A Resource Scheduling Optimization Method for Deep-Sea Mining Integrated Control System

Abstract

This invention discloses a resource scheduling optimization method for a deep-sea mining integrated control system, relating to the field of deep-sea equipment automation and control technology. By creating a virtual resource pool and elastic task blocks, this invention fundamentally solves the scheduling barriers caused by resource heterogeneity and task complexity in deep-sea mining systems. It encapsulates the attributes and states of various physical resources into a virtual resource pool with a unified service interface, allowing the scheduling system to query, reserve, and allocate resources without needing to concern itself with the physical implementation details of the resources, relying solely on standardized interfaces. Simultaneously, it decomposes macroscopic mining tasks and defines them as fine-grained elastic task blocks. Each task block clearly declares its required resource type, quantity, and target location. This decoupled architecture transforms the resource scheduling problem into a dynamic matching problem between task blocks and resource pools, enabling the rapid allocation of any task block to any resource pool with the corresponding capabilities.

Description

Technical Field

[0001] This invention relates to the field of deep-sea equipment automation and control technology, specifically a resource scheduling optimization method for a deep-sea mining integrated control system. Background Technology

[0002] Deep-sea mining is a key approach to obtaining mineral resources such as polymetallic nodules, cobalt-rich crusts, and hydrothermal sulfides from the seabed. Its operational system is a typical complex mega-system, usually consisting of surface support vessels, underwater relay stations, multiple heterogeneous seabed mining vehicles, and long-distance transportation systems. This system operates in extreme environments and faces inherent challenges such as high pressure, low temperature, darkness, and complex seabed topography. Furthermore, the high latency, narrow bandwidth, and poor reliability of underwater acoustic communication make it difficult for the ground control center to conduct real-time and precise control of the underground dynamics.

[0003] Currently, resource scheduling methods in this field mostly remain at the level of static planning or local optimization, resulting in systemic defects. First, at the resource abstraction level, existing technologies lack a unified description of heterogeneous resources. Surface vessels, relay stations, and different types of mining vehicles operate independently, and their energy, capacity, and operational capabilities cannot be uniformly recognized and allocated by the system, forming resource silos. Second, at the optimization decision-making level, existing methods are mostly single-objective optimizations, lacking a coordinated trade-off between mutually constraining objectives such as energy consumption, communication, and equipment wear and tear, failing to achieve optimal system-level efficiency. Furthermore, at the execution coordination level, multi-mining vehicle clusters typically employ strict centralized control, with all action commands relying on the control center to issue them via high-latency communication links. This mode is slow to respond and cannot cope with dynamic events such as sudden equipment failures underground or the discovery of new rich ore areas. Moreover, once the central command is interrupted, the entire cluster becomes paralyzed.

[0004] In summary, existing deep-sea mining resource scheduling technologies suffer from fundamental deficiencies in resource abstraction, global optimization, and cluster collaboration, severely restricting the efficiency, economy, and safety of deep-sea mining operations. Therefore, there is an urgent need in this field for a comprehensive control method capable of unified resource scheduling, multi-objective global optimization, intelligent cluster collaboration, and system resilience self-recovery. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a resource scheduling optimization method for a deep-sea mining integrated control system. This method fundamentally solves the scheduling barriers caused by resource heterogeneity and task complexity in deep-sea mining systems by creating virtual resource pools and elastic task blocks. It encapsulates the attributes and states of various physical resources into virtual resource pools with a unified service interface, allowing the scheduling system to query, reserve, and allocate resources without needing to concern itself with the physical implementation details. Simultaneously, it decomposes macroscopic mining tasks into fine-grained elastic task blocks, each clearly declaring its required resource type, quantity, and target location. This decoupled architecture transforms the resource scheduling problem into a clear dynamic matching problem between task blocks and resource pools, enabling flexible and rapid allocation of any task block to any resource pool with corresponding capabilities. This achieves a fundamental shift from planning tasks for a specific mining vehicle to finding the most suitable resources for a task.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, a resource scheduling optimization method for a deep-sea mining integrated control system, the specific steps of which are as follows:

[0007] S100, Resource and Task Standardization: All heterogeneous physical resources in the deep-sea mining integrated control system are assigned to a virtual resource pool with a unified interface definition, while macro-mining tasks are decomposed into multiple independent elastic task blocks.

[0008] S200, Global Optimization Decision: Based on the real-time status information of all virtual resource pools in S100 and the queue of all pending elastic task blocks, a multi-objective optimization model that comprehensively considers output, energy consumption and communication is constructed and solved to generate specific execution instructions;

[0009] S300, Cluster Autonomous Collaborative Execution: The mining vehicle cluster receives execution instructions issued by S200, shares its state and intent with neighboring intelligent agents through workshop communication, performs local path planning and behavior coordination, and completes its respective task blocks without conflict.

[0010] S400, System Status Monitoring and Fault Diagnosis: When the agent executes a task block, it monitors the health status indicators of all virtual resource pools in real time. When it is found that the health status value of any virtual resource pool is lower than the preset health status threshold or the device malfunctions, it immediately triggers the abnormal response process.

[0011] S500, Dynamic Rescheduling and Resilient Recovery: In response to anomalies, the system locks all ongoing and assigned task blocks associated with the faulty resource pool, marks them as pending reallocation and increases their priority, triggers S200 for rescheduling, and executes the rescheduling by the remaining mining vehicle cluster in S300 to achieve continuity and self-recovery of system operations.

[0012] Furthermore, in S100, the virtual resource pool includes four types of standardized resource pools:

[0013] Energy Pool: Integrates generators from surface ships, batteries from mining vehicles, and buffer energy from relay stations; the interface protocol supports real-time capacity queries and reservation requests.

[0014] Task capability pool: integrates robotic arm gripping force, crusher power, and mineral processing rate; interface protocol supports task capability query.

[0015] Capacity pool: combines ship cargo holds and mining vehicle mineral storage space, with interface protocols supporting space storage allocation;

[0016] Communication pool: Integrates relay station bandwidth and workshop communication links, and the interface protocol supports priority preemption scheduling.

[0017] Furthermore, in S100, the process of decomposing and defining the macro-mining task into elastic task blocks specifically involves:

[0018] Based on the seabed map and mineral abundance data from the exploration, the target mining area is divided into multiple regular operation grids, and the sub-tasks of each operation grid are defined according to the mission objectives.

[0019] For each initially defined subtask, the specific values ​​of various resource types consumed and occupied for its execution are quantified. The various resource types occupied correspond to the virtual resource pool categories, including the required energy, required operational capacity, expected cabin capacity occupied, and required communication bandwidth.

[0020] For each subtask, a standardized task descriptor is created, which includes a unique task identifier, task geographic coordinates, resource requirement vector, task priority, and latest completion time, thereby forming a unified elastic task block.

[0021] Furthermore, in S200, the process of constructing the multi-objective optimization model is as follows:

[0022] Define decision variables: The decision variables Boolean matrix , Represents a flexible task block Should it be allocated to a virtual resource pool? Execution, and continuous variables , respectively representing task blocks In the resource pool Start time on and end time ;

[0023] Establish a multi-objective optimization function: The multi-objective optimization function includes the objective of maximizing the total output of the system. Minimize the total energy consumption of the system 1. Maximize communication efficiency ;

[0024] Constraint integration: The constraints dynamically integrate information from multiple parties, including:

[0025] Resource constraints: Ensure that the total amount of resources allocated to all elastic task blocks does not exceed the real-time storage space of the corresponding virtual resource pool;

[0026] Spatiotemporal conflict-free constraint: Two tasks that are not spatiotemporally compatible cannot be executed simultaneously in the same resource pool;

[0027] Communication bandwidth constraint: The sum of the data flow rates of all concurrent tasks on any communication link must not exceed the instantaneous capacity of that link;

[0028] Task priority constraint: High-priority tasks must be completed before their deadline;

[0029] Constructing a multi-objective optimization model: Integrating multi-objective optimization functions and constraints to form a multi-objective optimization model;

[0030] Linked solution: Taking the state of the elastic task block queue and virtual resource pool generated by S100 as input, under the premise of satisfying all constraints, find the set of decision variable vectors X such that the objective vector F(X) reaches the Pareto optimal solution set.

[0031] Furthermore, in step S200, when solving the multi-objective optimization model in a coordinated manner, a non-dominated sorting genetic algorithm is used to generate a Pareto optimal solution set. The solution process of the non-dominated sorting genetic algorithm is as follows:

[0032] Initialization: Generate an initial population of size N, where each individual in the population represents a complete scheduling scheme, i.e., the encoding of a decision variable vector X;

[0033] Evaluation: Calculate the three objective function values ​​for each individual in the population, stratify all individuals in the population according to the non-dominated sorting rule, and calculate the crowding degree of each individual within the same stratum;

[0034] Selection, crossover, and mutation: Based on non-dominated ordering and crowding, parent individuals are selected through binary tournament selection, and simulated binary crossover and polynomial mutation operations are applied to generate offspring populations.

[0035] Elite preservation: The parent and offspring populations are merged, and the non-dominated ranking and crowding degree are recalculated in the merged large population. The top N best individuals are selected to form a new generation population.

[0036] Termination judgment: Repeat the evaluation to elite retention operation until the preset maximum number of generations is reached, and finally output the last set of individuals that represent the Pareto optimal frontier, that is, the Pareto optimal solution set. Each solution in this solution set represents a feasible scheduling scheme.

[0037] Optimal scheduling scheme decision: Select a final scheduling scheme from the Pareto optimal solution set based on the current job strategy;

[0038] Compiling and issuing scheduling instruction sets: The selected final scheduling scheme is decoded into a scheduling instruction set, and a specific execution instruction is generated for each allocated elastic task block. The content includes: execution resource pool ID, task block ID, start time, end time and expected resource consumption list. All execution instructions are grouped and packaged according to resource pools and issued to the corresponding mining vehicle cluster in S300 to start execution.

[0039] Furthermore, in S300, the specific process of performing local path planning and behavior coordination is as follows:

[0040] Command reception and parsing: Each mining vehicle acts as an intelligent agent. When executing the assigned task block, each intelligent agent in the cluster receives the issued execution command and parses out its own assigned task block sequence.

[0041] Local perception and communication: each agent It acquires local environmental information based on its own sensors, and periodically broadcasts its own status, including position, speed, and remaining tasks, through workshop communication, and receives information from neighboring agents. Status information;

[0042] Distributed collaborative decision-making: each agent In each decision cycle Within, for the received neighboring agents Risk of conflict in status information The assessment, the ,in, for In-time intelligent agent and Collision risk measurement between them and respectively intelligent agents and At any moment Location, This is the minimum safe collision avoidance radius; any distance less than this is considered a collision hazard. At that time, it was determined that there was a trajectory conflict;

[0043] Speed ​​negotiation and adjustment: When a trajectory conflict is detected, the relevant agents will negotiate to dynamically adjust their respective speeds and temporarily modify local waypoints in order to achieve avoidance and cooperative coverage.

[0044] Furthermore, in S500, the exception response process is specifically as follows:

[0045] When the S400 diagnoses a virtual resource pool failure, it locks all elastic task blocks that are being executed by that resource pool or that are scheduled to be executed by that resource pool.

[0046] Mark the locked task block status as pending reallocation and increase its priority to ensure that it is processed first in rescheduling;

[0047] The state of the remaining healthy resource pool after removing the faulty resource pool and the queue of task blocks to be reallocated are used as inputs to trigger the multi-objective optimization model of S200 and generate a new set of scheduling instructions adapted to the fault.

[0048] The newly generated scheduling instruction set is sent to the remaining healthy mining vehicle cluster in the S300, seamlessly taking over and executing the newly assigned task blocks.

[0049] Compared with existing technologies, this resource scheduling optimization method for a deep-sea mining integrated control system has the following advantages:

[0050] I. This invention fundamentally solves the scheduling barriers caused by resource heterogeneity and task complexity in deep-sea mining systems by creating virtual resource pools and elastic task blocks. It encapsulates the attributes and states of various physical resources into virtual resource pools with unified service interfaces, allowing the scheduling system to query, reserve, and allocate resources without needing to concern itself with the physical implementation details of resources. At the same time, it decomposes macro-level mining tasks and defines them as fine-grained elastic task blocks. Each task block clearly declares its required resource type, quantity, and target location. This decoupled architecture transforms the resource scheduling problem into a dynamic matching problem between clear task blocks and resource pools. It can flexibly and quickly allocate any task block to any resource pool with corresponding capabilities, thereby realizing a fundamental shift from planning tasks for a specific mining vehicle to finding the most suitable resources for a task.

[0051] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0053] Figure 1 A flowchart of a resource scheduling optimization method for a deep-sea mining integrated control system;

[0054] Figure 2 This is a flowchart illustrating the steps of a resource scheduling optimization method for a deep-sea mining integrated control system. Detailed Implementation

[0055] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below. Example 1

[0056] This embodiment provides the working principle of a resource scheduling optimization method for a deep-sea mining integrated control system, aiming to solve problems such as resource silos, limitations of single-objective optimization, slow cluster collaborative response, and insufficient fault recovery capabilities in existing deep-sea mining resource scheduling. Figure 2 As shown, this method constructs a unified scheduling foundation through resource and task standardization, generates multi-objective optimal scheduling schemes based on global optimization decisions, achieves conflict-free operations through cluster autonomous collaborative execution, identifies anomalies in a timely manner through status monitoring and fault diagnosis, and finally ensures continuous operation through dynamic rescheduling and resilient recovery, forming a standardized, optimized, executed, monitored, and recovered scheduling system, which effectively improves the efficiency, economy, and safety of deep-sea mining operations.

[0057] (1) Resource and task standardization stage (S100)

[0058] In deep-sea mining operations, the deep-sea mining integrated control system includes heterogeneous physical resources such as surface ship generators, mining vehicles of different types, and underwater relay stations. Moreover, the macro-mining tasks are complex and uncertain. If scheduling is carried out directly, problems such as chaotic resource allocation and disordered task execution may occur. Therefore, S100 lays a unified foundation for subsequent scheduling optimization through resource standardization and flexible task processing. Specifically, it consists of two parts: virtual resource pool construction and flexible task block decomposition.

[0059] Virtual resource pool construction: The essence of a virtual resource pool is to encapsulate the attributes, states, and functions of all heterogeneous physical resources within the system into a standardized resource set with a unified interface definition. This allows the scheduling system to complete resource querying, reservation, and allocation without needing to concern itself with the physical implementation details of the resources, simply through the standardized interface. In this embodiment, virtual resource pools are specifically divided into four categories, and the composition and interface protocol design of each type of resource pool are as follows:

[0060] Energy Pool: Integrates generators from surface ships, batteries from mining vehicles, and buffer energy from relay stations. The interface protocol supports real-time capacity queries and reservation requests, ensuring a precise match between energy supply and task execution needs.

[0061] Operational Capability Pool: Integrates robotic arm gripping force, crusher power, and mineral processing rate. The interface protocol supports task capability querying, providing a basis for allocating suitable operational equipment to task blocks.

[0062] Capacity pool: combines ship cargo holds and mining vehicle mineral storage space, with interface protocols supporting space storage allocation.

[0063] Communication pool: Integrates relay station bandwidth and workshop communication links, and the interface protocol supports priority preemption scheduling.

[0064] Flexible Task Block Decomposition: Macro-level mining tasks need to be decomposed into independent flexible task blocks. Each task block has clear resource requirements and execution objectives, ensuring that the deep-sea mining integrated control system can flexibly allocate and adjust individual task blocks. Specifically:

[0065] Operational grids are divided based on geographic and mineral data: First, the seabed map and mineral abundance data for exploration are obtained, and the target mining area is divided into multiple regular operational grids. Then, according to the overall mining task objectives, the sub-tasks of each operational grid are defined, so that each sub-task corresponds to a specific geographic area, avoiding task coverage overlap or omission.

[0066] Quantify the resource requirements of subtasks: For each initially divided subtask, based on the geographical features of the work grid, mineral data, and subtask objectives, quantify the specific values ​​of various resource types that need to be consumed and occupied for its execution. These resource types correspond one-to-one with the above-mentioned virtual resource pool categories, including: required energy, required operational capacity, expected cabin capacity occupied, and required communication bandwidth.

[0067] Create standardized task descriptors: A standardized task descriptor is created for each subtask. This descriptor is the core basis for the scheduling system to identify and schedule task blocks, and contains the following key information: unique task identifier, task geographic coordinates, resource requirement vector, task priority, and latest completion time. The macro-task is then transformed into a uniformly formatted elastic task block, forming a queue of elastic task blocks to be executed.

[0068] (2) Global optimization decision-making stage (S200)

[0069] Based on the real-time status information of the virtual resource pool generated by S100 and the elastic task block queue, a multi-objective optimization model is constructed and solved to generate the optimal execution instruction that takes into account output, energy consumption, and communication efficiency. The specific process is as follows:

[0070] Multi-objective optimization model construction

[0071] Multi-objective optimization models need to clearly define decision variables, optimization objectives, and constraints to ensure that the model can accurately reflect scheduling needs and conform to actual operational constraints.

[0072] Define decision variables: Decision variables are the core objects of model solving, used to describe the allocation relationship between task blocks and resource pools and the task execution time schedule. Specifically, they include two types of variables:

[0073] Boolean matrix : It is a Boolean value, when When, it indicates a flexible task block. Allocated to virtual resource pool Execute; when When, it indicates a task block. Not allocated to resource pool The allocation scheme of all task blocks and resource pools can be clearly described through a Boolean matrix.

[0074] Continuous variables :in Represents task block In the resource pool The start time of execution on Represents task block In the resource pool The end execution time is a continuous variable used to determine the time window for task execution, avoiding time conflicts between different tasks on the same resource pool.

[0075] Establishing a multi-objective optimization function: Deep-sea mining scheduling needs to simultaneously pursue three objectives: high output, low energy consumption, and high communication efficiency. Therefore, a multi-objective optimization function containing three sub-objectives is constructed. The specific definition is as follows:

[0076] Maximize the total output of the system Total production is determined by the sum of mineral production from all allocated and completed task blocks. Allocated to resource pool And in The previous completion, its output is ,but ( Task Block (latest completion time), among which This is an indicator function that returns 1 if the condition is met and 0 otherwise. The function aims to enable the system to mine as much mineral as possible within a specified time by reasonably allocating tasks and scheduling time.

[0077] Minimize total system energy consumption target Total energy consumption includes the total energy consumed by the resource pool to execute task blocks, such as the energy consumption of mining vehicles moving, the energy consumption of robotic arms and crushers operating, and the energy consumption of communication equipment. If the task block... Allocated to resource pool Its energy consumption is ,but This function aims to reduce system energy consumption and improve operational economy.

[0078] Maximize communication efficiency Communication efficiency is measured by the utilization rate of communication resources and the data transmission success rate. If the task block... Allocated to resource pool When its communication needs are met, the contribution value to communication efficiency is [value missing]. ,but This function aims to ensure efficient use of communication resources and avoid affecting task execution due to communication problems.

[0079] Constraint integration: Constraints are the boundaries of the model's feasible solutions, ensuring that the scheduling scheme conforms to actual job limitations. It dynamically integrates information from multiple sources, including resources, time and space, communication, and task priorities. Specifically, it includes:

[0080] Resource constraints: For each type of virtual resource pool All task blocks allocated to this resource pool The total amount of resources required must not exceed the resource pool. The real-time storage space, for example, in this embodiment, the total reserved energy of the energy pool must not exceed its real-time total capacity, as expressed mathematically as follows: Energy Pool The real-time total capacity is a constraint that prevents over-allocation of resources from causing tasks to fail.

[0081] Spatiotemporal conflict-free constraint: Same resource pool Two task blocks that are not spatiotemporally compatible cannot be executed simultaneously; that is, for any two different task blocks... , If all are allocated to the resource pool , and Then it must satisfy or This constraint prevents the same resource from being occupied by multiple tasks at the same time, ensuring that the operation is carried out in an orderly manner.

[0082] Communication bandwidth constraint: The sum of the data flow rates of all concurrent task blocks on any communication link in the communication pool must not exceed the instantaneous capacity of that link. Assigned to use communication link resource pool Its data stream rate is ,but Communication Link The instantaneous capacity is such that this constraint prevents data loss or delay caused by communication link congestion.

[0083] Task priority constraint: High-priority task blocks must be completed before their latest completion time. If allocated to the resource pool , Then it must satisfy Task Block The latest completion time is set, and in terms of resource allocation and time scheduling, high-priority task blocks take precedence over low-priority task blocks. This constraint ensures the timely completion of critical tasks.

[0084] Model Integration: Integrate the above multi-objective optimization functions and constraints to form a complete multi-objective optimization model, that is, to find the decision variable vector while satisfying all constraints. and The set of , such that the target vector To achieve optimal results.

[0085] Solving the multi-objective optimization model: Since the multi-objective optimization model involves multiple mutually constraining objectives, a unique optimal solution cannot be obtained through traditional single-objective optimization methods. Therefore, this embodiment uses a non-dominated sorting genetic algorithm to generate a Pareto optimal solution set. The specific solution process is as follows:

[0086] Initialization: Generation scale is The initial population, where each individual corresponds to a complete scheduling scheme, i.e., a vector of decision variables. and The encoding of the initial population must satisfy all constraints to ensure the feasibility of the initial solution.

[0087] Evaluation: Calculate the three objective function values ​​for each individual in the population. , , Then, according to the non-dominated sorting rule, all individuals are stratified: individuals that are not dominated by any other individuals are assigned to layer 1. After removing individuals from layer 1, individuals that are not dominated by any other individuals are assigned to layer 2, and so on, until all individuals are stratified. At the same time, the crowding degree of each individual in the same layer is calculated to measure the sparsity of the individual in the target space. The greater the crowding degree, the fewer solutions around the individual and the better the diversity, which provides a basis for subsequent selection operations.

[0088] Selection, crossover, and mutation:

[0089] Selection: Based on non-dominated sorting and crowding, a binary tournament selection operation is used to select parent individuals: two individuals are randomly selected from the population, with priority given to the individual with the lower stratum; if the stratum is the same, the individual with the higher crowding is selected, and this process is repeated until a parent population with the same size as the initial population is selected, ensuring that the parent individuals have both excellence and diversity.

[0090] Crossover: Applying a simulated binary crossover operation to the parent population, targeting a real-number encoded time variable. Randomly select two parent individuals, generate crossover points based on crossover probability, and then use the formula... , Generate offspring time variables, where As a crossover factor, it ensures that offspring individuals inherit the superior characteristics of their parents and introduce new variations.

[0091] Mutation: Applying a polynomial mutation operation to the parent population, targeting the time variable. : Randomly select the time variable from individuals, generate variable time lengths based on the probability of variation, and then use the formula... Update the time variable, where As a variable factor, , The range of values ​​for the time variable is defined to further increase population diversity and prevent the algorithm from getting trapped in local optima.

[0092] Elite preservation: Merge the parent and offspring populations, re-rank the non-dominated populations and recalculate crowding in the merged large population, and then select the top-ranked individuals from it. The best individuals form the next generation of the population.

[0093] Termination judgment: Repeat the evaluation, selection, crossover, mutation, and elite retention operations until the preset maximum number of generations is reached. At this point, the algorithm converges and outputs the last set of individuals that represent the Pareto optimal front, i.e., the Pareto optimal solution set. Each solution in this solution set is a feasible scheduling scheme.

[0094] Optimal scheduling scheme decision and instruction generation: From the Pareto optimal solution set, a final scheduling scheme is selected based on the current job strategy. The final scheduling scheme is then decoded into a set of scheduling instructions, generating a specific execution instruction for each allocated elastic task block. The instruction content includes: execution resource pool ID, task block ID, and start time. End time The expected resource consumption list is generated, and finally all execution instructions are grouped and packaged according to resource pools and sent to the corresponding mining vehicle cluster in S300 to start task execution.

[0095] (3) Cluster autonomous collaborative execution phase (S300)

[0096] In the S300 phase, after the mining vehicle cluster receives the execution instructions issued by S200, it shares information with neighboring agents through workshop communication to achieve local path planning and behavior coordination, completing their respective task blocks without conflict. This phase relies on a distributed collaborative mechanism to avoid the response delay problem of traditional centralized control. The specific process is as follows:

[0097] Command Reception and Parsing: Each mining vehicle, as an independent intelligent agent, has the ability to receive and parse commands. When the execution command packets issued by S200 arrive at the mining vehicle cluster, each intelligent agent in the cluster receives its own command packet, extracts its assigned task block sequence from the command, and clarifies the execution resource pool, start time, end time, geographical coordinates and resource consumption requirements of each task block, providing a clear execution basis for subsequent operations.

[0098] Local perception and communication: To achieve cluster collaboration, each agent needs to grasp its own and neighboring agents' status information in real time. This is accomplished through local perception and inter-agent communication. Local perception refers to each agent's... It acquires local environmental information based on its onboard sensors, including its current location. , For the current time and movement speed Remaining energy, progress of completed tasks, and the location of obstacles within the work area ensure that operational decisions are aligned with the actual environment; Workshop communication: Each intelligent agent Through the workshop communication link in the communication pool, it periodically broadcasts its own status information, while simultaneously receiving information from neighboring intelligent agents within the communication range. The state information and the setting of the communication cycle need to balance the real-time nature of the information and the consumption of communication resources to ensure that neighboring agents can obtain each other's dynamics in a timely manner, providing a basis for conflict judgment.

[0099] Distributed collaborative decision-making: each agent In each decision cycle Inside, The decision cycle duration is based on the agent's own and neighboring agents. Status information to assess the collision risk between the two. To determine whether there is a trajectory conflict, the aforementioned ,in, for Time-based intelligent agent and The collision risk metric between them, with a value range of [0, 1]. The closer the value is to 1, the closer the two objects are, and the higher the risk of a collision. The closer the distance is to 0, the greater the distance between the two objects, and the lower the risk of a collision. , respectively intelligent agents and exist Predicted position at any given time, based on the current position. , With speed , Based on the geographical coordinates of the task blocks, the following predictions were made. The minimum safe collision avoidance radius is determined based on the size of the mining vehicle and the extension range of the robotic arm during operation, ensuring that the distance between two intelligent agents is greater than [missing information]. At this time, no physical collision will occur. If the distance between the two is less than... If a collision is deemed to be at risk, a collision risk threshold is set. ,when At that time, determine the intelligent agent and exist There are constant trajectory conflicts, requiring the initiation of subsequent speed negotiation and adjustment procedures; when If no conflict is found, the agent executes the task as originally planned.

[0100] Speed ​​Negotiation and Adjustment: When a trajectory conflict is detected, the relevant agents initiate a negotiation mechanism through workshop communication to dynamically adjust their respective speeds and temporarily modify local waypoints to achieve avoidance and cooperative coverage. The specific process is as follows:

[0101] Negotiation Initiation: The agent that first detects the conflict sends a conflict negotiation request to the agent associated with the conflict. The request includes its current position, speed, target path and estimated arrival time.

[0102] Solution generation: Based on its own task plan, it generates multiple avoidance solutions, including: reducing its own speed, temporarily deviating from the original path, and adjusting the task execution order.

[0103] Solution optimization: The two intelligent agents exchange avoidance schemes through communication and select the final avoidance scheme based on the principle of minimum task delay.

[0104] Execution and Feedback: The relevant agents adjust their speed or path according to the selected avoidance scheme and provide real-time feedback on the adjusted status information to each other to ensure that no new conflicts are generated during the avoidance process. After the adjustment is completed, the agents resume the original task execution plan, continue to move to the geographical coordinates of the task block and execute the job, so as to achieve conflict-free collaborative operation of the cluster.

[0105] (4) System status monitoring and fault diagnosis phase (S400)

[0106] During the execution of task blocks by the intelligent agent, the system may experience a decline in the health status of the resource pool or equipment failure due to equipment wear and tear or extreme environmental influences. If this is not detected in time, it will lead to task interruption, resource waste, or even security incidents. Therefore, the S400 is used to monitor the health status of the virtual resource pool in real time, identify anomalies in a timely manner, and trigger response processes. To achieve quantitative monitoring of the virtual resource pool, a set of quantifiable health status indicators needs to be pre-set for each virtual resource pool. ,in To determine the number of health indicators for this resource pool, health status indicator data for each virtual resource pool is collected in real time, and the health level value of the resource pool is calculated using a weighted geometric average method. Set a preset health threshold for each virtual resource pool. ,when At that time, determine the resource pool Their health status is normal; continue to monitor changes in their health indicators. In case of equipment failure, determine the resource pool. If the health status is abnormal, the abnormal response process is immediately triggered, and the time of the abnormality, resource pool ID, and abnormal indicator data are recorded to provide a basis for subsequent fault analysis and rescheduling.

[0107] (5) Dynamic rescheduling and resilience recovery phase (S500)

[0108] When S400 diagnoses a virtual resource pool failure, it locks all task blocks associated with that resource pool, including task blocks currently being executed by the failed resource pool and task blocks scheduled for execution by the failed resource pool. Simultaneously, it freezes the current state of these task blocks, marks the locked task blocks from "in execution" or "pending execution" to "awaiting reallocation," and updates this status information in the task descriptor. This ensures the scheduling system can quickly identify task blocks requiring reallocation. Furthermore, to guarantee the execution progress of these task blocks, their priority is increased by one level, ensuring that these task blocks are allocated resources preferentially during subsequent rescheduling, reducing task delays. Using the remaining healthy resource pool status after removing the failed resource pool, the queue of task blocks awaiting reallocation, and the status of task blocks still executing normally as input, the multi-objective optimization model in S200 is retried. This only requires excluding the failed resource pool from the constraints and prioritizing the latest completion time of task blocks awaiting reallocation in the task priority constraints. This is achieved through non-dominated sorting genetic algorithms. After solving the problem, a new Pareto optimal solution set is generated to adapt to the fault. Then, a new final scheduling scheme is selected based on the current operation strategy to ensure that the new scheme can efficiently execute the task blocks to be reallocated using the healthy resource pool, while not conflicting with normal tasks. The new final scheduling scheme is decoded into a new scheduling instruction set. The instruction content still includes the execution resource pool ID, task block ID, new start time, end time, and expected resource consumption list. The new scheduling instruction set is grouped and packaged according to the healthy resource pool and sent to the remaining healthy mining vehicle clusters in S300. After receiving the instruction, the healthy mining vehicle cluster parses the instruction content. If it has completed the current task block, it immediately goes to the geographical coordinates of the new task block to execute the job; if it is still executing the current task block, it seamlessly switches to the new task block after completion, realizing seamless takeover of the fault-related task blocks. At the same time, the scheduling system monitors the task execution status of the healthy mining vehicle cluster in real time to ensure that the task blocks are completed smoothly according to the new instructions, ultimately achieving the continuity and self-recovery of system operations.

[0109] In summary, this embodiment effectively solves the problems of resource isolation, single-objective limitations, slow coordination, and insufficient fault recovery capabilities in existing deep-sea mining scheduling, significantly improving the efficiency, economy, and safety of deep-sea mining operations, and providing a complete and feasible technical solution for the practical application of a comprehensive deep-sea mining control system. Example 2

[0110] Building upon the technology of Embodiment 1, this embodiment provides a specific process for resource scheduling optimization in deep-sea mining using the resource scheduling method of a comprehensive deep-sea mining control system, such as... Figure 1 As shown, the process is as follows:

[0111] Phase 1: Resource and Task Abstraction

[0112] Task reception and parsing: Receive macro-mining tasks.

[0113] Task decomposition: Based on seabed maps and mineral abundance data, the target area is divided into multiple regular task grids, with each grid defining a sub-task.

[0114] Resource requirement quantification: For each subtask, analyze and quantify the types and quantities of resources required to complete it.

[0115] Generate elastic task blocks: Create a standardized descriptor for each subtask, including task ID, coordinates, resource requirement vector, priority, and deadline, forming an elastic task block.

[0116] Resource pool status monitoring: At the same time, continuously monitor and summarize the status of all heterogeneous resources, and abstract them into a unified virtual resource pool.

[0117] Phase Two: Global Optimization and Decision Making

[0118] Build an optimization model: Using the real-time status of all elastic task blocks and virtual resource pools as input, build a multi-objective optimization model with the goal of simultaneously maximizing total system output, minimizing total energy consumption, and maximizing communication efficiency.

[0119] Solution and scheme generation: Solve the multi-objective optimization model to obtain a set of optimal scheduling schemes.

[0120] Solution Decision: Based on the current work strategy, the operator selects a final execution plan from the set of plans.

[0121] Instruction compilation and distribution: Decode the selected scheme into a specific, executable instruction set and distribute it to the corresponding mining vehicle.

[0122] Phase 3: Distributed Cooperative Execution

[0123] Command reception: Each mining vehicle receives the task commands issued.

[0124] Local perception and communication: Each vehicle perceives its surroundings through its own sensors and shares its position, speed, and mission intent with neighboring vehicles through vehicle-to-vehicle communication.

[0125] Collaborative decision-making and obstacle avoidance: Each vehicle independently calculates its optimal path and behavior based on the received neighbor information, and executes its own task blocks without conflict.

[0126] Phase 4: Condition Monitoring and Resilience Recovery

[0127] Real-time health monitoring: Continuously monitors the health indicators of all devices.

[0128] Fault diagnosis: Once abnormal deviations are found in the health indicators, the equipment is immediately identified as faulty or degraded.

[0129] Task freeze and rescheduling trigger: Lock all relevant task blocks of the faulty device, mark their status as pending reallocation and increase their priority.

[0130] System rescheduling: Using the current remaining healthy resource pool and the task blocks to be reallocated as input, the second phase of global optimization and decision-making is triggered again to quickly generate a new scheduling scheme.

[0131] Task takeover and continuation of operations: The new plan instructions are issued to the remaining healthy mining vehicle cluster, which seamlessly takes over the tasks of the faulty equipment to ensure uninterrupted operation until all macro tasks are completed.

[0132] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A resource scheduling optimization method for a deep-sea mining integrated control system, characterized in that, The specific steps of this method are as follows: S100, Resource and Task Standardization: All heterogeneous physical resources in the deep-sea mining integrated control system are assigned to a virtual resource pool with a unified interface definition, while macro-mining tasks are decomposed into multiple independent elastic task blocks. S200, Global Optimization Decision: Based on the real-time status information of all virtual resource pools in S100 and the queue of all pending elastic task blocks, a multi-objective optimization model is constructed and solved to generate specific execution instructions; S300, Cluster Autonomous Collaborative Execution: The mining vehicle cluster receives execution instructions issued by S200, shares its state and intent with neighboring intelligent agents through workshop communication, performs local path planning and behavior coordination, and completes its respective task blocks without conflict. In S200, the construction process of the multi-objective optimization model is as follows: Define decision variables: The decision variables Boolean matrix , Represents a flexible task block Should it be allocated to a virtual resource pool? Execution, and continuous variables , respectively representing task blocks In the resource pool Start time on and end time ; Establish a multi-objective optimization function: The multi-objective optimization function includes the objective of maximizing the total output of the system. Minimize the total energy consumption of the system 1. Maximize communication efficiency ; Constraint integration: The constraints dynamically integrate information from multiple parties, including: Resource constraints: Ensure that the total amount of resources allocated to all elastic task blocks does not exceed the real-time storage space of the corresponding virtual resource pool; Spatiotemporal conflict-free constraint: Two tasks that are not spatiotemporally compatible cannot be executed simultaneously in the same resource pool; Communication bandwidth constraint: The sum of the data flow rates of all concurrent tasks on any communication link must not exceed the instantaneous capacity of that link; Task priority constraint: High-priority tasks must be completed before their deadline; Constructing a multi-objective optimization model: Integrating multi-objective optimization functions and constraints to form a multi-objective optimization model; Linked solution: Taking the state of the elastic task block queue and virtual resource pool generated by S100 as input, under the premise of satisfying all constraints, find the set of decision variable vectors X such that the objective vector F(X) reaches the Pareto optimal solution set.

2. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 1, characterized in that, In S100, the virtual resource pool includes four types of standardized resource pools: Energy Pool: Integrates generators from surface ships, batteries from mining vehicles, and buffer energy from relay stations; the interface protocol supports real-time capacity queries and reservation requests. Task capability pool: integrates robotic arm gripping force, crusher power, and mineral processing rate; interface protocol supports task capability query. Capacity pool: combines ship cargo holds and mining vehicle mineral storage space, with interface protocols supporting space storage allocation; Communication pool: Integrates relay station bandwidth and workshop communication links, and the interface protocol supports priority preemption scheduling.

3. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 1, characterized in that, In S100, the process of decomposing and defining the macro-mining task into elastic task blocks is as follows: Based on the seabed map and mineral abundance data from the exploration, the target mining area is divided into multiple regular operation grids, and the sub-tasks of each operation grid are defined according to the mission objectives. For each initially defined subtask, the specific values ​​of various resource types consumed and occupied for its execution are quantified. The various resource types occupied correspond to the virtual resource pool categories, including the required energy, required operational capacity, expected cabin capacity occupied, and required communication bandwidth. For each subtask, a standardized task descriptor is created, which includes a unique task identifier, task geographic coordinates, resource requirement vector, task priority, and latest completion time, thereby forming a unified elastic task block.

4. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 1, characterized in that, The method also includes: S400, System Status Monitoring and Fault Diagnosis: When the agent executes a task block, it monitors the health status indicators of all virtual resource pools in real time. When it is found that the health status value of any virtual resource pool is lower than the preset health status threshold or the device malfunctions, it immediately triggers the abnormal response process. S500, Dynamic Rescheduling and Resilient Recovery: In response to anomalies, the system locks all ongoing and allocated task blocks associated with the faulty resource pool, marks them as pending reallocation and increases their priority, triggers S200 for rescheduling, and is then executed by the remaining mining vehicle cluster in S300.

5. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 1, characterized in that, In S200, when solving the multi-objective optimization model in a coordinated manner, a non-dominated sorting genetic algorithm is used to generate a Pareto optimal solution set. The solution process of the non-dominated sorting genetic algorithm is as follows: Initialization: Generate an initial population of size N, where each individual in the population represents a complete scheduling scheme, i.e., the encoding of a decision variable vector X; Evaluation: Calculate the three objective function values ​​for each individual in the population, stratify all individuals in the population according to the non-dominated sorting rule, and calculate the crowding degree of each individual within the same stratum; Selection, crossover, and mutation: Based on non-dominated ordering and crowding, parent individuals are selected through binary tournament selection, and simulated binary crossover and polynomial mutation operations are applied to generate offspring populations. Elite preservation: The parent and offspring populations are merged, and the non-dominated ranking and crowding degree are recalculated in the merged large population. The top N best individuals are selected to form a new generation population. Termination judgment: Repeat the evaluation to elite retention operation until the preset maximum number of generations is reached, and finally output the last set of individuals that represent the Pareto optimal frontier, that is, the Pareto optimal solution set. Each solution in this solution set represents a feasible scheduling scheme. Optimal scheduling scheme decision: Select a final scheduling scheme from the Pareto optimal solution set based on the current job strategy; Compiling and issuing scheduling instruction sets: The selected final scheduling scheme is decoded into a scheduling instruction set, and a specific execution instruction is generated for each allocated elastic task block. The content includes: execution resource pool ID, task block ID, start time, end time and expected resource consumption list. All execution instructions are grouped and packaged according to resource pools and issued to the corresponding mining vehicle cluster in S300 to start execution.

6. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 1, characterized in that, In S300, the specific process of local path planning and behavior coordination is as follows: Command reception and parsing: Each mining vehicle acts as an intelligent agent. When executing the assigned task block, each intelligent agent in the cluster receives the issued execution command and parses out its own assigned task block sequence. Local perception and communication: each agent It acquires local environmental information based on its own sensors, and periodically broadcasts its own status, including position, speed, and remaining tasks, through workshop communication, and receives information from neighboring agents. Status information; Distributed collaborative decision-making: each agent In each decision cycle Within, for the received neighboring agents Risk of conflict in status information The assessment, the ,in, for In-time intelligent agent and Collision risk measurement between them and respectively intelligent agents and At any moment Location, To achieve the minimum safe collision avoidance radius, when At that time, it was determined that there was a trajectory conflict; Speed ​​negotiation and adjustment: When a trajectory conflict is detected, the relevant agents will negotiate to dynamically adjust their respective speeds and temporarily modify local waypoints in order to achieve avoidance and cooperative coverage.

7. The resource scheduling optimization method for a deep-sea mining integrated control system according to claim 4, characterized in that, The specific exception response process in S500 is as follows: When the S400 diagnoses a virtual resource pool failure, it locks all elastic task blocks that are being executed by that resource pool or that are scheduled to be executed by that resource pool. Mark the locked task block status as pending reallocation and increase its priority to ensure that it is processed first in rescheduling; The state of the remaining healthy resource pool after removing the faulty resource pool and the queue of task blocks to be reallocated are used as inputs to trigger the multi-objective optimization model of S200 and generate a new set of scheduling instructions adapted to the fault. The newly generated scheduling instruction set is sent to the remaining healthy mining vehicle cluster in the S300, seamlessly taking over and executing the newly assigned task blocks.

Images