Roadbed filling and digging operation intelligent scheduling method and system based on artificial intelligence assistance

Abstract

The invention discloses a roadbed filling and digging operation intelligent scheduling method and system based on artificial intelligence assistance, and the method comprises the steps: collecting water content distribution data in soil through a multispectral imager, obtaining laser point cloud data through a laser sensor, and fusing the water content distribution data and the laser point cloud data to generate a high-dimensional geologic feature tensor; dynamically weighting contribution degrees of different sensor data in the roadbed real-time state data by utilizing an attention mechanism; establishing a hybrid model, and inputting the cross-modal roadbed data into the hybrid model; and based on the geological stability thermodynamic diagram, utilizing an MOEA / D multi-objective evolutionary algorithm to initialize a population, utilizing a GWO grey wolf optimization algorithm to perform local search, and setting a dynamically adjusted convergence factor to obtain a roadbed filling and digging scheduling scheme. A scheduling scheme can be adjusted in time according to actual operation conditions, equipment is ensured to be always in an efficient collaborative operation state, and the construction efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of engineering construction technology, and in particular to an artificial intelligence-assisted intelligent scheduling method and system for roadbed excavation and filling operations. Background Art

[0002] In the field of road construction, traditional methods for scheduling roadbed excavation and filling operations, which rely heavily on manual experience and simple mechanical equipment, present numerous drawbacks. In the data collection and analysis stages, traditional methods struggle to comprehensively and accurately obtain roadbed geological information. Soil moisture testing often relies on local sampling, which fails to provide a comprehensive picture of the moisture distribution across the entire roadbed area. Topographic data collection is also limited to single-point measurements using equipment such as total stations, resulting in delayed data updates and difficulty reflecting real-time changes during construction. The scheduling of operations between equipment often relies on manual coordination and cannot be dynamically adjusted based on geological conditions and construction progress, resulting in low equipment utilization and inefficient construction. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an intelligent scheduling method for roadbed excavation and filling operations based on artificial intelligence assistance.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is as follows: further, in the above-mentioned intelligent scheduling method for roadbed excavation and filling operations based on artificial intelligence assistance, the intelligent scheduling method for roadbed excavation and filling operations includes the following steps: The soil moisture distribution data is collected by a multispectral imager, and the laser point cloud data is obtained by a laser sensor. The moisture distribution data and the laser point cloud data are fused to generate a high-dimensional geological feature tensor to obtain the real-time status data of the roadbed; Using an attention mechanism to dynamically weight the contribution of different sensor data in the real-time roadbed status data to obtain cross-modal roadbed data; Establishing a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model to generate a geological stability thermodynamic map; Based on the geological stability heat map, the MOEA / D multi-objective evolutionary algorithm is used to initialize the population, the GWO grey wolf optimization algorithm is used for local search, and a dynamically adjusted convergence factor is set to obtain the roadbed excavation and filling scheduling scheme.

[0005] Furthermore, in the above-mentioned AI-assisted intelligent scheduling method for roadbed excavation and filling operations, the method of collecting soil moisture distribution data by a multispectral imager and obtaining laser point cloud data by a laser sensor includes: Establish a calibration model between soil moisture content and spectral reflectance, and collect soil samples with different moisture contents in the construction area to obtain spectral data and actual moisture content; The image data collected by the multispectral imager is input into the inversion model to calculate the soil moisture content corresponding to each pixel point and generate the moisture content distribution data of the entire roadbed area; By measuring the time difference from laser emission to reception, the distance from the target object to the sensor is calculated. Combined with the scanning angle and rotation speed of the sensor, the three-dimensional coordinates of the target object surface are obtained to obtain laser point cloud data.

[0006] Furthermore, in the above-mentioned intelligent scheduling method for roadbed excavation and filling operations based on artificial intelligence assistance, the fusion of the moisture content distribution data and the laser point cloud data to generate a high-dimensional geological feature tensor to obtain real-time roadbed status data includes: Converting the moisture content distribution data and the laser point cloud data into a unified global coordinate; Extract moisture characteristics from moisture distribution data, including at least the mean, variance, and gradient of moisture content; extract terrain and landform characteristics from point cloud data, including at least elevation, slope, aspect, and curvature; The moisture characteristics, terrain characteristics and landform characteristics are dimensionally expanded to form multi-dimensional feature vectors, and the feature vectors are fused into high-dimensional geological feature tensor data using tensor product operations to obtain real-time roadbed status data.

[0007] Furthermore, in the above-mentioned AI-assisted intelligent scheduling method for roadbed excavation and filling operations, the method of dynamically weighting the contribution of different sensor data in the roadbed real-time status data using an attention mechanism to obtain cross-modal roadbed data includes: For each header in the roadbed real-time status data, the similarity between the query and the key is calculated, and a matrix is calculated using a dot product similarity formula to obtain a similarity matrix; Performing Softmax normalization on the similarity matrix to obtain an attention weight matrix, where the weight value range is between 0 and 1; The Value matrix is weighted and summed according to the attention weight to obtain the cross-modal feature vector of each spatiotemporal grid point.

[0008] Furthermore, in the above-mentioned AI-assisted intelligent scheduling method for roadbed excavation and filling operations, the establishment of a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model, and generating a geological stability thermodynamic map include: Establishing a Transformer-GNN hybrid model, and inputting the cross-modal roadbed data into the Transformer-GNN hybrid model; The Transformer encoder is used to segment the cross-modal roadbed data into time series, position-encode the data at each time step, and incorporate time information into the model. The position encoding is generated using sine and cosine functions. In the multi-head self-attention layer, self-attention calculation is performed on the cross-modal data of each time step to capture the dependency between different time steps. The FFN layer is used to perform nonlinear transformation on the attention output to extract temporal features. The construction equipment is regarded as nodes in the graph. The node attributes include the type, location, and working status of the equipment. The GCN graph convolutional network is used to perform graph convolution operations to obtain the equipment node features.

[0009] Furthermore, in the above-mentioned AI-assisted intelligent scheduling method for roadbed excavation and filling operations, the establishment of a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model, and generating a geological stability thermodynamic map further include: The temporal features output by the Transformer encoder and the device node features output by the GNN are fused and mapped through a fully connected neural network to obtain geological stability assessment indicators, including at least soil bearing capacity, settlement risk and sliding possibility.

[0010] Furthermore, in the above-mentioned AI-assisted intelligent scheduling method for roadbed excavation and filling operations, the method uses the geological stability thermodynamic map, initializes the population using the MOEA / D multi-objective evolutionary algorithm, performs local search using the GWO grey wolf optimization algorithm, and sets a dynamically adjusted convergence factor to obtain a roadbed excavation and filling scheduling plan, including: Establish a multi-objective optimization model with the objectives of maximizing cut-and-fill efficiency, minimizing costs, and optimizing quality; The roadbed excavation and filling scheduling scheme is represented as a chromosome. The chromosome encoding includes the allocation of construction equipment, operation path planning and allocation of excavation and filling volume. The population is initialized by random generation. The improved Grey Wolf Optimization Algorithm (GWO) is used to search the neighborhood of each individual. Neighborhood operations include equipment reallocation, operation path adjustment, and fill-cut volume fine-tuning. Each operation generates a neighborhood individual. The objective function value of the neighborhood individuals is calculated, the optimal individual is selected to replace the original individual, and the roadbed excavation and filling scheduling scheme is obtained through dynamic adjustment of the convergence factor through iteration.

[0011] Furthermore, in an artificial intelligence-assisted intelligent scheduling system for roadbed excavation and filling operations, the intelligent scheduling system for roadbed excavation and filling operations includes the following modules: A roadbed data acquisition module is used to collect soil moisture distribution data using a multispectral imager and obtain laser point cloud data using a laser sensor. The moisture distribution data and the laser point cloud data are then fused to generate a high-dimensional geological feature tensor to obtain real-time roadbed status data. a contribution calculation module for dynamically weighting the contribution of different sensor data in the roadbed real-time status data using an attention mechanism to obtain cross-modal roadbed data; A stability prediction module is used to establish a Transformer-GNN hybrid model, input the cross-modal roadbed data into the Transformer-GNN hybrid model, and generate a geological stability thermodynamic map; The scheduling scheme generation module is used to initialize the population based on the geological stability heat map using the MOEA / D multi-objective evolutionary algorithm, perform local search using the GWO grey wolf optimization algorithm, set a dynamically adjusted convergence factor, and obtain the roadbed excavation and filling scheduling scheme.

[0012] Furthermore, in a system for implementing the above-mentioned method for intelligent scheduling of roadbed excavation and filling operations assisted by artificial intelligence, the roadbed data acquisition module includes the following submodules: A submodule is established to establish a calibration model between soil moisture content and spectral reflectance. Soil samples with different moisture contents are collected in the construction area to obtain spectral data and actual moisture content. The calculation submodule is used to input the image data collected by the multispectral imager into the inversion model, calculate the soil moisture content corresponding to each pixel point, and generate the moisture content distribution data of the entire roadbed area; The acquisition submodule is used to calculate the distance from the target object to the sensor by measuring the time difference from laser emission to reception, and obtain the three-dimensional coordinates of the target object surface by combining the scanning angle and rotation speed of the sensor to obtain laser point cloud data.

[0013] Furthermore, in the system for implementing the above-mentioned method for intelligent scheduling of roadbed excavation and filling operations assisted by artificial intelligence, the roadbed data acquisition module further includes the following submodules: A conversion submodule, used to convert the moisture content distribution data and laser point cloud data into a unified global coordinate; The extraction submodule is used to extract moisture characteristics from the moisture distribution data, including at least the mean, variance, and gradient of the moisture content, and to extract terrain characteristics and landform characteristics from the point cloud data, including at least elevation, slope, aspect, and curvature; The fusion submodule is used to expand the dimensions of the moisture characteristics, terrain characteristics and landform characteristics to form a multi-dimensional feature vector, and use tensor product operation to fuse the feature vector into high-dimensional geological feature tensor data to obtain real-time roadbed status data.

[0014] Its beneficial effects are as follows: the moisture distribution data in the soil is collected by a multispectral imager, and the laser point cloud data is obtained by a laser sensor, and the moisture distribution data and the laser point cloud data are fused to generate a high-dimensional geological feature tensor to obtain the real-time status data of the roadbed; the contribution of different sensor data in the real-time status data of the roadbed is dynamically weighted using the attention mechanism to obtain cross-modal roadbed data; a Transformer-GNN hybrid model is established, and the cross-modal roadbed data is input into the Transformer-GNN hybrid model to generate a geological stability heat map; based on the geological stability heat map, the population is initialized using the MOEA / D multi-objective evolutionary algorithm, and a local search is performed using the GWO grey wolf optimization algorithm. A dynamically adjusted convergence factor is set to obtain a roadbed excavation and filling scheduling plan. 1. The geological stability heat map enables construction personnel to intuitively and clearly grasp the stability of each area of the roadbed, providing a visual basis for scientific decision-making. 2. It can quickly find Pareto optimal solutions that simultaneously meet multiple objectives, such as maximizing operational efficiency, minimizing costs, and optimizing quality, within a complex scheduling solution space. Compared to traditional single-objective optimization methods, this significantly improves the overall performance of scheduling solutions. 3. It can promptly adjust scheduling solutions based on actual operational conditions, ensuring that equipment is always in an efficient and collaborative state, effectively improving construction efficiency and reducing construction costs. Furthermore, by continuously optimizing construction process parameters, it ensures that the quality of roadbed excavation and filling operations remains consistently up to standard, laying the foundation for the long-term safe use of road projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0016] Figure 1 Schematic diagram of a first embodiment of an artificial intelligence-assisted intelligent scheduling method for roadbed excavation and filling operations according to an embodiment of the present invention; Figure 2 Schematic diagram of a second embodiment of an artificial intelligence-assisted intelligent scheduling method for roadbed excavation and filling operations according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a first embodiment of an artificial intelligence-assisted intelligent scheduling system for roadbed excavation and filling operations in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0019] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, an artificial intelligence-assisted intelligent scheduling method for roadbed excavation and filling operations includes the following steps: Step 101: Collect soil moisture distribution data using a multispectral imager, obtain laser point cloud data using a laser sensor, fuse the moisture distribution data and the laser point cloud data to generate a high-dimensional geological feature tensor, and obtain real-time roadbed status data; Specifically, in this embodiment, a calibration model between soil moisture content and spectral reflectance is established, and spectral data and actual moisture content are obtained by collecting soil samples with different moisture contents in the construction area; The image data collected by the multispectral imager is input into the inversion model to calculate the soil moisture content corresponding to each pixel point and generate the moisture content distribution data of the entire roadbed area; By measuring the time difference from laser emission to reception, the distance from the target object to the sensor is calculated. Combined with the scanning angle and rotation speed of the sensor, the three-dimensional coordinates of the target object surface are obtained to obtain laser point cloud data.

[0020] Convert moisture content distribution data and laser point cloud data into a unified global coordinate system; Extract moisture characteristics from moisture distribution data, including at least the mean, variance, and gradient of moisture content; extract terrain and landform characteristics from point cloud data, including at least elevation, slope, aspect, and curvature; The moisture characteristics, terrain characteristics and landform characteristics are expanded in dimension to form multi-dimensional feature vectors, which are then fused into high-dimensional geological feature tensor data using tensor product operations to obtain real-time roadbed status data.

[0021] Step 102: Using an attention mechanism to dynamically weight the contribution of different sensor data in the real-time roadbed status data to obtain cross-modal roadbed data; Specifically, in this embodiment, for each header in the roadbed real-time status data, the similarity between the Query and the Key is calculated, and the matrix is calculated using the dot product similarity formula to obtain a similarity matrix; Perform Softmax normalization on the similarity matrix to obtain the attention weight matrix, with the weight value range between 0 and 1; The Value matrix is weighted and summed according to the attention weight to obtain the cross-modal feature vector of each spatiotemporal grid point.

[0022] Step 103: Establish a Transformer-GNN hybrid model, input the cross-modal roadbed data into the Transformer-GNN hybrid model, and generate a geological stability heat map; Specifically, in this embodiment, a Transformer-GNN hybrid model is established, and the cross-modal roadbed data is input into the Transformer-GNN hybrid model; The Transformer encoder is used to segment the cross-modal roadbed data into time series, position-encode the data at each time step, and incorporate time information into the model. The position encoding is generated using sine and cosine functions. In the multi-head self-attention layer, self-attention calculation is performed on the cross-modal data of each time step to capture the dependency between different time steps. The FFN layer is used to perform nonlinear transformation on the attention output to extract temporal features. The construction equipment is regarded as nodes in the graph. The node attributes include the type, location, and working status of the equipment. The GCN graph convolutional network is used to perform graph convolution operations to obtain the equipment node features.

[0023] The temporal features output by the Transformer encoder and the device node features output by the GNN are fused and mapped through a fully connected neural network to obtain geological stability assessment indicators, including at least soil bearing capacity, settlement risk and sliding possibility.

[0024] Step 104: Based on the geological stability heat map, the population is initialized using the MOEA / D multi-objective evolutionary algorithm, a local search is performed using the GWO grey wolf optimization algorithm, a dynamically adjusted convergence factor is set, and a roadbed excavation and filling scheduling plan is obtained.

[0025] Specifically, in this embodiment, a multi-objective optimization model is established, with the objectives including maximizing the efficiency of cut-and-fill operations, minimizing costs, and optimizing quality; The roadbed excavation and filling scheduling scheme is represented as a chromosome. The chromosome encoding includes the allocation of construction equipment, operation path planning and allocation of excavation and filling volume. The population is initialized by random generation. The improved Grey Wolf Optimization Algorithm (GWO) is used to search the neighborhood of each individual. Neighborhood operations include equipment reallocation, operation path adjustment, and fill-cut volume fine-tuning. Each operation generates a neighborhood individual. Calculate the objective function value of the neighborhood individual, select the optimal individual to replace the original individual, and dynamically adjust the convergence factor to obtain the roadbed excavation and filling scheduling plan through iteration.

[0026] Its beneficial effects are: 1. It generates a geological stability heat map, allowing construction personnel to intuitively and clearly grasp the stability of each area of the roadbed, providing a visual basis for scientific decision-making. 2. It can quickly find the Pareto optimal solution that simultaneously meets multiple objectives such as maximizing operating efficiency, minimizing costs, and optimizing quality in a complex scheduling plan space. Compared with traditional single-objective optimization methods, it significantly improves the overall performance of the scheduling plan. 3. It can adjust the scheduling plan in a timely manner according to actual operating conditions to ensure that equipment is always in an efficient and collaborative working state, effectively improving construction efficiency and reducing construction costs. At the same time, by continuously optimizing construction process parameters, the quality of roadbed excavation and filling operations is guaranteed to be stable and meet standards, laying the foundation for the long-term safe use of road projects.

[0027] See also Figure 2 In an AI-assisted intelligent scheduling method for roadbed excavation and filling operations, soil moisture distribution data is collected using a multispectral imager, and laser point cloud data is obtained using a laser sensor. The following steps are included: Step 201: Establish a calibration model between soil moisture content and spectral reflectance, and obtain spectral data and actual moisture content by collecting soil samples with different moisture contents in the construction area; Step 202: Input the image data collected by the multispectral imager into the inversion model, calculate the soil moisture content corresponding to each pixel point, and generate the moisture content distribution data of the entire roadbed area; Step 203: Calculate the distance from the target object to the sensor by measuring the time difference between laser emission and reception, and obtain the three-dimensional coordinates of the target object surface by combining the scanning angle and rotation speed of the sensor to obtain laser point cloud data.

[0028] See also Figure 3 In an artificial intelligence-assisted intelligent scheduling system for roadbed excavation and filling operations, the intelligent scheduling system for roadbed excavation and filling operations includes the following modules: The roadbed data acquisition module is used to collect soil moisture distribution data using a multispectral imager and obtain laser point cloud data using a laser sensor. The moisture distribution data and laser point cloud data are then fused to generate a high-dimensional geological feature tensor to obtain real-time roadbed status data. The contribution calculation module is used to dynamically weight the contribution of different sensor data in the real-time roadbed status data using the attention mechanism to obtain cross-modal roadbed data; The stability prediction module is used to establish a Transformer-GNN hybrid model, input cross-modal roadbed data into the Transformer-GNN hybrid model, and generate a geological stability heat map; The scheduling scheme generation module is used to initialize the population based on the geological stability heat map using the MOEA / D multi-objective evolutionary algorithm, perform local search using the GWO grey wolf optimization algorithm, set a dynamically adjusted convergence factor, and obtain the roadbed excavation and filling scheduling scheme.

[0029] Specifically, this embodiment can also be implemented in the following ways:

[0030] (1) Multispectral imager data acquisition Moisture distribution data extraction: Spectral inversion algorithms, such as partial least squares (PLS) or neural network algorithms, are used to process image data from each band. First, a calibration model is established between soil moisture and spectral reflectance. Soil samples with varying moisture contents are collected within the construction area, and their spectral data and actual moisture contents are obtained. This is then used to train a high-precision inversion model. Image data collected by the multispectral imager is then fed into the inversion model, which calculates the soil moisture content for each pixel. This generates moisture distribution data for the entire roadbed area, with a resolution of 0.5-1 meter.

[0031] (2) Laser sensor data acquisition Point cloud data acquisition: The laser sensor emits laser pulses and measures the time difference (TOF) or phase difference between emission and reception to calculate the distance from the target object to the sensor. Simultaneously, the sensor's scanning angle and rotation speed are combined to obtain the three-dimensional coordinates (X, Y, Z) of the target object's surface, generating laser point cloud data. The scanning frequency is adjusted based on the operating speed and accuracy requirements of the construction equipment, generally ranging from 50-200Hz, to ensure that the point cloud data reflects real-time topographic changes on the roadbed surface.

[0032] (3) Data fusion to generate high-dimensional geological characteristic tensor data Spatiotemporal registration: First, the moisture distribution data acquired by the multispectral imager and the point cloud data acquired by the laser sensor are converted to a unified global coordinate system. For sensors on drones and construction equipment, the Global Positioning System (GPS) and Inertial Measurement Unit (IMU) are used to obtain their position and attitude information for coordinate conversion and attitude correction. For fixed sensors, their position and orientation in the global coordinate system are determined through manual measurement or equipment such as total stations.

[0033] Feature extraction: Extract moisture-related features from moisture distribution data, such as mean, variance, and gradient, to reflect the spatial distribution and changing trends of soil moisture. Extract terrain features from point cloud data, such as elevation, slope, aspect, and curvature, as well as geomorphic features such as surface roughness, height of protrusions, and depth of depressions. Furthermore, point cloud intensity features can be extracted to reflect differences in surface materials.

[0034] Tensor fusion: The extracted moisture content and point cloud features are dimensionally expanded to form multidimensional feature vectors. These feature vectors are then fused into high-dimensional geological feature tensor data using a tensor product operation. The tensor dimensions include spatial dimensions (X, Y), feature dimensions (such as moisture content, elevation, and slope), and temporal dimensions (different acquisition times). This constructs a real-time roadbed status data tensor containing spatiotemporal information and multi-source features. The dimensions are N × M × K × T, where N and M are the number of spatial grid cells, K is the number of features, and T is the length of the time series.

[0035] (1) Construction of attention mechanism model A multi-head self-attention model is employed. Its structure consists of three input matrices: query, key, and value. High-dimensional geological feature tensor data is first flattened into a two-dimensional matrix of (N×M×T)×K dimensions, where N×M×T is the number of spatiotemporal grids and K is the number of features. The input matrix is then mapped to the query, key, and value matrices through linear transformations. The dimensions of each matrix are K×d_k, where d_k is the dimension of each head, typically set to 64 or 128.

[0036] (2) Dynamic weighted calculation of contribution Similarity Calculation: For each header, calculate the similarity between the query and the key using the dot product similarity formula: Similarity = (Query・Key^T) / sqrt(d_k). The similarity matrix has dimensions (N×M×T)×(N×M×T), where each element represents the degree of correlation between two spatiotemporal grid points.

[0037] Attention weight generation: Softmax normalization is performed on the similarity matrix to generate an attention weight matrix. The weight values range from 0 to 1, indicating the contribution of each spatiotemporal grid point to the current grid point. For multimodal data, namely moisture content features and point cloud features, attention weights are calculated separately and then fused through weighted averaging or concatenation to obtain cross-modal attention weights.

[0038] Cross-modal data generation: The Value matrix is weighted and summed according to the attention weights to obtain a cross-modal feature vector for each spatiotemporal grid point. The outputs of all heads are concatenated and restored to their original feature dimensions through a linear transformation to generate the final cross-modal roadbed data. This data dynamically integrates important information from different sensor data, highlighting features that have a significant impact on the current roadbed state.

[0039] (1) Model architecture design The Transformer-GNN hybrid model consists of two parts: a Transformer encoder and a graph neural network (GNN). The Transformer encoder processes time series data to capture how roadbed conditions change over time; the GNN models the spatial topological relationships between construction equipment and analyzes their interactions.

[0040] (2) Transformer encoder processes time series data Input processing: The cross-modal roadbed data is segmented into time series. The input data dimensions for each time step are N×M×K, where N×M is the number of spatial grids and K is the number of features. The data for each time step is position-encoded to incorporate temporal information into the model. Position encoding is generated using sine and cosine functions. The formula is: PosEncoding(pos,2i)=sin(pos / 10000^(2i / K)),PosEncoding(pos,2i+1)=cos(pos / 10000^(2i / K)), where pos is the time step position and i is the feature dimension index.

[0041] Encoding Process: The Transformer encoder consists of multiple encoding layers, each of which incorporates a multi-head self-attention mechanism and a feedforward neural network (FFN), employing residual connections and layer normalization. In the multi-head self-attention layer, self-attention calculations are performed on cross-modal data at each time step, capturing dependencies between different time steps. The FFN layer performs nonlinear transformations on the attention output to further extract temporal features. Through multi-layer encoding, a high-level feature representation that incorporates temporal information is obtained.

[0042] (3) GNN modeling of the spatial topological relationship of construction equipment Graph Construction: Construction equipment is treated as nodes in a graph. Node attributes include equipment type, location, and operating status (load, speed, fuel consumption, etc.). When the spatial distance between equipment falls below a set threshold (50 meters), an edge connection is established. Edge attributes include distance, communication latency, etc. The graph structure is dynamically updated as equipment moves and its operating status changes.

[0043] Graph convolution operations: Graph convolution networks (GCNs) or graph attention networks (GATs) are used for graph convolution operations. Taking GATs as an example, each node calculates attention weights with its neighbors, aggregates information from neighboring nodes, and updates its own node features. The attention weight calculation considers both node and edge attributes, using the formula: α_ij = Softmax(LeakyReLU(a^T[Wh_i||Wh_j])), where a is the attention parameter, W is the weight matrix, and h_i and h_j are the feature vectors of nodes i and j. Through multi-layer graph convolution, the spatial topological relationships and collaborative operation patterns between devices are captured.

[0044] (IV) Generating a geological stability thermodynamic map The time series features output by the Transformer encoder are fused with the device features output by the GNN and mapped to geological stability assessment indicators such as soil bearing capacity, settlement risk, and slip potential through a fully connected neural network. The stability indicators for each spatial grid point are normalized and converted to grayscale or color values ranging from 0 to 255 to generate a geological stability heat map. The heat map uses different color codes to indicate stability levels, such as red for high-risk areas, yellow for medium-risk areas, and green for low-risk areas. With a resolution of 1×1 meter, it clearly displays the stability status of each area of the roadbed.

[0045] (1) Initialization of population using MOEA / D multi-objective evolutionary algorithm Objective function definition: A multi-objective optimization model is established with the following objectives: maximizing cut-and-fill efficiency (shortest completion time, highest equipment utilization), minimizing costs (fuel consumption, equipment wear, labor costs, etc.), and optimizing quality (roadbed flatness and compaction meet standards). The specific objective function is as follows: Efficiency target: f1=-1 / T, where T is the time required to complete the cut-and-fill operation; Cost target: f2 = Σ(C_fuel + C_wear + C_labor); Quality target: f3=Σ(1-|s_i-s_0| / s_max), where s_i is the actual flatness or compaction degree of the i-th grid point, s_0 is the standard value, and s_max is the maximum allowable deviation.

[0046] Population initialization: The roadbed excavation and fill scheduling plan is represented as a chromosome. The chromosome encodes the allocation of construction equipment (the number and model of excavators, loaders, and bulldozers), operation path planning (the route and operation sequence of each piece of equipment), and the allocation of excavation and fill volume (the amount of fill or cut at each grid point). The population is initialized using a random generation method, and the population size is typically set to 50-100 individuals to ensure diversity.

[0047] (2) GWO Improved Grey Wolf Optimization Algorithm Local Search Improvements to the basic GWO algorithm: Adaptive inertia weights and dynamic learning factors are introduced to the traditional Grey Wolf Optimization algorithm to enhance the algorithm's local search capabilities and convergence speed. The inertia weight w decreases linearly with the number of iterations, using the formula: w = w_max - (w_max - w_min) * t / T_max, where w_max = 0.9, w_min = 0.4, t is the current iteration number, and T_max is the maximum number of iterations. Dynamic learning factors c1 and c2 are adjusted using nonlinear functions to enhance the algorithm's search capabilities at different iteration stages.

[0048] Local search strategy: Based on the population individuals initialized by MOEA / D, a neighborhood search is performed for each individual. Neighborhood operations include equipment reallocation, work path adjustment, and fine-tuning of cut-fill volume. Each operation generates a certain number of neighborhood individuals (5-10). The objective function value of each neighborhood individual is calculated, and the optimal individual is selected to replace the original individual to achieve local optimization.

[0049] (3) Dynamically adjust the convergence factor The convergence factor a controls the search range in the GWO algorithm. It is initially set to 2 and decreases linearly to 0 with the number of iterations. The formula is: a = 2 - 2 * t / T_max. By dynamically adjusting the convergence factor, the algorithm has a strong global search capability in the early stages, while focusing on local, refined search in the later stages, improving solution quality and convergence speed. After multiple rounds of iterations (typically 100-200), a Pareto-optimal solution set is obtained, from which the solution with the best overall performance is selected as the roadbed excavation and filling scheduling plan.

[0050] (1) Implementation of intelligent scheduling for cut-and-fill operations Plan Communication: The generated roadbed excavation and fill scheduling plan is transmitted via wireless communication to the onboard terminal of the construction equipment. The terminal displays the equipment's work tasks, route planning, and required excavation and fill volumes. Simultaneously, the scheduling plan and equipment operating status are displayed in real time on a large screen at the construction site's monitoring center, facilitating overall monitoring and coordination for management personnel.

[0051] Equipment Collaboration: Construction equipment automatically adjusts operating parameters based on the scheduling plan, such as excavator digging depth, loader loading capacity, and bulldozer speed. Equipment exchanges information in real time via wireless communication networks, enabling collaborative operations. For example, an excavator can directly load excavated soil onto a designated loader, which then transports it to the fill area, where it is leveled by a bulldozer, improving operational efficiency.

[0052] (2) Real-time feedback result collection Sensor Data Collection: Various sensors, such as inclination sensors, pressure sensors, and position sensors, are installed on construction equipment to collect real-time operating parameters (excavation depth, load capacity, and travel speed) and the real-time status of the roadbed (flatness and compaction of the excavated and filled areas). Furthermore, drones or fixed monitoring equipment are used to collect moisture distribution data and point cloud data of the roadbed, which are then compared with pre-construction data to evaluate operational effectiveness.

[0053] Feedback Information Processing: Collected real-time data is transmitted to the central control system for data cleaning and preprocessing to remove noise and outliers. Data analysis and model prediction are then used to evaluate the effectiveness of the current scheduling plan and identify existing problems, such as deviations in fill and cut volume, low operational efficiency, and substandard quality.

[0054] (3) Generate target roadbed excavation and filling scheduling plan Plan Adjustment Strategy: Based on real-time feedback, a feedback control algorithm is used to adjust the original scheduling plan. If the fill-cut deviation exceeds the allowable range (±5%), the fill-cut volume for each grid point is recalculated and the equipment's work task allocation is adjusted. If work efficiency is low, the equipment's operation path and coordination mechanism are optimized. If quality does not meet standards, construction process parameters such as the number of compaction times and speed of compaction machinery are adjusted.

[0055] Iterative Optimization: The adjusted scheduling plan is used as the new input, re-entering the data collection, model processing, and optimization algorithm process to form a closed-loop control. Through multiple iterative optimizations, the accuracy and effectiveness of the scheduling plan are continuously improved, ultimately generating a target roadbed excavation and filling scheduling plan that meets construction requirements, ensuring the efficient and high-quality completion of the roadbed excavation and filling operations.

[0056] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-assisted intelligent scheduling method for roadbed excavation and filling operations, characterized in that: The intelligent scheduling method for roadbed excavation and filling operations comprises the following steps: The soil moisture distribution data is collected by a multispectral imager, and the laser point cloud data is obtained by a laser sensor. The moisture distribution data and the laser point cloud data are fused to generate a high-dimensional geological feature tensor to obtain the real-time status data of the roadbed; Using an attention mechanism to dynamically weight the contribution of different sensor data in the real-time roadbed status data to obtain cross-modal roadbed data; Establishing a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model to generate a geological stability thermodynamic map; Based on the geological stability heat map, the MOEA / D multi-objective evolutionary algorithm is used to initialize the population, the GWO grey wolf optimization algorithm is used for local search, and a dynamically adjusted convergence factor is set to obtain the roadbed excavation and filling scheduling scheme.

2. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 1, characterized in that: The method of collecting soil moisture distribution data by a multispectral imager and obtaining laser point cloud data by a laser sensor includes: Establish a calibration model between soil moisture content and spectral reflectance, and collect soil samples with different moisture contents in the construction area to obtain spectral data and actual moisture content; The image data collected by the multispectral imager is input into the inversion model to calculate the soil moisture content corresponding to each pixel point and generate the moisture content distribution data of the entire roadbed area; By measuring the time difference from laser emission to reception, the distance from the target object to the sensor is calculated. Combined with the scanning angle and rotation speed of the sensor, the three-dimensional coordinates of the target object surface are obtained to obtain laser point cloud data.

3. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 1, characterized in that: The step of fusing the moisture content distribution data and the laser point cloud data to generate a high-dimensional geological feature tensor and obtain real-time roadbed status data includes: Converting the moisture content distribution data and the laser point cloud data into a unified global coordinate; Extract moisture characteristics from moisture distribution data, including at least the mean, variance, and gradient of moisture content; extract terrain and landform characteristics from point cloud data, including at least elevation, slope, aspect, and curvature; The moisture characteristics, terrain characteristics and landform characteristics are dimensionally expanded to form multi-dimensional feature vectors, and the feature vectors are fused into high-dimensional geological feature tensor data using tensor product operations to obtain real-time roadbed status data.

4. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 1, characterized in that: The method of dynamically weighting the contribution of different sensor data in the roadbed real-time status data using the attention mechanism to obtain cross-modal roadbed data includes: For each header in the roadbed real-time status data, the similarity between the query and the key is calculated, and a matrix is calculated using a dot product similarity formula to obtain a similarity matrix; Performing Softmax normalization on the similarity matrix to obtain an attention weight matrix, where the weight value range is between 0 and 1; The Value matrix is weighted and summed according to the attention weight to obtain the cross-modal feature vector of each spatiotemporal grid point.

5. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 1, characterized in that: The step of establishing a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model, and generating a geological stability thermodynamic map includes: Establishing a Transformer-GNN hybrid model, and inputting the cross-modal roadbed data into the Transformer-GNN hybrid model; The Transformer encoder is used to segment the cross-modal roadbed data into time series, position-encode the data at each time step, and incorporate time information into the model. The position encoding is generated using sine and cosine functions. In the multi-head self-attention layer, self-attention calculation is performed on the cross-modal data of each time step to capture the dependency between different time steps. The FFN layer is used to perform nonlinear transformation on the attention output to extract temporal features. The construction equipment is regarded as nodes in the graph. The node attributes include the type, location, and working status of the equipment. The GCN graph convolutional network is used to perform graph convolution operations to obtain the equipment node features.

6. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 5, characterized in that: The step of establishing a Transformer-GNN hybrid model, inputting the cross-modal roadbed data into the Transformer-GNN hybrid model, and generating a geological stability thermodynamic map further includes: The temporal features output by the Transformer encoder and the device node features output by the GNN are fused and mapped through a fully connected neural network to obtain geological stability assessment indicators, including at least soil bearing capacity, settlement risk and sliding possibility.

7. The method for intelligent scheduling of roadbed excavation and filling operations based on artificial intelligence as claimed in claim 1, characterized in that: Based on the geological stability heat map, the MOEA / D multi-objective evolutionary algorithm is used to initialize the population, the GWO grey wolf optimization algorithm is used to perform local search, and a dynamically adjusted convergence factor is set to obtain the roadbed excavation and filling scheduling plan, including: Establish a multi-objective optimization model with the objectives of maximizing cut-and-fill efficiency, minimizing costs, and optimizing quality; The roadbed excavation and filling scheduling scheme is represented as a chromosome. The chromosome encoding includes the allocation of construction equipment, operation path planning and allocation of excavation and filling volume. The population is initialized by random generation. The improved Grey Wolf Optimization Algorithm (GWO) is used to search the neighborhood of each individual. Neighborhood operations include equipment reallocation, operation path adjustment, and fill-cut volume fine-tuning. Each operation generates a neighborhood individual. The objective function value of the neighborhood individuals is calculated, the optimal individual is selected to replace the original individual, and the roadbed excavation and filling scheduling scheme is obtained through dynamic adjustment of the convergence factor through iteration.

8. An artificial intelligence-assisted intelligent scheduling system for roadbed excavation and filling operations, characterized in that: The intelligent scheduling system for roadbed excavation and filling operations includes the following modules: A roadbed data acquisition module is used to collect soil moisture distribution data using a multispectral imager and obtain laser point cloud data using a laser sensor. The moisture distribution data and the laser point cloud data are then fused to generate a high-dimensional geological feature tensor to obtain real-time roadbed status data. a contribution calculation module for dynamically weighting the contribution of different sensor data in the roadbed real-time status data using an attention mechanism to obtain cross-modal roadbed data; A stability prediction module is used to establish a Transformer-GNN hybrid model, input the cross-modal roadbed data into the Transformer-GNN hybrid model, and generate a geological stability thermodynamic map; The scheduling scheme generation module is used to initialize the population based on the geological stability heat map using the MOEA / D multi-objective evolutionary algorithm, perform local search using the GWO grey wolf optimization algorithm, set a dynamically adjusted convergence factor, and obtain the roadbed excavation and filling scheduling scheme.

9. The intelligent scheduling system for roadbed excavation and filling operations based on artificial intelligence as claimed in claim 8, characterized in that: The roadbed data acquisition module includes the following submodules: A submodule is established to establish a calibration model between soil moisture content and spectral reflectance. Soil samples with different moisture contents are collected in the construction area to obtain spectral data and actual moisture content. The calculation submodule is used to input the image data collected by the multispectral imager into the inversion model, calculate the soil moisture content corresponding to each pixel point, and generate the moisture content distribution data of the entire roadbed area; The acquisition submodule is used to calculate the distance from the target object to the sensor by measuring the time difference from laser emission to reception, and obtain the three-dimensional coordinates of the target object surface by combining the scanning angle and rotation speed of the sensor to obtain laser point cloud data.

10. The intelligent scheduling system for roadbed excavation and filling operations based on artificial intelligence assistance according to claim 8, characterized in that: The roadbed data acquisition module also Includes the following submodules: A conversion submodule, used to convert the moisture content distribution data and laser point cloud data into a unified global coordinate; The extraction submodule is used to extract moisture characteristics from the moisture distribution data, including at least the mean, variance, and gradient of the moisture content, and to extract terrain characteristics and landform characteristics from the point cloud data, including at least elevation, slope, aspect, and curvature; The fusion submodule is used to expand the dimensions of the moisture characteristics, terrain characteristics and landform characteristics to form a multi-dimensional feature vector, and use tensor product operation to fuse the feature vector into high-dimensional geological feature tensor data to obtain real-time roadbed status data.

Images