Construction area compaction method and system

Abstract

Embodiments of the present application disclose a construction area compaction method and system, and a storage medium. The method comprises obtaining geological parameters of a construction area and setting target compaction parameters according to target engineering of the construction area; before compaction of the construction area, inputting the geological parameters, the target compaction parameters and construction parameters supported by a rolling device into an artificial intelligence (AI) model to obtain first rolling planning parameters; after the end of the nth stage, detecting actual compaction parameters; inputting the actual compaction parameters, a stage compaction target of the nth stage and construction parameters supported by a current rolling device into the AI model to determine a minimum number of stages N0 that need to be adjusted; and using a parameter adjustment strategy corresponding to the N0 value to complete all subsequent stage compaction targets of the original planning as a hard constraint, and at least adjust construction parameters of the nth+1 stage.

Description

Technical Field

[0001] This invention relates to the field of compaction construction technology, and in particular to a method and system for compacting a construction area. Background Technology

[0002] In infrastructure projects such as water conservancy (e.g., reservoir dams) and transportation engineering (e.g., roadbeds), compaction degree is a core quality indicator determining the stability, impermeability, and bearing capacity of the engineering structure. Compaction equipment (e.g., vibratory rollers) is the core device for compacting the media (soil, rock, roadbed fill, etc.) in the construction area. In actual construction, to meet the demands of large-area, high-efficiency compaction, multiple compaction devices are usually required to work together to form a compaction working surface covering the entire construction area. The current quality control model is process control plus result sampling. That is, parameters are set, rolling is carried out, and then the results are verified by testing to see if they meet the standards. If they do not meet the standards, remedial measures are taken. This is a passive and reactive control method. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method and system for compacting a construction area. The technical solution of the present invention is implemented as follows: A first aspect provides a method for compacting a construction area, characterized in that the method includes: acquiring geological parameters of the construction area and setting target compaction parameters according to the target project of the construction area; before compacting the construction area, inputting the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment into an artificial intelligence (AI) model to obtain first compaction planning parameters; the first compaction planning parameters include construction parameters for M stages; the construction parameters for each stage include a stage compaction target and stage construction parameters of the compaction equipment; after the nth stage, detecting the actual compaction parameters; inputting the actual compaction parameters, the stage compaction target of the nth stage, and the construction parameters supported by the current compaction equipment into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted; adopting a parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent stage compaction targets of the original plan as a hard constraint, adjusting the construction parameters of at least the (n+1)th stage.

[0004] The second aspect provides a construction area compaction system, characterized in that the system comprises: an acquisition module, used to acquire geological parameters of the construction area and set target compaction parameters according to the target project of the construction area; a planning module, used to input the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment into an artificial intelligence (AI) model before compaction of the construction area to obtain first compaction planning parameters; the first compaction planning parameters include construction parameters for M stages; the construction parameters for each stage include a stage compaction target and stage construction parameters of the compaction equipment; a detection module, used to detect the actual compaction parameters after the nth stage; a prediction module, used to input the actual compaction parameters, the stage compaction target of the nth stage, and the construction parameters supported by the current compaction equipment into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted; and an optimization module, used to adopt a parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent stage compaction targets of the original plan as a hard constraint, to adjust the construction parameters of at least the (n+1)th stage.

[0005] A third aspect provides a computer-readable storage medium storing computer-executable instructions; the computer-executable instructions, when executed by a processor, are capable of implementing the method provided by any of the technical solutions of the first aspect.

[0006] The technical solution provided in this disclosure has the following effects: Traditional methods rely on fixed parameters and human experience, making it difficult to cope with changes in geological conditions and prone to uneven quality. This disclosed embodiment uses an AI model to generate a scientific multi-stage compaction plan based on geological conditions and engineering objectives. More importantly, it introduces real-time monitoring and feedback after each stage, with the AI ​​model evaluating deviations and intelligently adjusting strategies based on the principle of minimizing disturbances and ensuring the final goal. This dynamic closed-loop control based on real-time data enables the construction process to continuously optimize towards the design objective, significantly improving the overall uniformity of compaction results and compliance with design requirements, fundamentally enhancing the long-term reliability of the project.

[0007] While ensuring quality, the AI ​​model's adjustment decisions are geared towards minimizing adjustment phases, striving to correct deviations at the lowest cost. It provides a holistic assessment and intelligently selects the most economical strategy, moving from single-step adjustments to global replanning, thus avoiding unnecessary total shutdowns or rework caused by localized issues. This not only ensures construction continuity but also reduces energy waste and equipment wear caused by over- or under-voltage conditions, achieving comprehensive optimization of efficiency, cost, and resource consumption.

[0008] This disclosure deeply embeds an AI model into the control loop, enabling it to undertake core planning and decision-making functions. The AI ​​model can predict the long-term impact of different adjustment strategies based on real-time data, making the optimal decision to ensure the final goal. This endows the construction system with a high degree of intelligence and adaptability, allowing it to proactively respond to on-site uncertainties, reducing reliance on human experience and the risk of human error, and promoting the standardization and intelligentization of construction management. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of a construction area compaction system provided in an embodiment of the present invention; Figure 2 A schematic flowchart of a construction area compaction method provided in an embodiment of the present invention; Figure 3 This is a structural schematic diagram of a construction area compaction system provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device for a light-construction area compaction system provided in an embodiment of the present invention. Detailed Implementation

[0010] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0011] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0012] like Figure 1As shown in the figure, this disclosure provides a method for compacting a construction area, which is executed by a compaction system including multiple compaction devices. Each compaction device is equipped with multiple sensors, specifically, these sensors may include a vibration sensor, a positioning module, and a compaction weight detection module. The vibration sensor is used to collect the excitation signal of the compaction device and the environmental vibration signal. The positioning module is used to locate the position of the compaction device. The compaction weight detection module is used to collect the actual compaction weight of the compaction device. The multiple compaction devices are connected by a network to form a collaborative sensing network.

[0013] In some embodiments, these compaction devices may be connected to a mobile communication network, and the following construction area compaction method may be executed by an edge computing server, or the construction area compaction method of any of the following embodiments may be executed by a local server that is locally connected to a collaborative sensing network.

[0014] like Figure 2 As shown in the embodiments of this disclosure, a method for compacting a construction area is provided, the method comprising: S1110: Obtain the geological parameters of the construction area and set the target compaction parameters according to the target project of the construction area.

[0015] In some embodiments, geological parameters are the core fundamental factors determining the compaction effect. Their acquisition requires a combination of on-site investigation and testing to ensure their authenticity and representativeness. In this embodiment, geological parameters obtained through drilling sampling, in-situ testing, or querying geological databases include, but are not limited to, the type of soil in the construction area (such as silty clay, gravelly soil), natural moisture content, dry density, void ratio, and soil layer distribution thickness.

[0016] The target compaction parameters are set according to the engineering specifications and usage requirements of the construction area. Taking highway subgrade construction as an example, the target compaction parameters specifically refer to the compaction degree standards for different depth areas of the subgrade. For example, the compaction degree of the 0-80cm area below the road surface is ≥96%, the compaction degree of the 80-150cm area is ≥94%, the compaction settlement limit, and the compaction uniformity requirements, etc. Geological parameters directly determine the suitability of the construction parameters of the compaction equipment (e.g., clay soil requires higher excitation force), while the target compaction parameters are the core basis for subsequent construction planning and quality judgment. Together, they constitute the basic input for compaction parameter planning.

[0017] S1120: Before compacting the construction area, the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment are input into the artificial intelligence (AI) model to obtain the first compaction planning parameters; the first compaction planning parameters include construction parameters for M stages; the construction parameters for each stage include the stage compaction target and the stage construction parameters of the compaction equipment.

[0018] In some embodiments, S1120 can achieve accurate pre-planning of compaction parameters through AI models, avoiding the problems of poor adaptability and large quality fluctuations in traditional experience-based planning.

[0019] The construction parameters supported by the compaction equipment are the inherent attributes and adjustable range parameters of the equipment, specifically including the model of the compaction equipment, the adjustable range of excitation force (such as 200-400kN), the adjustable range of compaction speed (such as 2-6km / h), the width of the compaction wheel, and the maximum number of compaction passes.

[0020] The AI ​​model used in this embodiment is a deep learning model trained on a large number of compaction project samples. Its training samples cover the correspondence between compaction parameters and compaction effects under different geological conditions and target compaction requirements. After inputting the above three types of parameters into the AI ​​model, the model learns the mapping rules between geological conditions, equipment parameters, and compaction effects from historical data, and outputs the first compaction planning parameters adapted to the current construction scenario. Considering the differences in compaction difficulty at different depths and in different areas of the construction zone, the first compaction planning parameters adopt a phased planning mode (i.e., M phases). M can be any positive integer greater than 3, and the specific value of M can be estimated based on the project type and the hardware parameters of the compaction equipment.

[0021] For example, for the compaction of a 120cm thick roadbed, M=3 stages are set. The first stage (0-40cm) has a compaction target of ≥90%, and the stage construction parameters are a rolling speed of 3km / h, a vibration force of 300kN, and 4 rolling passes. The second stage (41-80cm) has a compaction target of ≥93%, and the stage construction parameters are a rolling speed of 2.5km / h, a vibration force of 350kN, and 5 rolling passes. The third stage (81-120cm) has a compaction target of ≥96%, and the stage construction parameters are a rolling speed of 2km / h, a vibration force of 400kN, and 6 rolling passes. Through the phased progressive compaction, the final target compaction parameters are gradually approached.

[0022] In some embodiments, the AI ​​model can be various deep learning models, such as CNN, LSTM, and fully connected hybrid deep learning models. This model combines feature extraction and temporal prediction capabilities, adapting to the temporal nature of compacted parameters and the requirements for multi-dimensional feature fusion. The specific structure from input to output is as follows: Input layer: Receives three types of structured data: actual compaction parameters, stage-specific compaction target for stage n, and construction parameters supported by the current compaction equipment; Feature extraction layer: A two-layer CNN (Convolutional Neural Network) structure is used, with the first layer's kernel size larger than the second layer's kernel size, to extract local key features from various parameters (e.g., the correlation between compaction deviation and equipment excitation force). 3. Temporal modeling layer: A one-layer LSTM (Long Short-Term Memory) network with a number of hidden layer units is set up to capture the temporal correlation patterns of compaction parameters as construction progresses (e.g., the cumulative impact of early compaction deviation on subsequent stages). Fusion and Output Layer: The local features extracted by the CNN and the temporal features output by the LSTM are fused through a fully connected layer. The optimal adjustment scheme is output through the Softmax activation function. The final output dimension is the construction parameter adjustment value and the number of adjustment stages N0 corresponding to the stage to be adjusted.

[0023] In some embodiments, after the parameters are input into the model, they need to undergo multiple preprocessing and feature transformation steps. The specific process is as follows: Data preprocessing: The three types of input parameters are normalized (Min-Max normalization is used to map the parameters to the [0,1] interval) to eliminate the influence of differences in the magnitude of different parameters (such as the excitation force kN and the compaction percentage); the unstructured related data (such as vibration waveform data collected by sensors) are feature encoded and converted into structured feature vectors. Outlier filtering: Abnormal data (such as extreme data caused by momentary sensor failure) in the actual compaction parameters collected by the sensor are removed by the 3σ criterion to ensure the reliability of the input data; Feature extraction and fusion: The preprocessed parameters are first input into the CNN layer, and local feature matrices are extracted through convolution and pooling operations (using max pooling). Then, the feature matrices are input into the LSTM layer, and key temporal features are filtered and memorized through gating units (input gate, forget gate, output gate), and a temporal feature vector is output. Finally, the two types of features are fused through concatenation to obtain a comprehensive feature vector with a dimension of 512. Optimization and Output: The integrated feature vector is input into the fully connected layer and iteratively trained by the Adam optimizer (initial learning rate of 0.001, decaying by 0.1 every 10 rounds). The loss function that minimizes the deviation between the compaction effect corresponding to the adjusted parameters and the target value is minimized. Finally, the minimum number of adjustment stages N0 and the corresponding parameter adjustment range for each stage are output.

[0024] In some embodiments, the detection of actual compaction parameters at each stage is accomplished using integrated sensors integrated into the compaction equipment. Specifically, three types of sensors are integrated on the compaction wheel of the equipment: an integrated compaction degree sensor (installed inside the compaction wheel to obtain compaction degree by detecting stress wave signals from the contact between the compaction wheel and the soil), a vibration sensor (installed at the end of the compaction wheel shaft to collect vibration frequency and amplitude data in real time during the compaction process), and a displacement sensor (installed on the equipment body to monitor settlement displacement during the compaction process). These sensors communicate with the equipment's central control system in real time. After each stage, the system automatically summarizes the raw data collected by the sensors, performs simple filtering preprocessing, and generates actual compaction parameters. This eliminates the need for additional independent detection equipment, improving detection efficiency and data timeliness.

[0025] S1130: After the nth stage, check the actual compaction parameters.

[0026] In some embodiments, the detection of actual compaction effects provides data support for subsequent parameter adjustments, ensuring that construction quality remains within a controllable range. Here, n is a positive integer, and 1 ≤ n < M. This embodiment uses n=2 (i.e., the end of the second stage) as an example. The detection of actual compaction parameters must employ testing methods that comply with engineering specifications to ensure the accuracy of the test results. Specifically, this can be achieved using equipment such as compaction degree testers and settlement observation instruments. The testing range covers the entire area compacted in the second stage, using a random sampling method (sampling frequency not less than 3 test points per 100㎡). The actual compaction parameters obtained include the actual compaction degree, actual settlement, and / or compaction uniformity deviation value of each test point in this stage. After the test is completed, the actual compaction parameters are initially compared with the stage compaction target of the second stage to determine whether there is a compaction quality deviation, laying the foundation for the parameter input of the subsequent AI model.

[0027] S1140: Input the actual compaction parameters, the phased compaction target of the nth stage, and the construction parameters supported by the current rolling equipment into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted.

[0028] In S1140, the data analysis and optimization capabilities of the AI ​​model are used to minimize the adjustment range of construction parameters under the hard constraints of ensuring the total construction period remains unchanged and the final compaction target is achieved, thereby reducing the difficulty and cost of construction management. Specifically, the AI ​​model first calculates the degree of deviation and the reasons for the deviation between the actual compaction parameters and the stage-specific compaction target of the nth stage.

[0029] For example, in this embodiment, the average actual compaction degree of the second stage was detected to be 91%, which is lower than the stage compaction target of 93%. The deviation was analyzed by the model as insufficient excitation force and too few rolling passes. Subsequently, the AI ​​model, with the core constraint of completing the original rolling plan on schedule (i.e., the total construction period is consistent with the original plan), combined with the adjustable range of construction parameters supported by the current rolling equipment, simulated the compaction effect of subsequent stages under different adjustment schemes: if only the construction parameters of the third stage (n+1 stage) are adjusted, can the quality deviation of the second stage be made up for and the final goal be achieved? If only the third stage is adjusted, the requirements cannot be met, then the parameters of the third and fourth stages (if they exist) need to be simulated and adjusted, and so on.

[0030] Through simulation and iteration of multiple schemes, the AI ​​model ultimately selects the optimization scheme with the fewest stages to be adjusted, and determines the minimum number of stages N0 that need to be adjusted. For example, in this embodiment, the model simulation shows that adjusting only the construction parameters of the third stage can make up for the deviation and complete the original plan, so N0=1 is determined.

[0031] S1150: Adopt the parameter adjustment strategy corresponding to the value of N0, with the timely completion of all subsequent phase compaction targets of the original plan as a hard constraint, and adjust the construction parameters of at least the (n+1)th phase.

[0032] In some embodiments, a targeted adjustment strategy is formulated based on the specific value of N0. The core principle is to compensate for deviations in the early stages of construction by precisely adjusting construction parameters under the hard constraint of completing all subsequent stage compaction targets on schedule, while avoiding excessive adjustments that could lead to resource waste or project delays. In this embodiment, with N0=1, the corresponding parameter adjustment strategy is to adjust the construction parameters only for the (n+1)th stage (i.e., the 3rd stage), without affecting other subsequent stages (in this embodiment, M=3, and there are no other subsequent stages). Specific adjustments may include: adjusting the stage construction parameters of the 3rd stage based on the optimization results simulated by the AI ​​model, within the adjustable range of the compaction equipment's construction parameters. This ensures that after this stage is completed, not only will its own stage compaction target (compaction degree ≥96%) be achieved, but the compaction degree deviation of the 2nd stage will also be compensated, so that the overall compaction quality meets the overall target requirements. For example, suppose the construction parameters for stage 3 are adjusted to a compaction speed of 1.8 km / h, a vibration force of 400 kN, and 8 compaction passes. After adjustment, the AI ​​model verifies that these parameter settings can achieve a compaction degree of 97% or higher in stage 3, while simultaneously driving secondary compaction reinforcement in the compacted area of ​​stage 2, ultimately ensuring that the overall compaction quality meets the engineering requirements. If N0=2 (i.e., stages n+1 and n+2 need to be adjusted), the adjustment strategy must consider the synergy of the parameters in both stages. For example, the adjustment range of stage n+1 can be appropriately reduced, while the parameters of stage n+2 can be slightly optimized to ensure that the two stages work together to achieve the subsequent compaction goals, and the total construction period does not exceed the original plan.

[0033] In some embodiments, the AI ​​model solves for optimal compaction parameters by minimizing an objective function J based on geological parameters, target compaction parameters, and equipment parameters. These compaction parameters include at least compaction planning parameters, such as the number of compaction passes, compaction speed, vibration frequency, and / or amplitude.

[0034] In some embodiments, when a deviation between the actual compaction parameters and the target is detected after the nth stage, the AI ​​model will rerun the optimization algorithm, using the timely completion of all subsequent planned compaction targets as a hard constraint, and adjust the construction parameters for subsequent stages.

[0035] In some embodiments, the AI ​​model seeks the optimal balance between compaction quality (compaction degree, permeability coefficient, foundation bearing capacity, resilient modulus) and construction efficiency (compaction energy) by adjusting weighting coefficients, thereby generating a construction scheme that meets both quality requirements and is cost-effective.

[0036] In summary, the embodiments disclosed herein naturally form a complete digital closed loop covering planning, execution, feedback, and adjustment. Data throughout the process is quantitatively recorded and analyzed, laying the core methodological and data foundation for achieving precise traceability of construction quality, process optimization, and high-level intelligent construction based on digital twins.

[0037] In some embodiments, S1150 may include at least one of the following: When N0 equals 1, a single-step adjustment strategy is adopted, using the stage compaction target of stage n+1 as the optimization constraint to adjust the construction parameters of stage n+1 only.

[0038] This single-step adjustment strategy is suitable for scenarios where compaction deviations are small in the early stages, and only the next adjacent stage needs adjustment to compensate for the deviations and advance construction. It has a small adjustment range, is easy to operate, and minimizes interference with the original construction plan. Based on a general construction scenario, let n be any positive integer satisfying 1 ≤ n < M (i.e., the nth stage of compaction is complete). Through AI model analysis in step S1140, N0 = 1 is determined, meaning only the construction parameters for the (n+1)th stage need adjustment.

[0039] The specific implementation process is as follows: First, the optimization constraints are clearly defined—the phased compaction target of stage n+1 is the core optimization constraint, while also taking into account the potential need to compensate for the compaction deviation in stage n (the actual compaction parameters in stage n did not reach its phased compaction target, resulting in a slight deviation). Then, based on these constraints, the actual compaction parameters of stage n, the phased compaction target of stage n+1, and the construction parameters supported by the current compaction equipment (such as excitation force, adjustable range of compaction speed, etc.) are input into the AI ​​model. The model, through feature extraction and temporal optimization, outputs an adjustment scheme for the construction parameters specifically for stage n+1. The original construction parameters for stage n+1 are set according to the initial plan; the adjusted parameters are optimized values ​​to meet the deviation compensation needs, ensuring that the adjustment range is reasonable and controllable.

[0040] During construction, the integrated sensors (compaction degree sensor, vibration sensor, and displacement sensor) carried by the compaction equipment collect real-time compaction data for stage n+1, ensuring that the adjusted parameters accurately achieve the stage's objectives. The final test results must meet the actual compaction parameters for stage n+1, and the secondary reinforcement effect of the compaction operation essentially eliminates deviations in the compacted area of ​​stage n, achieving the effect of compensating for deviations with a single adjustment. In this strategy, because adjustments are only made for a single stage, there is no need to modify the original construction parameters for subsequent stages after stage n+1, ensuring that the overall construction schedule proceeds according to the original plan.

[0041] In some embodiments, S1150 may include: when N0 is greater than 1 and less than a preset threshold N1, adopting a finite reprogramming strategy, using the phased compaction target of the (n+N0)th stage as the optimization constraint, to jointly adjust the construction parameters of the N0 consecutive stages starting from the (n+1)th stage.

[0042] This finite reprogramming strategy is applicable to scenarios where the compaction deviation in the early stages is moderate, requiring adjustments to parameters in multiple consecutive stages (but not reaching the global reprogramming threshold) to compensate for the deviation, while the overall progress can still rely on the original planned goals for subsequent stages. The core logic is to avoid construction quality risks (such as over-compaction damaging soil structure) caused by excessive adjustments in a single stage through the collaborative optimization of parameters in multiple stages. Considering a general construction scenario, let n be any positive integer satisfying 1 ≤ n < M (i.e., the nth stage of compaction is complete). AI model analysis shows that the actual compaction parameters in the nth stage did not reach its stage compaction target, exhibiting a moderate deviation. This deviation cannot be compensated for by adjusting only the parameters in the (n+1)th stage; adjustments to N0 consecutive stages starting from the (n+1)th stage are required to achieve the subsequent goals, where 1 < N0 < N1 (N1 is a preset threshold). Therefore, a finite reprogramming strategy is adopted.

[0043] The specific implementation process is as follows: First, the core optimization constraints are defined, with the phased compaction target of stage n+N0 as the optimization constraint. The core purpose of jointly adjusting the parameters of the N0 consecutive stages starting from stage n+1 is to ensure that the phased target is accurately achieved by the end of stage n+N0, while gradually compensating for compaction deviations in stage n. Then, the actual compaction parameters of stage n, the phased compaction target of stage n+N0, the current compaction equipment construction parameters, and the original planned parameters from stage n+1 to n+N0 are input into the AI ​​model. The model performs joint optimization calculations based on the principle of multi-stage parameter collaborative adaptation.

[0044] The parameters for stages n+1 to n+N0 are set according to the initial plan. After joint optimization by the model, a coordinated adjustment scheme for these N0 consecutive stages is output. In this way, the preceding adjustment stages compensate for some deviations by moderately optimizing the construction parameters, and subsequent adjustment stages make minor optimizations based on the original parameters, ultimately achieving the stage compaction target for stage n+N0 while completely eliminating the deviations in stage n. During construction, sensors monitor the compaction data of each adjustment stage in real time to ensure that the parameters work synergistically. Ultimately, each adjustment stage accurately achieves its corresponding stage compaction target without affecting the original construction plan for subsequent stages after stage n+N0.

[0045] In some embodiments, S1150 may further include: when N0 is greater than or equal to the preset threshold N1, using... Complete Bureau-level planning strategy With the completion of all subsequent phased compaction targets of the original plan as a hard constraint, new compaction planning parameters are generated based on the current actual compaction parameters, target compaction parameters, and equipment parameters.

[0046] This global replanning strategy is applicable to scenarios where compaction deviations are extremely large in the early stages, and adjustments to multiple consecutive stages cannot compensate for them, or adjustments would prevent subsequent stages from achieving their targets on schedule. By completely reconstructing the subsequent construction plan, it ensures that the overall project quality and schedule goals are not affected. Based on a general construction scenario, let n be any positive integer satisfying 1 ≤ n < M (i.e., the nth stage of compaction is complete). Testing revealed that the actual compaction parameters in the nth stage did not meet its stage-specific compaction target, exhibiting a large deviation. AI model analysis indicated that adjustments to N0 consecutive stages starting from the (n+1)th stage are needed to compensate for the deviation, where N0 ≥ N1 (N1 is a preset threshold). Therefore, a global replanning strategy is adopted.

[0047] First, a hard constraint is clearly defined: all subsequent phased compaction targets in the original plan must be completed on schedule, and the total construction period must remain consistent with the original plan. Then, based on this hard constraint, the actual compaction parameters of phase n, the original target compaction parameters, the current compaction equipment construction parameters, and the original planned total construction period data are input into the AI ​​model. The model discards the original planning parameters for phase n+1 and all subsequent phases, and regenerates entirely new compaction planning parameters based on the principle of global optimization.

[0048] The number of subsequent phases in the new plan remains the same as that in the original plan (to ensure matching of construction periods), but the phased compaction targets and construction parameters for each subsequent phase are reset: the phased compaction targets for the preceding reconstruction phase focus on gradually correcting deviations (avoiding excessive compaction) and are matched with corresponding optimized construction parameters; the phased compaction targets for the subsequent reconstruction phases must meet the final target requirements of the original plan, and the construction parameters are optimized and adjusted as needed to ensure that each phase progresses step by step to achieve the overall goal.

[0049] During the implementation of the new plan, after each stage, sensors on the compaction equipment were used to detect the actual compaction parameters, which were then fed back to the AI ​​model in real time for effect verification. The final construction result must meet the following requirements: all redesigned stages must accurately achieve the adjusted stage compaction targets; the overall compaction quality must meet the original target compaction parameter requirements; and the total construction period must be controlled within the original plan, achieving the goal of completing the construction on schedule and with high quality even with significant deviations. Compared to the previous two strategies, although this strategy involves significant changes to the original plan, it ensures the achievement of the overall project goals through global replanning, avoiding rework caused by the expansion of local deviations.

[0050] In summary, by adopting differentiated adjustment strategies for different values ​​of N0, precise control can be achieved, where the smaller the deviation, the smaller the adjustment range, and the larger the deviation, the more comprehensive the adjustment. This not only ensures the compaction quality but also maximizes the adaptation to the construction period requirements, significantly improving the intelligence and reliability of compaction in the construction area.

[0051] In some embodiments, the global replanning strategy is used to re-perform multi-stage planning, outputting a second rolling planning parameter containing P stages, where P may be different from the initial M.

[0052] In some embodiments, the determination of the minimum number of stages N0 to be adjusted is performed by the AI ​​model through a model predictive control algorithm. This algorithm takes the timely completion of all subsequent phases of compaction targets in the original plan as a hard constraint and takes the minimum total adjustment amount of construction parameters or the minimum number of adjustment stages as the optimization objective to perform rolling optimization.

[0053] In some embodiments, the AI ​​model includes a first planning model and a second adjustment model working collaboratively; the first compaction planning parameters are output by the first planning model, which is a deep reinforcement learning model trained based on historical construction big data and used to output multi-stage compaction plans; the determination of the minimum number of stages N0 to be adjusted and the execution of the parameter adjustment strategy are completed by the second adjustment model, which is a dynamic adjustment model built based on online learning or adaptive model predictive control algorithms. This split model processing can improve the robustness of a single task while reducing the size of the model.

[0054] Furthermore, after adopting a finite replanning strategy or a global replanning strategy, the adjusted construction parameters and their actual compaction effect data are fed back to the first planning model for incremental training and updating. The optimization objective, based on reducing the construction parameters in the stage to be adjusted while achieving the original compaction plan on schedule, is specified as minimizing the adjustment range of construction parameters, and / or minimizing the additional energy consumption caused by the adjustment, and / or maximizing the uniformity of compaction quality in subsequent stages.

[0055] In some embodiments, after compaction is performed according to the adjusted construction parameters, the actual compaction parameters are compared with the corresponding phased compaction targets. If the deviation continues to exceed the allowable range, the preset threshold N1 is dynamically increased to trigger the global replanning strategy earlier.

[0056] In some embodiments, the parameter adjustment strategy corresponding to the value of N0, with the timely completion of all subsequent phased compaction targets of the original plan as a hard constraint, includes at least one of the following adjustments to the construction parameters for at least the (n+1)th phase: When determining the minimum number of stages N0 that need adjustment, the probability of occurrence and the magnitude of impact of one or more potential future interference events are predicted based on the AI ​​model. When N0 equals 1, a forward-looking single-step adjustment strategy is adopted. While meeting the compaction target of the (n+1)th stage, the adjusted construction parameters are also used to actively offset the future potential disturbances with the greatest impact in the prediction. When N0 is greater than 1 and less than the preset threshold N1, a robust finite reprogramming strategy is adopted. When optimizing the construction parameters of N0 consecutive stages, a robust optimization objective for predicting disturbance events is introduced, so that the probability that the adjusted multi-stage construction parameter sequence can still meet the stage compaction target when the disturbance occurs is higher than the preset threshold. When N0 is greater than or equal to the preset threshold N1, a resilient global replanning strategy is adopted. When generating new crushing planning parameters, one or more predicted future potential interference events are used as virtual adversarial training scenarios. The generated planning parameters have the best comprehensive performance evaluation in standard scenarios and multiple adversarial scenarios.

[0057] In some embodiments, based on the aforementioned parameter adjustment strategy, this embodiment further introduces the logic for predicting and responding to potential future interference events. An AI model is used to predict potential interference during construction and incorporate it into the adjustment strategy design, thereby improving the anti-interference capability and quality stability of the compaction construction. Specifically, the process of adjusting the construction parameters for at least the (n+1)th stage using a parameter adjustment strategy corresponding to the N0 value, with the timely completion of all subsequent planned compaction targets as a hard constraint, also includes the following implementation content related to interference prediction and response, which are described in detail in different scenarios below: First, it needs to be clarified that when determining the minimum number of stages N0 requiring adjustment, the AI ​​model will simultaneously execute a prediction process for potential future interference events: based on the environmental characteristics of the construction area, historical equipment operation data, and common construction interference types in the industry, it will predict the probability and impact of one or more potential future interference events. These potential future interference events include, but are not limited to, sudden weather changes in the construction area (such as short-term rainfall or strong winds), localized abrupt changes in geological conditions (such as hidden soft soil interlayers or culverts), and momentary malfunctions of compaction equipment (such as fluctuations in excitation force or jamming of the walking system). By analyzing the historical occurrence patterns and impact levels of various interference events, the AI ​​model outputs the probability of occurrence for each type of interference (such as the probability of short-term rainfall) and the corresponding impact magnitude (such as a percentage point deviation in compaction testing or a decrease in construction efficiency), providing a basis for the formulation of subsequent differentiated adjustment strategies.

[0058] This forward-looking single-step adjustment strategy adds forward-looking anti-interference capabilities to the original single-step adjustment. Through appropriate parameter redundancy optimization, it proactively avoids the impact of potential disturbances with the greatest influence on the compaction mass. It is suitable for scenarios with small deviations and a clear risk of high-impact disturbances. The specific implementation process is as follows: 1. Interference prediction and screening: After determining N0=1 (only the n+1 stage needs to be adjusted), the AI ​​model outputs a list of potential interference events that may affect the construction of the n+1 stage. By comparing the impact of each interference, the interference event with the greatest impact is screened out (such as short-term rainfall, with an impact of 2 percentage points on the compaction test deviation). 2. Constraint setting: Based on meeting the phased compaction target of stage n+1, a forward-looking constraint to actively offset the maximum impact interference is added. That is, the adjusted construction parameters must ensure that even if the maximum impact interference occurs, the actual compaction parameters of stage n+1 can still meet the target. 3. Parameter adjustment implementation: Input the actual compaction parameters of stage n, the stage compaction target of stage n+1, the equipment construction parameters, and the screened maximum impact interference information (occurrence probability, impact magnitude) into the AI ​​model. The model outputs a construction parameter adjustment scheme for stage n+1 that has both deviation compensation and anti-interference capabilities.

[0059] 4. Construction monitoring: During the n+1 stage of construction, the sensors carried by the compaction equipment collect compaction data and environmental data in real time. If the predicted maximum impact interference does not occur, the redundant parameters can be fine-tuned in real time (such as slightly increasing the compaction speed) to optimize construction efficiency. If interference occurs, the adjusted parameters are used to ensure that the compaction quality meets the standards.

[0060] This robust finite reprogramming strategy improves the anti-interference stability of multi-stage parameter sequences through robust optimization, preventing the spread of interference from a single stage from affecting the overall construction effect. It is suitable for scenarios with moderate deviations and multiple potential interferences. The specific implementation process is as follows: 1. Interference prediction and robustness target setting: After determining that 1 < N0 < N1, the AI ​​model predicts potential interference events (such as local geological changes and occasional equipment failures) covering N0 consecutive adjustment stages in the future, and clarifies the probability of occurrence and scope of impact of various interferences; at the same time, a robustness optimization target is set - the adjusted construction parameter sequence for N0 consecutive stages, in which the probability of achieving the stage compaction target of each stage is higher than the preset threshold when any predicted interference event occurs; 2. Construction of joint optimization constraints: Taking the achievement of the phased compaction target in stage n+N0 as the core constraint, and superimposing robust optimization objectives, a multi-constraint joint optimization system is formed to ensure that parameter adjustments can not only gradually make up for the deviation in stage n, but also resist potential interference.

[0061] 3. Parameter Coordination Adjustment: The actual compaction parameters of stage n, the stage-specific compaction target of stage n+N0, equipment parameters, and predicted interference information are input into the AI ​​model. Based on the principle of multi-stage parameter coordination and anti-interference redundancy, the model outputs construction parameter adjustment schemes for N0 consecutive stages. For example, in response to possible local geological abrupt changes, the model appropriately expands the adjustable redundancy range of the excitation force when optimizing the parameters of the preceding adjustment stages, and simultaneously optimizes the rolling path planning of each stage to ensure that even if a geological abrupt change occurs in a certain stage, the interference effect can be offset by parameter coordination in adjacent stages.

[0062] 4. Dynamic verification: During the N0 consecutive construction phases, the AI ​​model receives compaction data and actual interference occurrences from sensors in real time. If no interference occurs, the model can dynamically reduce parameter redundancy to improve efficiency. If interference occurs, the model ensures that the probability of meeting the standard is met through preset parameter redundancy, thus avoiding the expansion of deviations caused by interference.

[0063] This resilient global replanning strategy enhances the resilience of the global planning process through adversarial training-style planning, ensuring timely and high-quality completion of construction even in extreme scenarios with significant deviations and potential disturbances. It is suitable for scenarios with extremely large deviations and high disturbance risks. The specific implementation process is as follows: Interference scenario construction: After determining that N0≥N1, the AI ​​model will transform one or more predicted future potential interference events (such as continuous rainfall, large-scale geological changes) into virtual adversarial training scenarios. Each scenario specifies the time of occurrence, duration and impact of the interference. For example, if continuous rainfall occurs in the n+2 stage, lasting for 2 construction units, the compaction efficiency will decrease by 30%.

[0064] Multi-scenario optimization constraint setting: taking the timely completion of all subsequent phased compaction targets of the original plan as a hard constraint, adding a standard scenario (no interference) and multiple adversarial scenarios to optimize the overall performance of the target. The overall performance includes core indicators such as compaction quality compliance rate, construction efficiency, and equipment energy consumption.

[0065] 3. Resilience Global Planning Generation: The actual compaction parameters of stage n, the original target compaction parameters, equipment parameters, and multiple constructed adversarial training scenarios are input into the AI ​​model. The model generates entirely new compaction planning parameters through iterative simulations across multiple scenarios. These planning parameters are not limited to a single scenario; rather, they enable efficient construction in a standard, interference-free scenario and can withstand interference through adaptive parameter adaptation in various adversarial scenarios, ensuring optimal overall performance. For example, for a continuous rainfall adversarial scenario, the plan includes alternative compaction parameter schemes for the rainy season, clarifies the adjustment thresholds for excitation force and compaction speed during rainfall, and optimizes stage divisions to reserve time redundancy for rainy season construction.

[0066] 4. Full-process resilience verification: During the implementation of the new plan, after each stage, sensor data feedback is used to dynamically determine whether an adversarial scenario has been triggered. If it is triggered, the corresponding alternative parameters are activated; if it is not triggered, construction is carried out according to standard parameters. This ensures that the overall construction can proceed along the established goals regardless of whether interference occurs, achieving dual guarantees of schedule and quality under the dual challenges of major deviations and potential interference.

[0067] In some embodiments, the first planning model is trained and constructed based on historical construction big data and adopts a deep reinforcement learning architecture, possessing strong historical experience learning and complex scenario adaptation capabilities. It can mine deep correlation patterns between different geological conditions, equipment parameters, and compaction effects from massive historical construction data. Combining the exploration-exploitation characteristics of deep reinforcement learning, when outputting the first compaction planning parameters, it can not only draw on the mature experience of similar projects, but also perform personalized optimization for the geological parameters and target compaction parameters of the current construction area. This effectively avoids the problems of poor adaptability and large quality fluctuations that exist in traditional experience-based planning or simple algorithm planning, laying a precise planning foundation for subsequent construction.

[0068] In other embodiments, the second adjustment model is constructed using online learning or adaptive model predictive control algorithms, possessing real-time iterative updates and look-ahead optimization capabilities. On the one hand, the real-time iterative update capability of online learning allows the model to quickly absorb new information from the actual compaction parameters in the nth stage, dynamically correcting the decision logic for parameter adjustment, and ensuring timely responses to deviations that occur during construction (such as compaction deviations caused by uneven soil). On the other hand, the look-ahead optimization characteristics of adaptive model predictive control enable the model to predict in advance the impact of different adjustment schemes on the compaction effect of subsequent stages when determining the minimum number of adjustment stages N0. From a global perspective of achieving the original planning goals on schedule, the model selects the scheme with the fewest adjustment stages, minimizing interference with the original construction plan while ensuring the feasibility and optimization of the adjustment strategy.

[0069] In other embodiments, the collaborative working mode of the first planning model and the second adjustment model forms a closed-loop control logic of precise initial planning and dynamic deviation correction. The multi-stage compaction plan output by the first planning model provides the second adjustment model with clear stage target benchmarks and parameter adjustment boundaries, avoiding blindness in the adjustment process; the dynamic adjustment results of the second adjustment model can in turn verify the adaptability of the first planning model. If no major deviations occur in the subsequent adjustment process, it further confirms the rationality of the initial planning; if a major deviation requiring global replanning occurs, the adjustment requirements of the second adjustment model can be quickly fed back to the overall AI model framework, providing the first planning model with accurate current status data (actual compaction parameters, current equipment status, etc.) for regenerating new compaction plan parameters, ensuring that the replanned parameters still meet the actual needs of the project, and ultimately ensuring the realization of the core hard constraint of completing all subsequent stage compaction targets on schedule.

[0070] The specific implementation process is as follows: First, the optimization constraints are clearly defined—the phased compaction target of stage n+1 is the core optimization constraint, while also taking into account the potential need to compensate for the compaction deviation in stage n (the actual compaction parameters in stage n did not reach its phased compaction target, resulting in a slight deviation). Then, based on these constraints, the actual compaction parameters of stage n, the phased compaction target of stage n+1, and the construction parameters supported by the current compaction equipment (such as excitation force, adjustable range of compaction speed, etc.) are input into the AI ​​model. The model, through feature extraction and temporal optimization, outputs an adjustment scheme for the construction parameters specifically for stage n+1. The original construction parameters for stage n+1 are set according to the initial plan; the adjusted parameters are optimized values ​​to meet the deviation compensation needs, ensuring that the adjustment range is reasonable and controllable.

[0071] During construction, the integrated sensors (compaction degree sensor, vibration sensor, and displacement sensor) carried by the compaction equipment collect real-time compaction data for stage n+1, ensuring that the adjusted parameters accurately achieve the stage's objectives. The final test results must meet the actual compaction parameters for stage n+1, and the secondary reinforcement effect of the compaction operation essentially eliminates deviations in the compacted area of ​​stage n, achieving the effect of compensating for deviations with a single adjustment. In this strategy, because adjustments are only made for a single stage, there is no need to modify the original construction parameters for subsequent stages after stage n+1, ensuring that the overall construction schedule proceeds according to the original plan.

[0072] In some embodiments, the method further includes: The N1 is determined based on the size of M and the type of project in the construction area; If the values ​​of M are the same for different project types, when the project type is a dam project, N1 is located in the first value interval; when the project type is a building foundation project, N1 is located in the second value interval; when the project type is a roadbed project, N1 is located in the third value interval; wherein, the median of the first value interval is less than the median of the second value interval; and the median of the third value interval is less than the median of the second value interval.

[0073] In this embodiment, the threshold N1 is not a fixed constant but needs to be determined through dynamic adaptation in two dimensions. This means it must consider both the size of the total number of stages M and the characteristics of the construction area's project type. Ultimately, this ensures that N1 accurately meets the quality control needs of different projects, providing a scientific and reasonable basis for switching between subsequent single-step adjustment strategies, finite replanning strategies, and global replanning strategies. Specifically, when different project types have the same value for M, the range of N1 will be clearly distinguished based on the differences in the core requirements of each project type, thus satisfying the compaction requirements of each project type.

[0074] In some embodiments, the AI ​​model employs an adaptive multi-objective optimization algorithm based on project type to optimize construction parameters, and the objective function used during optimization is: Among them, the weighting coefficient to Dynamically adjusted according to project type: For dam projects: and Greater than ,and Greater than as well as ; For building foundation engineering: and Greater than ;and Greater than as well as ; For road foundation engineering: Greater than , or ; Greater than ; The actual compaction degree in stage i; Target compaction degree; Let be the water permeability coefficient in stage i; is the compaction uniformity deviation coefficient for the i-th stage; The target compaction uniformity deviation coefficient; The target water permeability coefficient; Let be the resilient modulus of the top surface of the roadbed in the i-th stage; The resilient modulus of the target roadbed top surface in the construction area; Let be the compaction energy of the j-th pass of the i-th stage; Let be the number of rolling passes in the i-th stage.

[0075] This embodiment proposes a technical solution for optimizing construction parameters using an AI model with an adaptive multi-objective optimization algorithm based on engineering type. Combining three typical scenarios—dam engineering, building foundation engineering, and road foundation engineering—the optimized construction parameters accurately match the core quality requirements of different projects by applying the same objective function and adjusting the engineering adaptability of the weight coefficients.

[0076] The multi-objective optimization function used by the AI ​​model to optimize construction parameters aims to achieve synergistic optimization of both quality compliance and optimal energy consumption by quantitatively evaluating the deviations between compaction quality indicators and construction energy consumption indicators at each stage. The specific meanings and functions of each parameter in the objective function are as follows: The target compaction degree in stage i is to quantify the deviation between the actual compaction effect and the quality standard. Compaction degree is the basic core indicator of compaction quality in various projects, and it directly determines the bearing capacity and stability of the foundation / dam body.

[0077] Controlling the permeability of compacted structures is particularly suitable for projects requiring impermeability (such as dams). The smaller the permeability coefficient, the better the impermeability.

[0078] The target roadbed top surface resilient modulus in the construction area is crucial for ensuring the rigidity and deformation resistance of the compacted material, and is essential for engineering projects such as road foundations and building foundations that bear dynamic or static loads.

[0079] The compaction energy of the j-th pass of the i-th stage.N Let represent the number of compaction passes in stage i. This project aims to quantify the energy consumption cost during the construction process and achieve energy-efficient construction by optimizing and reducing unnecessary compaction operations.

[0080] The weight coefficients in the objective function are dynamically adjusted according to the project type, so that the weight coefficients of the core quality indicators using a single objective function are higher, ensuring that the optimization direction is aligned with the core needs of the project. The simple calculation method also achieves the unification of objective functions for different construction projects.

[0081] The requirements for dam projects are strong seepage prevention and structural stability. The compaction quality directly affects the flood control safety of the dam body. This embodiment sets specific weighting coefficients. For example, the weighting coefficients are set as follows: (0.3, 0.3, 0.2, 0.1, 0.1). As a water-retaining structure, seepage prevention is the primary core requirement for dams, followed by structural stability ensured by compaction quality. The resilient modulus has a relatively small impact on the core performance of the dam body, and energy consumption optimization is only a secondary objective. Through this weighting allocation, the AI ​​model will prioritize ensuring that the water permeability coefficient and compaction quality meet the standards during optimization, and then consider the resilient modulus and energy consumption.

[0082] The core requirements of building foundation engineering are load-bearing stability and uniform compaction to ensure the effective transfer of the weight of the superstructure. Therefore, the weighting coefficients are adjusted according to the above-mentioned relationship. For example, weighting coefficients are set to 0.35, 0.15, 0.25, 0.1, and 0.15. The compaction degree of the building foundation directly determines the bearing capacity, the resilient modulus affects the deformation characteristics of the foundation (to avoid uneven settlement of the superstructure, corresponding to the second largest weight), while the permeability requirement of the building foundation is relatively low (smallest weight). Considering the construction period and cost control requirements, the weight of energy consumption optimization is slightly higher than that of dam engineering.

[0083] The core requirements for road subgrade engineering are adequate stiffness and uniform load-bearing capacity, which must be able to withstand the dynamic loads of vehicle traffic, thus giving high weight to the corresponding parameters. The resilient modulus of the road subgrade directly determines the bearing capacity and service life of the pavement. Compaction degree is the foundation for achieving the required resilient modulus. Road subgrades have relatively low requirements for permeability. At the same time, road construction is large-scale and energy consumption costs account for a high proportion, so the weight of energy consumption optimization is comparable to that of building foundation engineering.

[0084] For example, the weighting coefficients for dam projects can be 0.4, 0.4, 0.1, 0.005, and 0.005 respectively; the weighting coefficients for building foundation projects can be 0.3, 0.2, 0.3, 0.1, and 0.1 respectively; and the weighting coefficients for road foundation projects can be 0.2, 0.1, 0.2, 0.3, and 0.2 respectively.

[0085] Compaction energy The calculation formula is: The number of compaction passes (construction parameter) for the i-th stage; The compaction speed (construction parameter) for stage i; Let be the vibration frequency (construction parameter) of the i-th stage; The amplitude (construction parameter) for the i-th stage; Let be the crushing time in the i-th stage.

[0086] In some embodiments, the method further includes: determining key geological parameters of the geological parameters and setting weights for the key geological parameters based on the engineering type of the construction area; the key geological parameters are used for determining and adjusting construction parameters.

[0087] In this embodiment, after obtaining the geological parameters of the construction area, the system does not treat all parameters equally. Instead, it first identifies the engineering type of the construction area, such as determining whether it is a high earth-rock dam, a deep foundation for a high-rise building, or a high-grade highway subgrade. Then, based on the core quality requirements and key control points of this engineering type, the system automatically determines the key geological parameters from all geological parameters. For example, for dam core wall filling with seepage prevention as the core objective, the key geological parameters might include the optimum moisture content of the fill material, the particle size distribution curve (especially the fine particle content), and the maximum dry density; for foundation treatment of structures primarily focused on deformation control, the key geological parameters might emphasize the soil's compression modulus, preconsolidation pressure, and in-situ bearing capacity.

[0088] After identifying key geological parameters, the method dynamically assigns differentiated weights to each parameter based on the inherent logic of the project type and historical data. The weight directly reflects the importance of that parameter in the current engineering compaction quality control system. For example, in dam projects, the optimal moisture content might be assigned the highest weight because it directly controls compaction efficiency and seepage prevention performance; while in roadbed projects, the California bearing ratio (CBR) of the fill material might have a more prominent weight.

[0089] Finally, these weighted key geological parameters will, as a whole, deeply participate in and guide the subsequent AI model in determining (initial planning stage), and dynamically adjusting (process control stage) construction parameters (such as the number of compaction passes, driving speed, and vibration frequency). When performing multi-objective optimization, the AI ​​model will impose stronger constraints or give higher optimization priority to the quality indicators (such as compaction degree and permeability coefficient) associated with high-weighted key parameters, thereby ensuring that the decision-making of construction parameters is always closely centered on the core quality objectives of the current project, achieving an intelligent upgrade from "general control" to "key and precise control oriented towards engineering objects".

[0090] In some embodiments, the method further includes: controlling the compaction equipment during the compaction operation phase, including: The rolling planning parameters output by the AI ​​model are processed by the path smoothing module to generate a series of path points. Target heading angle Target speed The resulting compaction track; The position of the equipment's center of gravity is obtained by deploying a positioning module on the compaction equipment. Heading angle Roll angle and pitch angle ; The trajectory tracking controller is based on the current position with target path point Calculate the lateral position deviation and heading angle deviation ; Using model predictive control algorithm, front wheel steering angle and drive / brake torque To control the variables, with the objective of minimizing lateral and heading deviations while satisfying vehicle dynamics constraints, the optimal values ​​are calculated in real time. and ; The attitude compensation controller receives real-time roll angle. and pitch angle ; The attitude compensation controller uses real-time attitude angles and equipment geometry models to dynamically calculate the hydraulic suspension system compensation required to maintain the vertical excitation force of the vibrating wheel. and vibration shaft speed compensation amount ; Will , , as well as The transmission mechanism sends signals to the steering actuator, drive motor, hydraulic suspension solenoid valve, and vibration motor frequency converter of the compaction equipment, achieving millisecond-level coordination between driving and compaction actions.

[0091] This embodiment further integrates intelligent planning with high-precision collaborative control of underlying equipment, achieving a seamless connection from "digital instructions" to "precise physical execution." Specifically, this embodiment describes how the compaction equipment transforms AI planning parameters into precise and coordinated mechanical actions during the compaction operation execution phase.

[0092] After the compaction operation begins, the compaction planning parameters (e.g., where, at what speed, and in what posture) output by the upper-level AI decision-making system, containing complex spatial and temporal information, are first sent to a path smoothing module. This module is responsible for processing the discrete planning points into a continuous, smooth, and executable trajectory that conforms to the kinematic characteristics of the compaction equipment. This trajectory is concretized into a series of timestamped path points, each containing not only planar coordinates but also the target heading angle and target velocity at that point, forming a four-dimensional spatiotemporal "guidance path."

[0093] To accurately track this "guided path," this embodiment deploys a high-precision multi-source fusion positioning and attitude perception module on the compaction equipment. This module typically integrates GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit), enabling it to output the precise position of the equipment's center of mass in the geodetic coordinate system, the current heading angle, and the roll and pitch angles reflecting the vehicle's tilt on the slope in real time at high frequency. This real-time perception data is the cornerstone of closed-loop control.

[0094] Subsequently, the trajectory tracking controller is activated. It continuously receives information about the current position / heading of the device and the target path point, and calculates the key tracking errors in real time: lateral position deviation (the vertical distance from the vehicle's current position to the planned path) and heading angle deviation (the angle between the vehicle's current orientation and the tangent direction of the path at that point). To optimally eliminate these deviations, the controller employs a model predictive control algorithm. Based on a built-in dynamic model of the compaction equipment, the MPC algorithm performs rolling optimization within a future prediction time domain to solve for the optimal front wheel steering angle and drive / braking torque sequence, and outputs the first control variable to ensure that the tracking process is both fast and smooth, while satisfying physical constraints such as tire adhesion and steering rate.

[0095] Simultaneously, a parallel attitude compensation controller begins operation. It specifically processes the real-time roll and pitch angles acquired from the sensing module. When the equipment operates on cross or longitudinal slopes, vehicle tilt causes the excitation force direction of the vibratory wheels to deviate from the ideal vertical direction, severely reducing the effective transfer of compaction energy. Based on the equipment's three-dimensional geometric model, the controller calculates in real-time the compensation required to counteract the tilt angle effect and maintain the vibratory wheels' vertical force on the ground: hydraulic suspension system compensation (leveling the vibratory wheel frame by adjusting the hydraulic pressure on one side of the suspension) and vibratory shaft speed compensation (compensating for excitation force loss due to changes in transmission angle by adjusting the speed).

[0096] Ultimately, under the coordination of the vehicle domain controller, the optimal steering angle and torque commands calculated by the MPC, along with the suspension compensation and speed compensation commands calculated by the attitude compensation controller, are synchronously packaged and distributed in milliseconds to the corresponding actuators via a high-speed onboard network (such as CANFD): steering actuators, drive motors, solenoid valves of the hydraulic suspension system, and frequency converters of the vibration motors. This enables the compaction equipment to not only accurately "follow" the planned path but also automatically adjust its attitude on slopes, ensuring that the energy "compacted" is vertical and efficient. Ultimately, this achieves deep temporal and spatial synergy between trajectory control and compaction quality control, perfectly replicating the optimized planning of the digital world in physical construction operations.

[0097] This system is designed with seamlessly switchable automatic control mode, semi-automatic control mode and manual control mode, forming a multi-level and progressive human-machine collaborative operation system.

[0098] In automatic control mode, the system performs closed-loop autonomous operation. As described in the previous embodiment, the entire process, from trajectory generation, real-time perception, MPC control calculation to attitude compensation and multi-actuator coordination, is completed automatically by the system. The operator in the cockpit mainly acts as a supervisor, monitoring the execution of the planned path, real-time compaction quality cloud map, equipment status, and outputs of various controllers through a human-machine interface. This mode is suitable for construction in conventional areas with good working conditions and reliable planning, maximizing construction consistency, reducing operator workload, and ensuring the accurate realization of AI planning intentions.

[0099] In semi-automatic control mode, this mode aims to leverage the advantages of human-machine hybrid intelligence, and is suitable for areas with complex boundaries, near obstacles, or where temporary adjustments to the strategy are required. The system provides different levels of assistance: Path tracking assistance: The operator manually controls the equipment's steering and speed, but the path tracking controller remains active, only "smoothing" and "safety correction" the operator's input, such as preventing instability caused by oversteering, or automatically making fine adjustments to ensure precise overlap of the rolling wheel track. The core control is with the human.

[0100] Compaction parameter assistance: While operator-controlled movement of the equipment, the compaction system (vibration, frequency, amplitude) is entirely automated, controlled by an AI model based on real-time planned parameters matched to the current location. Simultaneously, the attitude compensation controller remains operational, ensuring effective compaction even under manual movement. This allows the operator to focus more on complex path manipulation without being distracted by adjusting compaction parameters.

[0101] Preset tracking assistance: The operator manually sets a simple desired path (such as bypassing potholes) in the interface, and the system will take over control and automatically complete the driving and compaction along this path until the path ends or the operator takes over.

[0102] Mode switching can be completed instantly via the mode selection knob in the cockpit or voice command. During switching, the control system achieves a seamless handover of control, ensuring construction continuity.

[0103] In manual control mode, all advanced auxiliary control functions (track tracking, MPC, automatic attitude compensation) are disabled. The operator directly controls steering, drive, braking, vibration start / stop, and frequency / amplitude using traditional joysticks, pedals, and buttons. However, the underlying basic electronic control and safety protection systems remain operational, such as anti-skid control, engine power protection, and critical parameter over-limit alarms. All sensor data (position, attitude, compaction degree) are still displayed in real-time on the interface, providing information support to the operator. This mode is primarily used for system debugging, emergency response, or operator-led exploratory construction.

[0104] In some embodiments, the system incorporates an intelligent context awareness module that can suggest or automatically trigger mode switching based on the following factors: Environmental awareness: When an unstructured environment or nearby dynamic obstacles are detected, the system can suggest switching to semi-automatic or manual mode. System status: When the positioning signal is lost or a critical sensor malfunctions, the system automatically downgrades to manual mode and issues an alarm. Operator intervention: If the operator's actively applied control force exceeds a threshold, the system determines it as a takeover request and smoothly transitions control. Construction quality deviation: When the real-time compaction quality detection value continuously deviates from expectations and automatic adjustment is ineffective, the system prompts the operator to intervene. Through this multi-level control architecture, the embodiments of this disclosure can achieve unmanned, high-precision automated construction under ideal conditions, maximizing efficiency and quality; and can leverage the advantages of human experience, judgment, and flexibility through flexible human-machine collaboration under complex working conditions; while ensuring corresponding safety redundancy and information support in any mode, achieving a balance between robustness, adaptability, and safety of the intelligent construction system.

[0105] In some embodiments, when the minimum number of stages N0 to be adjusted is determined, predictive anti-interference analysis is initiated simultaneously. Based on historical meteorological data, equipment operation logs, and real-time scanning data of ground-penetrating radar in the construction area, the probability of occurrence P_I and the intensity of influence S_I of potential interference events in the next H construction stages are predicted by a time-series prediction model. The potential interference events include: sudden rainfall causing a change in the moisture content of the filler, communication delay of the equipment group causing coordination error, and local weak interlayers causing abnormal compaction resistance. The prediction results (P_I, S_I) are input to the parameter adjustment strategy generator to output interference events or potential interference events.

[0106] In some embodiments, an agent is constructed for each compaction device, and all agents constitute a multi-agent system; Each agent maintains a local belief state based on its own sensor data and a shared global construction map; A multi-agent reinforcement learning algorithm is used for collaborative compaction planning, and its reward function R is designed as follows: R = α * R_quality (global compaction quality reward) + β * R_efficiency (group construction efficiency reward) - γ * R_collision (collision risk penalty) - λ * R_communication (communication cost penalty); The agents make decisions through a partially observable Markov decision process, and the Shapley value in game theory is used to fairly allocate each agent's contribution to the global quality as the basis for reward distribution. The central coordinator calculates the optimal group crushing trajectory based on the decisions of each agent and dynamically resolves potential resource conflicts. In this way, it achieves local and central coordinated decision-making, improves decision-making efficiency and accuracy, and solves complex conflict and cooperation problems in multi-machine operations by introducing multi-agent reinforcement learning and game theory. It is the core technology for realizing large-scale unmanned cluster construction and is innovative.

[0107] In some embodiments, a high-fidelity digital twin model of the construction area is constructed, which couples discrete element method to simulate the movement of filler particles, finite element method to calculate stress transfer, and multiphysics field to simulate water-thermal-mechanical coupling effect. During the compaction process, the actual compaction parameters, equipment status parameters and environmental parameters collected in real time are synchronized to the digital twin model to drive it to perform real-time simulation and predict the compaction evolution trend in the near future. Cross-validate the predictions from the digital twin model with the planning results from the AI ​​model: If the predicted trend is consistent, the original plan will proceed. If a significant deviation occurs, a model inaccuracy warning will be triggered, and a data assimilation algorithm will be activated. Using real-time observation data as a benchmark, key parameters of the digital twin model (such as filler constitutive model parameters) will be corrected in reverse to improve its prediction accuracy. The corrected digital twin model generates new high-confidence predictions to guide the AI ​​model in the next round of parameter adjustments and planning, forming a real-time closed-loop optimization system integrating the physical world, digital twin, and AI decision-making. Through digital twins, the decision-making basis of the control system is upgraded from data-driven to a fusion of mechanistic models and data-driven approaches. Real-time data assimilation and cross-validation ensure the accuracy and reliability of decisions, representing a cutting-edge direction in intelligent construction.

[0108] In some embodiments, detecting actual compaction parameters may include: acquiring detection data for the current cycle; the detection data includes time information, location information, and signal information; the signal information includes at least excitation signal information and environmental vibration signal information; determining the compaction grid where each compactor is located based on the location information; wherein the construction area to be compacted is divided into multiple compaction grids; calculating the local reference wave field formed by each compactor in the construction area based on the excitation signal information of each compactor and the AI ​​model; integrating the local reference wave fields to obtain the global reference wave field of the construction area based on the parameters of the excitation signals corresponding to the overlapping grids corresponding to the multiple local reference wave fields; determining the listening parameters of the excitation waves of other compactors detected by each compactor based on the environmental vibration signal information of each compactor; filtering non-specified environmental vibration signals based on the listening parameters to obtain a filtered signal; determining the transmission parameters of the vibration waves across compactors based on the filtered signal, and forming a supplementary wave field based on the transmission parameters; and determining the actual compaction distribution map of the compaction grids in the construction area based on the global reference wave field and the supplementary wave field.

[0109] This actual compaction distribution map can serve as an important component / representation of the aforementioned actual compaction parameters. It can be used to adjust construction parameters for the next one or more stages.

[0110] In some embodiments, the actual compaction parameters can be periodically detected, with the detection period being shorter than the time length between two adjacent stages. The operation of the compaction equipment is then periodically planned and controlled based on this detection period. For example, the priority of the grids to be compacted is determined according to the target compaction distribution map (e.g., a phased compaction target) and the actual compaction distribution map. The priority includes a first priority and a second priority. The difference between the actual compaction condition and the expected target compaction condition of the grids with the first priority is greater than the difference between the actual compaction condition and the expected target compaction condition of the grids with the second priority. The compaction route and compaction parameters for each compaction equipment in the next cycle are planned based on the location of each equipment, the compaction parameters, and the priority. The compaction route avoids grids that have already met the compaction standards and prioritizes nearby grids that have not yet met the standards. By accurately distinguishing between excitation vibration signals and environmental vibration signals using vibration sensors, and combining the actual operating parameters of the equipment collected by the compaction weight detection module, a local reference wavefield for a single piece of equipment is calculated based on an AI model. Multiple devices share data through a collaborative sensing network, integrating the global reference wavefield through overlapping grid excitation signal parameters, and then extracting cross-device vibration wave transmission parameters to form a supplementary wavefield, completely eliminating detection blind spots. Real-time data sharing avoids position and signal delay deviations, reducing compaction errors compared to traditional single-point detection, and dynamic detection adapts to the needs of large-area construction.

[0111] Furthermore, based on the parameters of the excitation signals corresponding to the overlapping grids of multiple local reference wavefields, the local reference wavefields are integrated to obtain the global reference wavefield of the construction area. This includes: extracting the grid identifiers of the compaction grids covered by each local reference wavefield and generating a multidimensional parameter set; each element of the multidimensional parameter set is a multidimensional array; the multidimensional array includes the following ordered parameters: grid identifier, theoretical wave velocity, measured wave velocity, acquisition timestamp, equipment compaction weight, and signal-to-noise ratio; by comparing the grid identifiers of the compaction grids covered by each local reference wavefield, it is determined that more than two compaction devices are involved. The overlapping grid is obtained by covering the compaction grid with a local reference wavefield. For each overlapping grid, the beam variation coefficient detected by different compaction devices is calculated, and the wavefield parameters of the overlapping grid are corrected according to the beam variation coefficient. For non-overlapping grids, the wavefield parameters of the non-overlapping grids in each local reference wavefield are corrected based on the corrected wavefield parameters or correction parameters of the overlapping grids adjacent to the non-overlapping grids. The corrected wavefield parameters of the overlapping grids and the corrected wavefield parameters of the non-overlapping grids are spliced ​​together to obtain the global wavefield parameters. In this way, accurate detection with low power consumption can be achieved.

[0112] Furthermore, for non-overlapping grids, based on the corrected wavefield parameters or correction parameters of the overlapping grids adjacent to the non-overlapping grids, the wavefield parameters of the non-overlapping grids in each of the local reference waves are corrected, including: based on the adjacency relationship of each compaction grid in the construction area, all overlapping grids within a preset distance range around each non-overlapping grid are selected; when a non-overlapping grid has multiple adjacent overlapping grids, the grid identifier, final wave velocity, attenuation coefficient, and straight-line distance d between the multiple adjacent overlapping grids and the non-overlapping grid are determined; the inverse distance weighting method is used to calculate the non-overlapping grid pairs of each overlapping grid in the multiple adjacent overlapping grids. The calibration weights of the overlapping grids are used to determine a first calibration value based on the calibration weights and the corrected wavefield parameters of the multiple adjacent overlapping grids. When the difference between the first calibration value and the original wavefield parameters of the non-overlapping grids is less than a first threshold, the corrected wavefield parameters of the non-overlapping grids are determined to be equal to the original wavefield parameters. When the difference between the first calibration value and the original wavefield parameters of the non-overlapping grids is greater than or equal to the first threshold, the corrected wavefield parameters of the non-overlapping grids are calculated based on the weights of each overlapping grid in the multiple adjacent overlapping grids and the corrected wavefield parameters. In this way, accurate detection with low power consumption can be achieved.

[0113] In some embodiments, for non-overlapping grids, the wavefield parameters of the non-overlapping grids in each local reference wave are corrected based on the corrected wavefield parameters or correction parameters of the overlapping grids adjacent to the non-overlapping grids. This further includes: when the non-overlapping grid has one adjacent overlapping grid, after completing the wavefield parameter correction for multiple adjacent overlapping grids, determining the grid identifier, final wave velocity, attenuation coefficient, and straight-line distance d between the multiple adjacent compacted grids and the non-overlapping grid; and calculating the calibration weight of each compacted grid relative to the non-overlapping grid using an inverse distance weighting method. A second calibration value is determined based on the calibration weights and the corrected wavefield parameters of the plurality of compacted grids. When the difference between the second calibration value and the original wavefield parameters of the non-overlapping grids is less than a second threshold, the corrected wavefield parameters of the non-overlapping grids are determined to be equal to the original wavefield parameters. When the difference between the second calibration value and the original wavefield parameters of the non-overlapping grids is greater than or equal to the second threshold, the corrected wavefield parameters of the non-overlapping grids are calculated based on the weights of each compacted grid in the plurality of compacted grids and the corrected wavefield parameters.

[0114] like Figure 3 As shown, this disclosure provides a construction area compaction system, the system comprising: The acquisition module 3110 is used to acquire the geological parameters of the construction area and set the target compaction parameters according to the target project of the construction area. The planning module 3120 is used to input the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment into the artificial intelligence (AI) model before compaction of the construction area to obtain the first compaction planning parameters. The first compaction planning parameters include construction parameters for M stages. The construction parameters for each stage include the stage compaction target and the stage construction parameters of the compaction equipment. The detection module 3130 is used to detect the actual compaction parameters after the nth stage. Prediction module 3140 is used to input the actual compaction parameters, the phased compaction target of the nth stage, and the construction parameters supported by the current compaction equipment into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted. The optimization module 3150 is used to adopt a parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent phase compaction targets of the original plan as a hard constraint, to adjust the construction parameters of at least the (n+1)th phase.

[0115] In summary, the construction area compaction system provided in this disclosure can achieve the construction area compaction method provided in any of the foregoing embodiments.

[0116] Combination Figure 4As shown in the illustration, this application provides an electronic device that can be a component of a compaction system, including a processor 10 and a memory 11. Optionally, the device may further include a communication interface 12 and a bus 9. The processor 10, communication interface 12, and memory 11 can communicate with each other via the bus 9. The communication interface 12 can be used for information transmission. The processor 10 can call logical instructions in the memory 11 to execute the construction area compaction method described in the above embodiment.

[0117] Furthermore, the logical instructions in the aforementioned memory 11 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0118] The memory 11, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this application. The processor 10 executes functional applications and data processing by running the program instructions / modules stored in the memory 11, thereby implementing the construction area compaction method in the above embodiments.

[0119] The memory 11 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 11 may include high-speed random access memory and may also include non-volatile memory.

[0120] The embodiments or examples disclosed in this application are not exhaustive, but merely illustrative of some embodiments or examples, and are not intended to limit the scope of protection of this disclosure. Unless contradictory, each step in a particular embodiment or example can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment or example can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment or example can be arbitrarily interchanged. Furthermore, optional methods or examples in a particular embodiment or example can be arbitrarily combined; moreover, embodiments or examples can be arbitrarily combined. For example, some or all steps of different embodiments or examples can be arbitrarily combined, and a particular embodiment or example can be arbitrarily combined with optional methods or examples of other embodiments or examples.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0123] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0125] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for compacting a construction area, characterized in that, The method includes: Obtain the geological parameters of the construction area and set the target compaction parameters according to the target project of the construction area; Before compacting the construction area, the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment are input into the artificial intelligence (AI) model to obtain the first compaction planning parameters. The first compaction planning parameters include construction parameters for M stages. The construction parameters for each stage include the stage compaction target and the stage construction parameters of the compaction equipment. After the nth stage is completed, the actual compaction parameters are checked. The actual compaction parameters, the phased compaction target of the nth stage, and the construction parameters supported by the current rolling equipment are input into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted. A parameter adjustment strategy corresponding to the value of N0 is adopted, with the completion of all subsequent phased compaction targets of the original plan as a hard constraint, and at least the construction parameters of the (n+1)th phase are adjusted.

2. The method according to claim 1, characterized in that, The parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent phased compaction targets of the original plan as a hard constraint, includes at least one of the following adjustments to the construction parameters of the (n+1)th phase: When N0 equals 1, a single-step adjustment strategy is adopted, using the phased compaction target of the (n+1)th stage as the optimization constraint to adjust the construction parameters of the (n+1)th stage only. When N0 is greater than 1 and less than the preset threshold N1, a finite reprogramming strategy is adopted, taking the phased compaction target of the (n+N0)th stage as the optimization constraint, and jointly adjusting the construction parameters of the N0 consecutive stages starting from the (n+1)th stage. When N0 is greater than or equal to the preset threshold N1, a global replanning strategy is adopted. The completion of all subsequent phased compaction targets of the original plan is taken as a hard constraint. Based on the current actual compaction parameters, target compaction parameters and equipment parameters, a brand-new compaction planning parameter is generated.

3. The method according to claim 1 or 2, characterized in that, The parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent phased compaction targets of the original plan as a hard constraint, includes at least one of the following adjustments to the construction parameters for at least the (n+1)th phase: When determining the minimum number of stages N0 that need adjustment, the probability of occurrence and the magnitude of impact of one or more potential future interference events are predicted based on the AI ​​model. When N0 equals 1, a forward-looking single-step adjustment strategy is adopted. While meeting the compaction target of the (n+1)th stage, the adjusted construction parameters are also used to actively offset the future potential disturbances with the greatest impact in the prediction. When N0 is greater than 1 and less than the preset threshold N1, a robust finite reprogramming strategy is adopted. When optimizing the construction parameters of N0 consecutive stages, a robust optimization objective for predicting disturbance events is introduced, so that the probability that the adjusted multi-stage construction parameter sequence can still meet the stage compaction target when the disturbance occurs is higher than the preset threshold. When N0 is greater than or equal to the preset threshold N1, a resilient global replanning strategy is adopted. When generating new crushing planning parameters, one or more predicted future potential interference events are used as virtual adversarial training scenarios. The generated planning parameters have the best comprehensive performance evaluation in standard scenarios and multiple adversarial scenarios.

4. The method according to claim 2, characterized in that, The method further includes: The N1 is determined based on the size of M and the type of project in the construction area; If the values ​​of M are the same for different project types, when the project type is a dam project, N1 is located in the first value interval; when the project type is a building foundation project, N1 is located in the second value interval; when the project type is a roadbed project, N1 is located in the third value interval; wherein, the median of the first value interval is less than the median of the second value interval; and the median of the third value interval is less than the median of the second value interval.

5. The method according to claim 1 or 2, characterized in that, The AI ​​model employs an adaptive multi-objective optimization algorithm based on project type to optimize construction parameters, and the objective function used during optimization is: Among them, the weighting coefficient to Dynamically adjusted according to project type: For dam projects: and Greater than ,and Greater than as well as ; For building foundation engineering: and Greater than ;and Greater than as well as ; For road foundation engineering: Greater than , or ; Greater than ; The actual compaction degree in stage i; Target compaction degree; Let be the water permeability coefficient in stage i; The target water permeability coefficient; is the compaction uniformity deviation coefficient for the i-th stage; The target compaction uniformity deviation coefficient; Let be the resilient modulus of the top surface of the roadbed in the i-th stage; The resilient modulus of the target roadbed top surface in the construction area; Let be the compaction energy of the j-th pass of the i-th stage; Let be the number of rolling passes in the i-th stage.

6. The method according to claim 1 or 2, characterized in that, The method further includes: Based on the engineering type of the construction area, the key geological parameters of the geological parameters are determined and the weights of the key geological parameters are set; the key geological parameters are used for the determination and adjustment of construction parameters.

7. The method according to claim 1 or 2, characterized in that, The method further includes: control of the compaction equipment during the compaction operation phase, including: The rolling planning parameters output by the AI ​​model are processed by the path smoothing module to generate a series of path points. Target heading angle Target speed The resulting compaction track; The position of the equipment's center of gravity is obtained by deploying a positioning module on the compaction equipment. Heading angle Roll angle and pitch angle ; The trajectory tracking controller is based on the current position with target path point Calculate the lateral position deviation and heading angle deviation ; Using model predictive control algorithm, front wheel steering angle and drive / brake torque To control the variables, with the goal of minimizing lateral and directional deviations while satisfying vehicle dynamics constraints, the optimal front wheel steering angle is calculated in real time. and drive / brake torque ; The attitude compensation controller receives real-time roll angle. and pitch angle ; The attitude compensation controller uses real-time attitude angles and equipment geometry models to dynamically calculate the hydraulic suspension system compensation required to maintain the vertical excitation force of the vibrating wheel. and vibration shaft speed compensation amount ; front wheel steering angle Drive / braking torque Hydraulic suspension system compensation amount and vibration shaft speed compensation amount The transmission mechanism sends signals to the steering actuator, drive motor, hydraulic suspension solenoid valve, and vibration motor frequency converter of the compaction equipment, achieving millisecond-level coordination between driving and compaction actions.

8. A compaction system for construction areas, characterized in that, The system includes: The acquisition module is used to acquire geological parameters of the construction area and set target compaction parameters according to the target project of the construction area; The planning module is used to input the geological parameters, target compaction parameters, and construction parameters supported by the compaction equipment into the artificial intelligence (AI) model before compaction of the construction area to obtain the first compaction planning parameters. The first compaction planning parameters include construction parameters for M stages. The construction parameters for each stage include the stage compaction target and the stage construction parameters of the compaction equipment. The detection module is used to detect the actual compaction parameters after the nth stage. The prediction module is used to input the actual compaction parameters, the phased compaction target of the nth stage, and the construction parameters supported by the current compaction equipment into the AI ​​model to determine the minimum number of stages N0 that need to be adjusted. The optimization module is used to adopt a parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent phase compaction targets of the original plan as a hard constraint, to adjust the construction parameters of at least the (n+1)th phase.

9. The system according to claim 8, characterized in that, The parameter adjustment strategy corresponding to the value of N0, with the completion of all subsequent phased compaction targets of the original plan as a hard constraint, includes at least one of the following adjustments to the construction parameters of the (n+1)th phase: When N0 equals 1, a single-step adjustment strategy is adopted, using the phased compaction target of the (n+1)th stage as the optimization constraint to adjust the construction parameters of the (n+1)th stage only. When N0 is greater than 1 and less than the preset threshold N1, a finite reprogramming strategy is adopted, taking the phased compaction target of the (n+N0)th stage as the optimization constraint, and jointly adjusting the construction parameters of the N0 consecutive stages starting from the (n+1)th stage. When N0 is greater than or equal to the preset threshold N1, a global replanning strategy is adopted. The completion of all subsequent phased compaction targets of the original plan is taken as a hard constraint. Based on the current actual compaction parameters, target compaction parameters and equipment parameters, a brand-new compaction planning parameter is generated.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions; when executed by a processor, the computer-executable instructions are able to implement the method described in any one of claims 1 to 7.

Images