Artificial intelligence-based inspection task real-time optimization scheduling method

Abstract

The application discloses a real-time optimization scheduling method for inspection tasks based on artificial intelligence, and relates to the technical field of data processing.The method comprises the following steps: for an inspection site, full-factor spatio-temporal perception mapping is performed to determine an inspection view; based on the inspection view, a scheduling optimization model is supervised and trained; an inspection task is received, collision analysis under high concurrency is performed, and an optimization instruction for inspection scheduling is generated; the scheduling optimization model is combined to perform balanced scheduling and distribution of the inspection task, a task scheduling strategy is determined and transmitted and placed to an inspection target, and execution and tracking management of the inspection task are performed.The application solves the technical problem that the conventional inspection task scheduling method in the prior art is usually static, cannot adapt to changes in the site environment in real time, and leads to inflexible task scheduling and low resource utilization, and achieves the technical effect that the task scheduling strategy is dynamically adjusted according to real-time data of the inspection site, and the real-time performance, flexibility and resource utilization efficiency of the inspection task are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a real-time optimization scheduling method for inspection tasks based on artificial intelligence. Background Art

[0002] In modern intelligent inspection systems, with the improvement of industrial automation and intelligence levels, real-time optimization and scheduling of inspection tasks has become a key technology to improve inspection efficiency, reduce labor costs, and improve equipment management levels.

[0003] However, most existing inspection task scheduling solutions utilize optimization methods based on static planning. These methods typically fail to fully consider factors such as the dynamic location of inspection targets (such as inspectors and robots), task urgency, and resource availability. Furthermore, traditional scheduling methods lack real-time monitoring and dynamic adjustment capabilities for task execution, making them unable to handle conflicts between tasks in high-concurrency scenarios. This results in low utilization of inspection resources and can even lead to task delays and resource waste. Summary of the Invention

[0004] This application provides a real-time optimization scheduling method for inspection tasks based on artificial intelligence, which is used to solve the technical problems in the existing technology that traditional inspection task scheduling methods are usually static and cannot adapt to changes in the on-site environment in real time, resulting in inflexible task scheduling and low resource utilization.

[0005] The present application provides a real-time optimization and scheduling method for inspection tasks based on artificial intelligence, which includes: performing full-element spatiotemporal perception mapping on the inspection site to determine an inspection view, wherein the inspection view includes invariant mapping and variable mapping, and the inspection view is updated in real time; based on the inspection view, supervising the training of a scheduling optimization model, and the scheduling optimization model performs data-driven training in a two-step scheduling optimization mode; receiving inspection tasks, performing task collision analysis under high concurrency, and generating optimization instructions for inspection scheduling; receiving the optimization instructions, executing balanced scheduling allocation of the inspection tasks in combination with the scheduling optimization model, and determining the task scheduling strategy; transmitting and delegating the task scheduling strategy to the inspection target, and performing execution and tracking management of the inspection tasks.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The present application provides an artificial intelligence-based real-time optimization and scheduling method for inspection tasks, which relates to the field of data processing technology. Through real-time updated inspection views and a two-step scheduling optimization model, based on spatiotemporal perception mapping and task collision analysis, the scheduling of inspection tasks is dynamically optimized, and data-driven training is used to generate a balanced scheduling strategy to ensure efficient task allocation and transmission to inspection targets. This solves the technical problem that traditional inspection task scheduling methods in the prior art are usually static and cannot adapt to changes in the on-site environment in real time, resulting in inflexible task scheduling and low resource utilization. This achieves the technical effect of dynamically adjusting task scheduling strategies based on real-time data from the inspection site, thereby improving the real-time performance, flexibility, and resource utilization efficiency of inspection tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A flowchart of a method for real-time optimization and scheduling of inspection tasks based on artificial intelligence provided in an embodiment of the present application;

[0010] Figure 2 A flowchart of determining a task scheduling strategy in the real-time optimization scheduling method for inspection tasks based on artificial intelligence provided in an embodiment of the present application. DETAILED DESCRIPTION

[0011] This application provides a real-time optimization scheduling method for inspection tasks based on artificial intelligence, which is used to solve the technical problems in the existing technology that traditional inspection task scheduling methods are usually static and cannot adapt to changes in the on-site environment in real time, resulting in inflexible task scheduling and low resource utilization.

[0012] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0014] Example 1, as Figure 1 As shown, the present application provides a real-time optimization scheduling method for inspection tasks based on artificial intelligence, the method comprising:

[0015] P10: For the inspection site, full-element spatiotemporal perception mapping is performed to determine the inspection view, wherein the inspection view includes invariant mapping and variable mapping, and the inspection view is updated in real time.

[0016] Furthermore, step P10 in the embodiment of the present application further includes:

[0017] P11: Set a preset migration amount, wherein the preset migration amount is the average change of the elements per unit time; P12: Divide the invariant elements and variable elements in the inspection site according to the preset migration amount; P13: Perform three-dimensional mapping on the invariant elements to determine a first-layer view, perform three-dimensional mapping on the variable elements to determine a second-layer view, fuse the first-layer view and the second-layer view to determine the inspection view.

[0018] It should be understood that for the inspection site, the inspection view is constructed and updated in real time through full-element spatiotemporal perception mapping to support the dynamic scheduling of inspection tasks.

[0019] Specifically, a preset migration amount is first set. This migration amount represents the average change in each element of the inspection site per unit time. Elements can be fixed facilities in the environment (such as electrical equipment and buildings) or dynamic objects (such as inspection personnel and inspection robots). Setting a preset migration amount helps quantify changes and serves as a basis for subsequent element classification and view construction. By dynamically adjusting this migration amount, it is possible to accurately determine which elements require frequent updates and which can maintain a lower update frequency.

[0020] Next, based on the preset migration amounts, all elements within the inspection site are divided into immutable and variable elements. Immutable elements generally refer to those that remain unchanged or change very slowly over time, such as fixed equipment and buildings. Volatile elements, on the other hand, are those that change significantly over time, such as dynamic objects like inspection personnel and robots. This categorization allows the system to update different types of elements in a targeted manner, optimizing resource utilization and improving the accuracy of inspection task scheduling.

[0021] For immutable elements, 3D mapping technology is used to construct a first-level view. This view displays the structural characteristics of fixed facilities and sites, ensuring that inspection tasks can be rationally planned based on stable, unchanging information. For variable elements, 3D mapping is also used, but a dynamic second-level view is constructed to take into account the spatial and temporal mobility of these elements. This second-level view reflects the real-time location, status, and movement paths of inspection personnel or robots.

[0022] Ultimately, the invariant and variable views are fused to form a comprehensive inspection view. This view can be updated in real time, ensuring that inspection tasks are not affected in changing environments. Through view fusion, the system obtains a real-time monitoring platform that comprehensively considers both static and dynamic elements, enabling task scheduling decisions based on comprehensive, dynamically changing data.

[0023] The key to this step is to accurately classify feature types, combine 3D mapping with view fusion technology, and update the spatiotemporal information of all elements at the inspection site in real time, providing accurate basic data support for subsequent scheduling optimization. The technical support for this approach includes 3D mapping algorithms, spatiotemporal awareness mapping models, and dynamic feature tracking and update mechanisms, ensuring that inspection tasks can be efficiently scheduled and executed in a constantly changing environment.

[0024] Furthermore, step P13 of the embodiment of the present application further includes:

[0025] P13-1: For the variable elements, dynamic data is transmitted back in real time, wherein the dynamic data includes position and posture information and observation data; P13-2: The second-layer view is constructed with the position and posture as nodes and the observation data as edges, wherein the second-layer view is updated in real time, and the first-layer view has cumulative drift updates under loop detection based on the second-layer view.

[0026] Optionally, the method of constructing the inspection view can be further refined, especially for the real-time update and integration of variable elements, to form a first- and second-layer dynamically updated inspection scenario.

[0027] First, sensors transmit real-time position and posture information and observation data for variable elements (such as inspection personnel and inspection robots). Position and posture information primarily includes the specific position, orientation, and posture of the inspection personnel or robot in three-dimensional space. This data is crucial for path planning and task scheduling. Observation data, on the other hand, includes environmental information and equipment status collected by sensors. This information helps the system perceive on-site changes in real time and update the execution plan for inspection tasks.

[0028] Next, based on the real-time dynamic data, a second-level view is constructed, using position and posture information as nodes and observation data as edges. This second-level view is a dynamically updated model that represents all changing inspection elements and their interrelationships. Each node represents a variable element (such as an inspection personnel or equipment), while edges represent their interactions or temporal continuity. Because these elements change during the inspection process, the second-level view must be continuously updated in real time to ensure the accuracy and timeliness of scheduling tasks.

[0029] To maintain the accuracy of the second-level view, it is frequently updated and adjusted based on the location and status of the inspector or robot, as well as dynamic changes in the environment. This continuously updated second-level view dynamically adapts to sudden changes in inspection tasks and allows for timely adjustments to scheduling strategies.

[0030] At the same time, the first-layer view (fixed scene elements such as buildings and equipment) may also experience cumulative drift. To address this, loop closure detection is required. This involves detecting the overlap between the current location and historical locations to identify and correct drift caused by sensor errors, environmental interference, and other factors. Whenever an inspector or device reaches a location previously inspected, the scene elements in the first-layer view are adjusted based on updated data from the second-layer view to ensure consistency with actual conditions. This process helps reduce errors and ensures accurate mission planning.

[0031] Ultimately, the first-layer and second-layer views together form a complete inspection scenario model. Changes in the second-layer view trigger necessary updates to the first-layer view, ensuring that the entire inspection scenario reflects the actual situation in real time. This dynamic view fusion enables the system to flexibly respond to on-site changes during inspection tasks, enabling real-time task scheduling and route planning.

[0032] Through the above steps, not only can the inspection scene be updated in real time, but it can also quickly respond to and adjust to dynamically changing inspection elements. The core of this process is to combine position and posture information with observation data to construct and update the second-level view in the form of a node-directed graph, effectively supporting the intelligent scheduling and path optimization of inspection tasks.

[0033] Furthermore, the cumulative drift update under the loop detection, step P13-2 of the embodiment of the present application further includes:

[0034] P13-21: Using the same-location inspection visit as the loop detection group, determine the first observation data and the second observation data, wherein the second observation data is real-time observation data, and the first observation data is the observation data of the upper time node; P13-22: For the invariant element, measure the element drift of the second observation data compared with the first observation data; P13-23: Based on the element drift, update the one-layer view.

[0035] Specifically, the accuracy and real-time performance of the inspection view can be ensured through cumulative drift updates under loop detection, especially in the update of immutable elements (such as fixed equipment, building structures, etc.).

[0036] First, a loop closure detection mechanism is used to identify multiple visits to the same location during the inspection process. At the inspection site, inspectors or inspection robots may visit the same location at different time points. In this case, these locations are considered loop closure groups. By comparing the first observation data (i.e., the observation data at the time of the last visit) with the second observation data (i.e., the real-time observation data at the current time point), it is possible to assess whether the state of the invariant element has changed. The second observation data is typically data collected by sensors in real time, while the first observation data is a previous record saved by the system based on historical data. Through this comparison, the loop closure detection mechanism can identify accumulated errors that occur during the inspection process, especially deviations related to environmental changes or sensor errors.

[0037] During loop closure comparison, it's necessary to measure element drift—the amount by which the second observation changes relative to the first. This drift can be calculated using differential analysis methods to measure the error or change between two time points. For invariant elements (such as fixed facilities and equipment), drift reflects actual changes due to sensor errors, environmental changes, or accumulated model errors.

[0038] The measurement of feature drift not only considers drift in position, but also changes in angle, attitude, and environmental characteristics. Accurately measuring this drift helps ensure the accuracy of invariant features and prevents error accumulation during subsequent view updates.

[0039] Then, based on the calculated element drift, the first-level view is updated. The first-level view primarily describes fixed elements in the inspection scene, which typically do not change frequently and are therefore updated less frequently. However, during the inspection process, factors such as environmental changes, equipment wear, or sensor errors may cause the fixed elements of the first-level view to drift. Through loop detection and drift calculation, the first-level view is dynamically corrected to ensure its accuracy and real-time performance. This update process not only maintains the accuracy of the inspection scene but also eliminates long-term drift caused by errors in inspection equipment or sensors, thereby improving the accuracy of task scheduling and the reliability of inspection task execution.

[0040] This cumulative drift update method under loop closure detection allows for fine-tuning of invariant elements, making inspection view updates more stable and reliable over time, and improving the intelligence of inspection task scheduling. This strategy is particularly suitable for long-running inspection tasks, ensuring that drift effects during data accumulation do not cause deviations in task scheduling.

[0041] P20: Based on the inspection view, supervise the training of the scheduling optimization model, and perform data-driven training on the scheduling optimization model in a two-step scheduling optimization mode.

[0042] It should be understood that the scheduling optimization model is trained based on the inspection view, and a two-step scheduling optimization method is used for data-driven training.

[0043] During inspection tasks, spatiotemporal data collected from the inspection view (including real-time data on both invariant and variable elements) provides training data for the scheduling optimization model. The inspection view contains detailed environmental information, equipment status, the locations of inspectors and robots, and real-time observation data. This multidimensional data is used as input for model training using supervised learning. This involves providing known, labeled data (such as ideal task scheduling results) so that the model can learn the relationship between the input data and the desired output. Supervised learning helps the system automatically extract patterns and regularities from the data, enabling the scheduling optimization model to make intelligent decisions based on real-time data.

[0044] The training process utilizes a two-step scheduling optimization approach for data-driven training. This two-step scheduling optimization optimizes the scheduling of inspection tasks in two phases. In the first phase, tasks are rationally allocated to different inspection personnel or equipment by analyzing characteristics such as task priority, urgency, and resource requirements. Task allocation not only considers resource availability but also dynamically adjusts task priority and urgency to ensure timely completion of high-priority tasks. The goal of this phase is to reduce task conflicts and idle time through rational resource allocation, thereby improving resource utilization.

[0045] In the second phase, the inspection personnel or robot's path planning is optimized to reduce unnecessary movement during the inspection process. Path planning not only considers the order of tasks but also the location and capabilities of the inspector or robot to ensure efficient execution of inspection tasks. By dynamically adjusting the path, ineffective movement of inspectors or equipment can be effectively avoided, energy consumption can be reduced, and work efficiency can be improved.

[0046] The two-step scheduling optimization model is trained using a data-driven approach, continuously optimizing the model using both historical and real-time data. This feedback mechanism gradually improves the scheduling strategy, making the scheduling model more accurate and efficient under varying inspection environments and task requirements. The model training process not only relies on traditional algorithm optimization but also utilizes technologies such as reinforcement learning to enable the model to continuously adapt and evolve during actual inspection tasks.

[0047] Through this series of steps, the system can adjust task allocation and path planning in real time during the actual execution of inspection tasks, ensuring the optimal execution of inspection tasks under resource constraints, improving overall efficiency and reducing unnecessary energy consumption.

[0048] P30: Receive inspection tasks, perform task collision analysis under high concurrency, and generate optimized instructions for inspection scheduling.

[0049] Optionally, after receiving the inspection task, a task collision analysis under high concurrency is performed, and an optimization instruction for the inspection scheduling is generated.

[0050] First, inspection task information is received from the inspection scheduling system. These inspection tasks include the equipment to be inspected, the inspection area, the task priority, the urgency, and the resources required for each task (such as personnel, equipment, and time). Each inspection task contains multiple key parameters, such as task type, location, and execution time window. This received task information provides input data for subsequent task collision analysis.

[0051] In actual inspections, multiple tasks may need to be performed within the same timeframe, which can lead to conflicts between tasks, especially when inspectors or robots are limited. For example, if multiple tasks involve the same inspection area or the same equipment, and resources are insufficient to meet the demand, conflicts will arise.

[0052] To address this issue, collision analysis is required for highly concurrent tasks. The goal of collision analysis is to predict and identify pairs of tasks that may compete for or conflict with resources, especially when resources are limited. The specific collision analysis process includes: First, by analyzing the time, space, and resource requirements of the tasks, detect which tasks may conflict at the same time and place. Then, automatically adjust the execution order of tasks based on the urgency, importance, and resource availability of the tasks. For high-priority tasks, adjust the time or location of other low-priority tasks to ensure that high-priority tasks can be completed first. In addition, in the case of task conflicts, the allocation of resources can be optimized based on the availability of inspection personnel or equipment to ensure that each task can be executed efficiently.

[0053] After completing the task collision analysis, optimization instructions are generated based on the analysis results to guide inspection personnel or inspection robots on how to perform the task. These instructions include task allocation strategies, priority adjustments, path planning, time scheduling, and other content. The goal of optimization instructions is to ensure that inspection tasks can be executed efficiently and without conflict within limited resources, thereby improving overall inspection efficiency. For example, if the inspection task involves the movement of inspection personnel or robots, path planning optimization can be performed based on real-time environmental data to reduce ineffective movement distance during the inspection process and avoid unnecessary resource waste. In addition, the timing of tasks can be optimized to reduce idle time between tasks and improve work efficiency.

[0054] P40: Receive the optimization instruction, perform balanced scheduling and allocation of the inspection task in combination with the scheduling optimization model, and determine a task scheduling strategy.

[0055] Further, such as Figure 2 As shown, step P40 in this embodiment of the application also includes:

[0056] P41: Determine a two-step scheduling optimization method, wherein the two-step scheduling optimization method includes a one-step scheduling optimization based on global random adjustment and a two-step scheduling optimization based on local directional adjustment; P42: Determine a first scheduling strategy, based on the two-step scheduling optimization method, with the first scheduling strategy as the initial strategy, perform iterative optimization until convergence, and select the maximum fitness strategy as the task scheduling strategy.

[0057] It should be understood that after receiving the optimization instruction, the balanced scheduling allocation of the inspection tasks is performed in combination with the scheduling optimization model, and the task scheduling strategy is determined.

[0058] Specifically, the two-step scheduling optimization method must first be determined based on the optimization instructions. This method is a scheduling optimization method that comprehensively considers both global and local information. Specifically, the two-step scheduling optimization is divided into two parts: a one-step scheduling optimization based on global random adjustments, and a two-step scheduling optimization based on local targeted adjustments.

[0059] The one-step scheduling optimization phase based on global random adjustment perturbs various parameters of the task scheduling policy through global random adjustment, thereby exploring possible scheduling solutions. This process increases the system's search diversity and flexibility by randomly selecting possible scheduling strategies from the entire scheduling space. The goal of global random adjustment is to avoid local optimal solutions and enhance the exploratory nature of the entire optimization process, ensuring that a more efficient scheduling solution is found.

[0060] The two-step scheduling optimization method based on local targeted adjustments optimizes the strategy based on local targeted adjustments after global random adjustments. At this point, the disturbed strategy points are corrected based on the inspection task's constraints (such as time limits and resource constraints). Local adjustments refine the strategy's adaptability, making the scheduling solution more responsive to actual needs. This local targeted adjustment process ensures the feasibility of the scheduling strategy while improving its adaptability and effectiveness. By combining global randomness with local targeted adjustments, the two-step scheduling optimization method effectively balances the exploratory and adaptable nature of the scheduling strategy, avoids falling into local optimal solutions, and improves the overall efficiency of task scheduling.

[0061] After determining the two-step scheduling optimization method, the first scheduling strategy is determined based on this optimization method. The first scheduling strategy is a preliminary task scheduling solution and can be the initial solution obtained through random perturbations and local adjustments in the first stage. Next, this initial strategy is continuously optimized using an iterative optimization method until convergence conditions are met.

[0062] Exemplarily, the first scheduling strategy is adjusted through multiple iterations to gradually improve the adaptability of the scheduling solution. In each iteration, the effectiveness of the current strategy is evaluated, and adjustments are made based on task execution and resource utilization. This process can utilize optimization algorithms, such as genetic algorithms and simulated annealing algorithms, to gradually find the optimal solution.

[0063] After multiple iterations, the maximum fitness strategy is selected as the final task scheduling strategy. Fitness is a metric that measures the effectiveness of a scheduling strategy and can include factors such as timeliness of task completion, full resource utilization, and conflict minimization. The maximum fitness strategy is the scheduling solution that best meets these optimization goals. This process ensures that the task scheduling strategy is not only optimal but also dynamically adapts to changes in inspection tasks and the environment.

[0064] Furthermore, step P42 of the embodiment of the present application further includes:

[0065] P42-1: Traverse the inspection tasks, combine the real-time positions of the inspection targets in the second-layer view, and determine the first inspection strategy based on the proximity principle, wherein the inspection targets include inspection personnel and inspection robots; P42-2: Perform task priority avoidance based on the real-time status of the inspection targets, and determine the first constraint condition, wherein the real-time status includes the idle state and the inspection state; P42-3: Perform task priority avoidance for the inspection tasks, and determine the second constraint condition; P42-4: Determine the first scheduling strategy based on the first inspection strategy, the first constraint condition and the second constraint condition.

[0066] In a possible embodiment of the present application, the generation process of the task scheduling strategy is further refined, especially by gradually determining the first scheduling strategy by combining the characteristics of the inspection task, the real-time status of the inspection target and the constraints.

[0067] Specifically, first, all inspection tasks are traversed, and the order of task execution is determined in combination with the real-time location of the inspection targets in the second-layer view. The inspection targets in the second-layer view include dynamic inspection objects such as inspection personnel and inspection robots. Their real-time location and status are crucial for task scheduling. In this process, the system applies the proximity principle, that is, it prioritizes assigning tasks to the task points closest to the current inspection target to reduce the ineffective movement distance during task execution and improve inspection efficiency. The application of the proximity principle ensures that task scheduling can effectively utilize existing resources (such as personnel and robots) and avoid idle time or ineffective waiting due to over-scheduling of resources. At the same time, it can also improve the flexibility of task scheduling, allowing the system to dynamically respond to sudden changes in inspection tasks.

[0068] After determining the primary inspection strategy, task prioritization is then performed based on the real-time status of the inspection target (e.g., whether the inspector or robot is idle or in inspection mode). Specifically, when the inspector or robot is idle, tasks are prioritized to ensure maximum resource utilization. When the inspector or robot is in inspection mode, the system adjusts the scheduling based on the urgency and priority of the task to avoid interference with ongoing tasks. The goal of this step is to ensure that resource utilization is conflict-free during scheduling and to effectively avoid task prioritization through the primary constraint, thereby maximizing inspection resource utilization and the timeliness of task completion.

[0069] In addition to considering the inspection target status, more refined task priority avoidance is required. This means further adjusting the scheduling strategy based on each task's priority and constraints. For example, if certain tasks are urgent, they should be prioritized, while other lower-priority tasks may need to be postponed or rescheduled. The key to this process is to prioritize tasks, ensuring that even with high concurrency, tasks can be processed in a reasonable order, preventing delays for higher-priority tasks.

[0070] Finally, a comprehensive optimization process based on the previously determined first inspection strategy, the first constraint (priority avoidance based on the real-time status of the inspection target), and the second constraint (avoidance based on task priority) results in the finalized first scheduling strategy. This strategy comprehensively considers the spatial distribution of tasks, resource allocation, and the priorities of each task to ensure balanced scheduling and reasonable allocation of tasks. After multiple rounds of optimization, resource utilization is maximized, task conflicts are minimized, and flexible adjustments based on real-time conditions are implemented, ensuring system efficiency and the smooth completion of inspection tasks.

[0071] Furthermore, step P42 of the embodiment of the present application further includes:

[0072] P42-5: Based on the fitness of the strategy point, set the optimization step size; P42-6: For the first inspection strategy, use the optimization step size to perform global strategy point random perturbation adjustment on the inspection strategy to determine a one-step optimization strategy; P42-7: Based on the first constraint condition and the second constraint condition, perform local strategy point directional adjustment on the one-step optimization strategy to determine a two-step optimization strategy; P42-8: Integrate the one-step optimization strategy and the two-step optimization strategy to determine a one-layer iterative optimization strategy.

[0073] Optionally, the specific process of iterative optimization with the first scheduling strategy as the initial strategy may be that the optimization step size needs to be set first according to the fitness of the strategy point. The strategy point refers to each specific decision or operation point in the scheduling strategy, such as task allocation, resource scheduling, path planning, etc. Each strategy point is associated with a specific inspection task execution process, representing a specific resource configuration or execution step. By evaluating these strategy points, the scheduling plan is continuously adjusted to achieve the optimal task execution effect. Fitness is an evaluation indicator of the execution effect of the scheduling strategy, which measures the gap between the current strategy and the optimal strategy. The higher the fitness, the closer the current strategy is to the optimal solution. The setting of the optimization step size determines the extent of the strategy adjustment during the optimization process. For example, a larger step size can speed up the optimization process, but may result in skipping the local optimal solution; a smaller step size can refine the adjustment, but may require more computing time.

[0074] Dynamically adjusting the optimization step size according to the fitness of the strategy point can maintain a balance between global exploration and local refinement, and improve the efficiency and accuracy of strategy optimization. For example, the fitness of each strategy point is first evaluated. Fitness is a quantitative description of the current strategy execution effect, including task completion time, resource utilization efficiency, task priority satisfaction, etc. Strategy points with high fitness indicate that the scheduling is relatively reasonable, resource allocation is effective, and task execution is smooth. Strategy points with low fitness indicate that the current scheduling effect is not good and needs to be adjusted. After calculating the fitness of each strategy point, the fitness gap between the current strategy point and the target optimal strategy can be evaluated. The larger the fitness gap, the greater the gap between the current scheduling strategy and the ideal state, and further optimization is needed. If the gap is small, it means that the current strategy is already relatively close to the optimal state and needs fine-tuning and optimization.

[0075] When the fitness gap between the current strategy point and the optimal strategy is large, a larger optimization step size can be set. At this point, the system will make extensive adjustments and explorations to rapidly change the current strategy toward a more optimal solution. A larger step size helps quickly escape from the local optimal solution and enter the possible global optimal region. When the fitness gap is small, indicating that the current strategy is close to the optimal solution, a smaller optimization step size can be set for fine-grained adjustments. This fine-tuning helps preserve existing strengths and eliminate minor deficiencies during the optimization process, further improving the strategy's accuracy and adaptability.

[0076] After setting the optimization step size, the first inspection strategy is subjected to a global random perturbation adjustment based on this step size, generating a one-step optimized strategy. This process involves a global exploration of task scheduling by perturbing the strategy points, identifying strategy points that are more likely to improve fitness. This perturbation adjustment can break through local optimal solutions, allowing the strategy to search across a wider scheduling space, increasing strategy diversity and exploratory potential. The key to perturbation adjustment lies in randomness. By randomly selecting different strategy points for adjustment, the scheduling strategy can be optimized from different angles, thus avoiding being trapped in local optimal solutions.

[0077] After the one-step optimization strategy is generated, it is necessary to make directional adjustments to the local strategy points to generate a two-step optimization strategy. This step mainly corrects potential problems in the one-step optimization strategy through fine-grained adjustments, especially local problems such as trajectory collision avoidance. Specifically, according to the actual constraints of the task (such as task priority, equipment location, personnel status, etc.), the strategy points are fine-tuned to ensure that there will be no task conflicts, resource contention and other problems during the execution process. For the trajectories of inspection personnel or robots, the system will pay special attention to avoid collisions between them, or conflicts with fixed facilities and obstacles. Through fine-grained directional adjustments, the two-step optimization strategy can better adapt to complex environments and constraints, ensuring the safety and smooth execution of local operations while ensuring global scheduling efficiency.

[0078] After generating the first- and second-step optimization strategies, they are combined to determine a final, iterative optimization strategy. By integrating the global perturbation and the locally adjusted strategies, an optimal scheduling solution is obtained that not only achieves global optimization but also addresses local details. This combination of global and local optimization, dynamic adjustment, and feedback correction makes the scheduling system more adaptable and flexible when facing high-concurrency tasks.

[0079] P50: The task scheduling strategy is transmitted and decentralized to the inspection target to execute and track the inspection task.

[0080] Furthermore, step P50 in the embodiment of the present application further includes:

[0081] P51: Traverse the task scheduling strategy, determine the scheduling group of inspection task-inspection target, and identify the task execution limit, which is the initial node for the execution of the inspection task; P52: Based on the task execution limit, set the timestamp constraint for task decentralization; P53: Based on the timestamp constraint, perform transmission management on the inspection task.

[0082] It should be understood that the optimized task scheduling strategy is transmitted to the inspection target (such as inspection personnel or inspection robots) so as to execute specific inspection tasks and track and manage the tasks.

[0083] First, the generated task scheduling policy is traversed. Based on the task information in the policy, the inspection target scheduling group corresponding to each inspection task is determined. This process primarily allocates tasks and inspection targets (such as inspectors, robots, etc.) as needed, forming a matching task and target group. Each task is assigned to the corresponding inspection target, and the allocation is made appropriately based on the target's resource status and the task's priority.

[0084] During this process, the system also identifies each task's execution deadline. This deadline is the point at which a task must begin execution and is typically related to its priority. High-priority tasks may be assigned earlier deadlines, while low-priority tasks may have longer execution windows. Identifying task deadlines is crucial for scheduling management, as it determines whether a task can begin execution within the specified timeframe.

[0085] After determining the task scheduling group and task execution deadline, a timestamp constraint is set for each task based on the task execution deadline. This timestamp constraint defines the maximum delay before a task is received and executed by the inspection target, meaning that the task must be transmitted to the inspection target before this timestamp. Setting timestamp constraints helps optimize task transmission timeliness, ensuring that tasks can be issued within the specified time limit and reducing task scheduling errors or delays caused by communication delays. By using appropriate timestamp constraints, tasks can be promptly transmitted to inspection personnel or equipment according to priority and deadline requirements, reducing waiting time and ineffective delays in task execution.

[0086] Finally, task transmission is managed based on the previously defined timestamp constraints. This process ensures that task information is delivered to the inspection target within the appropriate time window for efficient scheduling and execution. Transmission management involves not only delivering task instructions to inspection personnel or robots but also monitoring the task's transmission status to ensure that the task instructions reach the designated target accurately and without error.

[0087] In a highly concurrent environment, transmission management also needs to consider the system's communication load. Strategies such as task priority adjustment and batch transmission may be employed to optimize communication efficiency. Precise task transmission management can effectively coordinate the execution of multiple tasks within the same timeframe, ensuring efficient and orderly task scheduling.

[0088] In summary, the embodiments of the present application have at least the following technical effects:

[0089] This application uses full-element spatiotemporal perception to build maps, updating the inspection view in real time. Combining invariant and variable mapping, it provides dynamic data support for inspection task scheduling. Based on this view, a scheduling optimization model is supervised for training, employing data-driven training using two-step scheduling optimization. Optimization instructions are generated through high-concurrency task collision analysis, and balanced task allocation is achieved through the use of the scheduling optimization model. Finally, the scheduling policy is transmitted to the inspection target, ensuring efficient task execution and tracking management.

[0090] The technical effect of dynamically adjusting the task scheduling strategy according to the real-time data of the inspection site and improving the real-time performance, flexibility and resource utilization efficiency of the inspection tasks has been achieved.

[0091] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0092] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0093] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, to the extent such modifications and variations fall within the scope of the present application and its equivalents, the present application is intended to include such modifications and variations.

Claims

1. A real-time optimization and scheduling method for inspection tasks based on artificial intelligence, characterized in that: The method comprises: For the inspection site, full-element spatiotemporal perception mapping is carried out to determine the inspection view, wherein the inspection view includes invariant mapping and variable mapping, and the inspection view is updated in real time; Based on the inspection view, supervise the training of a scheduling optimization model, wherein the scheduling optimization model is data-driven trained in a two-step scheduling optimization mode, wherein the two-step scheduling optimization mode is a one-step scheduling optimization based on global random adjustment and a two-step scheduling optimization based on local directional adjustment; Receive inspection tasks, perform task collision analysis under high concurrency, and generate optimized instructions for inspection scheduling; Receive the optimization instruction, perform balanced scheduling and allocation of the inspection tasks in combination with the scheduling optimization model, and determine a task scheduling strategy; The task scheduling strategy is transferred to the inspection target to execute and track the inspection task; The full-element spatiotemporal perception mapping and inspection view determination include: Setting a preset migration amount, wherein the preset migration amount is an average change of the element per unit time; Dividing the invariant elements and variable elements in the inspection site according to the preset migration amount; Performing three-dimensional mapping on the invariant elements to determine a first-layer view, performing three-dimensional mapping on the variable elements to determine a second-layer view, and fusing the first-layer view and the second-layer view to determine the inspection view; Among them, 3D mapping is performed for variable elements to determine the second-level view, including: For the variable elements, dynamic data is transmitted back in real time, wherein the dynamic data includes position and posture information and observation data; The second-layer view is constructed with the position and posture as nodes and the observation data as edges, wherein the second-layer view is updated in real time, and the first-layer view is updated with accumulated drift under loop closure detection based on the second-layer view; The cumulative drift update under the loop detection includes: Using the same-location inspection visit as a loop detection group, determine the first observation data and the second observation data, wherein the second observation data is real-time observation data and the first observation data is observation data of the upper time node; For the invariant element, measuring the element drift of the second observation data compared with the first observation data; The layer view is updated based on the element drift amount.

2. The method for real-time optimization and scheduling of inspection tasks based on artificial intelligence according to claim 1, characterized in that: The balanced scheduling allocation of the inspection tasks is performed in combination with the scheduling optimization model to determine the task scheduling strategy, including: Determine the two-step scheduling optimization method; Determine a first scheduling strategy, based on the two-step scheduling optimization method, use the first scheduling strategy as the initial strategy, perform iterative optimization until convergence, and select the maximum fitness strategy as the task scheduling strategy.

3. The method for real-time optimization and scheduling of inspection tasks based on artificial intelligence according to claim 2, characterized in that: The determining of the first scheduling strategy includes: Traversing the inspection tasks, combining the real-time locations of the inspection targets in the second-layer view, and determining a first inspection strategy based on the proximity principle, wherein the inspection targets include inspection personnel and inspection robots; Performing task priority avoidance based on the real-time status of the inspection target and determining a first constraint condition, wherein the real-time status includes an idle state and an inspection state; Performing task priority avoidance for the inspection task and determining a second constraint condition; A first scheduling strategy is determined based on the first inspection strategy, the first constraint condition and the second constraint condition.

4. The method for real-time optimization and scheduling of inspection tasks based on artificial intelligence according to claim 3 is characterized in that: Based on the two-step scheduling optimization method, the first scheduling strategy is used as the initial strategy to perform iterative optimization, including: Based on the fitness of the strategy point, set the optimization step size; For the first inspection strategy, perform a global strategy point random perturbation adjustment on the inspection strategy with the optimization step size to determine a one-step optimization strategy; According to the first constraint condition and the second constraint condition, performing a local strategy point-oriented adjustment on the one-step optimization strategy to determine a two-step optimization strategy; The one-step optimization strategy and the two-step optimization strategy are integrated to determine a one-layer iterative optimization strategy.

5. The method for real-time optimization and scheduling of inspection tasks based on artificial intelligence according to claim 1, characterized in that: The task scheduling strategy is transferred to the inspection target, including: Traversing the task scheduling strategy, determining the inspection task-inspection target scheduling group, and identifying the task execution time limit, wherein the task execution time limit is the initial node of the inspection task execution; Based on the task execution time limit, set the timestamp constraint for task delegation; Based on the timestamp constraint, the inspection task is managed for transmission.

Images