Construction operation collaborative scheduling method and system based on multi-source data fusion

Abstract

The invention discloses a construction operation collaborative scheduling method and system based on multi-source data fusion, and relates to the technical field of operation scheduling. The method comprises the following steps: executing multi-source data acquisition, and establishing a multi-source time sequence state tensor; inputting the state tensor into a BIM model, and updating and generating a digital twinborn body synchronous with the construction state according to parameters of the BIM model; abstracting the operation unit into a plurality of cooperative agents; after the twinborn state and resource constraint of the intelligent agent are configured by using the digital twinborn body, game analysis is executed through a revenue function, and a local strategy is established; and aggregating the local strategies and carrying out construction operation collaborative scheduling. The technical problems that the construction state is difficult to simulate accurately in real time, cooperation among operation units lacks scientific strategy guidance, and conflicts and resource waste are prone to occurring are solved, and the purposes of providing scientific strategies for cooperation of the operation units, reducing conflicts, optimizing resource allocation and improving the working efficiency are achieved by constructing the digital twinborn bodies synchronously evolved with the construction state. And the construction cooperation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of job scheduling technology, specifically to a collaborative scheduling method and system for construction jobs based on multi-source data fusion. Background Technology

[0002] In the current field of underground space development and large-scale engineering construction, construction operations are rapidly developing towards large-scale, complex, and refined operations. The construction site, as a dynamic and complex system, is filled with multi-source, heterogeneous, and closely interconnected elements, such as constantly revealed geological conditions, the coordinated operation of numerous pieces of machinery, the precise scheduling of materials and personnel, and ever-changing environmental indicators. Traditional work scheduling methods heavily rely on pre-established plans and human experience for decision-making, making it difficult to perceive and respond to the rapidly changing global state of the site in real time. This lag often leads to information barriers between different work units, easily causing problems such as resource allocation conflicts, poor workflow coordination, and untimely safety risk warnings. This not only restricts further improvements in construction efficiency but also brings significant uncertainty and safety concerns to project management. Summary of the Invention

[0003] This application provides a construction operation collaborative scheduling method and system based on multi-source data fusion, which solves the technical problems of difficulty in simulating construction status in real time and accurately, lack of scientific strategy guidance for collaboration between work units, and easy occurrence of conflicts and resource waste.

[0004] The first aspect of this application provides a method for collaborative scheduling of construction operations based on multi-source data fusion, the method comprising: Multi-source data acquisition is performed, and a multi-source temporal state tensor is established using a spatiotemporal correlation mechanism. The multi-source data includes ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data. This multi-source temporal state tensor is input into a 3D underground space BIM model, and a digital twin is generated that evolves synchronously with the applied construction state based on the updated parameters of the 3D underground space BIM model. After reading the work unit, the work unit is abstracted into multiple collaborative intelligent agents, each possessing local perception, task priority evaluation, and conflict constraint learning capabilities. After configuring the local twin state and resource constraints of each collaborative intelligent agent using the digital twin, game analysis is performed on the collaborative intelligent agents through an immediate reward function to establish a local execution strategy. Global state aggregation is performed on the local execution strategy to establish a global correction result, and collaborative scheduling of construction operations is carried out based on the global correction result.

[0005] A second aspect of this application provides a construction operation collaborative scheduling system based on multi-source data fusion, the system comprising: Data Acquisition Module: Performs multi-source data acquisition, establishes a multi-source temporal state tensor using a spatiotemporal correlation mechanism, including ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data; Digital Twin Module: Inputs the multi-source temporal state tensor into a 3D underground space BIM model, and generates a digital twin that evolves synchronously with the applied construction state based on the parameters of the 3D underground space BIM model; Intelligent Agent Construction Module: After reading the work unit, abstracts the work unit into multiple collaborative intelligent agents, each of which has the ability to perceive locally, evaluate task priorities, and learn conflict constraints; Game Theory Analysis Module: After configuring the local twin state and resource constraints of each collaborative intelligent agent using the digital twin, performs game theory analysis on the collaborative intelligent agents through an immediate reward function to establish a local execution strategy; Job Scheduling Module: Aggregates the global state of the local execution strategy, establishes a global correction result, and performs collaborative scheduling of construction operations based on the global correction result.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, a multi-source temporal state tensor is constructed through multi-source data acquisition and spatiotemporal correlation mechanisms. Data sources include ground-penetrating radar, drilling feedback, construction equipment operation, material and personnel positioning, and environmental monitoring information. This data is then input into a 3D underground space BIM model. Through parameter updates of this model, a digital twin that evolves synchronously with the construction status is generated. Based on this, work units are abstracted into multiple collaborative agents. Each agent can perceive its local environment, assess task priorities, and learn how to handle conflict constraints. Next, the local state and resource constraints of each agent are configured through the digital twin, and game theory analysis is performed using an immediate reward function to formulate local execution strategies. Finally, all local execution strategies are aggregated, and global state aggregation and correction are performed to form a global scheduling strategy, ensuring the coordinated and optimized execution of construction operations. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram of the construction operation collaborative scheduling method based on multi-source data fusion provided in the embodiments of this application.

[0009] Figure 2A schematic diagram of the structure of a construction operation collaborative scheduling system based on multi-source data fusion provided in this application embodiment.

[0010] Figure labeling: Data acquisition module 11, Digital twin module 12, Intelligent agent construction module 13, Game analysis module 14, Job scheduling module 15. Detailed Implementation

[0011] 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.

[0012] Example 1, as Figure 1 As shown, this application provides a collaborative scheduling method for construction operations based on multi-source data fusion, wherein the method includes: Multi-source data acquisition is performed, and a multi-source temporal state tensor is established using a spatiotemporal correlation mechanism. The multi-source data includes ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data.

[0013] In this embodiment, multi-source data acquisition is first performed to obtain comprehensive construction site information. These data sources include, but are not limited to, ground-penetrating radar and drilling feedback data, operational data of construction machinery and equipment, location data of materials and personnel, and environmental monitoring data. Ground-penetrating radar and drilling feedback data can provide geological information of underground space, including key parameters such as underground soil structure, groundwater level, and collapse risk, helping to assess potential risks during construction. Operational data of construction machinery and equipment covers the usage status of various types of machinery and equipment, such as the operating progress, load status, and fault information of tunnel boring machines, mixing pile machines, ventilation and drainage equipment, etc. This data is of great significance for evaluating the availability and efficiency of machinery and equipment. Material and personnel location data provides real-time basis for work scheduling by tracking the storage of materials and the location of personnel at the construction site, ensuring efficient allocation of resources. Environmental monitoring data acquires environmental factors such as air quality, temperature, humidity, noise, and vibration in real time, helping to ensure that the construction site meets safety and environmental protection requirements and prevent accidents or delays in the construction period caused by environmental changes. Subsequently, a spatiotemporal correlation mechanism is activated to align these different time data according to their timestamps and perform necessary interpolation or supplementation to ensure complete data at the same point in time. Then, based on the known coordinate system of the construction site, such as total station or geographic information system, all data sources are converted into a unified spatial coordinate system to ensure that the spatial positions of different data sources can be compared within the same coordinate framework. After completing the alignment of timestamps and spatial coordinates, a multi-source temporal state tensor is formed. This multi-source temporal state tensor not only reflects the current state of the construction site but also reflects the dynamic changes of various resources and the environment. This enables subsequent work scheduling and resource management to be more accurate and efficient, thereby greatly improving the collaborative effect of construction operations, reducing resource waste and conflicts, and ensuring the smooth progress of the construction project.

[0014] The multi-source temporal state tensor is input into the three-dimensional underground space BIM model, and a digital twin that evolves synchronously with the applied construction state is generated based on the updated parameters of the three-dimensional underground space BIM model.

[0015] In one embodiment, multi-source temporal state inputs are obtained into a 3D underground space BIM model. This 3D underground space BIM model is a digital representation based on the actual construction environment, accurately describing the geometry, structural composition, and spatial relationships with various facilities of the underground space. By combining the multi-source temporal state tensor with the 3D underground space BIM model, the digital representation of the underground space can be dynamically updated, accurately reflecting the real-time status and changes of various resources during construction in the spatiotemporal dimension. Subsequently, based on the parameter names of the 3D underground space BIM model, the corresponding parts in the model are updated using the multi-source temporal state tensor, thereby generating a digital twin that evolves synchronously with the construction operation status. This digital twin is a virtual model driven by real-time data, capable of accurately presenting changes in the actual construction status in the digital world, simulating the interactive influence of different variables during construction, and providing accurate basic data support for subsequent intelligent scheduling, resource optimization, and risk warning.

[0016] After reading the task unit, the task unit is abstracted into multiple collaborative agents, each of which has the ability to perceive local conditions, evaluate task priorities, and learn conflict constraints.

[0017] In one embodiment, the work units involved in the current task are first read. These work units refer to the execution equipment required for the current task, such as excavators, bulldozers, and transport vehicles. After reading the work units, each work unit is abstracted into a collaborative intelligent agent. Each collaborative intelligent agent, when executing a task, possesses local perception, task priority evaluation, and conflict constraint learning capabilities. Local perception allows each collaborative intelligent agent to acquire real-time information about its surrounding environment, including equipment status, task progress, and resource consumption. For example, the agent corresponding to an excavator will perceive the equipment's load, fuel consumption, and location in real time. This perception capability enables each collaborative intelligent agent to adjust its behavior according to real-time environmental changes to avoid unnecessary resource waste or conflicts. Task priority evaluation allows each collaborative intelligent agent to self-judge and schedule tasks based on the urgency and importance of the construction task. For example, some work units may need to be executed first, such as emergency repairs or debugging of critical equipment, while other work units may be postponed. This capability ensures the rationality and efficiency of task scheduling, enabling important tasks to be completed in the shortest possible time. The conflict constraint learning capability enables each collaborative agent to learn and handle potential conflicts and resource competition issues encountered during task execution. For example, multiple devices may need to use the same space or resources simultaneously. Each agent can identify these conflicts and adjust according to constraints, learning how to avoid conflicts or optimize resource allocation strategies when conflicts occur. Through these three capabilities, collaborative agents can make autonomous decisions and coordinate efficiently in the overall construction task, ensuring maximum priority and resource utilization, thereby improving the efficiency and quality of the entire construction operation.

[0018] After configuring the local twin state and resource constraints of each collaborative agent using the digital twin, game analysis of the collaborative agents is performed through an immediate reward function to establish a local execution strategy.

[0019] In one embodiment, after obtaining multiple collaborative agents, the system continuously acquires status information of various resources and equipment during the construction process from the digital twins. It then extracts the virtual environment status information of the work unit corresponding to each collaborative agent, understanding its own equipment's working status, location, task progress, availability of surrounding resources, and potential environmental changes. Based on this, it configures the local twin state of each collaborative agent to describe its current working status at the construction site. Simultaneously, based on on-site resource usage data, it configures resource constraints for each collaborative agent, such as energy consumption, operation time, equipment load, and spatial location limitations. Subsequently, in policy control, game analysis is performed on these local twin states, resource constraints, and pre-built immediate reward functions for the collaborative agents. During game analysis, each collaborative agent interacts through game strategies, evaluates the payoff of its current candidate execution strategy based on the immediate reward function, obtains the corresponding local payoff evaluation result, and dynamically adjusts its own strategy by combining environmental state changes fed back by the digital twins and payoff and conflict constraint information shared by neighboring collaborative agents. Through the aforementioned iterative game process based on benefit assessment, information feedback, and strategy updates, each collaborative agent gradually develops local execution strategies that match the current construction environment and resource conditions. These local execution strategies not only consider the priority of the current task but also the actual capabilities of the equipment and resource conditions, thereby helping the collaborative agents to execute tasks efficiently and in a coordinated manner during construction, avoiding unnecessary delays and resource waste, and improving the overall scheduling efficiency of the construction site.

[0020] Furthermore, game analysis of collaborative agents is performed through immediate reward functions, including: For each collaborative agent, a set of state variables corresponding to the work unit is extracted from the digital twin. After normalizing the set of state variables, a local state feature vector is constructed. Resource usage data from the field is read, and resource constraints are established based on this data. The policy space of the collaborative agent is then defined according to these resource constraints. An immediate reward function is constructed using the local state feature vector, which includes a geological disturbance risk term, an equipment load rate term, a work progress deviation term, and an energy consumption cost term. The immediate reward function is used to evaluate the benefit of the fitted execution strategy for each collaborative agent within the policy space. Based on the communication topology, local benefits and conflict constraints are shared, and game iteration is performed based on the shared results to establish a local execution strategy.

[0021] Preferably, for each collaborative agent, the set of state variables corresponding to its work unit is first extracted from the digital twin. The digital twin provides a real-time virtual model of the construction site, containing the real-time operating status of each work unit. The extracted set of state variables includes information such as the working status of each executing device, task execution progress, resource consumption, and changes in the working environment, enabling a comprehensive understanding of the status of each work unit during construction. Subsequently, the extracted set of state variables is normalized using a maximum-minimum normalization method, adjusting the values ​​of each state variable to the same scale range, eliminating dimensional differences, and allowing different types of variables to be compared and calculated on the same scale. After normalization, these state variables are concatenated according to a preset order to construct a local state feature vector to reflect the actual execution status of the equipment. Simultaneously, resource usage data from the construction site is read, including the actual usage of construction equipment, personnel, materials, and other resources. Based on this resource usage data, resource constraints are established. These constraints limit the resources that the collaborative agents can use when performing tasks, such as equipment load capacity, working time, material inventory, and personnel availability, ensuring that the actions of the collaborative agents do not exceed the actual available resources. Then, using the normalized local state feature vectors, an immediate reward function is constructed. This immediate reward function is used to evaluate the effectiveness of each collaborative agent's current behavior and is typically in the form of a weighted function, including several important factors such as geological disturbance risk, equipment load rate, work schedule deviation, and energy cost. The specific form is as follows: ,in, These are the corresponding weighting coefficients, used to characterize the importance of each evaluation item in the current construction stage. They can be preset according to construction goals or changes in the external environment. For example, they can be set to 0.4, 0.25, 0.25, and 0.1 respectively. It represents the normalized evaluation value of geological disturbance risk items, assesses the potential impact of underground environmental changes, such as soil layer changes and groundwater level changes, on construction tasks, and reflects the geological risks that may be encountered during construction. This represents the normalized evaluation value of the equipment load rate, which is used to measure the load situation of the equipment during the execution of tasks. Excessive load may lead to equipment failure or reduced operating efficiency. The normalized evaluation value representing the work progress deviation item is the deviation between the planned schedule and the actual progress. The normalized evaluation value of the energy consumption cost item represents the energy consumption of the equipment. High energy consumption not only increases costs but may also affect equipment availability. Using the constructed immediate reward function, the benefit of each collaborative agent's fitted execution strategy in the policy space is evaluated, quantifying the short-term benefits of each strategy. Then, based on these local short-term benefits and conflict constraints, game iteration is conducted. Finally, through multiple rounds of game iteration, each collaborative agent generates a final local execution strategy. These strategies not only consider the local situation of the current work unit but also integrate the behaviors and resource constraints of other agents, thereby ensuring optimal overall scheduling and reducing conflict and resource waste.

[0022] Furthermore, based on the sharing of local benefits and conflict constraints in the communication topology, game iteration is performed according to the shared results, including: Based on the spatial location, operation type, equipment interference relationship, and resource coupling degree of each collaborative agent in the 3D underground space BIM model, a communication topology diagram is constructed. The edge weights of the communication topology diagram dynamically represent the information interaction intensity and resource competition intensity among the collaborative agents. After each collaborative agent calculates its benefit evaluation, the local benefit vector and conflict constraint parameters are broadcast in a neighborhood-level weighted manner through the communication topology. After receiving the neighborhood shared results, each collaborative agent performs an adaptive update of the immediate benefit function based on the neighborhood benefit volatility and resource competition gradient, and generates a local strategy search preference. The game is then iterated based on the adaptively updated immediate benefit function and strategy search preference.

[0023] Optionally, the spatial location, task type, equipment interference relationship, and resource coupling degree of each collaborative agent are first extracted from the 3D underground space BIM model. Spatial location determines whether collaborative agents will perform tasks in the same area. If multiple collaborative agents are located in adjacent or overlapping areas, resource competition and task conflicts may exist. This can be quantified based on the Euclidean distance between the center points of the work areas of the corresponding equipment of the collaborative agents. When the distance is less than a preset spatial threshold, spatial association is determined, and a connection is established. The smaller the distance, the greater the corresponding spatial association weight. Based on this positional relationship, connections can be established between collaborative agents, indicating that they are potential collaborators or competitors. Task type is used to reflect the collaboration or conflict that may arise from different types of tasks. For example, bulldozers and excavators may perform earthmoving operations in the same area. The text discusses various aspects of collaborative work, including equipment interference and resource coupling. It mentions that while the operations of transport vehicles and excavators may be complementary, equipment interference can lead to interference. The text further describes how the actions of robotic arms and excavators can interfere with each other, assigning higher weights to interfering equipment to represent conflict intensity. These weights are quantified by averaging the normalized proportions of overlapping operating radii and times. Higher interference results in higher weights, representing conflict intensity. Resource coupling refers to the dependence of different work units on resource usage. For example, if two work units share the same batch of materials, resource coupling is high. This can be quantified by averaging the normalized proportions of shared resource types and quantities, shared resource occupancy ratios, and resource usage time overlap ratios. Based on these relationships, connections are established between collaborative agents with cooperative or conflicting relationships, forming a communication topology. Each connection is assigned a weight, which is obtained by weighting the quantified values ​​of spatial location, work type, equipment interference, and resource coupling to describe the intensity of information interaction and resource competition among collaborative agents. Subsequently, after each cooperative agent calculates the benefit evaluation of the corresponding fitted execution strategy using the immediate benefit function, it generates a corresponding local benefit vector based on the calculation result. Combined with the conflict constraint parameters generated by the cooperative agents, the vector is broadcast in a neighborhood-level weighted manner through the communication topology. That is, the cooperative agents propagate their benefit and constraint information to neighboring agents according to the communication topology. Each cooperative agent obtains information about the behavior of other cooperative agents by receiving these neighborhood-shared results.Upon receiving the shared results from the neighborhood, each collaborative agent adjusts its strategy based on the neighborhood payoff volatility and resource competition gradient. Neighborhood payoff volatility reflects the impact of the fluctuations in the payoffs of other agents within the neighborhood on the current agent's decision-making. Large payoff fluctuations indicate unstable behavior, requiring the current agent to adjust its strategy to adapt to environmental changes. The resource competition gradient represents the intensity of resource competition among agents within the neighborhood. If an agent faces significant resource competition pressure, it may adjust its work plan or priorities to reduce resource conflicts. Changes in the resource competition gradient influence strategy selection, and collaborative agents strive to avoid conflicts with other agents over limited resources. Based on this information, each collaborative agent performs an adaptive update of its immediate payoff function. This involves adjusting the weight parameters and constraint penalty parameters in the immediate payoff function, representing the importance of different evaluation items, based on the neighborhood-shared information, to better adapt to the current environment and resource constraints. During adjustments, when the volatility of neighborhood returns is high, the weight parameters corresponding to the work progress deviation term or resource conflict-related constraint term are increased to enhance the suppression of unstable execution states. When an increase in resource competition gradient is detected, the weight parameters corresponding to the resource-related constraint penalty parameter or energy consumption cost term are increased to guide the collaborative agent to actively reduce resource occupancy intensity or adjust the work plan. When the neighborhood environment tends to be stable and the degree of conflict decreases, the weight ratio of the constraint penalty parameter is appropriately reduced to improve the flexibility of strategy search. These parameter adjustments are all made using preset ratios, and the adjustment range is constrained by preset parameter update thresholds or change rate upper limits to ensure the stability of the strategy update process. Through this adaptive update, the collaborative agent can dynamically adjust its behavior and generate local strategy search preferences to represent the selection tendency of each collaborative agent in the strategy space. These strategy search preferences, combined with the adaptively updated immediate return function, jointly drive the game iteration. During the game iteration process, the collaborative agent continuously adjusts its strategy according to changes in the neighborhood, return evaluation, and resource constraints, gradually approaching the optimal local execution strategy. Through multiple rounds of game iteration, each collaborative intelligent entity eventually forms a set of coordinated local execution strategies based on the global goal and the local environment. The optimization of these local execution strategies ensures that each work unit can work efficiently and collaboratively in a complex construction environment, maximizing resource utilization, reducing conflicts, and improving overall construction efficiency.

[0024] Furthermore, the game iteration is performed based on the adaptively updated immediate payoff function and policy search preferences, including: The updated immediate payoff function is used to re-evaluate the candidate policy set of the corresponding cooperative agent, and a payoff change gradient vector is established. Policy migration direction parameters are generated based on the payoff change gradient vector and policy search preferences. These parameters are shared in the communication topology, and a dual-channel game update is performed. The dual channels include a synchronous payoff migration channel and a conflict constraint feedback channel. The synchronous payoff migration channel is used to adjust the probability distribution of its own policy based on the neighborhood average payoff trend, and the conflict constraint feedback channel is used to compensate for changes in neighborhood conflict constraints. A local execution policy is established based on multiple rounds of dual-channel game iterations.

[0025] Optionally, after updating the immediate reward function, each collaborative agent re-evaluates the candidate strategy set using its updated immediate reward function. This candidate strategy set refers to a collection of pre-generated feasible execution strategies for the corresponding collaborative agent, satisfying current resource constraints and job rule constraints. These feasible execution strategies include at least one or more of the following: job timing adjustment strategy, job path or job area selection strategy, and resource usage intensity adjustment strategy. The immediate reward function comprehensively considers various factors, such as job progress, resource utilization efficiency, and equipment load, to calculate the immediate reward of each candidate strategy. By comparing the rewards of different strategies and calculating the difference between the immediate rewards of each candidate strategy and the immediate reward of the currently executed strategy, the reward change of each candidate strategy relative to the current strategy can be calculated. The reward changes are then combined according to the order of the candidate strategies in the strategy space to form a reward change gradient vector. This reward change gradient vector represents the changing trend between different strategies, helping the collaborative agent identify which strategies can bring higher rewards, thereby selecting the optimal action direction. Subsequently, a policy is generated using the gradient vector of the change in payoff and the previously determined policy search preference. The dominant direction of policy migration is determined by the direction with the largest increase in payoff in the gradient vector of the change in payoff. This direction is then weighted and modified in conjunction with the policy search preference to generate a policy migration direction parameter. This policy migration direction parameter indicates which direction the cooperative agent should adjust its policy in order to maximize payoff. Subsequently, utilizing the sharing mechanism of the communication topology, the obtained policy migration direction parameters are shared to each cooperative agent, performing a dual-channel game update. Specifically, the game update utilizes a synchronous payoff migration channel and a conflict constraint feedback channel within the cooperative agent. The synchronous payoff migration channel, which can be built on a neural network, adjusts the probability distribution of the current agent's policy based on the average payoff trend of other agents in the neighborhood. In a multi-agent cooperative environment, neighboring cooperative agents influence the behavior of each agent. Therefore, by transmitting the average payoff trend of neighboring agents, the synchronous payoff migration channel adjusts its own policy selection probability, tending to choose policies consistent with those of neighboring agents, thereby improving overall cooperative efficiency. The conflict constraint feedback channel, also built on a neural network, handles conflict constraint information between neighboring agents. By transmitting conflict constraint information within the neighborhood to each cooperative agent, it enables them to adjust their resource usage plan. For example, if an agent competes with other agents in the neighborhood for a certain resource, it dynamically adjusts its resource usage plan based on changes in the conflict constraint to avoid resource conflicts and ensure the smooth progress of the task. Through multiple rounds of dual-channel game iteration, the strategies of each collaborative agent are continuously optimized. In each round of the game, the agent adjusts its strategy migration direction based on feedback from its neighborhood, and continuously iterates and optimizes its execution strategy by combining changes in payoffs and conflict constraints.This process gradually converges through multiple rounds of game play, eventually leading each agent to find an optimal local execution strategy. Through this iterative game, each collaborating agent can better coordinate its behavior with other agents, maximizing resource utilization efficiency, reducing conflicts, and optimizing the overall construction progress.

[0026] Furthermore, the collaborative agent is equipped with a reward mechanism, which incorporates nonlinear target weights. These nonlinear target weights are updated based on the evolutionary characteristics of the external environment, and the collaborative agent, adjusted according to the reward mechanism, performs game analysis.

[0027] Preferably, each collaborative agent employs a reward mechanism to guide optimal decision-making during construction. The core of this reward mechanism lies in influencing the agent's decision-making process through rewards and penalties, enabling continuous optimization of execution strategies to achieve efficient resource utilization and successful task completion. The reward mechanism introduces non-linear objective weights, which are weight parameters used to adjust the relative importance of different objective items in the immediate reward function. These weight parameters do not have a linear relationship with the construction state variables but are adjusted non-linearly according to changes in the environmental state, thus adapting to the multi-objective optimization needs in complex construction environments. For example, in some cases, work progress may require higher priority, while in others, resource conservation may be more important. Specifically, non-linear objective weights can be constructed using piecewise functions, exponential functions, sigmoid functions, or threshold-triggered mapping functions. When the corresponding environmental indicator reaches a preset range or threshold, the objective weight changes non-linearly, thereby amplifying or suppressing the influence of the corresponding objective item in the immediate reward function. By setting non-linear objective weights, the agent's focus in multi-objective optimization can be adjusted in real time according to changes in different tasks and environments. Furthermore, the nonlinear target weights are updated based on the evolution characteristics of the external environment. These external environment evolution characteristics refer to the dynamic changes at the construction site, including changes in geological conditions, environmental conditions, equipment operating status, and other factors affecting construction progress and efficiency. The system periodically collects external environment evolution characteristic indicators and matches them with preset environmental state intervals. When a transition in the environmental state is detected, the corresponding nonlinear target weights are updated. This update method includes weight amplification and weight decay. For example, when a rapid increase in geological risk indicators is detected, the weight ratio of safety-related target items is increased through a nonlinear mapping function; when the construction environment tends to stabilize and resource constraints decrease, the weight of safety constraint targets is reduced while the weight of progress or energy consumption-related targets is increased, thus achieving dynamic switching of target focus. After the reward mechanism is adjusted, the collaborative agents perform the aforementioned game analysis based on the new reward mechanism to evaluate the returns of different strategies. Through the game process, they interact with other agents to achieve the optimal balance in resource allocation, task execution, and time management throughout the overall construction process, ensuring the efficient implementation of the construction project.

[0028] The local execution strategy is aggregated globally to establish a global correction result, and construction operations are coordinated and scheduled based on the global correction result.

[0029] In one embodiment, after obtaining all local execution strategies, a global state is constructed for these strategies within the digital twin. Based on the constructed global state, strategy consistency and resource conflict indices are calculated, and these indices are compared with corresponding thresholds. The thresholds are divided into a first preset threshold for comparison with the strategy consistency index and a second preset threshold for comparison with the resource conflict index. The first preset threshold can fluctuate between 0.6 and 0.8, and the second preset threshold can fluctuate between 0.3 and 0.5. Typically, the first preset threshold is set to 0.7, and the second preset threshold to 0.4. For regions with significant differences, a compensation optimization under a dynamic constraint alignment mechanism is performed. This involves rebalancing the resource allocation ratio of associated collaborative agents within the region, adjusting the execution order of conflicting work units, and correcting the task priorities corresponding to low-consistency strategies. This reduces resource conflict and improves strategy consistency, thereby aggregating these local execution strategies into a global correction result. This global correction result represents the optimal adjustment scheme for the overall state of the construction task and resource usage after comprehensively considering the execution strategies of all collaborative agents. Finally, task verification nodes are configured based on the global correction results, and collaborative execution status verification and construction operation scheduling are carried out based on these verification nodes. This ensures that each agent works efficiently and collaboratively within limited resources and time, avoids conflicts and delays, and improves overall construction efficiency.

[0030] Furthermore, the local execution strategy is subjected to global state aggregation to establish a global correction result, including: All local execution strategies are obtained, and a global state tensor is constructed based on the spatial dependency relationships and task temporal constraints within the digital twin. The global state tensor represents the spatiotemporal interference distribution and resource consumption mapping of each cooperating agent in the underground space. Based on the global state tensor, a strategy consistency index and a resource conflict index are calculated. Local regions where the strategy consistency index is lower than a first threshold or the resource conflict index is higher than a second threshold are marked as global deviation regions. Compensation optimization is performed on the global deviation regions under a dynamic constraint alignment mechanism to establish a global correction result.

[0031] Preferably, the local execution strategies of all collaborative agents are first obtained. These strategies reflect the optimal decisions of each agent within its work scope, including task allocation, resource usage plans, and job priorities. Subsequently, based on these local execution strategies, a global state tensor is constructed. This tensor integrates the spatiotemporal information of the construction site, resource usage, and job sequence. Combined with the virtual representation of the underground space in the digital twin, it reflects the spatiotemporal interference distribution and resource consumption mapping of each collaborative agent in the underground space. The spatiotemporal interference distribution refers to the positional relationships of different work units in the digital twin and their temporal task execution overlap, which is crucial for determining whether there are problems such as equipment interference or resource competition. The resource consumption mapping shows the types and quantities of resources used by each work unit, helping to assess whether resources are being used efficiently. Next, based on the data in the global state tensor, strategy consistency and resource conflict indices are calculated. The strategy consistency index measures the coordination of the execution strategies of various cooperating agents at the global level and can be calculated using a neural network model. If multiple agents execute similar tasks in similar times and spaces, their strategy consistency index is high. If agents within a certain region execute strategies with highly inconsistent results, it may lead to conflicts in job progress or resource waste; the strategy consistency index reflects the degree of this incoordination. The resource conflict index assesses the degree of conflict in resource allocation and can be calculated using a neural network model. A high conflict index indicates excessive resource concentration, which may lead to excessively long equipment idle times or task execution delays. Then, based on the calculated strategy consistency and resource conflict indices, the global state is analyzed. When the strategy consistency index is below a first threshold or the resource conflict index is above a second threshold, the region is marked as a global deviation region. These global deviation regions are typically areas where job progress delays or resource waste are caused by uneven resource allocation or unreasonable task scheduling. Once the global deviation region is identified, the system enters the dynamic constraint alignment mechanism phase for compensation and optimization. This phase adjusts resource allocation, task assignments, and time scheduling through gradient mapping and neighborhood propagation updates to eliminate conflicts and inconsistencies, thereby achieving efficient resource utilization and smooth task execution. After compensation and optimization by the dynamic constraint alignment mechanism, a global correction result is generated. This result is an optimized scheduling scheme that reflects the adjusted task unit execution strategy, resource allocation scheme, and time scheduling, enabling collaborative agents to work efficiently and collaboratively under resource-constrained and task-heavy environments, ultimately achieving optimal completion of the construction objectives.

[0032] Furthermore, the compensation optimization for the global deviation region under a dynamic constraint alignment mechanism includes: Extract the associated set of cooperative agent constraints within the global deviation region, establish a constraint coupling matrix based on the set of cooperative agent constraints, and calculate the constraint difference vector for each cooperative agent using the constraint coupling matrix; perform gradient mapping on the constraint difference vector, adaptively offset the constraint terms of the immediate reward function of the cooperative agent, and establish a constraint correction vector; use the constraint correction vector to update the neighborhood propagation of the communication topology, perform cooperative adjustment, and complete the compensation optimization.

[0033] Optionally, after identifying the global deviation regions, a constraint set of all cooperative agents related to these regions is extracted. Each cooperative agent's constraint set includes all resource limitations, time constraints, equipment load, and other restrictions faced by the agent when executing tasks. Based on these agent constraint sets, a constraint coupling matrix is ​​established. This constraint coupling matrix is ​​a two-dimensional matrix structure, whose elements characterize the strength of the mutual influence of constraints between different cooperative agents. The rows and columns of the matrix correspond to different cooperative agents. In this constraint coupling matrix, the matrix elements are constraint coupling strength values. This constraint coupling strength characterizes the degree of constraint association between two cooperative agents in terms of resource usage, task timing, or space occupancy. Its value can be quantified based on the shared resource ratio, task time overlap, spatial task area overlap ratio, and historical conflict frequency, and after normalization, it takes values ​​from 0 to 1, where a larger value indicates a higher degree of constraint coupling between the two cooperative agents. Subsequently, for each cooperative agent, the difference between the constraint state vectors of other cooperative agents and its own constraint state vector is calculated. This difference is then multiplied by the constraint coupling strength between the two cooperative agents in the constraint coupling matrix. By accumulating these products, a constraint difference vector for each cooperative agent is obtained. These difference vectors quantify the degree of deviation of the current cooperative agent from its associated cooperative agents in terms of constraints. Their direction represents the direction of constraint adjustment, and their magnitude represents the strength of the constraint adjustment, thus reflecting the constraint changes faced by the cooperative agent during task execution. Afterward, gradient mapping is applied to the calculated constraint difference vectors to determine the direction and magnitude of their changes. This process analyzes the constraint differences of each collaborative agent, calculates the gradient of constraint changes, and provides each agent with specific adjustment directions. This helps the agents understand how to adjust their strategies to reduce constraint conflicts and optimize resource usage under current resource and task constraints. Then, using the gradient-mapped information, the constraint terms of the immediate reward function are adaptively offset. For example, if an agent faces resource conflicts, the negative rewards caused by the conflict are reduced, encouraging it to adjust its strategy and choose resource usage methods with less conflict. Next, based on the adaptively offset reward function, a constraint correction vector is generated. This vector represents the agent's optimized adjustment direction in terms of resource and task constraints after correction, providing precise behavioral adjustment guidance for the collaborative agents. This allows them to choose the optimal operation plan to reduce constraint conflicts and improve efficiency in the event of resource conflicts or task inconsistencies. After obtaining the constraint correction vector, it is updated through neighborhood propagation based on the communication topology, ensuring that each agent can receive adjustment information from other agents in a timely manner, guaranteeing the coordination of global decisions.After information is propagated in the neighborhood, a dual-channel collaborative adjustment is used to optimize the strategy of each agent, ensuring the overall coordination of the operation. Through continuous iteration and optimization, compensation optimization is finally achieved, realizing the effective allocation of resources, the rational scheduling of tasks, and the minimization of conflicts, ensuring that the collaborating agents can work together efficiently and successfully complete the task.

[0034] Furthermore, the collaborative scheduling of construction operations based on the global correction results includes: Configure task verification nodes using global correction results; perform collaborative execution status verification on a node-by-node basis based on the task verification nodes to generate a node verification dataset; and perform collaborative scheduling compensation management based on the node verification dataset.

[0035] Preferably, after obtaining the global calibration results, to ensure effective task execution and reasonable resource allocation, task verification nodes are configured based on the global calibration results. Task verification nodes refer to key locations or stages during construction where their execution status needs to be monitored and verified in real time. These task verification nodes typically involve task progress verification, resource usage verification, and work coordination verification, used to ensure that the execution status of these key tasks or areas meets expectations, promptly identify potential problems, and make adjustments. After configuring the task verification nodes, the collaborative execution status of each node is verified one by one. During this process, it monitors whether the task is progressing according to the predetermined schedule, and quantifies the execution progress indicator by the difference between the actual and expected progress; it monitors whether the task is executed according to the planned resource allocation, and quantifies the resource consumption indicator by the difference between the actual and planned resource consumption; it quantifies the conflict and interference indicator by statistically analyzing the number of resource conflicts or spatial interferences between various work units. By summarizing these indicators, a node verification dataset is formed, providing a basis for subsequent scheduling decisions. Subsequently, these nodes are used to verify the dataset for task execution deviation analysis. Based on the analysis results, corresponding early warning signals are matched, and then collaborative scheduling compensation management is carried out based on these early warning signals. This allows users to be aware of existing scheduling problems in a timely manner, ensuring that all aspects of the construction process operate efficiently and avoiding the impact of small-scale task deviations on the progress and quality of the entire construction project.

[0036] Furthermore, collaborative scheduling compensation management based on the node verification dataset includes: Based on the node verification dataset, early warning signals for task execution deviations are matched; after configuring an early warning strategy using the matching results, early warning issuance management is implemented.

[0037] Optionally, after obtaining the node verification dataset, the metrics in the dataset are compared with the corresponding multi-level task execution deviation ranges to determine the current warning level, and then the corresponding warning signal is matched accordingly. Subsequently, this warning signal is matched with the warning strategy library to configure a suitable warning strategy. This warning strategy typically includes warning thresholds, response strategies, and warning levels, and can be adjusted and optimized in real time based on actual conditions to prevent further deterioration of the problem. After configuring the warning strategy, warning issuance management is implemented. That is, when the warning signal matches to a certain extent and reaches a preset threshold, a corresponding warning report is immediately generated, and relevant personnel are notified to take countermeasures, thereby improving overall construction efficiency, reducing costs, and ensuring that the construction project is completed on time and to the required quality.

[0038] In summary, the embodiments of this application have at least the following technical effects: First, multi-source data acquisition is performed, and a multi-source temporal state tensor is established using a spatiotemporal correlation mechanism. The multi-source data includes ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data. Next, the multi-source temporal state tensor is input into a 3D underground space BIM model. Based on the parameters of the 3D underground space BIM model, a digital twin is generated that evolves synchronously with the applied construction state. Then, after reading the work unit, the work unit is abstracted into multiple collaborative agents, each possessing local perception, task priority evaluation, and conflict constraint learning capabilities. Next, using the digital twin, the local twin state and resource constraints of each collaborative agent are configured. Game analysis of the collaborative agents is then performed through an immediate reward function to establish a local execution strategy. Finally, the local execution strategy is aggregated globally to establish a global correction result, and construction operations are collaboratively scheduled based on the global correction result. It solves the technical problems of difficulty in simulating the construction status in real time and the lack of scientific strategy guidance for collaboration between work units, which easily leads to conflicts and waste of resources. It achieves the technical effect of providing scientific strategies for collaboration between work units, reducing conflicts, optimizing resource allocation, and improving construction collaboration efficiency by constructing a digital twin that evolves synchronously with the construction status.

[0039] Example 2, based on the same inventive concept as the construction operation collaborative scheduling method based on multi-source data fusion in the previous examples, such as... Figure 2 As shown, this application provides a construction operation collaborative scheduling system based on multi-source data fusion, wherein the system includes: Data Acquisition Module 11: Performs multi-source data acquisition, establishes a multi-source temporal state tensor using a spatiotemporal correlation mechanism, the multi-source data including ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data; Digital Twin Module 12: Inputs the multi-source temporal state tensor into a three-dimensional underground space BIM model, and generates a digital twin that evolves synchronously with the applied construction state based on the parameters of the three-dimensional underground space BIM model; Intelligent Agent Construction Module 13: After reading the work unit, abstracts the work unit into multiple collaborative intelligent agents, each of which has the ability to perceive locally, evaluate task priorities, and learn conflict constraints; Game Analysis Module 14: After configuring the local twin state and resource constraints of each collaborative intelligent agent using the digital twin, performs game analysis of the collaborative intelligent agents through an immediate payoff function to establish a local execution strategy; Job Scheduling Module 15: Performs global state aggregation on the local execution strategy, establishes a global correction result, and performs collaborative scheduling of construction operations based on the global correction result.

[0040] Furthermore, the intelligent agent construction module 13 performs the following method: The collaborative agent is equipped with a reward mechanism, which incorporates nonlinear target weights. These nonlinear target weights are updated based on the evolutionary characteristics of the external environment. The collaborative agent, adjusted according to the reward mechanism, then performs game analysis.

[0041] Furthermore, the game analysis module 14 is used to perform the following methods: For each collaborative agent, a set of state variables corresponding to the work unit is extracted from the digital twin. After normalizing the set of state variables, a local state feature vector is constructed. Resource usage data from the field is read, and resource constraints are established based on this data. The policy space of the collaborative agent is then defined according to these resource constraints. An immediate reward function is constructed using the local state feature vector, which includes a geological disturbance risk term, an equipment load rate term, a work progress deviation term, and an energy consumption cost term. The immediate reward function is used to evaluate the benefit of the fitted execution strategy for each collaborative agent within the policy space. Based on the communication topology, local benefits and conflict constraints are shared, and game iteration is performed based on the shared results to establish a local execution strategy.

[0042] Furthermore, the game analysis module 14 is used to perform the following methods: Based on the spatial location, operation type, equipment interference relationship, and resource coupling degree of each collaborative agent in the 3D underground space BIM model, a communication topology diagram is constructed. The edge weights of the communication topology diagram dynamically represent the information interaction intensity and resource competition intensity among the collaborative agents. After each collaborative agent calculates its benefit evaluation, the local benefit vector and conflict constraint parameters are broadcast in a neighborhood-level weighted manner through the communication topology. After receiving the neighborhood shared results, each collaborative agent performs an adaptive update of the immediate benefit function based on the neighborhood benefit volatility and resource competition gradient, and generates a local strategy search preference. The game is then iterated based on the adaptively updated immediate benefit function and strategy search preference.

[0043] Furthermore, the game analysis module 14 is used to perform the following methods: The updated immediate payoff function is used to re-evaluate the candidate policy set of the corresponding cooperative agent, and a payoff change gradient vector is established. Policy migration direction parameters are generated based on the payoff change gradient vector and policy search preferences. These parameters are shared in the communication topology, and a dual-channel game update is performed. The dual channels include a synchronous payoff migration channel and a conflict constraint feedback channel. The synchronous payoff migration channel is used to adjust the probability distribution of its own policy based on the neighborhood average payoff trend, and the conflict constraint feedback channel is used to compensate for changes in neighborhood conflict constraints. A local execution policy is established based on multiple rounds of dual-channel game iterations.

[0044] Furthermore, the job scheduling module 15 is used to execute the following method: All local execution strategies are obtained, and a global state tensor is constructed based on the spatial dependency relationships and task temporal constraints within the digital twin. The global state tensor represents the spatiotemporal interference distribution and resource consumption mapping of each cooperating agent in the underground space. Based on the global state tensor, a strategy consistency index and a resource conflict index are calculated. Local regions where the strategy consistency index is lower than a first threshold or the resource conflict index is higher than a second threshold are marked as global deviation regions. Compensation optimization is performed on the global deviation regions under a dynamic constraint alignment mechanism to establish a global correction result.

[0045] Furthermore, the job scheduling module 15 is used to execute the following method: Extract the associated set of cooperative agent constraints within the global deviation region, establish a constraint coupling matrix based on the set of cooperative agent constraints, and calculate the constraint difference vector for each cooperative agent using the constraint coupling matrix; perform gradient mapping on the constraint difference vector, adaptively offset the constraint terms of the immediate reward function of the cooperative agent, and establish a constraint correction vector; use the constraint correction vector to update the neighborhood propagation of the communication topology, perform cooperative adjustment, and complete the compensation optimization.

[0046] Furthermore, the job scheduling module 15 is used to execute the following method: Configure task verification nodes using global correction results; perform collaborative execution status verification on a node-by-node basis based on the task verification nodes to generate a node verification dataset; and perform collaborative scheduling compensation management based on the node verification dataset.

[0047] Furthermore, the job scheduling module 15 is used to execute the following method: Based on the node verification dataset, early warning signals for task execution deviations are matched; after configuring an early warning strategy using the matching results, early warning issuance management is implemented.

[0048] 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 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 construction operation collaborative scheduling method based on multi-source data fusion, characterized in that, The method includes: Multi-source data acquisition is performed, and a multi-source temporal state tensor is established using a spatiotemporal correlation mechanism. The multi-source data includes ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data. The multi-source temporal state tensor is input into the three-dimensional underground space BIM model, and a digital twin that evolves synchronously with the applied construction state is generated based on the updated parameters of the three-dimensional underground space BIM model. After reading the task unit, the task unit is abstracted into multiple collaborative agents, each of which has the ability to perceive local situations, evaluate task priorities, and learn conflict constraints. After configuring the local twin state and resource constraints of each collaborative agent using the digital twin, game analysis of the collaborative agents is performed through an immediate reward function to establish a local execution strategy; The local execution strategy is aggregated globally to establish a global correction result, and construction operations are coordinated and scheduled based on the global correction result.

2. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 1, characterized in that, Game analysis of collaborative agents is performed using immediate payoff functions, including: For each collaborative agent, extract the set of state variables corresponding to the work unit from the digital twin, and construct a local state feature vector after normalizing the set of state variables; Read the resource usage data on site, establish resource constraints based on the resource usage data, and limit the policy space of the collaborative intelligent agent based on the resource constraints; An instantaneous revenue function is constructed using the local state feature vector, which includes a geological disturbance risk term, an equipment load rate term, an operation schedule deviation term, and an energy consumption cost term. The immediate reward function is used to evaluate the fitted execution strategy of each cooperative agent in the policy space, and local execution strategies are established by sharing local rewards and conflict constraints based on the communication topology and conducting game iteration based on the shared results.

3. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 2, characterized in that, Based on the sharing of local benefits and conflict constraints in the communication topology, game iteration is performed according to the sharing results, including: Based on the spatial location, operation type, equipment interference relationship and resource coupling degree of each collaborative intelligent agent in the three-dimensional underground space BIM model, a communication topology diagram is constructed. The edge weights of the communication topology diagram dynamically represent the information interaction intensity and resource competition intensity among the collaborative intelligent agents. After each cooperative agent calculates its benefit evaluation, the local benefit vector and conflict constraint parameters are broadcast in a neighborhood-level weighted manner through the communication topology. After receiving the shared results from the neighborhood, each collaborative agent adaptively updates the immediate payoff function based on the neighborhood payoff volatility and resource competition gradient, and generates a local policy search preference. The agent then iterates through the game based on the adaptively updated immediate payoff function and policy search preference.

4. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 3, characterized in that, The game iterative process is based on an adaptively updated immediate payoff function and policy search preferences, including: The updated immediate benefit function is used to re-evaluate the immediate benefit of the candidate policy set of the corresponding cooperative agent and establish a gradient vector of benefit change. Based on the gradient vector of the change in payoff and the policy search preference, a policy migration direction parameter is generated. The policy migration direction parameter is shared in the communication topology and a dual-channel game update is performed. The dual channels include a synchronous payoff migration channel and a conflict constraint feedback channel. The synchronous payoff migration channel is used to adjust the probability distribution of its own policy according to the average payoff trend of the neighborhood. The conflict constraint feedback channel is used to compensate the resource usage plan according to the change of neighborhood conflict constraints. Based on multiple rounds of dual-channel game iteration, a local execution strategy is established.

5. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 1, characterized in that, The local execution strategy is subjected to global state aggregation to establish a global correction result, including: All local execution strategies are obtained, and a global state tensor is constructed based on the spatial dependency relationships and task temporal constraints within the digital twin. The global state tensor represents the spatiotemporal interference distribution and resource consumption mapping of each cooperating agent in the underground space. Calculate the strategy consistency index and resource conflict index based on the global state tensor, and mark the local areas where the strategy consistency index is lower than the first threshold or the resource conflict index is higher than the second threshold as global deviation areas. The global deviation region is compensated and optimized under a dynamic constraint alignment mechanism to establish a global correction result.

6. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 5, characterized in that, The compensation optimization for the global deviation region under a dynamic constraint alignment mechanism includes: Extract the associated set of cooperative agent constraints within the global deviation region, establish a constraint coupling matrix based on the set of cooperative agent constraints, and use the constraint coupling matrix to calculate the constraint difference vector for each cooperative agent. Gradient mapping is performed on the constraint difference vector, and the constraint terms of the immediate reward function of the cooperative agent are adaptively offset to establish a constraint correction vector. The constraint correction vector is used to update the neighborhood propagation of the communication topology, perform cooperative adjustment, and complete the compensation optimization.

7. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 1, characterized in that, Based on the global correction results, collaborative scheduling of construction operations is performed, including: Configure the task verification node using the global calibration results; Based on the task verification nodes, the collaborative execution status of each node is verified to generate a node verification dataset; Cooperative scheduling compensation management is performed based on the node verification dataset.

8. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 7, characterized in that, Based on the node verification dataset, collaborative scheduling compensation management is performed, including: Match early warning signals for task execution deviations based on the node verification dataset; After configuring the early warning strategy using the matching results of the early warning signals, the early warning issuance management is executed.

9. The construction operation collaborative scheduling method based on multi-source data fusion as described in claim 1, characterized in that, The collaborative agent is equipped with a reward mechanism, which incorporates nonlinear target weights. These nonlinear target weights are updated based on the evolutionary characteristics of the external environment. The collaborative agent, adjusted according to the reward mechanism, then performs game analysis.

10. A construction operation collaborative scheduling system based on multi-source data fusion, characterized in that, The system is used to implement the construction operation collaborative scheduling method based on multi-source data fusion as described in any one of claims 1-9, the system comprising: Data acquisition module: performs multi-source data acquisition and establishes a multi-source time-series state tensor using a spatiotemporal correlation mechanism. The multi-source data includes ground-penetrating radar and drilling feedback data, construction machinery and equipment operation data, material and personnel positioning data, and environmental monitoring data. Digital twin module: Input the multi-source temporal state tensor into the three-dimensional underground space BIM model, and generate a digital twin that evolves synchronously with the applied construction state based on the updated parameters of the three-dimensional underground space BIM model; Intelligent agent construction module: After reading the task unit, the task unit is abstracted into multiple cooperative intelligent agents, each of which has the ability to perceive locality, evaluate task priority and learn conflict constraints. Game analysis module: After configuring the local twin state and resource constraints of each collaborative agent using the digital twin, the module performs game analysis of the collaborative agents through an immediate payoff function to establish a local execution strategy; The job scheduling module performs global state aggregation on the local execution strategy, establishes a global correction result, and performs collaborative scheduling of construction jobs based on the global correction result.

Images