangements.

[0050] In some embodiments of this disclosure, the device 200 can determine which projects need to be completed first by calculating the impact factor of each sub-project on the overall completion time during resource allocation. It can also convert resource usage into fuzzy values ​​using fuzzy logic, adjust project priorities based on these fuzzy values, and adjust the execution order and priority of projects according to load conditions. Furthermore, it can over-configure the resource requirements of multiple projects using virtualization technology, reserving resources for high-priority projects. When a high-priority project is executed, low-priority projects are suspended or delayed until the high-priority project is completed, at which point resources are returned to the low-priority projects. By parsing logs during project execution, it can detect abnormal resource utilization in real time. When abnormal resource utilization occurs, it can reorganize and adjust the dependencies between projects based on resource bottlenecks, construct a knowledge graph between projects, resources, and constraints, and adjust the resource allocation scheme through semantic analysis within the knowledge graph.

[0051] In some embodiments of this disclosure, the device 200 can construct a directed acyclic graph to describe the dependencies, resource allocation, and dynamic changes between projects, perform topological sorting on the project graph, and obtain the execution order of the projects; or, it can transform the objectives of minimizing completion time, reducing resource usage, and reducing energy consumption into constraints, and solve the optimal scheduling scheme based on mixed integer linear programming; or, it can encode project dependencies as chromosomes, perform crossover and mutation operations in a genetic algorithm, and optimize the execution order and resource allocation of projects by selecting chromosomes with high fitness.

[0052] In some embodiments of this disclosure, the device 200 can dynamically adjust the human intervention threshold based on the threat context, the confidence level of the machine's decision, and the matching degree of past experience; when the intervention threshold is triggered, the human modification and adjustment of the behavior plan is fed back to the machine, and the dynamic intervention threshold and decision-making mechanism are optimized using reinforcement learning and meta-learning algorithms; through situational awareness and data fusion, decision confidence assessment, autonomy boundary arbitration, feedback and knowledge update, a dynamic autonomous arbitration closed loop is formed.

[0053] In some embodiments of this disclosure, the device 200 can allow the machine to autonomously execute decisions in low-risk operations, while for high-risk operations, experts can adjust the machine's strategy or intervene directly through gesture recognition or natural language commands; the confidence level of the machine's decisions is evaluated in real time, and when the confidence level is high, the machine can execute autonomously, while when the confidence level is low, human intervention is initiated; the similarity between the current situation and past projects is analyzed based on past experience, and if the current project is similar to historical projects, the confidence level of the decision is increased and human intervention is reduced, otherwise the threshold for human intervention is increased.

[0054] In embodiments of this disclosure, processor 210 may be, for example, a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a processor based on a multi-core processor architecture, etc. Memory 220 may be any type of memory implemented using data storage technologies, including but not limited to random access memory, read-only memory, semiconductor-based memory, flash memory, disk storage, etc.

[0055] Furthermore, in embodiments of this disclosure, device 200 may also include input device 230, such as a keyboard, mouse, etc. Additionally, device 200 may also include output device 240, such as a display, etc.

[0056] In other embodiments of this disclosure, a computer-readable storage medium storing a computer program is also provided, wherein the computer program, when executed by a processor, is capable of performing the following functions: Figure 1 The steps of the behavior plan generation method 100 based on human-computer collaboration are shown.

[0057] In summary, the human-machine collaborative behavior plan generation method and apparatus according to the embodiments of this disclosure, through a behavior plan formulation mode of machine autonomous planning and human assistance, can effectively improve the efficiency and accuracy of project formulation and execution. By fully utilizing knowledge bases and intelligent reasoning, the machine plays a leading role in project analysis and decision-making, while humans perform evaluation and optimization in the final stage, jointly completing the efficient behavior plan formulation. By setting dynamic thresholds, the option of human intervention can be introduced during project execution, adjusting the plan according to external changes or unforeseen circumstances. This retains the efficiency of automatic plan generation by the machine while combining the flexibility of human intervention, making the generated behavior plan not only highly efficient in execution but also more aligned with the expected goals.

[0058] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatuses and methods according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0059] Unless otherwise expressly indicated by the context, the singular form of words used herein and in the appended claims includes the plural form, and vice versa. Thus, when referring to the singular, the plural form of the corresponding term is generally included. Similarly, the terms “comprising” and “including” shall be interpreted as including rather than exclusively. Likewise, the terms “including” and “or” shall be interpreted as including unless such interpretation is expressly prohibited herein. Where the term “example” is used herein, particularly when it follows a set of terms, “example” is merely exemplary and illustrative and should not be considered exclusive or extensive.

[0060] Further aspects and scope of adaptation become apparent from the description provided herein. It should be understood that various aspects of this application may be implemented individually or in combination with one or more other aspects. It should also be understood that the descriptions and specific embodiments herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0061] Several embodiments of this disclosure have been described in detail above. However, it is obvious that those skilled in the art can make various modifications and variations to the embodiments of this disclosure without departing from the spirit and scope of this disclosure. The scope of protection of this disclosure is defined by the appended claims.

Claims

1. A method for generating a behavior plan based on human-machine collaboration, characterized by, Comprise: receiving project information, performing semantic reasoning on the received project information based on an ontology library, and decomposing the project into multiple executable sub-projects; evaluating the current project environment, preconditions and constraints based on historical data in a case library, and recommending appropriate strategies and behavior arrangements for each sub-project; allocating resources required by each sub-project during project execution based on a resource library, and outputting a resource allocation scheme; generating an execution sequence for each sub-project according to the pre-and post-dependence relationships of the project, and dynamically adjusting the execution sequence according to state changes; and based on a dynamic artificial intervention threshold, adjusting the priority, resource allocation and execution strategy of the project through a feedback mechanism to generate a behavior plan.

2. The method of claim 1, wherein, The receiving project information, performing semantic reasoning on the received project information based on an ontology library, and decomposing the project into multiple executable sub-projects comprises: extracting project goals, time constraints, resource requirements and state requirements from the received project information through natural language processing technology; based on the unified ontology representation in the ontology library, standardizing the extracted information through ontology mapping; analyzing the internal relationship of the project through semantic reasoning, and decomposing the project into multiple executable sub-projects according to the complexity and specific goals of the project.

3. The method of claim 1, wherein, The evaluation of the current project environment, preconditions and constraints based on historical data in the case library, and the recommendation of appropriate strategies and behavior arrangements for each sub-project comprises: automatically recommending multiple feasible behavior arrangements for each sub-project based on similar cases in the historical case library; checking whether the current project environment meets the pre-state and post-state required by the recommended behavior arrangement, and if not, adjusting and optimizing the behavior arrangement.

4. The method of claim 1, wherein, The allocation of resources required by each sub-project during project execution based on the resource library, and the output of the resource allocation scheme comprises: when allocating resources, determine which projects need to be completed first by calculating the influence factor of each sub-project on the overall completion time, convert the use of resources into fuzzy values through fuzzy logic, adjust the priority of the project according to the fuzzy values, and adjust the execution order and priority of the project according to the load condition; through virtualization technology, over-allocate resources for multiple projects, reserve resources for high-priority projects, suspend or delay low-priority projects when high-priority projects are executed, and return resources to low-priority projects after high-priority projects are completed; and by analyzing the logs during project execution, real-time discovery of resource utilization anomalies, when the utilization of resources is abnormal, reorganize and adjust the dependence relationship between projects according to resource bottlenecks, construct a knowledge graph between projects, resources and constraints, and adjust the resource allocation scheme through semantic analysis in the knowledge graph.

5. The method of claim 1, wherein, The generation of an execution sequence for each sub-project according to the pre-and post-dependence relationships of the project, and the dynamic adjustment of the execution sequence according to state changes comprises: topologically sorting the project graph by constructing a directed acyclic graph to describe the dependence relationship between projects, resource allocation and its dynamic changes, and obtaining the execution order of the project; or The goal of minimizing the completion time, reducing resource usage, and reducing energy consumption is converted into a constraint condition, and the optimal scheduling scheme is solved based on mixed integer linear programming; or The project dependency relationship is encoded as a chromosome, and the crossover and mutation operations are performed in the genetic algorithm, and the execution order and resource allocation of the project are optimized by selecting chromosomes with high fitness.

6. The method of claim 1, wherein, The priority, resource allocation, and execution strategy of the project are adjusted through a feedback mechanism based on the dynamic artificial intervention threshold to generate a behavior plan, which includes: The artificial intervention threshold is dynamically adjusted according to the matching degree of the threat situation, the confidence of the machine decision, and the past experience; When the intervention threshold is triggered, the modification and adjustment of the behavior plan by the human are fed back to the machine, and the dynamic intervention threshold and decision mechanism are optimized using reinforcement learning and meta-learning algorithms; and Through situation awareness and data fusion, decision confidence evaluation, autonomy boundary arbitration, feedback and knowledge update, a dynamic autonomy arbitration closed loop is formed.

7. The method of claim 6, wherein, The artificial intervention threshold is dynamically adjusted according to the matching degree of the threat situation, the confidence of the machine decision, and the past experience, which includes: In low-risk operations, the machine autonomously executes the decision, and for high-risk operations, the expert adjusts the machine strategy or directly intervenes through gesture recognition or natural language instructions; The decision confidence of the machine is evaluated in real time, and when the confidence is high, the machine autonomously executes, and when the confidence is low, artificial intervention is started; Based on past experience, the similarity between the current situation and past projects is analyzed, if the current project is similar to the historical project, the confidence of the decision is increased, and the artificial intervention threshold is reduced, otherwise the artificial intervention threshold is enhanced.

8. The method of claim 6, wherein, Through situation awareness and data fusion, decision confidence evaluation, autonomy boundary arbitration, feedback and knowledge update, a dynamic autonomy arbitration closed loop is formed, which includes: The machine collects and fuses data in real time through a sensor network, log analysis, and threat intelligence platform tools to form a multi-dimensional situation awareness graph, and the human audits key nodes through a visual interface and supplements business context information; The machine uses Bayesian neural networks and Monte Carlo methods to evaluate the confidence of the decision, and quantifies the uncertainty of the prediction, and the human generates artificial confidence by comparing historical case libraries and business logic verification; An arbitration function is applied to determine the decision-making right, and the arbitration function is: ， wherein, , is a dynamic threshold parameter, is a human confidence, is a decision confidence, is an objective function, , , respectively represent machine, human decisions and historical decisions; The machine records each decision and execution result, and uses reinforcement learning to update the strategy network, and the human annotates the misjudgment cases to trigger incremental training of the case library and rule library.

9. A human-machine collaboration-based behavior plan generation device characterized by comprising: The device comprises: at least one processor; and at least one memory storing a computer program; When the computer program is executed by the at least one processor, the device executes the steps of the human-machine cooperation-based behavior plan generation method according to any one of claims 1 to 8.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the human-machine cooperation-based behavior plan generation method according to any one of claims 1 to 8.

Images