Intelligent scheduling and coordination system and method for urban reconstruction full life cycle

Abstract

The present application belongs to the technical field of smart city construction and intelligent construction management, and specifically relates to an intelligent scheduling and collaboration system and method for the whole life cycle of city reconstruction. The system includes an edge perception layer, a digital twin weaving module, a space-time brain module, a blockchain collaborative trust layer, and an XR collaborative interaction layer. The system collects multi-source data in real time through 5G edge devices on the construction site, uses graph convolutional neural networks to realize the fusion modeling of BIM, GIS and TIN models, and combines meta-reinforcement learning and dynamic graph neural networks to perform multi-objective scheduling optimization. At the same time, the system realizes the linkage of construction task verification and fund payment through blockchain smart contracts, and provides immersive collaboration and visual acceptance with the AR / VR platform. The present application solves the problems of model fragmentation, scheduling response lag, low data collaboration efficiency, and insufficient creditworthiness in the process of performance, and significantly improves the intelligent, transparent, and automated level of the construction process.

Description

Technical Field

[0001] This invention relates to the field of smart city construction and intelligent construction management technology, and in particular to an intelligent scheduling and collaboration system and method for the entire life cycle of urban reconstruction. Background Technology

[0002] With the rapid advancement of urban renewal and infrastructure renovation projects, traditional construction management models face a series of challenges, including fragmented multi-source data, delayed decision-making response, and difficulty in verifying execution processes. To achieve efficient operation of urban redevelopment projects under complex conditions, smart construction and digital supervision methods are gradually becoming development trends. While existing technologies have introduced BIM (Building Information Modeling), GIS (Geographic Information System), and IoT sensing devices, initially constructing a digital management framework for the entire construction process, significant shortcomings remain in the following aspects:

[0003] 1. Limited Model Fusion Capabilities: Currently, BIM, GIS, and TIN models are maintained independently, lacking effective topological fusion and semantic alignment mechanisms, making it difficult to support cross-system task collaboration within a unified space. Model boundary data conflicts occur frequently, especially in the process of 3D map modeling and design drawing registration in old urban areas, where fusion errors are particularly prominent, leading to mismatches in construction plans and waste of material resources.

[0004] 2. Slow and unintelligent scheduling response: Existing construction scheduling mostly adopts static planning methods, lacking the ability to respond in real time to dynamic construction environments (such as weather changes, personnel behavior, traffic congestion, etc.). Even when some AI models are introduced, their algorithms are generally optimized for single tasks and fail to form a strategic trade-off between multiple objectives (schedule, carbon emissions, safety, etc.), resulting in insufficient system robustness and inability to adapt to complex on-site changes.

[0005] 3. Low transparency in data collaboration and contract fulfillment: Due to data barriers among the various parties involved in construction, the progress and acceptance process heavily rely on manual confirmation and paper records, which easily leads to information delays, tampering risks, and frequent disputes. Although some projects have attempted to introduce blockchain technology for data on-chaining, most only achieve the "log recording" function and lack a reliable triggering mechanism that combines with the actual project status (such as LiDAR acceptance point cloud).

[0006] 4. Insufficient human-computer interaction efficiency: Traditional visualization platforms are disconnected from operational terminals, making it difficult for workers to obtain information such as drawings, tasks, and risk warnings through an intuitive interface. Some AR systems suffer from low spatial registration accuracy and high command interaction latency, limiting their practicality in complex construction sites.

[0007] 5. Lack of support for green and low-carbon construction: Most scheduling systems do not incorporate carbon emission indicators as a constraint variable into resource allocation optimization, and lack equipment carbon emission records and path planning mechanisms, which is detrimental to green construction. Furthermore, the interpretability of AI strategy models is poor, making it difficult for construction management units to conduct causal audits and compliance traceability of scheduling results.

[0008] To address the shortcomings of existing technologies, there is an urgent need for an intelligent scheduling and collaboration system for the entire lifecycle of urban redevelopment. This system should be able to integrate multi-source model data to build a unified digital twin environment; possess adaptive multi-task optimization scheduling capabilities; and significantly improve construction efficiency, management transparency, and contract fulfillment credibility through a reliable execution verification mechanism and efficient human-computer interaction methods. It should also support the continuous evolution of green and low-carbon goals and model transfer capabilities. Summary of the Invention

[0009] To address the common technical bottlenecks in existing urban redevelopment projects, such as fragmented model data, delayed scheduling response, low collaborative efficiency, and insufficient reliability of task execution, this invention proposes an intelligent scheduling and collaboration system and method that integrates multi-source sensing, digital twin modeling, graph neural network intelligent decision-making, blockchain performance verification mechanisms, and XR immersive interactive collaboration. This system, with "data-driven, model-unified, intelligent optimization, and reliable execution" as its core logic, constructs an intelligent, modular, and traceable closed-loop system covering the entire urban redevelopment process. It achieves real-time sensing and standardized fusion of construction site data, supports dynamic scheduling decisions for cross-model topology unification and multi-objective tasks, and, in conjunction with on-chain task verification and immersive collaborative feedback mechanisms, significantly improves the security, efficiency, transparency, and sustainability of urban infrastructure upgrades. Based on this, the invention proposes the following technical solutions.

[0010] In one possible implementation, an intelligent scheduling and coordination system for the entire lifecycle of urban redevelopment is provided, which adopts a layered modular structure and includes the following functional layers:

[0011] 1. Edge perception layer, comprising a cluster of 5G IoT devices and edge computing nodes deployed at the construction site, is used to collect 3D point clouds, equipment operating status, environmental parameters, and personnel trajectory information. The sampling frequency is no less than 10Hz, and the point cloud data is preprocessed using a Voxel Grid filter, with a compressed voxel size of 0.05m. Asynchronous data transmission and timestamp alignment are achieved through the ROS2 communication framework, with communication latency controlled within 30ms. Edge nodes integrate a Kalman filter module for IMU trajectory data denoising and attitude correction.

[0012] 2. The Digital Twin Network Module is used to integrate BIM (Building Information Modeling), GIS (Geographic Information System), and TIN (Triangulated Integrity Network) models to construct a unified topology map. This module uses a Graph Convolutional Neural Network (GCN) to process the topology structure, with node features having a dimension of 64, including spatial coordinates, semantic labels, and physical attributes. The GCN adopts a three-layer network architecture: the front end enhances semantic awareness through a Transformer substructure, and the back end uses a softmax weighted fusion mechanism to resolve model boundary conflicts. The model edge generation rule is based on the Euclidean distance threshold d < 0.3m, and the RBF kernel bandwidth σ in the topology alignment loss is determined by grid search with a step size of 0.1 in the interval [0.1, 1.0].

[0013] 3. The Spatiotemporal Brain module implements multi-task scheduling decisions based on meta-reinforcement learning and Dynamic Graph Neural Network (DGNN). The policy network is built on the RLlib training framework with a state space dimension of 23, and the PPO algorithm is used for training policies. During the pre-training phase, historical scheduling trajectories are loaded for imitation learning, and during the fine-tuning phase, online reinforcement learning is used to achieve policy adaptation. The DGNN model introduces a graph attention mechanism (GAT) to enhance edge feature representation, and during inference, an early-exit mechanism is used to return high-confidence results in advance, optimizing the average response time to 130ms.

[0014] 4. A blockchain collaborative trust layer, based on the Hyperledger Fabric consortium blockchain architecture, integrates Oracle nodes to verify the construction completion status. Point clouds of the site are acquired via a LiDAR module and compared with the BIM model. When the percentage of points with a matching error of less than 2.5cm reaches 97% and the overall difference rate is less than 3%, a credible acceptance certificate is generated, triggering the chaincode to execute payment logic and recording the hash and GPS timestamp. To enhance credibility, a two-factor acceptance mechanism combining design model hash verification and AI error prediction comparison is also introduced.

[0015] 5. The XR collaborative interaction layer supports AR navigation and VR sandbox functionality, creating an immersive collaborative environment. Spatial registration is achieved using AprilTag combined with SLAM (positioning accuracy ≤2cm). The VR sandbox supports multi-role collaboration and process playback, with construction drawings and risk warnings overlaid on the user's view in real time using graphics and text. Command control employs a voice + gesture fusion approach, equipped with an intent recognition module based on graph neural networks, achieving an accuracy rate of over 95%. Collaborative tasks support multi-user status synchronization, with a maximum interaction latency of no more than 100ms.

[0016] In one possible implementation, the system further includes:

[0017] The carbon emission optimization module is used to establish a mapping table between construction equipment and carbon emission factors, and dynamically generate low-carbon scheduling paths by combining task paths, equipment types and environmental conditions.

[0018] The causal analysis module uses SHAP values ​​to perform causal attribution analysis on input variables in the policy network, forming a task causal graph.

[0019] The federated upgrade module uses a differential privacy mechanism to protect data privacy, enabling secure aggregation of policy parameters and personalized optimization of local models across multiple projects. The differential privacy parameter ε is controlled between 0.5 and 1.

[0020] This invention dynamically adjusts the ε value (0.5–1.0) by classifying scenarios and training rounds to prioritize privacy protection under high-sensitivity data and improve model performance under low-sensitivity data, thus balancing security and performance.

[0021] In one possible implementation, the edge weight update of the DGNN model is controlled by the following coupling function:

[0022]

[0023] Where Wt is the edge weight matrix at time t, L is the joint loss function, α is the learning rate, β is the weight parameter of the coupling function, and Φ(St,Tt) is the spatiotemporal coupling factor, calculated as follows:

[0024]

[0025] Among them, S t Let the current state tensor be... To predict the state, T t λ is the timestamp, and λ is the time decay coefficient, with a default value of 0.5, which is adjusted to 0.8 when a high-risk event is detected.

[0026] In one possible implementation, the loss function for topology alignment in the GCN model is defined as:

[0027]

[0028] Where f θ γ is the node feature extraction function, γ is the balancing parameter, MMD (maximum mean difference) is used to measure the consistency of the distribution, and the kernel function bandwidth σ is determined by grid search in the interval [0.1, 1.0] to optimize training accuracy.

[0029] To further ensure the reproducibility of parameter σ selection, this invention divides the interval [0.1, 1.0] into 10 candidate values ​​with a step size of 0.1. For each candidate σ, the node classification accuracy is trained and evaluated on a fixed random seed and the same validation set. Finally, the σ with the highest average accuracy on the validation set is selected as the kernel bandwidth.

[0030] In one possible implementation, a method for achieving the above-mentioned function is provided, which includes the following steps:

[0031] S101: Utilizes 5G edge devices to collect multimodal data from the construction site and performs heterogeneous encapsulation via ROS2;

[0032] S102: Use GCN to perform topological fusion of BIM, GIS, and TIN models to construct a digital twin graph structure;

[0033] S103: Construct a policy network based on meta-reinforcement learning, and use DGNN to complete multi-objective inference and policy delivery;

[0034] S104: Trigger LiDAR scanning and compare with the BIM model. When the set difference threshold is met, automatically execute the blockchain contract.

[0035] S105: Provides visual navigation, immersive acceptance testing, and remote collaboration support through an XR platform;

[0036] S106: Implement policy transfer and local model update through federated learning mechanism.

[0037] In one possible implementation, a computer-readable storage medium is provided storing program instructions that, when executed by a processor, cause a device to perform the methods described in the steps above.

[0038] Based on the above technical solutions, the intelligent scheduling and collaboration system proposed in this invention, which addresses the entire lifecycle of urban redevelopment, effectively solves the following common problems in urban redevelopment by constructing a five-layer architecture (perception, modeling, decision-making, verification, and interaction) that integrates AI, BIM, GIS, blockchain, and XR technologies:

[0039] 1. Fragmented models and difficult data fusion: A GCN+Transformer fusion mechanism is proposed to achieve coordinated and consistent expression of heterogeneous model space and semantics;

[0040] 2. Lagging response of scheduling strategy: DGNN introduces graph attention mechanism and early-exit to optimize the policy inference process and shorten the policy response time;

[0041] 3. Lack of verifiability in acceptance and performance: Combining LiDAR point cloud comparison with a blockchain two-factor acceptance mechanism ensures credible execution and transparent fund linkage;

[0042] 4. Inefficient human-machine collaboration: Construct an immersive intelligent interactive environment by embedding semantic recognition and collaborative state synchronization mechanisms in AR / VR;

[0043] 5. Weak policy transferability and interpretability: Integrating federated learning and causal analysis modules enhances policy adaptability and AI auditability;

[0044] 6. Green and low-carbon goals are difficult to achieve: The built-in carbon emission path dynamic optimization module supports carbon emission control as a constraint condition to participate in scheduling decisions.

[0045] This invention not only possesses a complete technical closed loop of "perception-fusion-scheduling-execution-verification", but also significantly improves the level of intelligence, reliability and low carbon emissions through structural optimization and algorithm nesting, and has significant engineering application value.

[0046] Experimental Environment and Scheme

[0047] Hardware environment: CPU Intel Xeon Gold 6248R@3.0GHz, 64GB RAM; NVIDIA Tesla V100 GPU

[0048] Software environment: Ubuntu 20.04 + TensorFlow 2.6 + PyTorch 1.10 + Hyperledger Fabric 2.3 blockchain simulation platform

[0049] Comparison of options:

[0050] Traditional digital twin + single GCNN inference architecture (Baseline)

[0051] This invention utilizes a multi-model weighted fusion of a "spatiotemporal brain," edge reasoning, and blockchain parallel verification (Proposed).

[0052] Table 1: Comparison of Key Performance Indicators

[0053]

[0054]

[0055] Results Analysis

[0056] 1. Inference performance

[0057] After deploying lightweight models for parallel inference at edge nodes, the average latency decreased from the baseline of 200ms to 120ms, a reduction of 40%, significantly improving real-time performance and meeting the stringent low-latency requirements of large-scale urban redevelopment decisions.

[0058] Peak throughput increased from 50 times / second to 82 times / second, a 64% increase, enabling the system to support more parallel decision requests in high-concurrency scenarios and ensuring efficient operation even in complex environments.

[0059] 2. Prediction accuracy

[0060] By weighted fusion of Graph Convolutional Network (GCN) and Dynamic Graph Neural Network (DGNN) and the introduction of multi-source data, the model's event prediction accuracy was improved from 86.3% to 93.2%, an increase of 6.9%, effectively reducing false negatives and enhancing the reliability of the system.

[0061] 3. Energy consumption performance

[0062] With the help of model pruning and edge inference optimization, the average power consumption of a single node was reduced from 142W to 96W, a reduction of about 32.4%, which significantly reduced the energy burden and provided lower operation and maintenance costs for large-scale deployment.

[0063] 4. Trusted verification efficiency

[0064] By adopting parallel packaging and a lightweight consensus algorithm, the blockchain transaction confirmation latency was reduced from 110ms to 65ms, a decrease of approximately 40.9%. Under the premise of ensuring data immutability and traceability, it no longer became a system bottleneck.

[0065] 5. End-to-end response experience

[0066] The overall end-to-end latency from on-site perception to XR interaction feedback has been reduced from 410ms to 250ms, a reduction of 39%. Users can operate in the AR / VR interface with almost no noticeable lag, which greatly improves the smoothness of interaction and user experience.

[0067] In summary, through technical optimizations in multi-model fusion, collaborative scheduling, model pruning, and parallel verification, this system has achieved significant improvements in multiple dimensions, including latency, throughput, accuracy, energy consumption, and reliability, effectively meeting the stringent performance requirements of intelligent scheduling throughout the entire lifecycle of urban reconstruction. Attached Figure Description

[0068] To more clearly illustrate the technical solution of the present invention, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The accompanying drawings involved in the present invention include, but are not limited to:

[0069] Figure 1 System overall architecture diagram

[0070] This invention illustrates the overall hierarchical structure of an intelligent scheduling and collaboration system for the entire lifecycle of urban redevelopment, including an edge perception layer, a digital twin network module, a spatiotemporal brain module, a blockchain collaborative trust layer, and an XR collaborative interaction layer. Each layer sequentially constitutes a closed-loop system architecture from perception and acquisition, model fusion, intelligent reasoning, execution verification to immersive collaboration.

[0071] Figure 2 Data Acquisition and Preprocessing Flowchart

[0072] This diagram illustrates the data sources and preprocessing flow collected by the system during the edge perception stage. Input sources include BIM / GIS models, sensor arrays, historical work orders, and meteorological APIs. After cleaning, feature extraction, and format conversion, the data is used for subsequent digital twin modeling.

[0073] Figure 3 Digital Twin Web Module Structure Diagram

[0074] This demonstrates the workflow of BIM / GIS input data undergoing topology construction and attribute mapping before entering a three-layer graph convolutional network (GCN) for structural and semantic fusion. Each GCN layer sequentially extracts node spatial features, semantic labels, and physical attributes, ultimately outputting a unified topology diagram.

[0075] Figure 4 Edge sensing layer structure diagram

[0076] The diagram illustrates the perception and encapsulation process of multimodal data at the construction site, including point cloud, meteorological, and personnel trajectory data collected by lidar modules, environmental sensor modules, and worker positioning tags. The data is encapsulated by ROS2 and denoised by a Kalman filter, processed in a Jetson edge computing node, and then uploaded to the dispatch center.

[0077] Figure 5 Spacetime Brain Module AI Reasoning Structure Diagram

[0078] This paper demonstrates a meta-policy network built on the RLlib training platform, employing the PPO algorithm to train a multi-task scheduling model, and deploying it in a Dynamic Graph Neural Network (DGNN) accelerated by TensorRT. The diagram includes graph structure propagation, attention mechanism modules, and early-exit decision-making and response control modules, ensuring inference latency of less than 200 milliseconds.

[0079] Figure 6 Blockchain collaborative trust mechanism diagram

[0080] After the construction task is completed, point cloud data is obtained by LiDAR scanning and compared with the ICP difference rate of the BIM model. After verification by the Oracle node, a reliable acceptance certificate is generated, which automatically triggers the smart contract to execute payment and records the task hash and timestamp. Finally, the result is stored on the blockchain to realize the closed loop of acceptance and payment linkage.

[0081] Figure 7 XR Collaborative Interaction Flowchart

[0082] The demonstration shows that workers wearing AR devices enter the construction site, complete spatial registration through AprilTag and SLAM, and overlay construction drawings and risk warnings on the AR interface. The system uses graph neural networks to recognize gestures and voice commands to interpret operation intentions, and task feedback is recorded via the Fabric chain. The VR sandbox supports multi-role collaboration and process playback. Detailed Implementation

[0083] To better understand the technical solution of this invention, the following detailed description of the intelligent scheduling and collaboration system for the entire lifecycle of urban reconstruction, based on specific implementation scenarios, is provided. This invention can be applied to construction scheduling and management scenarios in various complex environments, such as urban rail transit renovation, municipal road reconstruction, and demolition and renovation of old buildings. Through a five-stage technical process of perception, modeling, reasoning, verification, and collaboration, a closed-loop intelligent management system is formed.

[0084] In specific applications, such as Figure 1 As shown, the following technical solutions can be adopted, including an intelligent scheduling and collaboration system for the entire lifecycle of urban redevelopment, consisting of the following modules: an edge perception layer, including a 5G IoT device cluster, used to collect 3D point clouds, environmental data, and personnel trajectories at the construction site, and to align multi-source heterogeneous data through ROS2; a digital twin weaving module, which uses graph convolutional neural networks (GCN) to perform topological alignment of BIM, GIS, and TIN models, thereby constructing a dynamic multi-scale digital twin; a spatiotemporal brain module, based on a meta-reinforcement learning framework, to perform multi-task decision-making for construction scheduling, traffic management, and carbon emission control; a blockchain collaborative trust layer, relying on the Hyperledger Fabric architecture, which verifies the construction status and triggers smart contracts through Oracle nodes; and an XR collaborative interaction layer, supporting AR navigation and VR sandbox functions, providing immersive collaboration and visual acceptance.

[0085] The system achieves functional collaboration among its modules through a layered architecture, such as... Figure 4As shown, the edge perception layer collects multimodal data from the construction site in real time, encapsulates and synchronizes it via ROS2, and uses Kalman filtering to remove IMU noise data; the digital twin weaving module processes model topology information through a three-layer GCN structure, integrates spatial coordinates, semantic labels and physical attributes, enhances semantic perception with Transformer, and uses a softmax weighting mechanism to resolve model boundary conflicts; the spatiotemporal brain module trains a PPO policy network based on RLlib, combines a dynamic graph neural network (DGNN) for multi-objective scheduling inference, and optimizes the inference latency to 130ms; the blockchain collaborative trust layer integrates a LiDAR point cloud comparison mechanism, and triggers payment by chaincode and records hash and timestamp when the error meets the threshold; the XR collaborative interaction layer realizes spatial registration through AprilTag and SLAM, and the command interaction is based on the intent recognition module of graph neural network, supporting multi-user collaboration.

[0086] This system achieves technological breakthroughs in several aspects: First, the GCN+Transformer structure significantly improves the fusion accuracy and topological consistency of BIM, GIS, and TIN models, solving the problem of model fragmentation; second, the DGNN and early-exit mechanism accelerate scheduling response, effectively addressing complex dynamic changes on-site; third, by combining blockchain and LiDAR point cloud dual-factor acceptance, a trusted execution verification mechanism is constructed, improving the transparency of the performance process; fourth, the XR interaction platform integrating AR / VR and semantic recognition enhances human-machine collaboration efficiency; in addition, the system also possesses advantages such as green and low-carbon scheduling capabilities and the transferability and interpretability of AI strategies, demonstrating high engineering practicality and scalability.

[0087] In terms of material selection, edge sensing devices can be replaced with a combination of industrial-grade cameras and LiDAR that are dustproof and waterproof, depending on the construction environment; the GCN network structure can be changed to a variant based on GraphSAGE to enhance model transfer capabilities; the policy network algorithm can also be replaced from PPO to SAC to enhance continuous motion control performance; in addition to Hyperledger Fabric, the blockchain platform can also adopt a consortium blockchain framework based on Ethereum; in the XR module, the spatial registration method can use a point cloud matching method based on deep learning to improve positioning accuracy; the system as a whole also supports modular deployment, adapting to the flexible configuration and expansion needs of different urban reconstruction scenarios.

[0088] like Figure 2 The diagram shows the data acquisition and preprocessing flowchart of the system of this invention. This flowchart covers multiple input channels and processing nodes, providing a unified, complete, and high-quality data foundation for subsequent modeling, scheduling, and inference work.

[0089] In the data acquisition phase, there are four types of data sources. The first is BIM models and GIS information, which can provide basic spatial topology and structural semantics. The second is sensor arrays, including on-site LiDAR scanning to obtain point cloud data, IMU sensors recording trajectories, and UWB technology for worker positioning, for real-time perception of the on-site situation. The third is historical work orders, including scheduling plan logs, task execution records, and maintenance work orders, reflecting past work situations. The fourth is meteorological APIs, which can obtain meteorological parameters such as temperature, humidity, and wind speed.

[0090] These input data enter the preprocessing stage, undergoing operations such as cleaning, feature extraction, and format conversion. During cleaning, outliers, null values, and duplicate information are removed to ensure data quality. In feature extraction, component numbers and attribute labels in the model are associated and bound to sensor data. Format conversion transforms data from different sources and in different formats; for example, raw BIM data is converted to a graph structure input format, point cloud data is converted to sparse tensor form, and spatial data from different sources is unified into the local coordinate system of the construction site. Furthermore, the entire preprocessing process includes data caching and incremental update mechanisms, supporting daily incremental uploads of data throughout the continuous construction period, thereby improving data integrity and real-time performance. Finally, the processed data results in a structurally unified and semantically clear input data stream, which can be directly used by subsequent graph neural network models and scheduling strategy models.

[0091] As an exemplary implementation, such as Figure 3 As shown, the graph convolutional neural network (GCN) node features used in the digital twin web module include spatial coordinates, semantic labels, and physical attributes. The GCN network structure has three layers, and a weighted fusion algorithm is used at the model boundary to handle topological conflicts.

[0092] This module addresses the spatial and semantic differences between BIM, GIS, and TIN models in urban redevelopment projects by constructing a unified topology map. Each graph node carries three types of features: spatial coordinates (e.g., X, Y, Z coordinates), semantic labels (e.g., "column," "wall," "ground"), and physical properties (e.g., material strength, thermal conductivity). Feature extraction and propagation are performed through a three-layer GCN. The first two layers complete the structured encoding of node feature vectors, while the third layer is used for full-map semantic aggregation and topology alignment. To address topology conflicts at model boundaries, the system introduces a weighted fusion algorithm. Based on softmax-normalized feature similarity, it calculates the fusion priority of adjacent nodes to form a transitional structure, avoiding geometric offsets or semantic mismatches.

[0093] This module significantly improves the fusion quality and topological consistency between heterogeneous models, resolving the frequent conflicts in boundary handling between traditional BIM and GIS / TIN models. It is particularly suitable for scenarios with complex 3D old city spaces and numerous overlapping model areas. Through a weighted fusion algorithm, it can preserve the semantic information of the original model while achieving spatial transitions, enhancing the coherence of the overall model structure and providing a more accurate digital twin foundation for subsequent scheduling and visualization.

[0094] In terms of node feature composition, in addition to spatial coordinates, semantic labels, and physical attributes, timestamps or construction stage identifiers can be introduced to support dynamically updated model version management; the GCN structure can be replaced with Graph Attention Network (GAT) to improve the selectivity of feature aggregation; the weighted fusion algorithm can be replaced based on Euclidean distance, cosine similarity, or fusion score based on neural network learning; in addition, the topology alignment processing flow can also be encapsulated as a modular plugin to support docking and customized development with different modeling platforms (such as Revit, ArcGIS, etc.).

[0095] As an exemplary implementation, such as Figure 5 As shown, the meta-policy network constructed by the spatiotemporal brain module is based on the RLlib training framework, with a state space dimension of 23 and the training algorithm being PPO. During inference, a dynamic graph neural network (DGNN) deployed by TensorRT is called, and the inference response time does not exceed 200 milliseconds.

[0096] This module constructs a meta-policy network with generalization capabilities to perform multi-objective optimization tasks such as construction scheduling, traffic management, and carbon emission control. The system represents dynamic information of the construction site in a 23-dimensional state space, including equipment location, operational status, personnel distribution, weather conditions, and road traffic levels. The meta-policy network is built on RLlib and uses the Proximal Policy Optimization (PPO) algorithm for iterative training. In the pre-training phase, it utilizes historical work trajectory imitation learning, and in the fine-tuning phase, it introduces online reinforcement learning for policy adaptation. In the inference phase, the DGNN model is accelerated using the TensorRT engine, and a graph attention mechanism (GAT) is introduced to enhance the modeling ability of the influence relationships between nodes and edges in the graph. An early-exit mechanism is used to output decisions in advance under high-confidence conditions, thereby controlling the average inference latency to within 200 milliseconds.

[0097] This module boasts high real-time performance and strong generalization capabilities, enabling it to adapt to dynamic changes in urban reconstruction construction site environments and rapidly respond to scheduling demands. A reinforcement learning training framework combining RLlib and the PPO algorithm effectively addresses the conflict and optimization balance issues among multi-objective policies. The introduction of TensorRT to deploy DGNN significantly shortens inference time, meeting the real-time decision-making requirements under complex conditions. Graph structure modeling and attention mechanisms enhance the rationality and interpretability of scheduling decisions, improving the stability and security of the policy.

[0098] The state space dimension can be adjusted according to the specific construction scenario, such as adding dimensions like carbon emission load and equipment energy consumption level to enhance green scheduling capabilities; the PPO algorithm can be replaced by other reinforcement learning algorithms such as A3C or SAC to adapt to different needs of continuous or discrete action spaces; the DGNN inference module can be replaced with frameworks such as ONNX or TVM for cross-platform deployment; in addition to TensorRT, inference acceleration methods can also use CUDA kernel optimization, XLA compiler, etc. to improve processing performance; the threshold of the early-exit mechanism can be set based on different construction tasks to balance response speed and decision accuracy.

[0099] As an exemplary implementation, such as Figure 6 As shown, the Oracle node in the blockchain collaborative trust layer compares the difference rate between the LiDAR scanned point cloud and the design model. When the difference is less than 5%, a trusted proof is generated, and the chaincode automatically executes the payment operation and records the hash value and GPS timestamp.

[0100] This module aims to automate the verification of construction status and link payment execution. After construction is completed, the Oracle node uses LiDAR sensors to perform a 3D point cloud scan of the target area, acquiring a high-precision site model. The system compares this measured point cloud with the original design model, using a minimum error registration algorithm (such as ICP) to calculate the difference rate. When the overall difference rate between the measured point cloud and the design model is less than 5%, and the matching degree of key structures reaches a set threshold (such as 95% point-to-point overlap), the system determines that the task has been completed and meets the requirements. At this point, the Oracle node generates a trusted acceptance certificate, and the chaincode automatically executes the payment logic, while recording the operation hash value and the corresponding GPS timestamp to ensure the verification process is traceable and tamper-proof.

[0101] This module effectively improves the objectivity and automation of construction task acceptance, avoiding errors and disputes arising from traditional reliance on manual inspection and paper reports. Combining the immutability of blockchain with the high-precision comparison mechanism of LiDAR, it achieves reliable linkage and transparency in the performance process, enhancing the trust of all participants in the project's execution status. The payment process is automatically triggered, reducing manual review costs and shortening the cycle, thereby improving overall construction efficiency.

[0102] The difference rate calculation can use various geometric evaluation indicators such as point cloud Hausdorff distance and Chamfer distance, and the threshold can be flexibly adjusted according to different construction accuracy requirements; LiDAR equipment can be replaced with a multimodal camera system with depth sensing function (such as an RGB-D camera) to reduce costs; Oracle nodes can be designed to support multiple data source inputs (such as environmental sensors, image recognition results, etc.) to enhance the robustness of the acceptance strategy; the payment chaincode execution logic can be customized according to contract details, including phased payments, dynamic progress ratio settlement, etc.; the timestamp mechanism can also incorporate satellite synchronous verification and multi-node consensus time to improve anti-counterfeiting capabilities.

[0103] As an exemplary implementation, such as Figure 7 As shown, the AR navigation function of the XR collaborative interaction layer includes: AprilTag combined with SLAM to achieve spatial registration, construction information is superimposed on the real scene in the form of pictures and text, and workers upload progress information and trigger blockchain records through gesture operation.

[0104] The AR navigation module, based on computer vision and graph neural network recognition technology, provides construction workers with an augmented reality user interface. First, the system uses AprilTag tagging combined with SLAM (Simultaneous Localization and Mapping) algorithms to achieve spatial positioning and environmental mapping, ensuring that AR content is accurately anchored to designated components or areas in the real-world scene. Construction task information, risk warnings, and drawing details are presented to the user through an overlay of text and images. Users control the interaction process using natural gestures (such as swiping, clicking, and pointing). The system employs an intent recognition module trained on a graph neural network for gesture parsing, achieving an accuracy rate exceeding 95%. Once workers upload construction progress information (such as task completion and problem reporting), the system uploads the relevant data to the blockchain via a predefined smart contract, creating a traceable progress record.

[0105] This module significantly improves the efficiency of human-machine collaboration and the intuitiveness of information interaction. Spatial registration technology ensures the accuracy of virtual-real alignment and avoids information drift issues; image and text overlay provides clear task guidance and risk warnings, improving operational accuracy and safety; gesture interaction reduces reliance on traditional terminal devices, freeing up construction workers' hands and improving on-site operational flexibility; the blockchain recording mechanism ensures that uploaded data is tamper-proof and synchronized in real time, providing a real, transparent, and reliable task tracking mechanism for construction management.

[0106] The spatial registration method can be replaced with a visual inertial navigation (VIO) mechanism based on RGB-D point clouds or ARKit / ARCore to enhance adaptability; the image and text overlay information can be expanded to multimodal content such as 3D graphics and voice broadcast to enhance immersion; the gesture recognition module can introduce infrared depth cameras or millimeter-wave radar to enhance detection accuracy; in addition to gesture operation, it can also integrate various human-computer interaction methods such as voice recognition and head motion tracking; the blockchain recording mechanism can be configured as a multi-level triggering strategy based on task type and role permissions to achieve more granular data governance.

[0107] As an exemplary implementation, the system further includes: a carbon emission optimization module for establishing a carbon emission mapping relationship for construction equipment and realizing dynamic planning of carbon emission paths; a causal analysis module for using SHAP values ​​to analyze the impact of state variables in multi-objective decision-making; and a federated upgrade module for using a differential privacy mechanism to realize cross-project strategy migration and model localization optimization.

[0108] The carbon emission optimization module constructs a carbon emission factor database, linking the carbon emission characteristics of different construction equipment (such as energy consumption and emissions per unit of operation time) with factors such as work tasks, routes, and climate. It then uses graph search and route planning algorithms to generate low-carbon scheduling routes. This module introduces carbon emission constraints into task scheduling as one of the optimization objectives.

[0109] The causal analysis module introduces the SHAP (SHapley Additive exPlanations) algorithm to evaluate the causal impact of state variables (such as equipment location, load, weather, etc.) input to the reinforcement learning policy network. It calculates the contribution of each variable to the probability of task completion and the change in decision, and draws a causal graph to provide interpretability support for the scheduling strategy.

[0110] The federated upgrade module employs a differential privacy mechanism to securely aggregate policy network parameters across multiple projects without exposing specific construction data, and pushes the optimization results to the local DGNN model, achieving cross-project knowledge transfer and localization adaptation. This module sets the privacy protection strength parameter ε to the range of 0.5 to 1, ensuring data sensitivity through noise perturbation.

[0111] This extension module integrates three major technological directions: green construction, AI auditing, and intelligent adaptation. The carbon emission optimization module achieves green scheduling goals and improves the system's responsiveness to "dual carbon" policies; the causal analysis module enhances model transparency and interpretability, helping construction units review scheduling logic and implement compliance supervision; and the federated upgrade module supports continuous model evolution in different project scenarios, avoiding repeated training costs, improving algorithm generalization ability and deployment flexibility, while ensuring data privacy and security.

[0112] Carbon emission factors can be expanded based on local environmental standards or internal assessment indicators of construction companies; in addition to AI and Dijkstra, path planning algorithms can use reinforcement learning or graph optimization methods to achieve dynamic path replanning; causal analysis methods can use LIME, Counterfactuals, etc. to supplement SHAP for joint interpretation; federated learning frameworks can be replaced by architectures such as Flower and FedAvg that support multiple communication protocols, and differential privacy mechanisms can also adjust the noise intensity and gradient pruning strategy according to the sensitivity level.

[0113] As an exemplary implementation, the edge weight update of the Dynamic Graph Neural Network (DGNN) satisfies the following formula:

[0114]

[0115] Where Wt is the edge weight matrix at time t, L is the joint loss function, α is the learning rate, β is the weight parameter of the coupling function, and Φ(S) is the spatiotemporal coupling factor. t ,T t The calculation method for ) is as follows:

[0116]

[0117] Among them, S t Let the current state tensor be... To predict the state, T t This is the timestamp, with a default time decay coefficient λ of 0.5, which is adjusted to 0.8 when a high-risk event is detected.

[0118] The DGNN model introduces a dynamic coupling effect between time and state differences through the aforementioned edge weight update mechanism, enabling the network structure to adaptively adjust as the construction status changes. The joint loss function L combines scheduling accuracy and carbon emission costs, ensuring the network controls resource consumption while maintaining task completion quality. The coupling factor Φ(S) t ,T t This reflects the deviation between the current state and the predicted state, amplifying the weight adjustment magnitude when state changes drastically or risk events occur, and ensuring the model's limited memory of historical information through a time decay term. This mechanism enhances DGNN's ability to model spatiotemporal change patterns in complex dynamic construction environments.

[0119] After adopting this update mechanism, DGNN can perceive multi-source dynamic factors such as task status, equipment movement, and risk events in the construction scenario in real time, improving the model's response speed and prediction accuracy. The state error term introduced by the coupling factor ensures the scheduling strategy's sensitivity to construction deviations, while time decay control avoids excessive disturbances and model instability, effectively improving the model's robustness and adaptability in long-term operation.

[0120] The coupling function can be adjusted based on different construction task characteristics. For example, cosine similarity and KL divergence can be introduced to replace Euclidean distance to improve semantic alignment. The time decay coefficient can be set as a task sensitivity function for dynamic adjustment, giving higher priority tasks a faster response speed. Constraint regularization terms can be added to the joint loss function to control resource scheduling balance. If the data frequency is high in the construction scenario, the state tensor update frequency can be combined with the decay factor to form an adaptive step size adjustment strategy.

[0121] As an exemplary implementation, the loss function for GCN topology alignment is defined as:

[0122]

[0123] Among them, f θ is the feature extraction function, MMD is the maximum mean discrepancy, γ is the balance parameter, and the bandwidth parameter σ of the RBF kernel function is determined in the interval [0.1, 1.0] through grid search.

[0124] This loss function is designed to guide the GCN model in achieving alignment and consistent fusion of BIM and GIS models in the feature space. The first term measures the distance between feature vectors at the node using Euclidean distance, ensuring geometric and semantic consistency at the single-point level. The second MMD term measures the structural deviation between the two models in the latent representation space using the overall distribution difference measure, thereby optimizing global consistency. The balance coefficient γ is used to weigh the contribution between local feature alignment and overall distribution fusion. The RBF kernel function is used to compute the kernel embedding map in the MMD, and its bandwidth parameter σ is determined through grid search to obtain optimal distribution matching performance.

[0125] This loss function effectively improves the fusion accuracy of BIM and GIS models in topological alignment, resolving issues such as model geometric misalignment and semantic label inconsistency. Local point-pair feature differences control detail errors, while global MMD distribution matching enhances overall spatial coordination, improving the topological stability and task execution reliability of the entire digital twin model and providing solid data support for subsequent scheduling and visualization operations.

[0126] The feature extraction function fθ can be implemented using graph neural network architectures such as GCN, GAT, or GraphSAGE based on different task scenarios; the MMD metric can be replaced with more complex distribution distance functions such as Sinkhorn distance and Wasserstein distance; the RBF kernel function can also be replaced with a multi-kernel combination strategy or a deep learning-driven kernel function adaptive mechanism; the γ value can be dynamically adjusted according to the performance of the validation set, or it can be set as a function of the number of training iterations to enhance the model's early convergence stability and later fusion accuracy.

[0127] In one embodiment, an intelligent scheduling and coordination method for the entire lifecycle of urban redevelopment includes the following steps:

[0128] S101: Collects 3D point cloud data, equipment operating status and environmental parameters of the construction site through 5G edge devices, and uses ROS2 packaging for synchronization;

[0129] S102: Utilize GCN to perform topological fusion of BIM, GIS, and TIN models to generate a unified digital twin structure;

[0130] S103: Based on meta-reinforcement learning to train policy networks, using DGNN for multi-objective inference, with a response latency of less than or equal to 200ms;

[0131] S104: Call LiDAR to scan the task area, detect the model difference rate, and automatically trigger the blockchain smart contract when the threshold condition is met.

[0132] S105: Provides construction navigation and collaborative acceptance functions through an AR / VR platform;

[0133] S106: Under the condition of satisfying scenario matching, a federated learning mechanism is used to complete policy transfer and model update.

[0134] This method constructs a complete closed loop for intelligent construction management. First, in S101, high-frequency 5G equipment is used to achieve comprehensive perception of the site status, while the ROS2 framework ensures heterogeneous data synchronization and maintains timestamp alignment. In S102, three types of modeling data are fused using a GCN neural network to establish a spatially consistent digital twin structure. S103 trains a meta-policy network based on RLlib and calls a DGNN deployed with TensorRT for real-time multi-objective task inference. In S104, the difference rate is verified by comparing LiDAR point clouds with the design model, triggering a blockchain contract to ensure the credibility of the execution status. S105 provides an XR-supported human-computer interaction interface, enabling workers to visually obtain task information and progress. Finally, in S106, a federated learning mechanism is used to achieve model migration and localization adjustments between multiple construction projects.

[0135] This method achieves closed-loop automated management from data acquisition, model construction, strategy reasoning, execution verification to collaborative interaction, improving the intelligence and controllability of the construction process. GCN integration enhances the quality of the digital twin model, DGNN strategy optimization achieves balanced response to multiple task objectives, blockchain mechanism enhances the transparency and trust foundation of the acceptance process, XR platform improves personnel operation efficiency and task collaboration, and federated learning mechanism strengthens the model's cross-project adaptability and protects data privacy.

[0136] In addition to 5G devices, data acquisition methods can also combine Wi-Fi 6 or LoRaWAN technologies to achieve low-power communication; during the modeling process, GCN can be replaced with dynamic graph neural networks or multi-scale network structures to enhance the fusion of models at different levels; the policy network training algorithm can also use algorithms such as DDPG and SAC to improve policy convergence in continuous control space; the blockchain platform can choose consortium chain, public chain or hybrid chain structure according to the project scale and the relationship between participants; the federated learning communication framework can adopt FLUTE, FedProx and other technologies to enhance robustness and heterogeneous collaboration capabilities.

[0137] In the following embodiments, a computer-readable storage medium is provided, having instructions stored thereon that, when executed by a processor, cause a computing device to perform all the steps of the above-described method, including:

[0138] Collect multimodal data from the construction site and encapsulate it heterogeneously using ROS2;

[0139] By integrating BIM, GIS, and TIN models using GCN, a digital twin map structure is constructed.

[0140] Multi-objective task inference based on meta-reinforcement learning and DGNN;

[0141] Trigger LiDAR scanning and comparison with the BIM model, and automatically execute the blockchain smart contract based on the difference rate;

[0142] Leveraging the XR platform for navigation, acceptance testing, and multi-user collaboration;

[0143] A federated learning mechanism is applied to achieve policy transfer and model update.

[0144] The computer-readable storage medium can be a physical or virtual carrier such as an SSD, eMMC, TF card, or cloud storage platform. The embedded software program includes components such as a multi-threaded control module, a graph neural network inference engine, a blockchain interface program, XR interface rendering logic, and a federated learning communication protocol stack. After the processor reads and executes the program, it can achieve end-to-end automated task processes within the urban reconstruction construction management system. Key modules include encapsulation of edge data streams, graph structure learning, real-time policy deployment, smart contract-based verification and payment linkage, and efficient task collaboration among multiple users.

[0145] This computer-readable storage medium integrates end-to-end management logic and intelligent algorithm models to provide standardized deployment capabilities and flexible application interfaces for intelligent urban construction systems. It is widely adaptable to various construction control platforms (such as edge gateways, mobile terminals, cloud servers, etc.), significantly reducing system integration difficulty and improving execution efficiency, ensuring the system possesses multiple application values ​​including intelligence, reliability, and low carbon footprint.

[0146] The program modules in this medium can be deployed across platforms using Docker or Kubernetes containers; the processor can be an embedded ARM chip, an x86 server, or an FPGA acceleration platform; the graph neural network module can support the ONNX format to be compatible with third-party AI engines; the blockchain interface can be compatible with Hyperledger Fabric, Ethereum, or other mainstream smart contract platforms; the program can also interface with third-party BIM / GIS platforms via API to achieve task linkage and platform integration applications.

[0147] Application Example 1:

[0148] In a typical implementation, the intelligent scheduling and coordination system described in this invention is applied to the reconstruction project of a multi-functional three-dimensional transportation hub in the old city area of ​​a provincial capital. This project includes simultaneous construction tasks across multiple sections, such as subway station expansion, elevated bridge structural reinforcement, public transport interchange integration, and intelligent reconstruction of the surrounding road system. The construction site is characterized by confined space, a tight schedule, and frequent multi-party collaboration, making it highly representative of the project's technical features.

[0149] In the initial construction phase, the system deployed an edge sensing layer network based on 5G communication modules, combined with LiDAR, thermal imaging equipment, environmental sensors, and vision units, to achieve multimodal real-time acquisition of the 3D structure of the work area, equipment operating status, personnel behavior trajectories, and on-site meteorological data. All sensing nodes were heterogeneously encapsulated based on ROS2 middleware, and a unified data synchronization standard was established through a timestamp mechanism. The edge computing nodes adopted the NVIDIA Jetson AGX Xavier platform, and containerized deployment included data filtering, target tracking, and point cloud preprocessing modules, ensuring that the sensing data completed preliminary calculations and was uploaded to the scheduling center within 30 milliseconds.

[0150] Next, the system enters the model construction phase. The construction party provides BIM 3D component models, the traffic management platform provides GIS traffic plot models, and the surveying unit provides TIN topographic grids. The system uses Graph Convolutional Neural Networks (GCNs) to extract the spatial location, semantic labels, and physical attributes of the three types of models as node features and constructs corresponding topological graphs. In the three-layer GCN processing structure, the system uses a softmax fusion algorithm to resolve model boundary conflicts and introduces a Transformer semantic attention mechanism to enhance semantic consistency. The system further dynamically generates cross-model edge structures based on Euclidean distance (<0.3m) and semantic cosine similarity (>0.85), and uses maximum mean difference (MMD) as the alignment loss to measure the global feature distribution of heterogeneous model fusion, ultimately constructing a digital twin map of the transportation hub with multi-scale structural consistency.

[0151] After model construction, the system models various tasks such as construction scheduling, equipment operation, carbon emission balance, and safety management as joint multi-objective Markov decision processes. The state tensor includes 23 dimensions such as real-time equipment utilization, operational risk level, path occupancy, energy intensity, and traffic impact index. The system uses the PPO reinforcement learning algorithm to train the meta-policy network on the RLlib platform and loads a historical construction case experience pool for pre-training and fine-tuning. The policy inference stage is executed by the Dynamic Graph Neural Network (DGNN) module deployed in edge computing nodes. This module introduces a graph attention mechanism to enhance the ability to model inter-edge dependencies. The inference process is compiled and deployed through TensorRT, achieving the generation of optimal task sequences and the push of control parameters within <200ms.

[0152] During the construction task execution phase, the system is configured with a blockchain collaborative trust layer based on the Hyperledger Fabric consortium blockchain. Each key construction node is bound to independent task contract logic, defining completion conditions, data source verification, and funding triggering logic. When a component is completed, the system calls upon the LiDAR installed on the device to perform a high-precision point cloud scan of the target area and compares it with the BIM design model using the ICP registration algorithm. If the matching point ratio is ≥95% and the maximum deviation is <2.5cm, an "Acceptance Certificate" is issued by the Oracle node. The system executes the chaincode according to the contract terms, automatically triggering fund disbursement and recording the task ID, contract hash, and GPS timestamp to ensure that the entire process data is traceable and verifiable.

[0153] At the human-machine collaboration level, the system integrates AR navigation and VR sandbox to construct an XR collaborative interactive environment. Workers wearing HoloLens 2 enter the work area, and the system achieves sub-centimeter-level spatial registration through AprilTag and SLAM mechanisms, automatically loading construction drawings, process guidelines, and safety warnings for the current component. Task information is presented in real-time in a graphic overlay format. After the operation is completed, task feedback can be triggered through natural language confirmation or gesture interaction. This feedback event is automatically stored on the blockchain and synchronized to the VR sandbox, allowing managers to remotely replay, inspect, and interactively adjust and audit the task based on errors, pace deviations, etc.

[0154] At the system security level, the system employs multiple security measures for both the policy model and on-chain contracts. During deployment, the policy network introduces a SHA256 model signature mechanism and model access control identifiers to ensure model version traceability and prevent unauthorized access. In addition to basic encryption, critical operations on on-chain data are recorded with node identity and two-factor timestamp-based tamper-proof logs. Furthermore, a fault-tolerant consensus mechanism is enabled across multiple nodes to ensure Fabric ledger consistency and chaincode transaction integrity.

[0155] Through the above integration, this system has realized a closed-loop technical process of "multi-source perception - model fusion - strategy decision-making - reliable execution - interactive collaboration" in the transportation hub reconstruction project, which significantly improves construction efficiency, scheduling response speed and multi-party collaboration transparency, and fully verifies the applicability, versatility and high reliability of the system in complex urban reconstruction projects.

[0156] Application Example 2:

[0157] During the construction of an underground integrated utility tunnel in a new urban area, the construction team introduced the system of this invention to perform real-time digital twin alignment modeling and construction conflict prediction for the access section. In the initial stage, the design unit provided a BIM utility tunnel model in IFC format and an urban underground GIS model in CityGML format to guide the excavation path and equipment layout.

[0158] The system first calls the GCN module to align the graph structures constructed by the two models. The BIM model contains 3467 nodes (representing air ducts, water pipes, cable trays, etc.), and the GIS model contains 2184 underground feature nodes. The feature vectors of the nodes have a dimension of 128 and include spatial coordinates (x, y, z), semantic type encoding (one-hot), and component physical attributes (such as inner diameter and material density). After initializing the weights of the graph convolutional network, it is trained and optimized using the following loss function:

[0159]

[0160] The MMD (Maximum Mean Difference) term measures the overall embedding distribution difference of the model, with γ set to 0.3. The kernel function is the RBF kernel, and the bandwidth parameter σ is optimized through grid search between 0.1 and 1.0, ultimately set to σ = 0.45.

[0161] After GCN training converged to a validation error of less than 0.05, the system automatically identified a 42mm center offset between a duct (component ID: W-245) and an underground rainwater pipeline node in a certain area. The system adjusted the GCN weights based on the geometric accuracy priority principle (weight 0.6), recommended duct adjustment paths, and provided real-time conflict highlighting via AR.

[0162] Ultimately, the adjusted path offset was controlled within ±10mm, avoiding the risk of on-site rework; the operation record was uploaded to the blockchain via an Oracle node, binding the component ID, change time, and responsible person information to ensure data closure and reliability.

[0163] Application Example 3:

[0164] During the demolition and reconstruction of a viaduct, the system deployed edge sensing nodes and 5G edge boxes at the construction site to collect multi-source data in real time, including rainfall, wind speed, equipment status, and operational emissions information. At 15:24 on a certain day, the system detected that the local rainfall intensity reached 14.3 mm / h and the wind speed increased to 10.1 m / s, approaching the preset risk threshold.

[0165] The system's spatiotemporal brain module immediately initiates a dynamic scheduling inference process, constructing a Markov decision state tensor St, which includes the following indices:

[0166] Crane utilization rate: 0.87

[0167] Vehicle traffic density: 0.52

[0168] PM2.5 concentration: 78 μg / m³ 3

[0169] Current rainfall: 14.3 mm / h

[0170] Carbon emission weighting factor: 0.5 (default)

[0171] At this point, the DGNN graph structure has 1327 nodes and 2716 edges. The system uses the following weight update formula to calculate the scheduling priority weights between tasks:

[0172]

[0173] The coupling factor is calculated as follows:

[0174]

[0175] Where α = 0.01, β = 0.05, and the attenuation coefficient λ is automatically adjusted to 0.8 based on the high-risk events detected by the system (increased from the default value of 0.5) to enhance time sensitivity.

[0176] Calculations show that the material transport mission "MT-038" has an emission intensity of 1.13 kg CO2 / km, but possesses the highest repositionability. The system completes strategy reconfiguration within 47 ms and switches the mission from the original path T3 to the backup path T5, reducing the overall carbon emissions to 0.79 kg CO2 / km.

[0177] The relevant scheduling adjustment results are recorded through the Fabric chain smart contract and AR navigation update instructions are sent to the front-line worker terminals. The entire response process takes less than 5 seconds.

[0178] Application Example 4:

[0179] In a municipal underground space expansion project, due to the high degree of enclosure of the construction work surface and the many risk factors, the project team used the XR collaborative interaction layer of this invention for task distribution, construction navigation and real-time feedback.

[0180] Workers wearing HoloLens 2 smart glasses enter the work area. The system performs initial positioning using AprilTag QR codes placed on the wall and calls the SLAM module for spatial map correction, with registration accuracy controlled within ±1.8cm.

[0181] In the XR collaborative interaction system of this invention, construction tasks are graphically distributed using components as the smallest operational unit. Taking the pipe gallery node "JX-W322" as an example, the system retrieves the standard construction process and auxiliary information corresponding to the node from the database according to the task number, and pushes it to the terminal of the front-line workers in a visual form.

[0182] During the operation at this node, the construction process includes, in sequence: drilling, inserting sleeves, structural fixing, and backfilling. The required construction tools include an impact drill, an electric hammer, and a support sleeve assembly. The impact drill is used for precise drilling on the concrete surface, the electric hammer assists in striking the foundation structure, and the support sleeve assembly is used for pipe positioning and stability assurance.

[0183] Given that a 10kV high-voltage cable is buried near the work area, in order to ensure work safety and personnel protection, the system provides safety warning information simultaneously, explicitly prohibiting the use of non-insulated power tools for any contact operations, and all construction operations must comply with the power facility protection specifications.

[0184] The aforementioned task information is presented to the workers in real time through AR overlay, including the work sequence, tool matching, and safety risk warnings. During operation, workers trigger task feedback through a dual confirmation method of gesture recognition and voice commands. The system then generates a construction record and synchronizes it to the task management and blockchain recording modules, achieving closed-loop management of task execution and data traceability.

[0185] During execution, the system calls the following data structure to bind on-site operations and task blockchain entries:

[0186] protobuf

[0187] message TaskExecution{

[0188] string user_id = 1;

[0189] string task_id = 2;

[0190] string component_id = 3;

[0191] float completion_rate = 4;

[0192] repeated string media_evidence=5;

[0193] string hash_id = 6;

[0194] int64 timestamp = 7;

[0195] }

[0196] in:

[0197] ˋcompletion_rate=1.0ˋ(100%);

[0198] The media evidence includes on-site photos and screenshots of infrared temperature measurements.

[0199] The hash_id is calculated using SHA256 and bound to the Fabric ledger.

[0200] ˋtimestamp=2025-05-1209:31:52ˋ.

[0201] Meanwhile, the VR sand table updates the status of the twin simultaneously, allowing project commanders to remotely replay and interactively inspect the work process in real time, greatly improving collaboration efficiency and construction visualization transparency.

[0202] Application Example 5:

[0203] In a certain affordable housing renovation project, in order to improve the interpretability of the construction scheduling system and the transparency of the project decision-making process, the construction party activated the causal analysis module in the system of this invention based on the "spatiotemporal brain module" to analyze and explain the internal behavior of the AI ​​strategy based on the SHAP (SHapley Additive exPlanations) value.

[0204] During the 42nd construction cycle, the system needs to optimize the sorting of the following three scheduling subtasks:

[0205] 1. Template setup (Task ID: TP-201);

[0206] 2. Concrete pouring (TP-202);

[0207] 3. Material transportation (TP-203).

[0208] The current state tensor S of the scheduling model t Includes the following key dimensions:

[0209] Concrete temperature: 22.5℃;

[0210] Concrete age: 8 hours;

[0211] Construction worker density: 0.68;

[0212] On-site light intensity: 4300 lx;

[0213] CO2 emission threshold limit: 40 kg / h.

[0214] Based on the deployed Dynamic Graph Neural Network (DGNN) inference model, the AI ​​system comprehensively evaluates the current construction status and outputs the optimal scheduling execution order as follows: first, execute the material transportation task (TP-203), followed by the formwork erection task (TP-201) and the concrete pouring task (TP-202).

[0215] To perform causal attribution analysis, as shown in Table 2, the system generates a SHAP explanation matrix for the decision sequence, and outputs the following results:

[0216] Table 2: Partial Results of the System's Output of the SHAP Explanation Matrix for the Decision Sequence

[0217] Feature Term SHAP value Contribution direction Age (8h) +0.42 Accelerate TP-202 <![CDATA[CO2 threshold limit]]> -0.61 Postponing TP-203 Population density 0.68 +0.25 Accelerate TP-201

[0218] After visualizing the distribution of all SHAP values, the system automatically constructed a causal graph as shown below:

[0219] Node: Task decision result;

[0220] Edge: SHAP explains causal edges;

[0221] Edge weight: w ij =|SHAP i -SHAP j |

[0222] Visual layout: Automatic layout using the Fruchterman-Reingold algorithm.

[0223] The graph results are submitted as a structured JSON record to the project monitoring interface, allowing monitors to view the causal basis of the system in specific scheduling decisions and assess whether they are reasonable or if there are any abnormal biases.

[0224] If scheduling anomalies or unrealistic strategy weights are detected, the system also allows regulators to manually adjust the influencing factor weights, triggering the strategy retraining process in real time and recording the changes in the regulatory channel ledger.

[0225] This invention constructs an intelligent scheduling and coordination system for the entire lifecycle of urban redevelopment. It achieves millisecond-level data acquisition (≤200ms latency) through a 5G edge sensing layer, constructs a dynamic digital twin based on GCN's multi-model topology alignment technology (BIM-GIS-TIN fusion error <2.5cm), and uses a meta-reinforcement learning framework (PPO algorithm) and a spatiotemporally coupled DGNN (λ = 0.5-0.8 dynamically adjusted) to achieve multi-objective joint optimization decision-making. The system innovatively combines LiDAR difference rate verification (threshold 95%) with the Hyperledger Fabric smart contract automatic triggering mechanism, along with SLAM+AprilTag XR collaborative interaction (positioning accuracy 1.8cm), forming a closed-loop technology system of "perception-modeling-decision-verification-interaction". Verified in actual engineering projects, this system improves construction scheduling response speed by 40%, reduces carbon emissions by 22%, and shortens the acceptance and payment cycle from the traditional 3-5 days to within 10 minutes, significantly solving industry pain points such as heterogeneous model alignment error (>5cm), emergency response delay (>30s), and inefficient multi-party collaboration.

[0226] It should be understood that the various technical solutions disclosed in this invention are not limited to the specific forms listed in the specification and embodiments. Without departing from the core concept of this invention, those skilled in the art can make various equivalent transformations and substitutions to the structure, algorithm, module combination or parameter selection, all of which should fall within the protection scope of this invention.

Claims

1. An intelligent scheduling and coordination system for the entire lifecycle of urban redevelopment, characterized in that, include: The edge perception layer, including a cluster of 5G IoT devices, is used to collect 3D point clouds, environmental data and personnel trajectories at the construction site, and achieves multi-source heterogeneous data alignment through ROS2. The digital twin weaving module is used to perform topological alignment of BIM, GIS, and TIN models based on graph convolutional neural networks (GCN) to construct dynamic multi-scale digital twins. The spatiotemporal brain module is based on a meta-reinforcement learning framework to realize multi-task decision-making for construction scheduling, traffic management, and carbon emission control. The meta-policy network constructed by the spatiotemporal brain module is based on the RLlib training framework, with a state space dimension of 23 and a training algorithm of PPO. During inference, a dynamic graph neural network (DGNN) deployed by TensorRT is called, and the inference response time does not exceed 200 milliseconds. The blockchain collaborative trust layer, based on the Hyperledger Fabric architecture, verifies the construction completion status and triggers smart contract execution through Oracle nodes; The XR collaborative interaction layer provides AR navigation and VR sandbox functions, supporting immersive collaboration and visual acceptance.

2. The system according to claim 1, characterized in that, The digital twin web module uses a graph convolutional neural network (GCN) whose node features include spatial coordinates, semantic labels, and physical attributes. The GCN adopts a three-layer network structure and uses a weighted fusion algorithm at the model boundary to handle topological conflicts.

3. The system according to claim 1, characterized in that, In the blockchain collaborative trust layer, the Oracle node compares the difference rate between the LiDAR scan point cloud and the design model. When the difference is less than 5%, a trusted proof is generated, and the chaincode automatically executes the payment operation and records the hash value and GPS timestamp.

4. The system according to claim 1, characterized in that, The AR navigation function of the XR collaborative interaction layer includes: AprilTag combined with SLAM to achieve spatial registration, construction information is superimposed on the real scene in the form of pictures and text, and workers upload progress information through gesture operation and trigger blockchain record.

5. The system according to any one of claims 1 to 4, characterized in that, Also includes: The carbon emission optimization module is used to establish the carbon emission mapping relationship of construction equipment and realize dynamic planning of carbon emission paths; The causal analysis module uses SHAP values ​​to analyze the impact of state variables in multi-objective decision-making. The federated upgrade module employs a differential privacy mechanism to enable cross-project policy migration and model localization optimization.

6. The system according to claim 1, characterized in that, The edge weight update of the Dynamic Graph Neural Network (DGNN) satisfies the following formula: ； Among them, W t Let L be the edge weight matrix at time t, and L be the joint loss function. For learning rate, Here, Φ(St,Tt) represents the weight parameters of the coupling function, and Φ(St,Tt) represents the spatiotemporal coupling factor, calculated as follows: ; Among them, S t Let the current state tensor be... To predict the state, T t λ is the timestamp, and λ is the time decay coefficient, with a default value of 0.5, which is adjusted to 0.8 when a high-risk event is detected.

7. The system according to claim 2, characterized in that, The loss function for GCN topology alignment is: ； Among them, f θ is the feature extraction function, MMD is the maximum mean difference, γ is the balance parameter, and the bandwidth σ of the RBF kernel function is determined in the interval [0.1, 1.0] through grid search.

Images