Intelligent scheduling system and scheduling method for engineering equipment

Abstract

The invention provides an engineering equipment intelligent scheduling system and scheduling method, and belongs to the technical field of engineering equipment intelligent control. The method comprises the steps that a multi-source sensing terminal reads and uploads standardized engineering equipment data to an edge node through a heterogeneous adaptation module; the cloud end constructs a three-dimensional terrain and a digital twinborn scene by using the unmanned aerial vehicle point cloud, and virtual-real calibration is completed through an iterative nearest point algorithm; the intelligent task distribution system decomposes a task into atomization operation units, and an optimal scheduling scheme is screened by adopting an improved genetic algorithm in combination with discrete event simulation; the edge decision node performs conflict resolution and energy consumption optimization cooperative control based on the four-dimensional space-time grid; and when an abnormality is detected, triggering an emergency response mechanism using a DLite algorithm to re-plan the path. According to the invention, the problems of difficult compatibility, scheduling stiffness and multi-machine conflict of heterogeneous equipment are solved, and real-time accurate scheduling and safe closed-loop control of the engineering equipment cluster are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling of engineering equipment, and in particular to an intelligent scheduling system and method for engineering equipment. Background Technology

[0002] In large-scale construction scenarios such as mines, ports, and municipal engineering projects, the scheduling efficiency of engineering equipment such as excavators, loaders, and dump trucks directly determines the construction progress, cost, and safety. With the development of industrial internet and intelligent manufacturing technologies, intelligent scheduling of engineering equipment has become an inevitable trend in the industry. However, the current field of engineering equipment scheduling still faces many technical bottlenecks. First, there is the compatibility problem of heterogeneous equipment. The communication protocols of different brands and models of engineering equipment are not unified, with multiple protocols such as CAN bus, Modbus, and Ethernet coexisting, making data acquisition difficult and hindering the formation of a unified control network. Second, task allocation is often static and rigid. Traditional scheduling methods lack the ability to dynamically respond to changes in real-time equipment operating conditions and working environment, making it difficult to achieve global optimization. Third, when multiple machines work together, especially when large-scale heterogeneous equipment clusters are involved, path planning is often independent, lacking a unified spatiotemporal coordination mechanism, leading to frequent conflicts in cross-operations and significant safety hazards.

[0003] In the prior art, patent number CN118246310A discloses a digital twin modeling method for remote intelligent monitoring and operation and maintenance of equipment. This technology achieves comprehensive monitoring and correlation analysis of equipment operating conditions, process logistics, and processing quality by constructing a hierarchical industrial flow data acquisition system architecture and utilizing a multimodal geometric model construction method and a dynamic event-state knowledge graph. However, this technology mainly focuses on digital twin modeling at the "monitoring" and "operation and maintenance" levels, emphasizing how to accurately map the state of the physical world for display and analysis, and lacks an "active scheduling" strategy for large-scale heterogeneous equipment clusters. Although it establishes a data acquisition architecture, it does not elaborate on the in-depth analysis and unified standardization adaptation of the underlying heterogeneous protocols (such as non-standard CAN messages) of different brands of engineering equipment, and does not involve dynamic task allocation and multi-machine collaborative path conflict resolution mechanisms based on intelligent optimization algorithms such as genetic algorithms, making it difficult to directly apply to complex on-site construction scheduling scenarios.

[0004] Furthermore, prior art patent CN120125180A discloses a digital engineering scheduling optimization method and device based on discrete event simulation technology. This technology constructs a digital twin model of project scheduling, uses discrete event simulation technology to simulate the execution process of construction tasks, and combines multi-objective optimization algorithms (such as genetic algorithms and particle swarm optimization algorithms) to optimize task scheduling and predict project duration. Although this solution introduces simulation and optimization algorithms to solve scheduling problems, in practical applications, it does not adequately consider the hardware deployment and real-time requirements of the "end-edge-cloud" three-level collaborative architecture. In particular, regarding real-time decision-making capabilities at the edge of the work site, this technology lacks specific implementation methods for millisecond-level path conflict detection and resolution, and has not established a refined occupancy model based on a four-dimensional spatiotemporal grid. At the same time, in terms of energy consumption optimization, this technology mainly focuses on macro-level resource allocation and lacks refined closed-loop control logic that adjusts the power output of equipment engines based on the physical properties of the work object (such as soil hardness), thus failing to achieve ultimate energy saving of engineering equipment at the micro-level.

[0005] In summary, existing technologies either lean towards static monitoring and modeling or macroscopic task scheduling simulation, lacking an intelligent scheduling system capable of deeply integrating underlying data from heterogeneous devices, possessing edge-cloud collaboration capabilities, resolving multi-machine spatiotemporal conflicts in real time, and dynamically optimizing energy consumption based on environmental physical properties. Especially when facing complex and ever-changing construction environments, how to avoid scheduling schemes getting trapped in local optima by improving genetic algorithms, how to utilize digital twins for high-precision virtual-real mapping calibration, and how to achieve real-time obstacle avoidance based on spatiotemporal grids and energy consumption control based on soil hardness at the edge remain pressing issues that current technologies need to address. Summary of the Invention

[0006] The main objective of this invention is to provide an intelligent scheduling system and method for engineering equipment, which solves the problems faced by existing scheduling systems, such as data silos caused by incompatibility of communication protocols between heterogeneous equipment, the problem that traditional genetic algorithms are prone to getting stuck in local optima in task allocation and lack simulation verification, the safety hazards caused by the lack of a refined spatiotemporal conflict resolution mechanism in multi-machine collaborative operations, and the problem of lacking real-time energy consumption closed-loop control based on the physical properties of the working environment.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an intelligent scheduling method for engineering equipment, comprising the following steps: S1. The edge-side multi-source sensing terminal reads the CAN bus and sensor data of the engineering equipment through the heterogeneous adaptation module, uses the protocol parsing algorithm to complete the data standardization, collects the operating data of the engineering equipment in real time and uploads it to the edge decision node. S2. The cloud-based dispatch center uses point cloud data of the work area acquired by drones to generate a 3D terrain model through statistical outlier removal filtering and Delaunay triangulation mesh construction algorithm. It also combines engineering equipment parameters to construct a digital twin scene and completes virtual-real mapping calibration through iterative nearest point algorithm. S3, the cloud-based intelligent task allocation system decomposes the total amount of construction tasks into discrete atomized work units based on the rated operating capacity of the engineering equipment. It uses an improved genetic algorithm with an adaptive crossover and mutation mechanism to generate candidate scheduling schemes and inputs the schemes into a digital twin scenario. Through discrete event simulation, it infers the operating status of each piece of equipment, records the completion time and energy consumption data, selects the optimal scheduling scheme, and sends it to the edge decision nodes. S4. The edge decision node receives the optimal scheduling scheme, establishes a spatiotemporal occupancy table through the conflict resolution module, and performs multi-machine collaborative control of the engineering equipment in combination with the load mapping logic of the energy consumption optimization unit. S5. When the multi-source sensing terminal on the edge detects abnormal data of the working environment or engineering equipment that exceeds the safety threshold, the emergency response mechanism is triggered. The cloud scheduling center uses a dynamic path planning algorithm to perform global path replanning and updates the control commands through the edge decision nodes.

[0008] In the preferred embodiment, in step S1, the heterogeneous adaptation module pre-configures a communication protocol library for multi-brand engineering equipment. The communication protocol library includes CAN 2.0 and Modbus RTU protocols. It identifies the device type through the message header identifier and calls the corresponding decoding rules to convert non-standard CAN messages or Modbus data into unified JSON format data. The status sensing unit collects data at a frequency of 20Hz. The data includes GNSS position data, hydraulic system pressure data, and triaxial vibration data of the engineering equipment. After smoothing the GNSS position data through a Kalman filter algorithm, it is transmitted to the edge decision node through the 5G C-V2X network.

[0009] In the preferred embodiment, in step S2: S21. Use a drone equipped with a lidar to scan the work area along a preset flight path at a height of 5-10m with a scanning frequency of 10Hz to obtain raw point cloud data, and use a statistical outlier removal algorithm to remove noise. S22. The Delaunay triangulation algorithm is used to convert the processed point cloud data into an irregular triangular mesh model, extract the elevation and slope features of the terrain, and generate a navigation mesh for passable areas. S23. Map the physical dimensions, kinematic parameters, and material distribution attributes of engineering equipment to a three-dimensional virtual scene to construct a multi-level virtual environment that includes terrain, equipment, and material layers. S24. Obtain the measured GNSS coordinates of more than 3 fixed reference points on site, use the iterative nearest point algorithm to calculate the rotation and translation matrix between the corresponding feature points in the virtual scene and the measured coordinates, and perform rigid body transformation calibration on the coordinate system of the 3D virtual scene.

[0010] In the preferred embodiment, the specific method for decomposing the construction task into atomized work units in step S3 is as follows: obtain the total material volume V of the task area and the average bucket capacity of the excavating equipment participating in the operation. Calculate the number of theoretical assignments ,in It is a rounding function; The atomic task allocation rule is as follows: based on the real-time load rate and operating efficiency of each piece of engineering equipment, a round-robin allocation combined with a load balancing strategy is adopted to allocate N atomic task units containing four action sequences of excavation, loading, transportation and unloading to each piece of equipment, ensuring that the atomic task sequence of a single piece of equipment is continuous and without idle intervals.

[0011] In the preferred scheme, step S3, which involves generating candidate scheduling schemes using an improved genetic algorithm, includes: S31. Initialize the population using a real number encoding method based on task sequences. Set the population size to 50-100. Randomly generate task allocation sequences to ensure that the total number of tasks for each chromosome is equal to the total number of atomic task units. S32. Define the fitness function. The specific expression of the fitness function is as follows: ,in , , , The weights for time, energy consumption, and load variance are determined using the analytic hierarchy process (AHP) based on construction priority. To maximize the completion time, Total energy consumption, For equipment load variance; S33. Evolutionary operation: During the evolution process, based on the current iteration number k and the maximum iteration number... Dynamically adjust the cross rate and variability Adjust the formula to , ,in The initial crossover rate is 0.8-0.9. The termination crossover rate is 0.4-0.5. The initial mutation rate was 0.05-0.1. The termination mutation rate is 0.01-0.02; S34. Simulation screening: For the generated candidate scheduling schemes, a virtual clock is started in the digital twin scenario. The action sequence of atomic tasks is used as the event unit. The time parameters of each action are set in combination with the rated parameters of the engineering equipment. The equipment operation is simulated through event queue scheduling. The completion time and energy consumption data at the end of the simulation are recorded as the basis for fitness evaluation. The evolution termination condition is that the number of iterations reaches a certain threshold. If the maximum value of the fitness function does not improve for 10 consecutive generations, select the scheduling scheme with the best fitness.

[0012] In the preferred embodiment, the specific method for control via the conflict resolution module in step S4 is as follows: S41. Construct a four-dimensional spatiotemporal grid for the work area. Where i, j, k, l are grid indices, the grid resolution of the spatial dimension (x, y, z) is set to 0.5m × 0.5m × 0.5m, and the time slice of the time dimension t is set to 0.1s, mapping the planned path of each piece of engineering equipment into a series of spatiotemporal voxel occupancy requests; S42. Collision Detection, Safety Threshold Based on the maximum width of the engineering equipment + a safety margin of 0.3m, if two pieces of equipment are detected operating at the same time... Internal requests occupy the same spatial element Or the spatial distance is less than the safety threshold If so, it is determined to be a path conflict; S43. Conflict resolution: Calculate the priority weight of each conflicting equipment mission. ,in To assess the urgency of the task, As for the importance of the task, Keep the paths of high-priority equipment unchanged, and apply a time lag to low-priority equipment. Where d is the distance between the conflict point and the current position of the low-priority equipment, v is the rated driving speed of the equipment, its spacetime occupancy request is updated until the conflict disappears, and a conflict-free control command is generated and issued.

[0013] In the preferred embodiment, the control method for the energy consumption optimization unit in step S4 is as follows: S44. Real-time acquisition of pump outlet pressure data P and flow data Q from the hydraulic system of the excavating equipment, using the formula... The current soil hardness value H is estimated in reverse, where k is a correction factor: k=0.8 for sand and k=1.2 for clay. The operating angular velocity; S45. Obtain the current task's transport distance L and road gradient. ; S46. Query the preset 3D MAP map of "hardness, transport distance, and power". The horizontal axis of the 3D MAP map is the earthwork hardness H, the vertical axis is the transport distance L, and the vertical axis is the engine power P. Supplement the unpreset operating parameters through linear interpolation, determine the engine's optimal fuel economy operating point under the current operating conditions, and output the corresponding target speed command to the engine electronic control unit. Adjust the fuel injection quantity through the PID control algorithm to match the target power output curve. The fuel injection quantity adjustment range corresponding to the target speed command is preset according to the engine parameters.

[0014] In the preferred embodiment, in step S5, the multi-source sensing terminal includes a triaxial accelerometer, and the emergency response mechanism comprises the following steps: S51. Perform Fast Fourier Transform (FFT) on the time-domain vibration signal acquired by the triaxial accelerometer to extract frequency domain feature values; S52. Preset frequency domain characteristic thresholds for different fault types. For bearing wear, the amplitude threshold for the 100-200Hz frequency band is 5g-8g, and for hydraulic leakage, the amplitude threshold for the 50-100Hz frequency band is 3g-5g. Through calibration with historical fault data, if the amplitude of a certain frequency band exceeds the corresponding preset fault characteristic threshold, the local rule engine will immediately generate an alarm signal and upload it. After receiving an alarm, the S53 cloud-based dynamic priority engine locks down the associated devices around the faulty device. All engineering equipment within a circular area with a radius of 5-8m centered on the faulty device is identified as associated devices, their original tasks are suspended, and the location of the faulty device is marked as a dynamic obstacle. S54, Adopting D The Lite dynamic path planning algorithm, with its heuristic function set as follows: ,in Current position For safe evacuation points, the location of the faulty equipment is marked as a high-cost area. The navigation grid of the passable area is updated synchronously. Avoidance paths are searched in the updated cost map, and an emergency control command sequence containing steering angle and speed is generated. The maximum turning angle of the replanned path does not exceed 30°.

[0015] An intelligent scheduling system for engineering equipment, the system comprising: The multi-source sensing terminal is installed on engineering equipment. The hardware includes a CAN bus adapter, a GNSS positioning module, a three-axis accelerometer, and a hydraulic sensor. The software includes a protocol parsing plugin and a Kalman filter program. It contains a heterogeneous adaptation module and a state sensing unit, which are used to perform protocol parsing, data filtering, standardized uploading, and anomaly detection in steps S1 and S51-S52. The edge decision node is deployed on the edge server at the work site. The edge server is an industrial-grade edge server. The software includes a real-time operating system, a conflict resolution algorithm plugin, and an energy consumption control program, which includes a local rule engine, a conflict resolution module, and an energy consumption optimization unit. It is used to perform spatiotemporal conflict detection, dynamic adjustment of engine power, and issuance of control commands for steps S4, S41-S43, and S44-S46. The cloud-based dispatch hub is deployed on a cloud server cluster. The software includes a digital twin platform, a genetic algorithm engine, and a path planning module. It includes a digital twin platform, an intelligent task allocation system, and a security monitoring module. It is used to perform point cloud data processing, scene construction, genetic algorithm task allocation, global path replanning, and emergency process monitoring for steps S2-S3, S31-S34, and S53-S54. The multi-source sensing terminal communicates with the edge decision-making node via a 5G C-V2X network, while the edge decision-making node communicates with the cloud scheduling center via fiber optic Ethernet. The data exchange format is uniformly JSON, and the transmission frequency is 20Hz.

[0016] In the preferred embodiment, the intelligent task allocation system is configured to perform the construction task decomposition and improvement genetic algorithm steps; The digital twin platform is configured to perform point cloud processing and virtual-to-real calibration steps; The conflict resolution module is configured to perform a four-dimensional spatiotemporal grid conflict detection step; The energy optimization unit is configured to execute energy consumption adjustment control steps; The local rules engine works in conjunction with the multi-source sensing terminal to execute anomaly detection and alarm procedures; The security monitoring module is configured to receive alarm signals, monitor the emergency response process, and synchronously update the device status in the digital twin scenario.

[0017] This invention provides an intelligent scheduling system and method for engineering equipment. The invention adopts a three-level collaborative architecture of "end-edge-cloud". Through the heterogeneous adaptation module of multi-source sensing terminals, it can pre-set multi-brand communication protocol libraries and call corresponding decoding rules, realizing standardized collection and interconnection of engineering equipment data of different brands and different protocols, effectively breaking down information silos and improving the system's compatibility and scalability.

[0018] This invention utilizes UAV point cloud data to construct a high-precision digital twin scene and combines iterative nearest-point algorithm for virtual-real mapping calibration, providing a precise digital foundation for scheduling decisions. In cloud task allocation, an improved genetic algorithm with an adaptive crossover and mutation mechanism is adopted, which can dynamically adjust algorithm parameters according to the iteration process, effectively avoiding premature convergence of the algorithm. Combined with discrete event simulation, it ensures that the generated scheduling scheme achieves global optimization in multiple dimensions such as load balancing, completion time, and energy consumption, significantly improving the scientific nature of task allocation and the reliability of pre-simulation.

[0019] This invention introduces a conflict resolution mechanism based on a four-dimensional spatiotemporal grid at the edge. By establishing a spatiotemporal occupancy table and combining it with task priority weights, it can detect and resolve path conflicts between engineering equipment in real time, achieving millisecond-level safe avoidance and significantly reducing the safety risks of multi-machine collaborative operations. At the same time, the energy consumption optimization unit innovatively establishes a reverse estimation model of earthwork hardness based on hydraulic system pressure and flow data, and dynamically adjusts the fuel injection quantity of the engine electronic control unit by querying a three-dimensional MAP map. This achieves adaptive adjustment of equipment power output based on the physical properties of the working environment, significantly reducing fuel consumption while ensuring operational efficiency.

[0020] Furthermore, this invention also establishes a comprehensive emergency response mechanism. Through end-side fast Fourier transform analysis of vibration signals, it can promptly identify fault characteristics such as bearing wear and hydraulic leakage, and combine this with D... The Lite dynamic path planning algorithm quickly generates avoidance paths, ensuring operational safety and system resilience in the event of sudden anomalies, and realizing intelligent, safe, and green scheduling of the entire process of engineering equipment clusters. Attached Figure Description

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the overall architecture of the intelligent scheduling system for engineering equipment of the present invention; Figure 2 This is a schematic diagram of the task allocation process of the improved genetic algorithm of this invention; Figure 3 This is a schematic diagram of the spatiotemporal reservation-based path conflict resolution process of the present invention; Figure 4 This is a schematic diagram of the emergency response dispatch process of the present invention. Detailed Implementation

[0022] Example 1 like Figure 1-4 As shown, an intelligent scheduling method for engineering equipment includes the following steps: S1. The edge-side multi-source sensing terminal reads the CAN bus and sensor data of the engineering equipment through the heterogeneous adaptation module, uses the protocol parsing algorithm to complete the data standardization, collects the operating data of the engineering equipment in real time and uploads it to the edge decision node. S2. The cloud-based dispatch center uses point cloud data of the work area acquired by drones to generate a 3D terrain model through statistical outlier removal filtering and Delaunay triangulation mesh construction algorithm. It also combines engineering equipment parameters to construct a digital twin scene and completes virtual-real mapping calibration through iterative nearest point algorithm. S3, the cloud-based intelligent task allocation system decomposes the total amount of construction tasks into discrete atomized work units based on the rated operating capacity of the engineering equipment. It uses an improved genetic algorithm with an adaptive crossover and mutation mechanism to generate candidate scheduling schemes and inputs the schemes into a digital twin scenario. Through discrete event simulation, it infers the operating status of each piece of equipment, records the completion time and energy consumption data, selects the optimal scheduling scheme, and sends it to the edge decision nodes. S4. The edge decision node receives the optimal scheduling scheme, establishes a spatiotemporal occupancy table through the conflict resolution module, and performs multi-machine collaborative control of the engineering equipment in combination with the load mapping logic of the energy consumption optimization unit. S5. When the multi-source sensing terminal on the edge detects abnormal data of the working environment or engineering equipment that exceeds the safety threshold, the emergency response mechanism is triggered. The cloud scheduling center uses a dynamic path planning algorithm to perform global path replanning and updates the control commands through the edge decision nodes.

[0023] In the preferred embodiment, in step S1, the heterogeneous adaptation module pre-configures a communication protocol library for multi-brand engineering equipment. The communication protocol library includes CAN 2.0 and Modbus RTU protocols. It identifies the device type through the message header identifier and calls the corresponding decoding rules to convert non-standard CAN messages or Modbus data into unified JSON format data. The status sensing unit collects data at a frequency of 20Hz. The data includes GNSS position data, hydraulic system pressure data, and triaxial vibration data of the engineering equipment. After smoothing the GNSS position data through a Kalman filter algorithm, it is transmitted to the edge decision node through the 5G C-V2X network.

[0024] During the operation of the multi-source sensing terminal, to address the data silo problem caused by inconsistent communication protocols among different brands of engineering equipment, the heterogeneous adaptation module pre-installs a multi-brand engineering equipment communication protocol library, including CAN 2.0 and Modbus RTU. Upon receiving the raw data stream, this module first reads the message header identifier of the data frame. Because different manufacturers' devices define different message IDs and data formats in the CAN bus protocol, the system can automatically identify the brand and model of the currently connected device by comparing the message header identifier. Once identified successfully, the system immediately calls the corresponding decoding rule table pre-stored in the protocol library. For non-standard CAN messages, the decoding rules specify the physical meaning and conversion coefficients represented by specific data bits; for Modbus data, they specify the mapping relationship between register addresses and physical quantities. Subsequently, the parsed heterogeneous data is uniformly converted into JSON (JavaScript Object Notation) format data packets. JSON format is lightweight, self-describing, and easy to parse, shielding underlying hardware differences and ensuring that data has a unified structural standard when transmitted to edge decision nodes, facilitating subsequent cross-platform processing and storage.

[0025] The status sensing unit is responsible for high-frequency and precise digital capture of the operating status of the engineering equipment. This unit is set to a data acquisition frequency of 20Hz, meaning it collects data 20 times per second, a frequency sufficient to capture instantaneous changes in the equipment during dynamic operation. The types of data collected include GNSS position data reflecting the equipment's spatial location, hydraulic system pressure data reflecting the equipment's workload, and triaxial vibration data reflecting the equipment's mechanical health. To eliminate positioning errors caused by environmental interference, signal obstruction, or sensor noise, the system incorporates a Kalman filter algorithm to process the raw GNSS position data.

[0026] The core of the Kalman filter algorithm lies in using the system's state equation and observation equation to estimate the system's optimal state through two stages: prediction and update. Specifically, assuming... The state vector of the system at time t is (Including position and velocity information), the state prediction equation is: ,in This represents the current state predicted based on the state at the previous time step. Here is the state transition matrix. To control the input matrix, To control the input vector, the prediction error covariance matrix is ​​calculated simultaneously. ,in Let be the process noise covariance matrix. During the update phase, calculate the Kalman gain. ,in For the observation matrix, To observe the noise covariance matrix, the predicted state is corrected using Kalman gain to obtain the optimal estimate. ,in This uses the current observations (i.e., the raw GNSS data). Finally, the error covariance matrix is ​​updated. ,in The matrix is ​​the identity matrix. Through the above iterative calculations, Gaussian white noise can be effectively filtered out, making the output positioning trajectory smoother and more realistic. The processed high-precision data is finally transmitted to the edge decision node via a 5G C-V2X (Cellular Vehicle-to-Everything) network. 5G C-V2X technology has high bandwidth, low latency, and high reliability communication capabilities, ensuring that massive amounts of real-time data can be uploaded in milliseconds, providing a reliable data foundation for subsequent real-time conflict resolution and energy consumption optimization.

[0027] The beneficial effects of the above technical solution are as follows: By pre-installing multiple protocol libraries and a unified JSON format conversion, it achieves "plug-and-play" access to heterogeneous equipment from multiple brands at the work site, eliminating the need for expensive hardware modifications to existing equipment, greatly reducing system deployment costs and improving compatibility. The use of 20Hz high-frequency acquisition combined with a Kalman filter algorithm effectively solves the problems of GPS signal drift and jitter in complex construction environments, significantly improving the robustness and accuracy of equipment positioning and ensuring the accuracy of virtual-real mapping in digital twin scenarios. Utilizing 5G C-V2X network transmission ensures the real-time performance and integrity of data when a large number of devices communicate concurrently, avoiding data packet loss due to network congestion, thereby guaranteeing the timeliness of dispatch instructions and operational safety.

[0028] In the preferred embodiment, in step S2: S21. Use a drone equipped with a lidar to scan the work area along a preset flight path at a height of 5-10m with a scanning frequency of 10Hz to obtain raw point cloud data, and use a statistical outlier removal algorithm to remove noise. S22. The Delaunay triangulation algorithm is used to convert the processed point cloud data into an irregular triangular mesh model, extract the elevation and slope features of the terrain, and generate a navigation mesh for passable areas. S23. Map the physical dimensions, kinematic parameters, and material distribution attributes of engineering equipment to a three-dimensional virtual scene to construct a multi-level virtual environment that includes terrain, equipment, and material layers. S24. Obtain the measured GNSS coordinates of more than 3 fixed reference points on site, use the iterative nearest point algorithm to calculate the rotation and translation matrix between the corresponding feature points in the virtual scene and the measured coordinates, and perform rigid body transformation calibration on the coordinate system of the 3D virtual scene.

[0029] In step S21, to construct a high-precision digital base, the system uses a drone equipped with a LiDAR to perform a low-altitude scanning task. Setting a low-altitude flight altitude of 5 to 10 meters ensures that the LiDAR has extremely high spatial resolution when acquiring ground texture and undulation details. Combined with a scanning frequency of 10Hz, i.e., 10 full-field scans per second, high-density raw point cloud data can be generated. Since the raw point cloud data inevitably contains outlier noise caused by air dust, birds, or equipment vibration, the system introduces a statistical outlier removal algorithm for preprocessing. This algorithm first calculates the number of outliers in each point in the point cloud and their neighborhood. The average distance between the nearest neighbors Then calculate the average distance of all points. and standard deviation The system sets a distance threshold. ,in This is the standard deviation factor. If the average distance of a point... Greater than this threshold If any such noise is detected, it is identified as noise and removed. This process effectively cleans the data source, preventing false sharp protrusions or holes from appearing in subsequent modeling, and laying the foundation for building a smooth terrain model.

[0030] In step S22, the Delaunay triangulation algorithm is used to transform the discrete point cloud data into a continuous triangular mesh model. The core characteristic of Delaunay triangulation is the empty circumcircle property, meaning that the circumcircle of any triangle in the mesh does not contain any other points from the point set. This property minimizes the generation of elongated triangles, ensuring the smoothness and numerical stability of the terrain model during interpolation. Based on the generated TIN model, the system further calculates the normal vector of each triangular facet. Through formula Calculate the slope angle When the slope angle When the slope angle is less than the preset maximum climbing angle for engineering equipment, the area is marked as a passable navigation grid. This step transforms the geometric terrain data into a navigation map with semantic information, providing physical constraints for subsequent path planning.

[0031] Step S23 involves constructing a multi-layered virtual environment, aiming to achieve a full-element mapping from the physical world to the digital world. This step not only imports terrain data but also uses parametric modeling technology to map the physical dimensions of engineering equipment into 3D bounding boxes, their kinematic parameters into the degrees of freedom constraints of virtual joints, and the material distribution of the work area into a volumetric data field. By constructing terrain, equipment, and material layers, the system achieves decoupled management of static environments and dynamic objects. The terrain layer provides the load-bearing foundation, the equipment layer is responsible for motion simulation, and the material layer is used to simulate volume changes during excavation and stacking. This layered architecture makes the digital twin scene not only a visual reproduction but also a precise simulation of physical properties and logical relationships, thereby supporting complex excavation resistance calculations and collision detection.

[0032] Step S24 ensures the spatial consistency of the digital twin system through virtual-real mapping calibration. At least three reference points with fixed locations and known absolute coordinates are selected at the actual work site, and their measured GNSS coordinates are used as the target point set by RTK equipment. Simultaneously, corresponding feature points are extracted from the virtual scene as the source point set. The system uses the iterative nearest-point algorithm to find the rotation matrix that minimizes the error between the source and target point sets. Translation vector Its objective function is to minimize the mean square error. Through iterative calculations until the error converges, the solved rigid body transformation matrix is ​​applied to the coordinate system transformation of the entire 3D virtual scene. This calibration process eliminates drift caused by initial positioning deviations or scale errors during virtual scene construction, ensuring that the scheduling path coordinates generated in the cloud can be accurately executed by on-site physical equipment.

[0033] The beneficial effects of the above technical solution are as follows: By combining the statistical outlier removal algorithm with Delaunay triangulation, environmental noise can be effectively removed while preserving key terrain features, generating a high-quality navigation mesh, and significantly reducing the risk of vehicles getting stuck due to terrain data errors in path planning. The use of multi-level virtual environment construction technology achieves deep mapping of physical attributes, enabling simulation pre-playing to realistically reflect the physical interaction between equipment and the environment, thus improving the feasibility of scheduling schemes. Rigid body transformation calibration using the iterative nearest point algorithm achieves sub-meter alignment between the virtual coordinate system and the real-world GNSS coordinate system, solving the common problem of virtual-real spatial misalignment in digital twin systems, and ensuring the accurate placement and execution of remote scheduling commands in physical space.

[0034] In the preferred embodiment, the specific method for decomposing the construction task into atomized work units in step S3 is as follows: obtain the total material volume V of the task area and the average bucket capacity of the excavating equipment participating in the operation. Calculate the number of theoretical assignments ,in It is a rounding function; The atomic task allocation rule is as follows: based on the real-time load rate and operating efficiency of each piece of engineering equipment, a round-robin allocation combined with a load balancing strategy is adopted to allocate N atomic task units containing four action sequences of excavation, loading, transportation and unloading to each piece of equipment, ensuring that the atomic task sequence of a single piece of equipment is continuous and without idle intervals.

[0035] In step S3, the cloud-based intelligent task allocation system first performs discretization processing on continuous work objects, transforming the macroscopic earthwork construction operation into the smallest scheduling unit that can be processed by computer algorithms. The system first calculates the total material volume of the current work area using terrain differences in the digital twin scenario. This value represents the total amount of earthwork that needs to be removed or filled. Simultaneously, the system reads the equipment parameters of the excavating equipment currently involved in the operation and extracts its average bucket capacity. As a standard unit of quantification. Based on these two key parameters, the system utilizes the formula Calculate the number of theoretical assignments In this formula, the symbol This represents the floor function, which mathematically maps the value within the parentheses to the smallest integer greater than or equal to that value. The physical significance of using floor calculation is that it ensures that even the remaining material, less than a full bucket, is classified as an independent task, thus guaranteeing that the total volume of the decomposed tasks completely covers the target volume and preventing any omissions.

[0036] Each calculated The value represents an independent atomic task unit, which is not a single action but encapsulated as an indivisible, complete chain of operations. Each atomic task unit strictly contains a sequence of four actions with strict temporal constraints: digging, loading, transporting, and unloading. This encapsulation allows the scheduling system to treat complex, continuous construction processes as a series of discrete, countable, and standardized objects, laying the mathematical foundation for subsequent combinatorial optimization using genetic algorithms. After clarifying... After each atomic task unit is completed, the system enters the task allocation phase. This phase introduces a dynamic polling and load balancing strategy based on equipment capabilities. The system reads the real-time load rate and operational efficiency data of each piece of engineering equipment in real time. The real-time load rate reflects the estimated total time of the currently assigned tasks, while the operational efficiency is determined by the historical average cycle time of the equipment.

[0037] The core logic of the atomic task allocation rule lies in balancing the performance differences between heterogeneous equipment through algorithms. The system allocates these... When assigning atomic task units, instead of a simple average allocation, the next atomic task unit is preferentially assigned to the equipment node with the lowest current cumulative load estimate. This mechanism can dynamically compensate for differences in power performance between different brands or models of equipment, preventing high-performance equipment from waiting and low-performance equipment from accumulating. Simultaneously, the allocation algorithm sets a continuity constraint: within the task queue of the same equipment, adjacent atomic task units must be consecutive on the timeline. The system automatically eliminates unnecessary waiting time windows, ensuring that a single piece of equipment is in a continuous operating state when executing atomic task sequences.

[0038] The beneficial effects of the above technical solution are as follows: By using a discretization decomposition method based on bucket capacity, the continuous-domain earthmoving scheduling problem is transformed into a discrete-domain combinatorial optimization problem, enabling the genetic algorithm to accurately optimize a massive number of micro-operation units. The use of an up-rounding function ensures the accuracy of the quantity calculation, avoiding construction shortages caused by the accumulation of tail errors. A polling allocation strategy based on real-time load rate and operational efficiency achieves global load balancing for heterogeneous equipment clusters, significantly reducing equipment idling rate and waiting time. Encapsulating the excavation to unloading process into atomic units and forcing continuous execution conforms to the actual physical operation logic of construction machinery, effectively increasing the proportion of effective engine operating time, thereby maximizing overall operational energy efficiency.

[0039] In the preferred scheme, step S3, which involves generating candidate scheduling schemes using an improved genetic algorithm, includes: S31. Initialize the population using a real number encoding method based on task sequences. Set the population size to 50-100. Randomly generate task allocation sequences to ensure that the total number of tasks for each chromosome is equal to the total number of atomic task units. S32. Define the fitness function. The specific expression of the fitness function is as follows: ,in , , , The weights for time, energy consumption, and load variance are determined using the analytic hierarchy process (AHP) based on construction priority. To maximize the completion time, Total energy consumption, For equipment load variance; S33. Evolutionary operation: During the evolution process, based on the current iteration number k and the maximum iteration number... Dynamically adjust the cross rate and variability Adjust the formula to , ,in The initial crossover rate is 0.8-0.9. The termination crossover rate is 0.4-0.5. The initial mutation rate was 0.05-0.1. The termination mutation rate is 0.01-0.02; S34. Simulation screening: For the generated candidate scheduling schemes, a virtual clock is started in the digital twin scenario. The action sequence of atomic tasks is used as the event unit. The time parameters of each action are set in combination with the rated parameters of the engineering equipment. The equipment operation is simulated through event queue scheduling. The completion time and energy consumption data at the end of the simulation are recorded as the basis for fitness evaluation. The evolution termination condition is that the number of iterations reaches a certain threshold. If the maximum value of the fitness function does not improve for 10 consecutive generations, select the scheduling scheme with the best fitness.

[0040] In step S31, the intelligent task allocation system first establishes the algorithm's initialization environment. The system employs a real-number encoding strategy based on task sequences. Each individual's chromosome consists of a sequence of real numbers, where each gene position corresponds to a specific atomic task unit. The gene's position within the chromosome directly represents the execution priority or order of that task unit. To balance the breadth of the algorithm's search space with computational efficiency, the system strictly limits the population size to between 50 and 100. During the initialization phase, the system generates the first-generation population using a random generation algorithm and incorporates a built-in verification mechanism to ensure that the total number of genes in each generated chromosome is strictly equal to the total number of atomic task units N calculated in the previous steps. This guarantees the integrity of the scheduling scheme and prevents task omissions or duplicate allocations.

[0041] In step S32, to evaluate the performance of the scheduling scheme, the system defines a multi-objective fusion fitness function. This function transforms the three optimization objectives that originally needed to be minimized into fitness values ​​that need to be maximized through a weighted sum of reciprocals. The specific expression is as follows: In the formula, This represents the maximum completion time, which is the final completion time of the entire project. Represents the total energy consumption of all participating equipment; This represents the variance of equipment load, used to measure the degree of workload balance among various pieces of equipment. To accommodate the different focuses of various construction scenarios, the formula introduces three weighting coefficients. , , And satisfy The constraints are defined. In practical applications, the Analytic Hierarchy Process (AHP) is used to assign weights based on the current construction priority; for example, increasing the weights in a tight schedule scenario. The value increases in cost-conscious scenarios. The value.

[0042] In step S33, to address the problem of traditional genetic algorithms easily getting trapped in local optima, the system introduces a linear adaptive crossover and mutation probability adjustment mechanism. During the evolutionary iteration process, the crossover rate... and variability It is no longer a fixed constant, but rather changes with the current iteration number. With maximum number of iterations The ratio changes dynamically. The adjustment formulas are as follows: as well as Among them, the initial crossover rate Setting the threshold in the high range of 0.8 to 0.9 aims to maintain population diversity in the early stages of iteration and facilitate extensive global searches; the termination crossover rate... The value is set between 0.4 and 0.5 to preserve excellent gene patterns in the later stages of iteration. Similarly, the mutation rate is gradually reduced from the initial 0.05 to 0.1 to 0.01 to 0.02. This dynamic adjustment strategy from high to low enables the algorithm to have strong exploration capabilities in the early stages to escape local optima, while having good convergence stability in the later stages to accurately lock in the global optimum.

[0043] In step S34, the system deeply couples the genetic algorithm with digital twin simulation technology to achieve dynamic and accurate evaluation of fitness. For each candidate scheduling scheme generated by the genetic algorithm, the system does not use static formulas to estimate the results, but instead inputs them into the digital twin scenario for discrete event simulation. The system starts a virtual clock, using the four standard action sequences of digging, loading, transporting, and unloading included in the atomic task as the smallest event unit, and combines physical parameters such as the rated power and movement speed of the engineering equipment to simulate the actual operation process in the virtual environment. The system simulates the running trajectory and interaction logic of the equipment through an event queue scheduling mechanism, and finally records the actual completion time and energy consumption data at the end of the simulation, and substitutes them into the fitness function for calculation. When the number of iterations reaches the preset $K_{max}$ or the maximum value of the fitness function of the population no longer increases after 10 consecutive generations, the algorithm is considered to have converged, and the system automatically outputs the scheduling scheme with the optimal fitness as the final execution instruction.

[0044] The beneficial effects of the above technical solution are as follows: By constructing a multi-objective fitness function, comprehensive optimization of construction period, energy consumption, and equipment load balancing is achieved, avoiding resource waste or excessive equipment wear caused by single-objective optimization. The adoption of a linear adaptive crossover and mutation mechanism effectively balances the algorithm's global search capability and local convergence capability, significantly improving the algorithm's solution efficiency and stability, and solving the problem of premature convergence in traditional genetic algorithms when dealing with large-scale scheduling problems. Most importantly, using discrete event simulation instead of traditional mathematical model estimation to calculate fitness fully considers the dynamic disturbances and temporal constraints of the construction site, ensuring that the generated scheduling scheme is realistically feasible in the physical world and greatly reducing the deviation between theoretical scheduling and actual construction.

[0045] In the preferred embodiment, the specific method for control via the conflict resolution module in step S4 is as follows: S41. Construct a four-dimensional spatiotemporal grid for the work area. Where i, j, k, l are grid indices, the grid resolution of the spatial dimension (x, y, z) is set to 0.5m × 0.5m × 0.5m, and the time slice of the time dimension t is set to 0.1s, mapping the planned path of each piece of engineering equipment into a series of spatiotemporal voxel occupancy requests; S42. Collision Detection, Safety Threshold Based on the maximum width of the engineering equipment + a safety margin of 0.3m, if two pieces of equipment are detected operating at the same time... Internal requests occupy the same spatial element Or the spatial distance is less than the safety threshold If so, it is determined to be a path conflict; S43. Conflict resolution: Calculate the priority weight of each conflicting equipment mission. ,in To assess the urgency of the task, As for the importance of the task, Keep the paths of high-priority equipment unchanged, and apply a time lag to low-priority equipment. Where d is the distance between the conflict point and the current position of the low-priority equipment, v is the rated driving speed of the equipment, its spacetime occupancy request is updated until the conflict disappears, and a conflict-free control command is generated and issued.

[0046] In step S41, to achieve refined control over the operational trajectory of the engineering equipment cluster, the conflict resolution module constructs a four-dimensional spatiotemporal grid model containing spatial and temporal dimensions. This model uses... Let represent the discrete grid indices along the X, Y, Z axes, and time axis, respectively. To adapt to the complex unstructured terrain of mines or construction sites and the large physical volume of engineering equipment, the system sets the spatial dimension grid resolution to 0.5 meters by 0.5 meters by 0.5 meters. This scale accurately encompasses the outline of the excavator or dump truck while effectively filtering computational noise caused by minor terrain undulations. Simultaneously, to match the high-frequency control requirements of edge decision nodes, the time dimension... The time slice resolution is set to 0.1 seconds. Based on this, the system discretizes the pre-planned continuous path of each piece of engineering equipment, mapping it to a series of spatiotemporal voxel occupancy requests. Each request represents the equipment's position within a specific time slice. Internal to specific spatial elements Exclusive right to use.

[0047] In step S42, the system executes real-time conflict detection logic. First, a safety threshold is determined. This value is not fixed, but is calculated based on the maximum physical width of the current engineering equipment plus a safety margin of 0.3 meters, thus constructing a virtual safety shield outside the physical collision boundary. The detection logic includes two parallel judgment conditions: First, the system traverses all spatiotemporal voxel occupancy requests; if two different pieces of equipment are found in the same time slice... Internal requests occupy the same spatial element First, it is determined to be a direct collision; second, the Euclidean distance between the center points of the two pieces of equipment within the same time frame is calculated. If this distance is less than the set safety threshold... Even if they do not occupy the same voxel, they are still judged as path conflicts with a risk of collision.

[0048] In step S43, once a conflict is detected, the system immediately initiates a priority-based conflict resolution mechanism. First, using the formula... Quantitative calculation of the mission priority weights of each conflict-affected piece of equipment In the formula, Represents the urgency of the task, used to measure the impact of task delays on the overall project duration; This represents the importance of the task and is used to measure the value attribute of the task itself. (Coefficient) and satisfy The normalization constraint is used to adjust the emphasis strategy in different scenarios. The weights of the conflicting parties are compared. The system maintains the original planned paths and schedules for high-priority equipment, ensuring the continuity of core tasks such as obstacle removal and critical transportation. For low-priority equipment, the system applies a time lag. This lag is expressed by the formula. The calculation shows that, among which For low-priority equipment, specify the path length from the current location to the predicted conflict point. This is the rated travel speed of the equipment. The system postpones the time-space occupancy requests of low-priority equipment on the timeline and re-performs conflict detection until all conflicts disappear, finally generating conflict-free control commands and issuing them to each piece of equipment for execution.

[0049] The beneficial effects of the above technical solution are as follows: Employing four-dimensional spatiotemporal grid technology transforms the complex continuous spatial collision detection problem into an efficient discrete grid index matching problem, significantly reducing the algorithm's time complexity and supporting millisecond-level real-time conflict calculation at the edge. High-resolution grid partitioning of 0.5 meters and 0.1 seconds ensures scheduling accuracy in high-risk areas such as narrow work surfaces or intersections, effectively avoiding missed detections or misjudgments caused by coarse models. A priority resolution strategy based on multi-factor weighting avoids the global efficiency degradation caused by simple "first-come, first-served" or "random avoidance," ensuring priority passage for high-value tasks. Using time lag rather than path replanning to resolve conflicts avoids increased energy consumption and wasted computing resources due to frequent path changes, ensuring the stability and robustness of the multi-machine collaborative operation system.

[0050] In the preferred embodiment, the control method for the energy consumption optimization unit in step S4 is as follows: S44. Real-time acquisition of pump outlet pressure data P and flow data Q from the hydraulic system of the excavating equipment, using the formula... The current soil hardness value H is estimated in reverse, where k is a correction factor: k=0.8 for sand and k=1.2 for clay. The operating angular velocity; S45. Obtain the current task's transport distance L and road gradient. ; S46. Query the preset 3D MAP map of "hardness, transport distance, and power". The horizontal axis of the 3D MAP map is the earthwork hardness H, the vertical axis is the transport distance L, and the vertical axis is the engine power P. Supplement the unpreset operating parameters through linear interpolation, determine the engine's optimal fuel economy operating point under the current operating conditions, and output the corresponding target speed command to the engine electronic control unit. Adjust the fuel injection quantity through the PID control algorithm to match the target power output curve. The fuel injection quantity adjustment range corresponding to the target speed command is preset according to the engine parameters.

[0051] In step S44, the energy consumption optimization unit performs physical attribute identification of the work object based on working condition back-calculation. The system first collects real-time pump outlet pressure data of the hydraulic system using a high-precision sensor installed at the outlet of the hydraulic main pump of the excavating equipment. With traffic data These two parameters directly reflect the real-time state of the hydraulic system's work on external loads, with pressure being one of them. Corresponding to excavation resistance, flow rate Corresponding to the work speed. To quantify the current soil hardness, the system introduces an estimation model based on power conservation, with the specific calculation formula as follows: In the formula, This represents the earthwork hardness value, and its physical meaning is the energy required per unit cutting depth. The instantaneous angular velocity of the boom or bucket during operation is used to eliminate the influence of operating speed on hardness assessment. This is a correction factor used to compensate for differences in friction coefficient and cohesion between different soil types. The system presets the correction factor for sandy soil conditions. The correction factor is 0.8 for clay conditions. The value is 1.2. Using this formula, the system can perceive changes in the geological hardness of the working face in real time using existing hydraulic parameters without adding dedicated geological exploration sensors.

[0052] In steps S45 and S46, the system constructs a multi-parameter coupled power matching control strategy. The system obtains the transportation distance in the current scheduling task. and road slopes in digital twin maps Combined with the soil hardness value calculated in step S44 These serve as the input index for querying the pre-set 3D MAP. This 3D MAP is an energy efficiency lookup table generated based on the engine's universal characteristic curves and a large amount of bench test data. Its horizontal axis is defined as soil hardness. The vertical axis is defined as the transportation distance. The vertical axis (i.e., the mapped value) is defined as the engine target power. Since actual operating conditions are continuously changing, the system uses a bilinear interpolation algorithm to calculate non-preset operating parameters falling between discrete grid points, thereby determining the optimal fuel economy operating point where the engine can maintain the lowest fuel consumption rate under the current operating conditions.

[0053] Determine target power The system then parses the target engine speed command into a corresponding target speed command and sends it to the engine electronic control unit (ECU). The ECU integrates a PID proportional-integral-derivative (PID) control algorithm, which calculates and adjusts the fuel injection quantity in real time based on the deviation between the target speed and the actual engine speed. The fuel injection quantity adjustment range corresponding to the target speed command strictly adheres to the boundaries of the MAP chart provided by the engine manufacturer to prevent stalling or black smoke due to over-adjustment.

[0054] The beneficial effects of the above technical solution are as follows: it innovatively proposes a soft measurement method based on hydraulic parameters to infer the earthwork hardness, solving the problem of coarse power matching caused by relying solely on experience to set gears in traditional dispatching systems. This is achieved by introducing a correction coefficient. With angular velocity This system achieves normalization for different soil types and operating habits, significantly improving the accuracy of hardness identification. Utilizing 3D MAP maps and linear interpolation technology, it achieves precise matching between engine power output and operating load (hardness) and transport load (transport distance), ensuring the engine always operates in its high-efficiency range. This avoids fuel waste caused by overpowered engines or insufficient power caused by underpowered engines, achieving ultimate energy efficiency optimization for engineering equipment under complex and variable working conditions.

[0055] In the preferred embodiment, in step S5, the multi-source sensing terminal includes a triaxial accelerometer, and the emergency response mechanism comprises the following steps: S51. Perform Fast Fourier Transform (FFT) on the time-domain vibration signal acquired by the triaxial accelerometer to extract frequency domain feature values; S52. Preset frequency domain characteristic thresholds for different fault types. For bearing wear, the amplitude threshold for the 100-200Hz frequency band is 5g-8g, and for hydraulic leakage, the amplitude threshold for the 50-100Hz frequency band is 3g-5g. Through calibration with historical fault data, if the amplitude of a certain frequency band exceeds the corresponding preset fault characteristic threshold, the local rule engine will immediately generate an alarm signal and upload it. After receiving an alarm, the S53 cloud-based dynamic priority engine locks down the associated devices around the faulty device. All engineering equipment within a circular area with a radius of 5-8m centered on the faulty device is identified as associated devices, their original tasks are suspended, and the location of the faulty device is marked as a dynamic obstacle. S54, Adopting D The Lite dynamic path planning algorithm, with its heuristic function set as follows: ,in Current position For safe evacuation points, the location of the faulty equipment is marked as a high-cost area. The navigation grid of the passable area is updated synchronously. Avoidance paths are searched in the updated cost map, and an emergency control command sequence containing steering angle and speed is generated. The maximum turning angle of the replanned path does not exceed 30°.

[0056] In step S51, the system utilizes a triaxial accelerometer installed in a multi-source sensing terminal at a key part of the engineering equipment to capture the time-domain vibration signal generated during equipment operation in real time. Since the original time-domain signal often contains a superposition of multiple components such as environmental noise, normal engine vibration, and potential fault vibration, direct analysis makes it difficult to extract fault characteristics. Therefore, the system introduces a Fast Fourier Transform (FFT) algorithm to process the time-domain signal. This algorithm converts the time-varying vibration waveform into a frequency-varying spectrum, thereby extracting frequency-domain feature values ​​containing amplitude and phase information. This process decouples the complex mixed signal into a single frequency component, enabling clear identification of the inherent frequency anomalies of specific mechanical components.

[0057] In step S52, the system pre-sets frequency domain feature threshold criteria for different fault types in the local rule engine on the edge side. Based on mechanical dynamics principles and historical fault data calibration results, the system sets the high-frequency band of 100Hz to 200Hz to correspond to bearing wear faults, with its normal amplitude safety range set to 5g to 8g, where g is the unit of gravitational acceleration; and sets the mid-frequency band of 50Hz to 100Hz to correspond to hydraulic system leakage or pump cavitation faults, with its amplitude threshold set to 3g to 5g. The local rule engine compares the current frequency domain feature values ​​with the preset thresholds in real time. Once it detects that the amplitude of a certain feature frequency band exceeds the corresponding fault feature threshold, it determines that the equipment has experienced a specific type of mechanical or hydraulic fault, and immediately generates an alarm signal containing a fault code and severity level, which is uploaded to the cloud. The system can trigger a local first-level response without waiting for instructions from the cloud.

[0058] In step S53, upon receiving an alarm signal uploaded from the edge side, the cloud-based dynamic priority engine immediately initiates the associated device locking procedure. The system defines a circular area with a radius of 5 to 8 meters as a danger isolation zone, centered on the coordinates of the faulty device. All other engineering equipment located within this circular area is identified as associated equipment. The system forcibly suspends the original tasks currently being performed by these associated devices, forcing them into standby or emergency avoidance states to prevent secondary collisions caused by obstructed vision or communication delays. Simultaneously, in the digital twin scenario, the system marks the grid location of the faulty device as a dynamic obstacle. This state not only signifies impassability but also indicates that the area carries a risk of outward spread.

[0059] In step S54, in order to quickly generate safe evacuation or detour routes, the system uses D... The Lite dynamic path planning algorithm. This algorithm, compared to the static A / B algorithm... Algorithms can handle scenarios where environmental costs change dynamically more efficiently. The core of an algorithm lies in the construction of its heuristic function; the system sets the heuristic function as follows: In the formula, Represents the position coordinates of the current search node. This represents the coordinates of a preset safe evacuation point or target point. The formula calculates the Euclidean distance between the two points, guiding the search direction to quickly approach the target point. The system updates the high-cost areas marked by faulty equipment to the global cost map and simultaneously updates the navigation grid for navigable areas. The algorithm re-searches for the optimal avoidance path in the updated map and generates a series of emergency control command sequences containing specific steering angles and speed parameters. To adapt to the kinematic constraints of heavy engineering equipment and prevent rollovers due to sharp turns, the system strictly limits the rate of change of heading between any adjacent nodes in the planned path, ensuring that the maximum turning angle does not exceed 30 degrees.

[0060] The beneficial effects of the above technical solution are as follows: Through end-side FFT frequency domain analysis, early identification and accurate classification of equipment sub-health conditions are achieved, avoiding catastrophic downtime accidents caused by "operating with defects." The use of a frequency-band threshold judgment mechanism effectively distinguishes between bearing wear and hydraulic failures, providing clear guidance for subsequent maintenance. Based on a radius-locked associated equipment management strategy, a physical isolation barrier is quickly established, preventing single-point failures from escalating into multi-machine chain accidents. Utilizing D... The Lite algorithm and Euclidean heuristic function enable millisecond-level path replanning in dynamic environments. Combined with steering angle constraints, it ensures that the generated escape path is not only theoretically the shortest but also conforms to the actual operating performance of heavy machinery, greatly improving operational safety and system robustness under extreme conditions.

[0061] Example 2 To further illustrate with reference to Embodiment 1, an intelligent scheduling system for engineering equipment includes: The multi-source sensing terminal is installed on engineering equipment. The hardware includes a CAN bus adapter, a GNSS positioning module, a three-axis accelerometer, and a hydraulic sensor. The software includes a protocol parsing plugin and a Kalman filter program. It contains a heterogeneous adaptation module and a state sensing unit, which are used to perform protocol parsing, data filtering, standardized uploading, and anomaly detection in steps S1 and S51-S52. The edge decision node is deployed on the edge server at the work site. The edge server is an industrial-grade edge server. The software includes a real-time operating system, a conflict resolution algorithm plugin, and an energy consumption control program, which includes a local rule engine, a conflict resolution module, and an energy consumption optimization unit. It is used to perform spatiotemporal conflict detection, dynamic adjustment of engine power, and issuance of control commands for steps S4, S41-S43, and S44-S46. The cloud-based dispatch hub is deployed on a cloud server cluster. The software includes a digital twin platform, a genetic algorithm engine, and a path planning module. It includes a digital twin platform, an intelligent task allocation system, and a security monitoring module. It is used to perform point cloud data processing, scene construction, genetic algorithm task allocation, global path replanning, and emergency process monitoring for steps S2-S3, S31-S34, and S53-S54. The multi-source sensing terminal communicates with the edge decision-making node via a 5G C-V2X network, while the edge decision-making node communicates with the cloud scheduling center via fiber optic Ethernet. The data exchange format is uniformly JSON, and the transmission frequency is 20Hz.

[0062] In the preferred embodiment, the intelligent task allocation system is configured to perform the construction task decomposition and improvement genetic algorithm steps; The digital twin platform is configured to perform point cloud processing and virtual-to-real calibration steps; The conflict resolution module is configured to perform a four-dimensional spatiotemporal grid conflict detection step; The energy optimization unit is configured to execute energy consumption adjustment control steps; The local rules engine works in conjunction with the multi-source sensing terminal to execute anomaly detection and alarm procedures; The security monitoring module is configured to receive alarm signals, monitor the emergency response process, and synchronously update the device status in the digital twin scenario.

[0063] The following details the specific implementation of the intelligent scheduling system for engineering equipment as described in claims 9 and 10, aiming to fully disclose the technical solution and provide clear support.

[0064] This intelligent scheduling system for engineering equipment adopts a three-tiered architecture of physical layering and logical collaboration, aiming to solve the challenges of data perception, real-time decision-making, and global optimization in complex operating environments for large-scale heterogeneous equipment clusters. The first tier of the system consists of multi-source sensing terminals installed on each excavator, loader, or dump truck. These terminals are highly integrated at the hardware level, physically connected to the equipment's electronic control unit via a CAN bus adapter, directly reading underlying operational messages; a GNSS positioning module provides high-precision latitude, longitude, and elevation information; a three-axis accelerometer captures vibration acceleration signals in the X, Y, and Z directions; and hydraulic sensors monitor the pressure and flow status of the main pump and actuators in real time. At the software level, the heterogeneous adaptation module incorporates a protocol parsing plugin, capable of identifying proprietary protocols defined by different manufacturers and converting them into standard data formats, while a Kalman filter is used to denoise the positioning data in real time. The multi-source sensing terminal is not only a data collection point but also a frontline station for anomaly detection. It is responsible for executing step S1 in claim 1 and steps S51 to S52 in claim 8, achieving a closed-loop process from underlying protocol parsing, data filtering, standardized uploading to frequency domain analysis-based anomaly detection. The beneficial effect of this implementation is that, through high hardware integration and edge-processing in software, it greatly reduces the bandwidth pressure on the communication network, ensuring that all data uploaded to upper-layer nodes is highly reliable and valid.

[0065] The second layer of the system consists of edge decision nodes deployed at the work site. These nodes utilize industrial-grade edge servers with high computing power and high protection levels, running a real-time operating system to ensure millisecond-level task response capabilities. The edge decision nodes integrate conflict resolution algorithm plugins and energy consumption control programs. A local rule engine handles sudden logical judgments, the conflict resolution module constructs a four-dimensional spatiotemporal grid and executes conflict detection logic, and the energy consumption optimization unit performs power matching calculations based on soil hardness. This node primarily undertakes the calculation tasks of step S4 and its sub-steps S41 to S46, directly performing spatiotemporal conflict detection, dynamic engine power adjustment, and the generation and issuance of control commands on the field side. The advantage of this implementation is that it pushes the real-time-critical collision avoidance control and energy consumption adjustment logic down to the edge side, avoiding control lag caused by cloud communication delays. This ensures that even in extreme situations such as network fluctuations or outages, the field equipment can still maintain basic safe operation and energy efficiency control capabilities.

[0066] The third layer of the system is a cloud-based scheduling hub deployed in a remote data center. This hub is built on a cloud server cluster, possessing massive storage space and parallel computing capabilities. The software platform integrates a digital twin platform, a genetic algorithm engine, and a path planning module. The digital twin platform utilizes UAV point cloud data to construct a high-fidelity 3D virtual scene and is responsible for virtual-real mapping calibration; the intelligent task allocation system uses an improved genetic algorithm to handle complex combinatorial optimization problems and generate globally optimal scheduling schemes; and the security monitoring module is responsible for emergency monitoring and status updates throughout the entire process. The cloud-based scheduling hub focuses on executing steps S2 to S3 and their sub-steps S31 to S34, as well as steps S53 to S54 in step S5. The beneficial effect of this implementation is that it utilizes the unlimited computing power resources of the cloud to solve computationally intensive tasks such as time-consuming terrain modeling, large-scale population evolution, and global path replanning, achieving an organic combination of globally optimal macro-scheduling and real-time precise micro-control.

[0067] In terms of communication links and data interaction, the multi-source sensing terminals and edge decision nodes communicate via a 5G C-V2X network. This network features low latency, high reliability, and support for high-speed mobility, perfectly meeting the dynamic operational needs of engineering equipment. The edge decision nodes and the cloud-based scheduling center are connected via fiber optic Ethernet, ensuring stable transmission of massive point cloud data and scheduling commands. The entire system uses the JSON standard for data interaction, eliminating data barriers between heterogeneous systems. The system is set to a transmission frequency of 20Hz, meaning a full state update is performed every 50 milliseconds, ensuring the synchronization accuracy of the digital twin scenario. In the optimized scheme, each functional module and specific algorithm step is strictly configured and bound: the intelligent task allocation system specifically executes the construction task decomposition and improved genetic algorithm steps; the digital twin platform focuses on point cloud processing and virtual-real calibration; the conflict resolution module performs four-dimensional spatiotemporal grid detection; the energy consumption optimization unit performs adjustment control based on the MAP graph; the local rule engine works with the sensing terminals to perform anomaly detection; and the safety supervision module is responsible for the global coordination of emergency response. This modular and service-oriented system design not only improves the maintainability and scalability of the system, but also ensures that each technical feature has a specific physical embodiment, making the technical solution protected by the claims fully engineering feasible.

[0068] Example 3 Further explanation in conjunction with Example 1, such as Figure 1 As shown, this embodiment provides an intelligent scheduling system and method for engineering equipment based on digital twins and improved genetic algorithms, applicable to large-scale open-pit mines or land reclamation projects. The system adopts a three-level collaborative architecture of edge-cloud, such as... Figure 1The diagram shows the overall architecture of the intelligent dispatching system for engineering equipment. The system's physical and logical layers correspond to each other, encompassing multi-source sensing terminals installed on the engineering equipment, edge decision nodes deployed at the work site, and a cloud-based dispatching hub deployed in a remote data center. At the edge, each excavator or dump truck is equipped with a multi-source sensing terminal. This terminal integrates a CAN bus adapter, a GNSS positioning module, a three-axis accelerometer, and a hydraulic sensor. For equipment from different brands used on-site, the heterogeneous adaptation module within the terminal is pre-configured with CAN2.0 and Modbus RTU communication protocol libraries. The module automatically identifies the device type by reading the message header and calls the corresponding decoding rules, converting non-standard underlying data into a unified JSON format. The status sensing unit collects data in real time at a frequency of 20Hz. The collected data includes high-precision GNSS position data smoothed by a Kalman filter algorithm, hydraulic main pump pressure and flow data, and three-axis vibration data of the vehicle body. This data is uploaded to the edge decision nodes with low latency via a 5G C-V2X network. Edge decision nodes utilize industrial-grade edge servers running a real-time operating system. They receive high-frequency data from the terminal side and interact with the cloud via fiber optic Ethernet, primarily responsible for millisecond-level real-time control. The cloud scheduling hub is deployed in a server cluster, handling computationally intensive tasks.

[0069] Before system operation or when significant terrain changes occur, the cloud-based dispatch center constructs the operational scenario. The system uses a drone equipped with a LiDAR to scan the work area along a preset flight path at a height of 5 to 10 meters, with a scanning frequency set to 10Hz. The acquired raw point cloud data is first processed using a statistical outlier removal algorithm to eliminate noise caused by airborne dust. Then, the Delaunay triangulation algorithm is used to convert the purified point cloud into an irregular triangular mesh model and extract the terrain's elevation and slope features, generating a navigation mesh for navigable areas with physical semantics. The system maps the physical dimensions and kinematic parameters of the engineering equipment to the virtual scene. To ensure… Figure 1 The digital twin platform is consistent with the coordinates of the physical world. The system obtains the measured GNSS coordinates of more than three fixed reference points on site, calculates the rotation and translation matrix using the iterative nearest point algorithm, and performs rigid body transformation calibration on the virtual scene coordinate system to eliminate positioning deviation.

[0070] like Figure 2 The diagram shown illustrates the improved genetic algorithm task allocation process. The cloud-based intelligent task allocation system first obtains the total volume of materials in the work area. and the average bucket capacity of excavating equipment Through formula Calculate the theoretical number of assignments and break down the task into... Each system comprises atomic task units containing sequences of actions for digging, loading, transporting, and unloading. The system uses a real-number encoding method based on task sequences to randomly generate an initial population of 50 to 100, with each chromosome representing an assignment sequence for one atomic task. The system defines a fitness function. The candidate solutions are input into the digital twin scenario, and a virtual clock is started to simulate discrete events and calculate the completion time. Total energy consumption and load variance The fitness value is calculated. The algorithm enters the crossover and mutation phase, and the system calculates the fitness value based on the current iteration number. Dynamically adjust the cross rate and variability Adjust the formula to and Initially, a high crossover rate is maintained to expand the search, while later the crossover rate is reduced to promote convergence. This continues until the maximum number of iterations is reached. The algorithm terminates when the fitness does not improve for 10 consecutive generations, and outputs the optimal scheduling scheme to the edge node.

[0071] Edge decision nodes perform real-time control based on the issued plan; this process involves... Figure 3 The conflict resolution process and energy consumption optimization logic are described. Edge nodes construct a four-dimensional spatiotemporal mesh with a resolution of 0.5 meters by 0.5 meters by 0.5 meters and a time slice of 0.1 seconds. The system maps the paths of each piece of equipment to spatiotemporal occupancy requests. It monitors in real time whether two pieces of equipment request to occupy the same voxel or if the spatial distance is less than the safety threshold determined by adding a 0.3-meter safety margin to the equipment width. If a conflict is found, then priority weights are calculated. Keep the paths of high-priority equipment unchanged, and impose restrictions on low-priority equipment based on the conflict distance. Divide by the rated speed Calculated time lag .like Figure 3 As shown, the updated time-space reservation request eliminated conflicts and generated the final control command for issuance. Regarding energy consumption optimization, the edge node acquires the excavator pump outlet pressure in real time. and traffic Using the formula Back-propagation soil hardness Among them, the sand correction factor Take 0.8 as the clay correction factor. The value is set to 1.2. The system then queries the preset three-dimensional MAP map of hardness, transport distance, and power, determines the optimal power point under the current operating condition through linear interpolation, and adjusts the engine fuel injection quantity through a PID algorithm to achieve closed-loop energy consumption control based on operating conditions.

[0072] When an exception occurs during the operation, the system triggers, such as Figure 4 The emergency response process is shown. The vibration signal from the triaxial accelerometer is subjected to a Fast Fourier Transform (FFT) at the end. If the amplitude in the 100Hz to 200Hz frequency band exceeds the bearing wear characteristic threshold of 5g to 8g, or the amplitude in the 50Hz to 100Hz frequency band exceeds the hydraulic leakage characteristic threshold of 3g to 5g, the local rule engine immediately generates an alarm and uploads it. Upon receiving the alarm, the cloud-based dynamic priority engine locks down related equipment within a 5-8 meter radius of the faulty equipment, suspends their existing tasks, and marks the fault point as a high-cost dynamic obstacle in the digital twin map. The cloud uses D... The Lite dynamic path planning algorithm uses Euclidean distance as a heuristic function. The system searches for avoidance paths in the updated cost map. The generated paths are strictly limited to a maximum turning angle of 30 degrees. Ultimately, emergency control commands are issued to relevant equipment, directing them to safely evacuate or detour. Through this process, the system achieves a closed-loop process from microscopic sensor data acquisition to macroscopic cluster scheduling optimization. Figures 1 to 4 The interaction logic and business processes of each module of the system are clearly demonstrated.

[0073] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for intelligent scheduling of engineering equipment, characterized by the following steps: include: S1. The edge-side multi-source sensing terminal reads the CAN bus and sensor data of the engineering equipment through the heterogeneous adaptation module, uses the protocol parsing algorithm to complete the data standardization, collects the operating data of the engineering equipment in real time and uploads it to the edge decision node. S2. The cloud-based dispatch center uses point cloud data of the work area acquired by drones to generate a 3D terrain model through statistical outlier removal filtering and Delaunay triangulation mesh construction algorithm. It also combines engineering equipment parameters to construct a digital twin scene and completes virtual-real mapping calibration through iterative nearest point algorithm. S3, the cloud-based intelligent task allocation system decomposes the total amount of construction tasks into discrete atomized work units based on the rated operating capacity of the engineering equipment. It uses an improved genetic algorithm with an adaptive crossover and mutation mechanism to generate candidate scheduling schemes and inputs the schemes into a digital twin scenario. Through discrete event simulation, it infers the operating status of each piece of equipment, records the completion time and energy consumption data, selects the optimal scheduling scheme, and sends it to the edge decision nodes. S4. The edge decision node receives the optimal scheduling scheme, establishes a spatiotemporal occupancy table through the conflict resolution module, and performs multi-machine collaborative control of the engineering equipment in combination with the load mapping logic of the energy consumption optimization unit. S5. When the multi-source sensing terminal on the edge detects abnormal data of the working environment or engineering equipment that exceeds the safety threshold, the emergency response mechanism is triggered. The cloud scheduling center uses a dynamic path planning algorithm to perform global path replanning and updates the control commands through the edge decision nodes.

2. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: In step S1, the heterogeneous adaptation module pre-configures a communication protocol library for multi-brand engineering equipment. The communication protocol library includes CAN 2.0 and Modbus RTU protocols. It identifies the device type through message header identification and calls the corresponding decoding rules to convert non-standard CAN messages or Modbus data into unified JSON format data. The status sensing unit collects data at a frequency of 20Hz. The data includes GNSS position data, hydraulic system pressure data, and triaxial vibration data of the engineering equipment. After smoothing the GNSS position data through a Kalman filter algorithm, it is transmitted to the edge decision node through the 5G C-V2X network.

3. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: in step S2: S21. Use a drone equipped with a lidar to scan the work area along a preset flight path at a height of 5-10m with a scanning frequency of 10Hz to obtain raw point cloud data, and use a statistical outlier removal algorithm to remove noise. S22. The Delaunay triangulation algorithm is used to convert the processed point cloud data into an irregular triangular mesh model, extract the elevation and slope features of the terrain, and generate a navigation mesh for passable areas. S23. Map the physical dimensions, kinematic parameters, and material distribution attributes of engineering equipment to a three-dimensional virtual scene to construct a multi-level virtual environment that includes terrain, equipment, and material layers. S24. Obtain the measured GNSS coordinates of more than 3 fixed reference points on site, use the iterative nearest point algorithm to calculate the rotation and translation matrix between the corresponding feature points in the virtual scene and the measured coordinates, and perform rigid body transformation calibration on the coordinate system of the 3D virtual scene.

4. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: In step S3, the specific method for decomposing the construction task into atomized work units is as follows: obtain the total material volume V of the task area and the average bucket capacity of the excavating equipment participating in the operation. Calculate the number of theoretical assignments ,in It is a rounding function; The atomic task allocation rule is as follows: based on the real-time load rate and operating efficiency of each piece of engineering equipment, a round-robin allocation combined with a load balancing strategy is adopted to allocate N atomic task units containing four action sequences of excavation, loading, transportation and unloading to each piece of equipment, ensuring that the atomic task sequence of a single piece of equipment is continuous and without idle intervals.

5. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: Step S3, which involves generating candidate scheduling schemes using an improved genetic algorithm, includes: S31. Initialize the population using a real number encoding method based on task sequences. Set the population size to 50-100. Randomly generate task allocation sequences to ensure that the total number of tasks for each chromosome is equal to the total number of atomic task units. S32. Define the fitness function. The specific expression of the fitness function is as follows: ,in , , , The weights for time, energy consumption, and load variance are determined using the analytic hierarchy process (AHP) based on construction priority. To maximize the completion time, Total energy consumption, For equipment load variance; S33. Evolutionary operation: During the evolution process, based on the current iteration number k and the maximum iteration number... Dynamically adjust the cross rate and variability Adjust the formula to , ,in The initial crossover rate is 0.8-0.

9. The termination crossover rate is 0.4-0.

5. The initial mutation rate was 0.05-0.

1. The termination mutation rate is 0.01-0.02; S34. Simulation screening: For the generated candidate scheduling schemes, a virtual clock is started in the digital twin scenario. The action sequence of atomic tasks is used as the event unit. The time parameters of each action are set in combination with the rated parameters of the engineering equipment. The equipment operation is simulated through event queue scheduling. The completion time and energy consumption data at the end of the simulation are recorded as the basis for fitness evaluation. The evolution termination condition is that the number of iterations reaches a certain threshold. If the maximum value of the fitness function does not improve for 10 consecutive generations, select the scheduling scheme with the best fitness.

6. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: In step S4, the specific method for control via the conflict resolution module is as follows: S41. Construct a four-dimensional spatiotemporal grid for the work area. Where i, j, k, l are grid indices, the grid resolution of the spatial dimension (x, y, z) is set to 0.5m × 0.5m × 0.5m, and the time slice of the time dimension t is set to 0.1s, mapping the planned path of each piece of engineering equipment into a series of spatiotemporal voxel occupancy requests; S42. Collision Detection, Safety Threshold Based on the maximum width of the engineering equipment + a safety margin of 0.3m, if two pieces of equipment are detected operating at the same time... Internal requests occupy the same spatial element Or the spatial distance is less than the safety threshold If so, it is determined to be a path conflict; S43. Conflict resolution: Calculate the priority weight of each conflicting equipment mission. ,in To assess the urgency of the task, As for the importance of the task, Keep the paths of high-priority equipment unchanged, and apply a time lag to low-priority equipment. Where d is the distance between the conflict point and the current position of the low-priority equipment, v is the rated driving speed of the equipment, its spacetime occupancy request is updated until the conflict disappears, and a conflict-free control command is generated and issued.

7. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: In step S4, the control method for the energy consumption optimization unit is as follows: S44. Real-time acquisition of pump outlet pressure data P and flow data Q from the hydraulic system of the excavating equipment, using the formula... The current soil hardness value H is estimated in reverse, where k is a correction factor: k=0.8 for sand and k=1.2 for clay. The operating angular velocity; S45. Obtain the current task's transport distance L and road gradient. ; S46. Query the preset "hardness, transport distance, power" 3D MAP. The horizontal axis of the 3D MAP is earthwork hardness H, the vertical axis is transport distance L, and the vertical axis is engine power P. Supplement the unpreset operating parameters through linear interpolation, determine the engine's optimal fuel economy operating point under the current operating conditions, and output the corresponding target speed command to the engine electronic control unit. Adjust the fuel injection quantity through PID control algorithm to match the target power output curve. The fuel injection quantity adjustment range corresponding to the target speed command is preset according to the engine parameters.

8. The intelligent scheduling method for engineering equipment according to claim 1, characterized in that: In step S5, the multi-source sensing terminal includes a three-axis accelerometer, and the emergency response mechanism consists of the following steps: S51. Perform Fast Fourier Transform (FFT) on the time-domain vibration signal acquired by the triaxial accelerometer to extract frequency domain feature values; S52. Preset frequency domain characteristic thresholds for different fault types. For bearing wear, the amplitude threshold for the 100-200Hz frequency band is 5g-8g, and for hydraulic leakage, the amplitude threshold for the 50-100Hz frequency band is 3g-5g. Through calibration with historical fault data, if the amplitude of a certain frequency band exceeds the corresponding preset fault characteristic threshold, the local rule engine will immediately generate an alarm signal and upload it. After receiving an alarm, the S53 cloud-based dynamic priority engine locks down the associated devices around the faulty device. All engineering equipment within a circular area with a radius of 5-8m centered on the faulty device is identified as associated devices, their original tasks are suspended, and the location of the faulty device is marked as a dynamic obstacle. S54, Adopting D The Lite dynamic path planning algorithm, with its heuristic function set as follows: ,in Current position For safe evacuation points, the location of the faulty equipment is marked as a high-cost area. The navigation grid of the passable area is updated synchronously. Avoidance paths are searched in the updated cost map, and an emergency control command sequence containing steering angle and speed is generated. The maximum turning angle of the replanned path does not exceed 30°.

9. An intelligent scheduling system for engineering equipment, characterized in that, The system adopts a three-level collaborative architecture of end, edge, and cloud, and is used in conjunction with the intelligent scheduling method for engineering equipment as described in any one of claims 1-8. The system includes: The multi-source sensing terminal is installed on engineering equipment. The hardware includes a CAN bus adapter, a GNSS positioning module, a three-axis accelerometer, and a hydraulic sensor. The software includes a protocol parsing plugin and a Kalman filter program. It contains a heterogeneous adaptation module and a state sensing unit, which are used to perform protocol parsing, data filtering, standardized uploading, and anomaly detection in steps S1 and S51-S52. The edge decision node is deployed on the edge server at the work site. The edge server is an industrial-grade edge server. The software includes a real-time operating system, a conflict resolution algorithm plugin, and an energy consumption control program, which includes a local rule engine, a conflict resolution module, and an energy consumption optimization unit. It is used to perform spatiotemporal conflict detection, dynamic adjustment of engine power, and issuance of control commands for steps S4, S41-S43, and S44-S46. The cloud-based dispatch hub is deployed on a cloud server cluster. The software includes a digital twin platform, a genetic algorithm engine, and a path planning module. It includes a digital twin platform, an intelligent task allocation system, and a security monitoring module. It is used to perform point cloud data processing, scene construction, genetic algorithm task allocation, global path replanning, and emergency process monitoring for steps S2-S3, S31-S34, and S53-S54. The multi-source sensing terminal communicates with the edge decision-making node via a 5G C-V2X network, while the edge decision-making node communicates with the cloud scheduling center via fiber optic Ethernet. The data exchange format is uniformly JSON, and the transmission frequency is 20Hz.

10. The intelligent scheduling system for engineering equipment based on digital twins and improved genetic algorithms according to claim 9, characterized in that, The intelligent task allocation system is configured to perform construction task decomposition and improvement genetic algorithm steps; The digital twin platform is configured to perform point cloud processing and virtual-to-real calibration steps; The conflict resolution module is configured to perform a four-dimensional spatiotemporal grid conflict detection step; The energy optimization unit is configured to execute energy consumption adjustment control steps; The local rules engine works in conjunction with the multi-source sensing terminal to execute anomaly detection and alarm procedures; The security monitoring module is configured to receive alarm signals, monitor the emergency response process, and synchronously update the device status in the digital twin scenario.

Images