Method for human-machine task decomposition and scheduling based on dynamic matching of dam underwater defect repair

Abstract

The application discloses a kind of based on dynamic matching library dam underwater defect repair man-machine task decomposition and scheduling method.The method includes: obtaining defect data to be repaired, two-stage disassembly is carried out to the repair operation to obtain atomic task set and its capability demand vector, and task dependency graph is constructed;Real-time state data of operation subject is obtained, and the time-varying capability supply vector of diver and underwater robot is constructed;Based on capability demand vector and time-varying capability supply vector, feasibility determination is carried out, and optimization solving is carried out in combination with task dependency graph, to generate initial scheduling scheme;According to dynamic feedback information, trigger local rescheduling, to obtain updated scheduling scheme;Scheduling instruction is issued.The application can quantify the capability of operation subject, guarantee the safety of underwater operation, consider material timeliness and repair quality, and can be used to improve the robustness of reservoir dam repair operation in dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of underwater engineering maintenance and intelligent scheduling technology, and in particular to a human-machine task decomposition and scheduling method for underwater defect repair of reservoirs and dams based on dynamic matching. Background Technology

[0002] As an important water conservancy infrastructure, the structural integrity of reservoir and dam systems is crucial to the operational safety of water conservancy projects and the safety of downstream areas. Under the long-term effects of high water pressure, water flow erosion, and complex physical and chemical environments, dam bodies and ancillary structures often develop defects such as cracks, erosion, and leakage.

[0003] Existing underwater repair solutions for reservoirs and dams typically rely on independent operations by human divers or remotely operated underwater robots, or on a division of labor between humans and machines based on a pre-defined static list. In practical applications, existing scheduling methods often employ an experience-based, fixed allocation model, setting the capabilities of the main operators as a constant baseline.

[0004] In actual engineering projects, the underwater repair environment is complex and has many uncertainties. For example, the physiological state of the workers in the deep-water high-pressure environment, the solidification cycle of the repair materials, and the hardware performance of the underwater equipment will all fluctuate as the operation time goes by. This makes it easy for the pre-planned operation to have insufficient capacity coverage in the middle and later stages of actual execution. Summary of the Invention

[0005] Purpose of the invention: To provide a human-machine task decomposition and scheduling method for underwater defect repair of reservoir dams based on dynamic matching, in order to solve the above-mentioned problems in the prior art.

[0006] Technical solution: A method for human-machine task decomposition and scheduling for underwater defect repair of reservoir dams based on dynamic matching, including:

[0007] Obtain the defect data to be repaired, decompose the repair work to obtain the set of atomic tasks and their capability requirement vectors, and construct the task dependency graph;

[0008] Acquire and process the real-time status data of the main body of the operation to construct the time-varying capacity supply vector of the main body of the operation;

[0009] Feasibility determination is performed based on the capacity demand vector and the time-varying capacity supply vector, and optimization is performed in conjunction with the task dependency graph to generate an initial scheduling scheme.

[0010] Based on the dynamic feedback information collected in real time, the initial scheduling scheme or the updated scheduling scheme currently in execution is selected. Combined with the updated time-varying capacity supply vector, local rescheduling is triggered to obtain the updated scheduling scheme.

[0011] Distribute scheduling instructions based on the initial scheduling scheme or the updated scheduling scheme.

[0012] Beneficial effects: It quantifies the capabilities of the main operators, ensures the safety of underwater operations, takes into account the timeliness of materials and the quality of repair, and improves the robustness of reservoir and dam repair operations in dynamic environments. Attached Figure Description

[0013] Figure 1 This is a flowchart of the human-machine task decomposition and scheduling method for underwater defect repair of reservoir dams based on dynamic matching, as described in this application.

[0014] Figure 2 A flowchart for constructing the task dependency graph for this application.

[0015] Figure 3 A flowchart of the time-varying capability supply vector generated for this application.

[0016] Figure 4 A flowchart for generating the time-varying capability supply vector for the underwater robot in this application.

[0017] Figure 5 This is a flowchart of the local emergency fine-tuning of the underwater terminal in this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a predetermined order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] In addition to the problems mentioned in the background technology, underwater repair materials have a preset process time limit, and environmental parameters such as flow rate and visibility fluctuate transiently. Traditional static scheduling mode is difficult to respond in real time to performance loss and environmental interference during the operation, resulting in a decline in repair quality or interruption of the operation link.

[0021] To solve these problems, combined with Figures 1 to 5 The present invention will be specifically described through the following embodiments.

[0022] According to one aspect of this application, this embodiment provides a method for human-machine task decomposition and scheduling based on dynamic matching of underwater defect repair in reservoir dams, the method comprising:

[0023] Step 101: Obtain the defect data to be repaired, decompose the repair job to obtain the set of atomic tasks and their capability requirement vectors, and construct the task dependency graph.

[0024] The data on defects to be repaired includes parameters such as the width, depth, and length of cracks on the surface of hydraulic structures, as well as the area of ​​eroded regions, and contains a preset classification code for the defect. After acquiring this data, the repair work is broken down into basic, indivisible operational units, i.e., atomic task sets, based on pre-stored industry repair standards or historical engineering procedures. For example, the grouting repair work on a dam panel crack can be progressively broken down into atomic tasks performed sequentially, such as surface cleaning, grooving, laying grouting holes, pipe connection, and high-pressure grouting.

[0025] Accordingly, the operational requirements of each atomic task are quantitatively assessed across dimensions such as operational precision, working force or torque, required duration, safety risk level, environmental tolerance, and degree of autonomous operation. The assessment results are then combined to form a capability requirement vector. This vector represents the entry threshold conditions required to complete the preset technical operations. Based on the spatial interference relationships and the sequence of repair processes for each basic operation, a task dependency graph with a directional topology is constructed, serving as the logical benchmark for the advancement of all repair tasks along the time axis.

[0026] In some optional implementations, each dimension of the capability requirement vector can be normalized. The system uses a unidirectional semantic definition for all dimensions, meaning that the closer the value is to 1, the higher the requirement for the corresponding capability of the subject, and the closer the value is to 0, the lower the capability requirement.

[0027] Step 102: Obtain and process the real-time status data of the main operation entity to construct the time-varying capacity supply vector of the main operation entity.

[0028] The operational entities include divers performing underwater repair work in deep-water environments, as well as remotely operated underwater vehicles (ROVs). Real-time status data consists of various basic operational parameters and physiological monitoring indicators reported by the operational entities in complex underwater operating environments. For example, for divers, this data includes the current absolute diving depth, the duration of continuous underwater operations, and the surrounding water temperature; for ROVs, this data includes the real-time state of charge of the control cabin battery, the pressure deviation rate of the robotic arm's hydraulic system, and the current length of the working umbilical cable deployed into the water.

[0029] Specifically, by using a pre-defined model of the physiological load degradation law of the human body and a model of the nonlinear degradation characteristics of the physical performance of heavy-duty equipment, the real-time status data is coupled and processed to transform the original monitoring data into a six-dimensional evaluation vector aligned with the dimension of the capacity demand vector, i.e., a variable capacity supply vector.

[0030] This feature vector can objectively and dynamically quantify the monotonic decline in the operator's ability due to the accumulation of nitrogen load in the body and the loss of water temperature, as well as the degradation of the underwater robot's positioning and control capabilities after long-term heavy-duty operation due to minor damage to hydraulic components and dragging by water flow resistance of long-distance cables.

[0031] Furthermore, in underwater resource-constrained scenarios, to ensure the real-time operation of the computing system, the dwell time nodes in the national standard table for diving safety decompression are extracted in advance, and a high-speed query hash index is established. During real-time processing, the system reads the data from the table and combines it with a nonlinear exponential decay equation to periodically output a time-varying capacity supply vector carrying a real-time timestamp.

[0032] Step 103: Perform a feasibility assessment based on the capacity demand vector and the time-varying capacity supply vector, and perform optimization solution in conjunction with the task dependency graph to generate an initial scheduling scheme.

[0033] After completing the digital mapping of capabilities at both ends, the capability requirement vectors of each atomic task are cross-referenced with the time-varying capability supply vectors of the candidate task subjects. Based on the barrel effect criterion, when all examined dimension components of the subject's supply vector are not lower than the corresponding task requirement dimension components, the matching combination of the task subjects is deemed feasible. For feasible solution space combinations that meet the basic judgment conditions, the system uses the task dependency graph as a topological constraint for temporal collision avoidance and process steps, and imports it into the optimization algorithm module.

[0034] Accordingly, in the optimization of the objective function, the matching degree of supply and demand capacity margins for each task, the time window constraint of the curing failure of the repair material during the hydration reaction, and the handover and conversion time cost caused by personnel rotation or equipment relocation are comprehensively considered. The optimization algorithm searches in a targeted manner from the feasible spatiotemporal solution tree to find a global task allocation sequence and a time node arrangement table refined to the minute level that meet the safety red line of personnel and equipment, so that the total maintenance operation time approaches the lower bound, and outputs the initial scheduling scheme.

[0035] In some embodiments, a multi-objective hierarchical solution can be adopted when performing the optimization solution phase. For example, spatial pruning and filtering can be performed layer by layer according to the descending priority lexicographical logic of ensuring safety red lines, maximizing operation matching quality margins, and minimizing completion scheduling span. Under the constraints of this model, the algorithm model prohibits the system from exchanging the overall project schedule for a compromise of tolerating minor safety risks or reducing the intensity of basic repairs.

[0036] Step 104: Based on the real-time collected dynamic feedback information, select the currently executing initial scheduling scheme or the updated scheduling scheme, and combine it with the updated time-varying capacity supply vector to trigger local rescheduling and obtain the updated scheduling scheme.

[0037] In actual operation, the conditions at the bottom of deep reservoir dams are complex, with unpredictable disturbances that can easily cause purely static, one-time pre-allocation commands to fail during implementation. In this embodiment, the system continuously collects dynamic feedback information transmitted in real time by underwater acoustic sensors, optical vision modules, and physiological monitoring wristbands at high frequency. The dynamic feedback information includes at least the following: a sudden decrease in bottom current visibility in the operating area; a surge in water flow velocity caused by dam discharge; or the actual geometric depth of cracks exposed after surface high-pressure water jet cleaning exceeding the original model detection value.

[0038] When the deviation rate of model parameters caused by dynamic feedback exceeds the preset tolerance boundary dead zone, or when the actual rate of physical exhaustion of the underwater operator is identified as faster than the theoretical exponential model prediction curve, a soft interrupt is triggered, immediately freezing the current queued downstream task allocation status. The frozen historical data is stripped away, and using the latest damaged capability status and adverse environmental parameters as boundary initial conditions, a new branch-bound optimization operation is performed on the remaining, unstarted atomic task subset affected by the event. An updated scheduling scheme is generated through local reconstruction.

[0039] Optionally, this local triggering reconstruction mechanism, compared to global recalculation, can avoid the time lag effect caused by huge computing power overhead, and achieve the transition and correction of the on-site allocation scheme with minimal task instruction switching cost.

[0040] Step 105: Issue scheduling instructions based on the initial scheduling scheme or the updated scheduling scheme.

[0041] Under normal operating conditions without any abnormal triggers, the shore-based main control center encodes and compresses task blocks containing parameters such as the operation subject identification code, expected start and end times, and spatial coordinate reference, according to the initial scheduling plan. When an emergency occurs, the shore-based central hub performs local replanning, updates the scheduling plan, removes redundant information, and issues incremental patch instructions to the affected preset execution nodes to refresh their local timing tables.

[0042] Furthermore, to avoid the multipath effect and long-term packet loss risk caused by obstruction in deep-water acoustic communication, the scheduling instructions carry the plaintext of the operation execution sequence and can also include a lightweight local emergency degradation rule tree. When the underwater terminal's heartbeat packet detects a timeout due to a physical communication link interruption, it makes a decision based on the pre-set degradation rule tree. According to the predetermined priority logic, it drives the diver or robot to remain in place with its hull on standby, actively jettison its buoyancy and return to base, or skip high-risk communication-dependent sections to execute parallel tasks in other areas.

[0043] In one possible implementation, the following steps are included:

[0044] Step 101a: Extract the type and geometric features from the defect data to be repaired, and perform a two-level decomposition of the repair operation based on the pre-built standard process template library to obtain a set of atomic tasks.

[0045] Furthermore, the operational scenarios corresponding to the defect data to be repaired include crack grouting repair, erosion surface repair, and leakage sealing repair of hydraulic structures; the pre-built standard procedure template library contains pre-set standard procedure chains with parameterized codes that match each operational scenario.

[0046] Specifically, the defect data to be repaired includes defect type codes, spatial location coordinates, geometric feature parameters, and severity levels. Defect type codes uniquely identify the physical attributes of the defect, and the work scenarios include grouting repair of cracks in hydraulic structures, repair of eroded surfaces, and sealing of leaks. Geometric feature parameters include the width, depth, and length of cracks, or the area and depth of eroded regions. Using a pre-built standard procedure template library, matching standard procedure chains are retrieved based on the defect type code. The standard procedure template library is compiled according to the technical specifications for the repair of hydraulic structures and can parameterize the repair process for different scenarios into logically ordered procedure nodes.

[0047] For example, taking the crack grouting repair scenario, the retrieved standard process chain includes defect verification and location, crack surface cleaning, crack grooving and removal, sealing material application, grouting hole layout and drilling, grouting pipeline connection, grout preparation and injection, pressure maintenance, curing and waiting, and grouting effect detection. The task is decomposed according to a two-level decomposition strategy: the first level decomposes the repair operation into a process-level task sequence according to process nodes; the second level decomposes each process-level task into indivisible operational action units, thereby generating an atomic task set.

[0048] Step 101b: For each atomic task in the atomic task set, quantitatively evaluate the operational precision, operation duration, safety risk level, environmental tolerance, and degree of autonomous operation, and summarize them to form a capability requirement vector.

[0049] Specifically, for each atomic task in the atomic task set, a six-dimensional capability requirement vector is defined based on its process attributes and geometric feature parameters. The components of the capability requirement vector are normalized, and the corresponding expressions are as follows:

[0050] d _i =[d _i_pre ,d _i_for ,d _i_dur ,d _i_saf ,d _i_env ,d _i_aut ];

[0051] Where, d _i Let d be the capability requirement vector for the i-th atomic task. _i_pre To meet operational precision requirements, characterize the task's requirements for the positioning accuracy and operational dexterity of the end effector; d _i_for Force or torque requirement, characterizing the level of working force or torque required for the task; d _i_dur The task duration requirement represents the proportion of the minimum continuous execution time required for the task to the total planned time; d _i_saf To meet safety risk level requirements, it characterizes the severity of risks that may result from mission failure; d _i_env To meet environmental tolerance requirements, the feasibility requirements of the task under conditions of low visibility and strong water flow disturbance are characterized; d _i_aut The requirement for autonomous operation level represents the degree to which the task requires the operator to independently complete the operation without real-time command interaction.

[0052] In some optional implementations, the set of atomic tasks can be dynamically updated based on the defect distribution. When the repair target includes multiple defects, each defect is decomposed separately and then merged into the same set. If the tool type or material type corresponding to an atomic task changes, the weight components of the capability requirement vector are updated to recalibrate the task difficulty.

[0053] Step 101c: Extract the preceding and following constraint rules carried by each atomic task in the atomic task set; construct a directed acyclic graph with atomic tasks as nodes and preceding and following constraint rules as directed edges, and mark nodes without directed edges as parallelizable states.

[0054] Within the set of atomic tasks, the temporal logic between atomic tasks is extracted based on the process number of each atomic task and the pre- and post-constraint rules in the standard process chain. The pre- and post-constraint rules define the order of each operation unit; for example, the application of sealing material must be initiated after the crack grooving and removal task is completed.

[0055] Specifically, a directed acyclic graph (DAG) is constructed with atomic tasks as nodes and forward / backward constraints as directed edges. In this graph, the direction of the directed edges indicates the order of task execution. If a directed edge exists from task A to task B, it means that task A needs to be completed before task B begins. For subsets of tasks from different defect regions, if there is no dependency relationship in physical space or repair procedures, no directed edge connection is established in the task dependency graph, and the system marks them as parallelizable, improving scheduling efficiency.

[0056] Furthermore, based on the feedback from the site regarding the completion status of the processes, the task dependency graph is pruned in real time. Once an atomic task is completed and passes acceptance, that node and all directed edges originating from it are removed, and the topology of subsequent tasks to be processed is dynamically updated.

[0057] Optionally, critical paths in the task dependency graph are identified, and atomic tasks located on these critical paths are marked as high-priority tasks. During subsequent scheduling, these tasks are preferentially assigned to work entities with higher time-varying capacity supply vector margins, thereby reducing the risk of project delays.

[0058] Based on the above embodiments, this embodiment also provides a capability modeling method for when the operator is a diver, specifically including:

[0059] Step 201a: Extract diving operation parameters from real-time status data and determine the decompression time node sequence by combining them with the diving decompression specification database. That is, determine the decompression time node sequence that matches the current operating environment by combining the pre-stored diving decompression specification database.

[0060] Specifically, real-time status data includes diving operation parameters such as the absolute water depth of the operating area, the diver's entry time, and the bottom water temperature. The diving decompression specification database is a static mapping table pre-constructed based on the standard safety requirements for conventional underwater air diving. After obtaining these operation parameters, multidimensional interpolation is performed in the diving decompression specification database to output the corresponding decompression time node sequence. This sequence represents the theoretical limit on the diver's segmented bottom dwell time under the current water depth conditions, and the level of decompression obligation to be fulfilled after crossing the preset time node. The decompression time node sequence consists of monotonically increasing time nodes, with the last node corresponding to the maximum allowable underwater dwell time at that water depth.

[0061] Step 202a: Combining the diver's cumulative underwater working time from the diving operation parameters, calculate the step-type decompression constraint factor, which characterizes the degree of compression of the remaining available working time. That is, calculate the step-type decompression constraint factor based on the decompression time node sequence and the cumulative underwater working time.

[0062] Specifically, the system calculates the cumulative underwater operation time from the start of the diver's entry into the water to the current moment in real time and compares it with the decompression time node sequence. Within the interval before the diver triggers the first decompression obligation node, the diver can use all of their allotted underwater time for repair work; correspondingly, the system assigns a stepped decompression constraint factor of 1.0. When the cumulative underwater operation time crosses a certain decompression time node, the diver must reserve corresponding ascent decompression time, thus compressing the remaining time window actually available for repair work.

[0063] Calculate the ratio of the remaining available working time to the total available working time for the given node, and update this ratio as the step decompression constraint factor for the current stage. As diving operations continue, this factor exhibits a discrete step-like decreasing trend. When the maximum underwater dwell time is reached, the step decompression constraint factor is set to 0, indicating that the diver must terminate the operation.

[0064] Step 203a: Based on the pre-stored initial physical condition of the diver, cumulative underwater working time, and dimensional decay rate, calculate the continuous physiological decay factor characterizing the degradation of multidimensional operational capabilities.

[0065] Specifically, different decay rates are configured for the various operational capabilities of divers, with the decay rate for the operational precision dimension set to the highest, and the decay rate for the working force or torque dimension set to the second highest. Because the physiological degeneration mechanism of the human body exhibits multidimensional heterogeneity under cold water immersion or continuous load conditions, this embodiment extracts pre-configured decay rates for different capability dimensions of divers. The operational precision dimension is set to be sensitive to low-temperature water environments, and its corresponding decay rate value is set. The working force or torque dimension is set to be dominated by skeletal muscle fatigue accumulation, and its decay rate value is set, which should be lower than the decay rate of the operational precision dimension. For indicators such as environmental tolerance and autonomous operation level, which are relatively indirectly affected by physiological load, lower decay rates are configured based on test constants. In this embodiment, the specific values ​​of the dimension decay rates can be adaptively adjusted according to the actual water depth, water temperature, and other conditions of the operating water area; those skilled in the art can determine appropriate values ​​without creative effort.

[0066] In some optional implementations, the system is also equipped with an external environment dynamic compensation module. This module acquires real-time monitoring data on changes in water depth and temperature. For example, for every 5-meter increase in water depth, the system increases the attenuation rate of each dimension by 5% to 10%; for every 5°C decrease in water temperature, it increases the attenuation rate of the operational precision dimension by 10% to 20%, thereby compensating for the increased nonlinear physiological load caused by high pressure and low temperature.

[0067] Based on the matched dimensional decay rate, the continuous physiological decay factor corresponding to each capability dimension is calculated using a nonlinear time decay law. In this embodiment, a nonlinear computational model is used to process continuous time variables, outputting each continuous physiological decay factor. The computational relationship is as follows:

[0068] η _phy_k (t)=exp(-λ _k ×t);

[0069] Where, η _phy_k (t) represents the continuous physiological decay factor of the diver in the k-th capability dimension at time t, exp is an exponential function with the natural constant as the base, and λ _k The dimensional decay rate is configured for the k-th capability dimension, where t is the diver's cumulative underwater working time. An exponential function is used to simulate the continuous soft degradation process caused by fatigue accumulation.

[0070] Step 204a: Extract the diver's initial capability value, and couple the initial capability value, the step-wise decompression constraint factor, and the continuous physiological decay factor to generate the corresponding time-varying capability supply vector.

[0071] Specifically, the pre-stored initial ability values ​​of divers in each dimension are extracted. The divers' skill assessments and physical fitness tests are obtained and mapped to generate initial ability values ​​for each dimension; a multiplicative coupling operation is performed, with the corresponding calculation formulas as follows:

[0072] s _h_k (t)=s _h_0_k ×η _dec (t)×η _phy_k (t);

[0073] Among them, s _h_k (t) represents the time-varying capability of the diver in the k-th dimension at time t, s _h_0_k η represents the corresponding initial ability value for the diver. _dec (t) is the step-wise decompression constraint factor at time t, η _phy_k (t) represents the continuous physiological decay factor at time t. The components of each dimension are iterated and combined to output the complete time-varying capability supply vector.

[0074] By coupling the step-by-step decompression constraint factor with the continuous physiological decay factor, the hard constraint of the diver's available working time and the soft physiological degradation are uniformly represented, enabling the scheduler optimizer to predict the full-time evolution trajectory of the operator's capabilities during the planning stage, thus avoiding the risk of work interruption caused by mid-operation capability depletion from the source.

[0075] In another embodiment, this embodiment also provides a capability modeling method when the operating entity is an underwater robot, specifically including:

[0076] Step 201b: Extract the underwater robot's equipment operation parameters from the real-time status data, including battery state of charge, cumulative number of heavy-load operations, and cable release length.

[0077] The underwater robot obtains its battery state of charge through its internal battery management system; it extracts the cumulative number of heavy-load operations through the motion counter register of the robotic arm controller; and it reads the current cable release length through the encoder of the surface vessel winch. These equipment operating parameters reflect the underwater robot's status under long-term operating conditions.

[0078] For example, the cumulative number of heavy-load operations records the frequency of cycles in which the underwater robot performs heavy-load tasks such as high-pressure chiseling or powerful cutting. By continuously extracting this type of parameter, a quantitative mapping benchmark between physical equipment and the degradation of multi-dimensional operational capabilities can be established.

[0079] Step 202b: Calculate the energy degradation factor based on the deviation between the current battery state of charge and the safe return threshold.

[0080] The underwater robot's energy reserves also need to retain sufficient redundant power to overcome water resistance and return to the surface mother ship. For example, a preset safe return threshold can be set at 20% of the total battery capacity. The current battery state of charge is extracted, and the energy decay factor is calculated using a linear processing unit. The corresponding calculation formula is as follows:

[0081] φ _SOC (t)=max(0,(SOC(t)-SOC _min) / (SOC _0 -SOC _min) );

[0082] Where, φ _SOC (t) is the energy decay factor at time t, max is the function to maximize the value, and SOC(t) is the current state of charge of the battery at time t. _min SOC _0 The battery is initially fully charged. As the remaining power approaches the safety threshold, the energy decay factor decreases and is truncated to 0 when it is equal to or below the preset safe return threshold. This prevents the allocation of new tasks before physical resources are exhausted. The safe return threshold can be adaptively adjusted according to the equipment model in the actual operating waters, and those skilled in the art can determine the appropriate value without creative effort.

[0083] Step 203b: Calculate the hydraulic efficiency factor based on the cumulative number of heavy-load operations and the fatigue limit of the hydraulic system.

[0084] After frequent high-load operations, the hydraulic system of an underwater robot may experience a deviation in actual pressure output due to increased hydraulic oil temperature and minor wear of internal seals. By extracting the calibrated preset fatigue limit of the hydraulic system and quoting it with the current cumulative operating frequency, the degree of performance degradation can be quantified. The corresponding calculation formula is:

[0085] φ _hyd (t)=1-γ×(n _hv (t) / n _hv_max );

[0086] Where, φ _hyd (t) represents the hydraulic efficiency factor at time t, γ is the preset hydraulic attenuation coefficient, and n _hv (t) represents the cumulative number of overloaded jobs acquired at time t, n _hv_max This represents the preset fatigue limit for the hydraulic system. The preset hydraulic attenuation coefficient can be set to 0.2, characterizing the degradation sensitivity of the hydraulic pipeline under high-frequency heavy-load conditions. The specific value of the hydraulic attenuation coefficient in this embodiment can be adaptively adjusted according to the equipment model in the actual operating water area; those skilled in the art can determine a suitable value without creative effort.

[0087] Step 204b: Calculate the cable constraint factor based on the ratio of the cable release length to the total pre-stored cable length and the hydrodynamic resistance influence parameters.

[0088] As the underwater robot moves away from the mother ship, its towed cable experiences lateral hydrodynamic drag due to the bottom current. This drag restricts the robot's degrees of freedom of movement and increases the control error of the attitude-keeping algorithm.

[0089] In this embodiment, the ratio of the currently released cable length to the total cable length is calculated, and a preset hydrodynamic resistance influence parameter is introduced for exponential scaling to calculate the cable constraint factor. The calculation relationship is as follows:

[0090] φ _cab (t)=1-(l _out (t) / l _total ) μ ;

[0091] Where, φ _cab (t) is the cable constraint factor at time t, l _out (t) represents the cable release length at the current moment, l _total Where is the total length of the cable, and μ is the preset hydrodynamic resistance influence parameter. The value of the preset hydrodynamic resistance influence parameter is determined by the measured flow velocity of the operating water area and the cross-sectional diameter of the cable itself.

[0092] Step 205b: Extract the pre-stored initial capability values ​​of the underwater robot and couple them with three factors to generate the corresponding time-varying capability supply vector. Specifically, couple the initial capability values ​​with the energy attenuation factor, hydraulic efficiency factor, and cable constraint factor to generate the time-varying capability supply vector of the underwater robot.

[0093] Based on the manufacturer's technical specifications and the positioning calibration test results before launching, the baseline parameters of the underwater robot in six capability dimensions, such as operational precision, working force, or torque, were obtained, normalized, and used as initial capability values.

[0094] Accordingly, the matrix multiplication module is invoked to align the initial capability value with the three independent decay factors in terms of dimensions, and then merge them to output a time-varying capability supply vector that represents the current comprehensive work capability.

[0095] Furthermore, to describe the impact of different physical degradation sources on the preset operational dimension, this embodiment introduces a pre-configured dimension selection indicator variable for targeted intervention during the coupling process, specifically including:

[0096] By utilizing pre-configured dimension selection indicator variables, the hydraulic efficiency factor is exclusively mapped to the working force or torque dimension, and the corresponding dimension component in the initial capability value is attenuated. Minor leaks in the hydraulic system cause a decrease in absolute output force, but this does not interfere with the signal transmission of the control circuit. In this embodiment, independent indicator variables are configured for each capability dimension. For the working force or torque dimension, its indicator variable is set to 1; for the other five capability dimensions, the indicator variable is set to 0. During coupled calculation, the hydraulic efficiency factor is multiplied only by the dimension feature set to 1, ensuring that the degradation of hydraulic efficiency only affects the force output index.

[0097] By utilizing pre-configured dimension selection indicator variables, the cable constraint factor is exclusively mapped to the operational precision dimension and the environmental tolerance dimension, and the components of the corresponding dimensions in the initial capability value are attenuated. The random disturbance torque generated by dragging the cable can easily damage the hovering accuracy of the robotic arm's end effector and increase the probability of entanglement in complex reinforced concrete environments. In this embodiment, the indicator variables for the operational precision dimension and the environmental tolerance dimension are set to 1, while the indicator variables for the other dimensions are set to 0. Through this directional correlation mapping, during calculation, the cable constraint factor is multiplied by the corresponding parameters of the operational precision dimension and the environmental tolerance dimension, allowing the scheduling center to perceive the loss of accuracy during long-distance operations.

[0098] The energy attenuation factor is used as a global attenuation variable to proportionally attenuate all capability dimensions of the underwater robot. Unlike the degradation of preset physical components, the depletion of the main power module's power can easily cause the thrusters, robotic arm servo motors, and main control chip to simultaneously enter a low-power, frequency-reducing mode, thus leading to a decrease in overall performance. Therefore, the system does not set any exclusive mask for the energy attenuation factor. Instead, the energy attenuation factor is defined as a global multiplier variable, which is multiplied by the initial values ​​of the six capability dimensions during the coupling processing phase.

[0099] According to one aspect of this application, this embodiment also provides a method for feasibility determination and matching degree calculation, specifically including:

[0100] Step 301a: For the combination of each atomic task and each operating entity, calculate the expected completion time of the atomic task based on the standard execution time and planned start time of the atomic task, and determine whether, at the expected completion time, each dimension component of the time-varying capability supply vector of the operating entity is not lower than the corresponding dimension component of the capability demand vector.

[0101] Iterate through the unassigned atomic tasks and available job subjects to construct candidate combinations of tasks and subjects. For each candidate combination, obtain the standard execution time of the atomic task, and calculate the estimated completion time of the atomic task by combining it with the pre-scheduled planned start time.

[0102] Accordingly, the expected completion time is used as a time variable and substituted into the multi-factor coupled capacity degradation model to extract the time-varying capacity supply vector of the work entity at that moment. A dimension-by-dimensional comparison is performed to determine whether the six dimensions of the supply vector are greater than or equal to the corresponding dimensions of the capacity demand vector. If a supply component in any dimension is lower than the demand component, the combination is marked as infeasible and removed from the candidate solution space.

[0103] This judgment mechanism sets the time reference point at the expected completion time, forming a rigorous underlying engineering constraint for this system. Because the working capacity of the main body of deep-water operations decreases monotonically and non-linearly over time, if the capacity threshold check is only performed at the beginning of the task plan, the system is prone to misjudgment, leading to exhaustion of physical or electrical energy, causing the main body to lose control of its actions midway through the operation.

[0104] Step 302a: For combinations that meet the judgment conditions, a weighted sum is performed based on the normalized margins of the time-varying capacity supply vector relative to the capacity demand vector in each dimension to obtain the matching degree that characterizes the operational quality assurance capability, which can be used for subsequent optimization solutions.

[0105] For candidate combinations that pass the feasibility verification, the sufficiency of the operational entity's capabilities in covering task requirements is further quantitatively evaluated. Specifically, the absolute capability margin in each dimension is calculated, and the difference between the dimensional components of the time-varying capability supply vector and the dimensional components of the capability demand vector is obtained. The difference is divided by the initial capability value of the corresponding dimension for that operational entity to obtain the normalized margin. The pre-configured weight coefficients for each dimension are extracted, and the normalized margin is multiplied by the corresponding weight coefficient. The products of all dimensions are summed to output the matching degree, and the corresponding formula is:

[0106] m _ij =∑(w _k ×(s _j_k (t _i_e )-d _i_k ) / s _j_0_k );

[0107] Where, m _ij Let w represent the matching degree of the task subject j performing atomic task i, ∑ represent the summation operation performed on all six capability dimensions of k from 1 to 6, and w represent the matching degree of the task subject j performing atomic task i. _k For the pre-configured weight coefficients of the k-th dimension, s _j_k (t _i_e ) provides the k-th dimension capability supply component for task subject j at the expected completion time of atomic task i, d _i_k Let s be the k-th dimension of the capability requirement component for atomic task i. _j_0_k Let be the initial capability value of the k-th dimension of the task subject j.

[0108] The pre-configured weight coefficients are assigned values ​​based on the degree to which each capability in the engineering operation affects the success or failure of the repair, and the sum of the weight coefficients of all dimensions is limited to 1.0.

[0109] In one optional implementation, the system configuration parameter system follows a quantitative principle that prioritizes safety over repair quality, and repair quality over time efficiency. Accordingly, weighting coefficients are set for the safety risk level dimension directly related to personnel survival; weighting coefficients are set for the operational precision dimension and the working force or torque dimension that determine repair intensity; weighting coefficients are set for the environmental tolerance dimension and the degree of autonomous operation dimension that affect environmental adaptability; and weighting coefficients are set for the operation duration dimension. Through this weighted summation, the dispersed multidimensional spatial vectors are mapped into a scalar form of matching degree.

[0110] Based on the above embodiments, this embodiment also provides a method for constraining the time window of associated repair materials, specifically including:

[0111] Step 301b: Based on the pre-stored material performance parameter library, obtain the normal operable window duration and the extreme operable window duration of the repair materials associated with each atomic task in the atomic task set.

[0112] Among them, the associated repair materials include epoxy grouting material or polyurethane foam material; the normal operable window duration and the extreme operable window duration are pre-calibrated based on the characteristics of the rheological properties of the associated repair materials deteriorating over time, to ensure that the allocation of tasks is forcibly blocked after the extreme operable window duration is exceeded.

[0113] After formulation, the polymer material undergoes an irreversible cross-linking and curing reaction, resulting in a continuous deterioration of its rheological properties over time, manifested as an accelerated increase in viscosity and a rapid decline in underwater injection performance. Pre-calibrated time parameters were extracted from the material database. The normal operable window duration is defined as the duration for which the viscosity remains within a preset proportion of the initial baseline value after material formulation. Within this time period, the material exhibits optimal working performance.

[0114] The critical time point at which the material viscosity reaches a level that causes it to lose pumpability or effective permeability is defined as the ultimate operational window duration. Both time parameters are pre-calibrated based on the characteristics of the rheological properties of the associated repair material deteriorating over time.

[0115] Step 302b: Based on the time span between the planned preparation completion time of the associated repair material and the expected completion time of the atomic task, a preset piecewise function is used to calculate the material time window penalty value that increases with the time span. When the time span exceeds the limit of the operable window duration, the material time window penalty value is set to infinity.

[0116] Specifically, the time variable is obtained from the candidate task allocation scheme calculated by the scheduling engine. The preparation completion time of the associated repair material and the expected completion time of the atomic task to which the material belongs are extracted, and the difference between the two is calculated to obtain the time span. A preset piecewise function is used to quantify the negative impact of material degradation on repair quality, specifically:

[0117] P _i_mat (Δt _mat )=0, Δt _mat ≤W _i_norm ;

[0118] P _i_mat (Δt _mat )=β×((Δt _mat -W _i_norm ) / (W _i_max -W _i_norm )) 2 W _i_norm <Δt _mat ≤W _i_max ;

[0119] P _i_mat (Δt _mat )=+∞,Δt _mat >W _i_max ;

[0120] Among them, P _i_mat (Δt _mat ) represents the material time window penalty value corresponding to the i-th atomic task, Δt _mat To correlate the time span between the completion time of the repair material preparation and the expected completion time of the atomic mission, W _i_norm W represents the normal, operable window duration. _i_max β represents the maximum operable window duration, and β is the penalty coefficient.

[0121] Furthermore, the penalty coefficient β is used to control the relative magnitude of the material overtime penalty in the overall quality objective function. The value of β can be adaptively adjusted according to the equipment model and material batch in the actual operating water area. Those skilled in the art can determine the appropriate value without creative effort.

[0122] In the first stage interval, when Δt _mat Not greater than W _i_norm When the material properties are in a stable period, the material time window penalty value is set to 0. In the second stage interval, when Δt... _mat Greater than W _i_norm and not greater than W _i_max When the material properties degrade more rapidly, a quadratic nonlinear function is used to calculate the penalty value for this interval, making the penalty amplitude increase nonlinearly and monotonically with the increase of the timeout. In the third stage interval, when Δt_mat More than W _i_max When the material loses its feasibility for construction, the penalty value is set to +∞.

[0123] Based on the above embodiments, this embodiment provides a method for the optimization solution and scheduling scheme generation stage, specifically including:

[0124] A lexicographical hierarchical optimization framework is adopted, based on matching degree, material time window penalty value and task dependency graph, and the solution is performed in layers according to the priority order of safety, repair quality and operation efficiency. In the first layer of optimization, the safety risk index is calculated according to the gap between the capacity demand vector and the time-varying capacity supply vector in the dimension of safety risk level. With minimizing the safety risk index as the objective, the safe feasible region that meets the preset safety threshold and the directed topological constraints of the task dependency relationship is determined.

[0125] Underwater defect repair operations are characterized by high irreversibility and physical risks. Traditional weighted multi-objective optimization algorithms, through weight trade-offs, may sacrifice underlying safety for shorter project durations. To eliminate this risk, this embodiment constructs a lexicographically ordered hierarchical optimization framework. This framework sets a descending priority sequence, requiring that constraints on higher-priority objectives in the upper layers are absolutely valid when solving lower-priority objectives in the lower layers, and prohibiting any weight transfer or numerical compromise between layers.

[0126] Accordingly, the capability gap values ​​of the safety risk level dimension are extracted, and additional penalties are added to the pre-set decompression critical state of the diver to calculate an independent safety risk index.

[0127] Furthermore, for each combination of atomic tasks and operational entities, the capability gap value in the safety risk level dimension is calculated. That is, when the component of the time-varying capability supply vector in the safety risk level dimension is lower than the component of the capability demand vector in that dimension, the difference between the two is taken as the safety gap value for that combination; otherwise, it is assigned a value of zero. The safety gap values ​​of all combinations are summed to obtain the safety risk index.

[0128] In one optional implementation, the safety risk index can be defined as the weighted sum of the capability gap values ​​of all tasks in the safety risk level dimension and the additional penalty term for the decompression critical state. In this embodiment, an appropriate weighting method and penalty level can be selected according to the actual engineering safety level requirements.

[0129] Accordingly, a first-level optimization search is performed to filter out all task subjects whose safety risk index is lower than or equal to a preset safety threshold by assigning matrices. The preset safety threshold is compatible with reasonable minor measurement errors caused by system measurements. If the solution finds that the safe feasible region is an empty set, subsequent iterations are terminated, and a warning for adding or delaying device scheduling is output to the upper-layer interface. In this embodiment, the specific value of the safety threshold can be adaptively adjusted according to the actual water depth, water temperature, and equipment model of the operating area. Those skilled in the art can determine a suitable value without creative effort.

[0130] As an feasible approach, if two atomic tasks with a sequential dependency are assigned to different work entities for execution, then the start time of the subsequent atomic task is determined to be no earlier than the sum of the completion time of the preceding atomic task and the preset handover time cost of the work entity; the handover time cost of the work entity includes the time spent on personnel evacuation and realignment of equipment spatial reference; if two atomic tasks are assigned to the same work entity for execution, then the start time of the subsequent atomic task is determined to be no earlier than the completion time of the preceding atomic task.

[0131] Specifically, when defining the safe and feasible region, the directed edges of the task dependency graph are parsed synchronously. When the preceding and subsequent atomic tasks connected by the two ends of a directed edge are assigned to heterogeneous task subjects by the system decision, a buffer block is inserted on the timeline. This buffer block represents the subject handover time cost, and the corresponding calculation formula is as follows:

[0132] t _q_s ≥t _p_e +c _ho ;

[0133] Among them, t _q_s For the start time of the subsequent atomic task q, t _p_e c is the completion time of the preceding atomic task p. _ho The main handover time cost.

[0134] Furthermore, the handover time cost is quantified. For example, in the basic mode, it is set as a static time constant, which can be set to 10 minutes. In this embodiment, the value of the static time constant can be adaptively adjusted according to the equipment model in the actual operating water area, and those skilled in the art can determine a suitable value without creative effort.

[0135] In another dynamic alternative, the three-dimensional Euclidean distance between the two working entities is obtained based on an underwater acoustic positioning array. The time consumed by the maneuver transfer between the two points is calculated and used as the dynamic handover time cost of the entities. This cost is then substituted into the constraint equation to improve the accuracy of time prediction.

[0136] In the second-level optimization, within the safe and feasible region, the goal is to maximize the difference between the matching degree and the material time window penalty value to obtain the optimal quality value.

[0137] After establishing the safety boundary, the second-layer optimization logic is activated. In this embodiment, maximizing the overall repair quality is the sole search objective. A joint objective function is constructed, extracting the cumulative matching degree of all task combinations in the allocation scheme as a positive benefit contribution; and extracting the cumulative material time window penalty value of all tasks as a negative attenuation deduction term. Using a solver, all solution nodes belonging to the safe feasible region are traversed in the multidimensional discrete variable space, and the absolute difference between the two is calculated. The value that makes this difference reach the global maximum peak is recorded and locked as the optimal quality value.

[0138] Through this stage of solution, the selected scheme has the highest operational margin while meeting the safety baseline and minimizing quality defects caused by material rheological curing.

[0139] In the third-level optimization, the goal is to minimize the completion time of all tasks, and an initial scheduling scheme is output. Specifically, the quality tolerance parameter is determined based on the product of a preset proportional coefficient and the absolute value of the optimal quality value. Under the constraint that the current solution quality is not lower than the difference between the optimal quality value and the quality tolerance parameter, the task allocation and timing arrangement combination that minimizes the completion time is searched as the initial scheduling scheme.

[0140] Entering the third optimization stage at the bottom layer, the optimization goal shifts to improving operational efficiency, i.e., reducing the completion time of all task combinations. A quality tolerance parameter is introduced, defined as a value extracted as a percentage of absolute values. This avoids excessive material timeout penalties that could lead to negative optimal quality values. The corresponding quantification logic expression is:

[0141] ε _Q =ρ×|Q _opt |;

[0142] Where, ε _Q Here, ρ is the quality tolerance parameter, and |Q is the set proportionality coefficient. _opt This is a calculation operation that takes the absolute value of the optimal quality. The value of the proportionality coefficient ρ can be adaptively adjusted according to the conditions of the actual operating water area, and those skilled in the art can determine the appropriate value without creative effort.

[0143] During the third-level algorithm iteration, a nonlinear constraint equation is constructed, namely, the quality function value of the current evaluation branch is greater than or equal to the difference between the optimal quality value and the quality tolerance parameter. Under the dual locking conditions of the floating quality lower bound and the safe feasible region, the process interleaving arrangement calculation is performed to find the solution that minimizes the overall scheduling span.

[0144] As an feasible approach, when the number of tasks does not exceed a preset scale threshold, the branch and bound method is used to perform exact solution and the lower bound of the critical path is used for bounding and pruning. When the number of tasks exceeds the preset scale threshold, a greedy heuristic algorithm is used to generate an initial feasible solution and a neighborhood improvement solution is performed in combination with a tabu search algorithm. Throughout the search process, the hard constraint of the safe feasible region is always maintained.

[0145] In this embodiment, the total number of atomic tasks to be assigned is extracted and compared with a preset scale threshold. The preset scale threshold is set based on the upper limit of the computing power of the currently equipped computing chip. When the number of tasks to be assigned is small, the branch and bound method is activated for global traversal. The topological sorting result of the task dependency graph is used as the basis for branching the state tree, and the lower bound of the time consumption of the currently arranged paths is calculated in real time. When it is detected that the lower bound is greater than the currently known global optimal completion time, a bounding and pruning operation is performed to cut off redundant computing power consumption. In this embodiment, the preset scale threshold can be adaptively adjusted according to the equipment model of the actual operating water area, and those skilled in the art can determine a suitable value without creative effort.

[0146] Correspondingly, when the scale of repairs is large and the number of tasks exceeds a preset threshold, precise solutions can easily lead to dimensionality explosion, resulting in response timeouts. To address this, the system employs a large-scale mixed-integer programming approximation strategy. Specifically, an initial solution vector is constructed by executing a greedy heuristic algorithm. In other words, according to the topological sorting order of atomic tasks in the task dependency graph, the system sequentially selects the task entity with the highest current matching degree and satisfying final state feasibility for each atomic task, thus constructing an initial feasible allocation solution.

[0147] Based on this, the solution vector is input into the tabu search algorithm, which iteratively executes the exchange actions of the single-task task to generate a new neighborhood solution group. After each exchange, the verification engine is called to determine whether it exceeds the safe feasible region, and any exchange actions that violate the safety baseline are eliminated. Then, within the millisecond-level computation time constraint, a scheduling sequence that approximates the optimal solution is output.

[0148] Furthermore, this embodiment provides a method for dynamic monitoring and emergency control mechanisms, specifically including:

[0149] The system monitors dynamic feedback information and generates trigger event records when abnormal operating conditions are identified. Abnormal operating conditions include deviations in defect parameters, abnormal degradation of the main body's capabilities, related repair materials entering the critical operable range, or changes in operating environment parameters exceeding preset threshold values.

[0150] Specifically, for deviations in defect parameters, the measured crack geometry is compared with the original parameters. When the deviation exceeds a preset spatial tolerance threshold, a judgment mechanism is activated. This preset spatial tolerance threshold can be set based on the measurement accuracy of the underwater detection equipment and the sensitivity of the defect repair process to the geometric parameters. For actual degradation rates exceeding preset degradation tolerances, the capacity supply value estimated using the physiological degradation formula is compared with the actual vital signs or electrical readings obtained by the current sensors. If the actual degradation rate exceeds the preset degradation tolerance, an anomaly is determined. For associated repair materials entering the critical operable range, the material's residence time after preparation is monitored. When the remaining normal operable window duration is less than the lower bound of the expected time for the associated task to be executed, an event is triggered. For changes in operating environment parameters exceeding a preset jump threshold, the underlying flow velocity or visibility index is extracted in real time. If the jump magnitude of this index within the monitoring time window exceeds the initial safety assumption, the system immediately captures the sudden change.

[0151] When any of the four scenarios is identified, the system generates a trigger event record containing a timestamp, anomaly category code, and deviation magnitude, and sends it to the highest priority channel of the central control queue. During underwater repair operations, continuous dynamic feedback information is collected through various types of sensors, and four types of parallel event-driven triggering logic are configured.

[0152] When no trigger event is recorded, based on a preset evaluation period, the predicted target value calculated from the currently executing initial scheduling scheme or an existing updated scheduling scheme is compared with the theoretically optimal target value recalculated using the latest calculated time-varying capacity supply vector. When the deviation between the two exceeds a preset improvement threshold, a periodic rescheduling signal is generated. The preset improvement threshold can be set by those skilled in the art based on the need to balance the scheme update frequency and computing resource overhead in actual operational scenarios.

[0153] To prevent subtle cumulative errors from causing hidden performance degradation in long-term operations, this embodiment establishes a periodic evaluation mechanism independent of event-driven processes. For example, one-third or one-half of the average working time of all atomic tasks is used as a preset evaluation period. Whenever this evaluation period node is reached, the latest capability status, such as equipment battery level and personnel downtime, is extracted, and the global matching degree matrix is ​​recalculated.

[0154] Accordingly, the algorithm interface is invoked to calculate the theoretically optimal quality function value achievable in the latest state, and the predicted quality function values ​​for the remaining paths of the currently executing initial scheduling scheme are extracted. The potential optimization space is quantified by calculating the difference between the two values; the calculation formula is as follows:

[0155] ΔQ=Q _new -Q _current ;

[0156] Where ΔQ is the quality deviation of the scheme, Q _new To re-optimize the theoretically optimal target value obtained from the latest capability state, Q _current This is the predicted target value of the current execution plan. ΔQ is compared with a preset improvement threshold. If the deviation is greater than the threshold, it indicates that the current plan is not suitable for the current slowly changing supply and demand situation, and the system immediately generates a periodic rescheduling signal.

[0157] In response to trigger event logs or periodic rescheduling signals, the allocation status of the set of completed atomic tasks is locked, and the optimization solution is re-executed for the remaining set of atomic tasks that have not yet started based on the updated capability status calculated using the latest real-time status data and the environmental parameters extracted from dynamic feedback information, generating an updated scheduling scheme.

[0158] In this embodiment, the task feedback flag of each working entity is checked, and the atomic task entities marked as completed or being executed are locked, so that their corresponding allocation decision variables are constrained to constants.

[0159] Accordingly, the latest measured environmental flow rate, visibility, and updated main multidimensional capability supply benchmarks are extracted. The remaining atomic tasks marked as not yet started in the queue are then input again into the safety-first lexicographical three-layer hierarchical optimization engine for solution. By reconstructing the local residual topology graph, an updated scheduling scheme adapted to the current actual working conditions is obtained while meeting millisecond-level computational latency requirements, and this scheme is then replaced in the on-site execution sequence.

[0160] Step 401: When the communication link between the shore-based and underwater control terminals is detected to be interrupted, the control operation entity continues to execute according to the locally stored scheduling instructions.

[0161] Underwater acoustic channels suffer from narrow bandwidth, severe multipath interference, and susceptibility to physical obstruction and disconnection. In this embodiment, the control terminals of each underwater operator are equipped with heartbeat packet monitoring modules. When the duration of continuous absence of shore-based acknowledgment frames exceeds a preset disconnection threshold, a communication link interruption is determined. After establishing the disconnection status, the underwater control terminal takes over the execution control of the operator, extracts the scheduling instruction sequence that was fully downloaded to the local buffer during the previous normal communication period, and drives the diver or underwater robot to proceed with the repair process according to the predetermined spatiotemporal logic. In this embodiment, the preset disconnection threshold can be set based on the typical round-trip time and packet loss rate characteristics of the underwater acoustic communication link; those skilled in the art can determine a suitable value without creative effort.

[0162] Step 402: During the communication link interruption, if a local anomaly is identified based on dynamic feedback information, the graded emergency fine-tuning rules are matched and activated in descending priority order: personnel safety protection, equipment protection, mission continuity maintenance, and material timeliness protection. The graded emergency fine-tuning rules are pre-stored in the underwater control terminal.

[0163] A lightweight decision tree model is pre-programmed into the underwater control terminal to handle unforeseen circumstances where shore-based computing power is unavailable during disconnection. When underwater environmental sensors or physiological feature extraction units detect parameters exceeding limits, diagnostics are performed by traversing the system in descending order of priority.

[0164] Specifically, scenarios involving abnormal heart rate of divers or pressure drop in breathing gas supply will be matched to the highest level of personnel safety protection rules; scenarios involving underwater robot battery capacity dropping below 5% or hydraulic main pipeline depressurization will be matched to the next highest level of equipment protection rules; scenarios where local water flow interference causes a single process to be blocked, provided that no personnel or equipment damage is involved, will be matched to the task continuity maintenance rules; and scenarios where it is detected that the time window for carrying grouting material is about to exceed the normal operating time will be matched to the material aging protection rules.

[0165] Step 403: Based on the activated hierarchical emergency fine-tuning rules, control the operation subject to perform fine-tuning actions such as local safe standby, autonomous return, task sequence jump or accelerated execution, and report the fine-tuning status after the communication link is restored to trigger global rescheduling.

[0166] In this embodiment, different levels of fine-tuning rules are mapped to specific execution actions. Specifically, when the personnel safety protection rule is activated, all operations being performed by the diver are immediately blocked, instructing them to maintain their posture in place or ascend along the guide rope to the first decompression stop depth for safe standby. When the equipment protection rule is activated, the heavy-duty hydraulic output is cut off, the backup battery pack is invoked, and the propeller is driven to perform autonomous return along the recorded historical trajectory. When the task continuity maintenance rule is activated, and the current pending task is blocked, a query is performed to find parallel branch atomic tasks with no dependency constraints in the local topology sequence, and a task sequence jump is executed. When the material aging protection rule is activated, the upper limit threshold of the equipment power output is adjusted, and an accelerated execution operation is performed to seize the curing time window.

[0167] When the heartbeat packet re-establishes the connection, the underwater control terminal packages all state changes, position displacements, and task status transition logs that occurred during the disconnection period and reports them to the shore-based computing center. This log is extracted and the global operational environment profile is reconstructed, triggering the algorithm engine to execute a system-wide global rescheduling, and the latest global convergence command is then reissued.

[0168] For example, if the underwater control terminal detects that the diver's heart rate remains above a preset safety limit during a communication interruption, it will match the highest level of personnel safety protection rules, immediately instructing the diver to cease operations and ascend along the guide rope to the first decompression stop depth to await further instructions. Once communication is restored, this status will be reported to the shore-based center, triggering a global rescheduling of the remaining tasks.

[0169] Based on the above embodiments, this embodiment provides a method for offline calibration of system parameters, specifically including:

[0170] Step 501: Extract historical engineering procedures for the repair of defects in hydraulic structures, parameterize and encode the repair process of typical defects, and obtain a pre-built standard procedure template library.

[0171] Specifically, offline data mining and parameterization transformation are performed before the underwater repair operation begins.

[0172] This study acquires textual data of current technical guidelines for defect repair and reinforcement in water conservancy projects, extracting standardized operational procedures for defects such as crack grouting and erosion surface repair. Using parametric coding technology, the repair procedures described in natural language are converted into a computer-readable directed graph structure sequence of nodes. A unique equipment identification code and a corresponding estimated resource consumption baseline are assigned to each process node. The entire encoded data is then stored, and a pre-built standard process template library is output.

[0173] Step 502: Based on diving decompression medical literature and underwater equipment factory test standards, under the set typical water depth and water temperature reference conditions, the characteristic decay time constants of divers and underwater robots in each capability dimension are measured respectively, and the pre-configured dimensional decay rate is obtained for online calculation of continuous physiological decay factor and degradation constraint.

[0174] To ensure the physical and medical realism of the core parameters in the attenuation model, a multi-dimensional parameter calibration matrix was established during the offline phase. Specifically, standard test benchmarks were set, such as a typical water depth benchmark of 15 to 20 meters and a typical water temperature benchmark of 10°C to 15°C. Under these benchmark conditions, time series data of hand dexterity tests and isomotor muscle strength tests of divers in a cold water simulation chamber were acquired. The characteristic time scales of the attenuation of the ability indicators were extracted and defined as the characteristic attenuation time constant.

[0175] Next, the actual dimensional decay rate is calculated using this characteristic decay time constant. The calculation formula is as follows:

[0176] λ _k =1 / τ _k ;

[0177] Where, λ _kLet τ be the dimensional decay rate of the k-th capability dimension obtained by mapping. _k Let be the measured characteristic decay time constant of the k-th capability dimension. Substitute the measured characteristic decay time constant into this division operation.

[0178] Furthermore, to address situations where the actual construction water environment deviates from the calibration benchmark, this embodiment includes a dynamic parameter compensation rule table. This involves pre-setting compensation conditions; for example, when the actual operating water depth exceeds the upper limit of the benchmark, and for every 5-meter increase in water depth, the system instruction calculation unit multiplies the attenuation rate of each dimension by a preset correction coefficient. When the actual bottom water temperature is below the lower limit of the benchmark, and for every 5°C decrease in water temperature, the attenuation rate of the operational precision dimension is separately added and multiplied by a preset temperature compensation coefficient. In this embodiment, the correction coefficient and temperature compensation coefficient can be adaptively adjusted according to the actual operating water depth, water temperature, and other conditions; those skilled in the art can determine appropriate values ​​without creative effort. The calibrated parameter matrix and compensation rules are encapsulated as a whole and stored in the calculation module.

[0179] This invention, by introducing a multi-factor coupled time-varying capability supply model and a final-state feasibility verification mechanism, can eliminate the high-risk hidden danger of mid-course capability depletion during the task allocation phase; through a lexicographical hierarchical optimization framework and a material time window penalty mechanism, it can balance repair quality and material timeliness while ensuring a safety baseline; and through hybrid trigger rescheduling and disconnection hierarchical autonomous rules, it can improve the robustness of the operational formation in dynamic underwater environments. Those skilled in the art will understand that the specific performance improvement may vary depending on the operating conditions, equipment configuration, and environmental parameters.

[0180] This embodiment provides a human-machine task decomposition and scheduling system for underwater defect repair of reservoir dams based on dynamic matching, including a memory and a processor. The memory is used to store computer programs, and the processor is used to execute the computer programs to implement the steps of any of the methods in the above embodiments.

[0181] In one alternative deployment method, the processor and memory are deployed in a shore-based master control computing unit, which is connected to the control terminal of the underwater operation entity via an underwater acoustic communication link or an umbilical fiber optic link.

[0182] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for human-machine task decomposition and scheduling based on dynamic matching for underwater defect repair of reservoir dams, characterized in that, include: Obtain the defect data to be repaired, decompose the repair work to obtain the set of atomic tasks and their capability requirement vectors, and construct the task dependency graph; Acquire and process the real-time status data of the main body of the operation to construct the time-varying capacity supply vector of the main body of the operation; Feasibility determination is performed based on the capacity demand vector and the time-varying capacity supply vector, and optimization is performed in conjunction with the task dependency graph to generate an initial scheduling scheme. Based on the dynamic feedback information collected in real time, the initial scheduling scheme or the updated scheduling scheme currently in execution is selected. Combined with the updated time-varying capacity supply vector, local rescheduling is triggered to obtain the updated scheduling scheme. Dispatch instructions are issued based on the initial or updated scheduling scheme; In constructing the time-varying capability supply vector for the main operator, when the main operator is a diver, it specifically includes: Diving operation parameters are extracted from real-time status data and combined with the diving decompression standard database to determine the decompression time node sequence; By combining the diver's cumulative underwater working time in the diving operation parameters, a step-type decompression constraint factor is calculated to characterize the degree of compression of the remaining available working time. Based on pre-stored initial physical condition of divers, cumulative underwater working time and dimensional decay rate, calculate the continuous physiological decay factor characterizing the degradation of multidimensional operational capabilities; Extract the initial capability value of the diver, and couple the initial capability value, the step decompression constraint factor and the continuous physiological decay factor to generate the corresponding time-varying capability supply vector. In constructing the time-varying capability supply vector of the operating entity, when the operating entity is an underwater robot, it specifically includes: Extract the underwater robot's equipment operation parameters from real-time status data, including battery state of charge, cumulative number of heavy-load operations, and cable release length. The energy degradation factor is calculated based on the deviation between the current battery state of charge and the safe return threshold. The hydraulic efficiency factor is calculated based on the cumulative number of heavy-load operations and the fatigue limit of the hydraulic system. The cable constraint factor is calculated based on the ratio of the cable release length to the total pre-stored cable length and the hydrodynamic resistance influence parameters. Extract the pre-stored initial capability values ​​of the underwater robot, couple them with three factors, and generate the corresponding time-varying capability supply vector; Perform a feasibility assessment and, in conjunction with the task dependency graph, perform optimization to generate an initial scheduling scheme, including: For each combination of atomic tasks and each operating entity, the expected completion time of the atomic tasks is calculated based on the standard execution time and planned start time of the atomic tasks. It is then determined whether each dimension of the time-varying capacity supply vector of the operating entity is not lower than the corresponding dimension of the capacity demand vector at the expected completion time. For combinations that meet the judgment criteria, the matching degree, which characterizes the operational quality assurance capability, is obtained by weighted summation based on the normalized margins of the time-varying capacity supply vector relative to the capacity demand vector in each dimension.

2. The method according to claim 1, characterized in that, Obtain the defect data to be repaired, and decompose the repair task to obtain a set of atomic tasks and their capability requirement vectors, including: Extract the type and geometric features from the defect data to be repaired, and perform a two-level decomposition of the repair operation based on a pre-built standard process template library to obtain a set of atomic tasks; For each atomic task in the atomic task set, the operational precision, operation duration, safety risk level, environmental tolerance, and degree of autonomous operation are quantitatively evaluated, and the results are summarized to form a capability requirement vector.

3. The method according to claim 2, characterized in that, The operational scenarios corresponding to the defect data to be repaired include crack grouting repair, erosion surface repair, and leakage sealing repair of hydraulic structures; The pre-built standard process template library contains pre-set standard process chains with parameterized codes that match various work scenarios.

4. The method according to claim 1, characterized in that, Construct a task dependency graph, including: Extract the pre- and post-constraint rules carried by each atomic task in the atomic task set; Construct a directed acyclic graph with atomic tasks as nodes and forward and backward constraint rules as directed edges, and mark nodes without directed edges as parallel-assignable states.

5. The method according to claim 1, characterized in that, The process includes performing a feasibility assessment, optimizing the solution based on the task dependency graph, generating an initial scheduling scheme, and also includes: Based on a pre-stored library of material performance parameters, the normal operable window duration and the extreme operable window duration of the repair materials associated with each atomic task in the atomic task set are obtained. Based on the time span between the planned preparation completion time of the associated repair material and the expected completion time of the atomic task, a preset piecewise function is used to calculate the material time window penalty value that increases with the time span. When the time span exceeds the limit of the operable window duration, the material time window penalty value is set to infinity.

6. The method according to claim 1, characterized in that, After issuing dispatch instructions based on the initial or updated dispatch scheme, the process also includes local emergency fine-tuning of the underwater terminal, specifically including: When the communication link between the shore-based and underwater control terminals is interrupted, the control operation entity continues to execute according to the locally stored scheduling instructions; During a communication link interruption, if a local anomaly is identified based on dynamic feedback information, the graded emergency fine-tuning rules are matched and activated in descending order of priority: personnel safety protection, equipment protection, task continuity maintenance, and material timeliness protection. Based on the activated hierarchical emergency fine-tuning rules, the control unit performs fine-tuning actions such as local safe standby, autonomous return, task sequence jump, or accelerated execution, and reports the fine-tuning status to trigger global rescheduling after the communication link is restored.

7. The method according to claim 1, characterized in that, Before acquiring the defect data to be repaired and breaking down the repair job into a set of atomic tasks, an offline build step is also included: Historical engineering procedures for the repair of defects in hydraulic structures are extracted, and the repair process for typical defects is parameterized and coded to obtain a pre-built standard procedure template library. Based on diving decompression medicine literature and underwater equipment factory test standards, under the set typical water depth and water temperature benchmark conditions, the characteristic decay time constants of divers and underwater robots in each capability dimension were measured respectively, and the pre-configured dimensional decay rate was obtained for online calculation of continuous physiological decay factor and degradation constraint.

Images