Multi-source error fusion compensation control method in logistics line docking process

Abstract

The invention relates to the technical field of logistics management, and particularly discloses a multi-source error fusion compensation control method in a logistics line docking process, which comprises the following steps: acquiring environmental parameters and equipment operation data in real time by deploying a sensor network, and constructing an agent model in combination with a Gaussian process regression method; predicting energy consumption and environment quality performance under different control strategies; according to the invention, the system uses a particle swarm optimization algorithm to realize multi-objective collaborative optimization of a motor, illumination and a fresh air system, and the system also has an online updating capability based on incremental learning, can dynamically adapt to the use mode change of a logistics line, and improves the optimization precision through a closed-loop feedback mechanism; in an abnormal condition, the system can identify a sensor or equipment fault based on a predicted residual error and automatically switch to a standby control strategy to guarantee stable operation of the system; according to the invention, intelligentization, self-adaption and high robustness of logistics line docking error regulation and control are realized.

Description

Technical Field

[0001] This invention relates to the field of logistics management technology, specifically to a multi-source error fusion compensation control method during the connection process of logistics lines. Background Technology

[0002] With the acceleration of urbanization and the continuous increase in energy consumption in logistics lines, logistics line energy management systems are playing an increasingly important role in achieving energy conservation and emission reduction goals. In recent years, with the development of the Internet of Things (IoT), artificial intelligence, and big data analytics, intelligent logistics line control systems have gradually shifted from traditional rule-based static control to data-driven dynamic optimization and regulation. Currently, most automated logistics line systems have the ability to monitor docking errors in real time based on environmental parameters and equipment operating status, and can execute preset strategies through the logistics line controller to adjust docking errors.

[0003] The existing technology has the following shortcomings: Existing particle swarm optimization (PSO) algorithms for optimizing logistics line docking errors suffer from problems such as unreasonable target weight settings and a lack of dynamic trade-off mechanisms when facing multi-objective collaborative control. Specifically, multiple optimization objectives typically need to be considered comprehensively; however, traditional PSO algorithms use static weight configuration, making it difficult to dynamically adjust the priority of each objective according to environmental changes and user needs. Therefore, there is an urgent need to provide an optimization and control method that can achieve dynamic trade-offs among multiple objectives to improve the level of intelligent management of logistics lines. Summary of the Invention

[0004] To address the aforementioned technical problems, a multi-source error fusion compensation control method is provided for the logistics line docking process. This technical solution solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A multi-source error fusion compensation and control method for logistics line docking process includes: A sensor network deployed within the logistics line collects environmental parameters, equipment operating status, and error data in real time. The environmental parameters include temperature, humidity, and belt tension in the logistics line. The collected data is cleaned and normalized, and a surrogate model is constructed to predict control strategy errors. The particle swarm optimization algorithm combined with a surrogate model is used to collaboratively optimize the operating parameters of motors in the logistics line; Periodically use newly added operational data to incrementally train the surrogate model, enabling it to adapt to changes in logistics line usage patterns; The optimized control parameters are sent to the logistics line automation system, and the system's operating status is monitored in real time to form a closed-loop optimization control. Based on the surrogate model, residuals are predicted to identify sensor anomalies or equipment failures, and a backup control strategy is switched when an anomaly occurs.

[0006] As a further aspect of the present invention: the construction of the surrogate model for predicting control strategy errors specifically includes: Feature extraction is performed on the cleaned and normalized multi-source historical data to extract input feature vectors related to energy consumption, current, and motor speed; A nonlinear mapping relationship is established based on the Gaussian process regression method. The input feature vector is used as input, and the output is the corresponding docking coordinate error, docking placement shape error, and docking completion time error. The parameters of the Gaussian process kernel function are automatically tuned using the Bayesian optimization algorithm to improve the generalization ability of the surrogate model under different working conditions; The trained proxy model is encapsulated into a callable interface, which can be called in real time by the particle swarm optimization algorithm to evaluate the multi-objective performance of candidate solutions.

[0007] As a further aspect of the present invention: the proxy model further includes: After each particle swarm optimization iteration, the newly generated control strategy and its corresponding actual operation feedback data are added to the training sample set; A sliding window mechanism is used to retain valid data from the most recent N days, while outdated or abnormal samples are removed to prevent model drift. An incremental learning strategy is used to update the Gaussian process regression model online, enabling the surrogate model to dynamically adapt to changes in logistics line usage patterns. By combining the confidence intervals predicted by the model, an exploration and utilization balance mechanism is introduced during the optimization process.

[0008] As a further aspect of the present invention: the method of using particle swarm optimization algorithm combined with a surrogate model to collaboratively optimize the operating parameters of motors in the logistics line specifically includes: A multi-dimensional decision variable space containing the motor set current and start / stop time is constructed as the search space for the particle swarm optimization algorithm; Based on the surrogate model, the docking coordinate error, docking placement shape error, and docking completion time error are quickly predicted for each particle under the corresponding control strategy. A multi-objective fitness function is designed, which comprehensively considers docking coordinate error, docking placement error, and docking completion time error, and dynamically adjusts the weights of each objective through fuzzy logic. The optimal combination of control parameters is iteratively searched using the particle swarm optimization algorithm under the guidance of the multi-objective fitness function, and the result is output to the control execution module.

[0009] As a further aspect of the present invention: the collaborative optimization of the operating parameters of the motors in the logistics line includes: A local exploration mechanism is introduced in each optimization iteration to perform a fine-grained search for control strategies in the neighborhood of the current optimal solution, thereby improving the accuracy of the solution. A memory database is established by combining historical optimization results to record the logistics line load and environmental response patterns in different time periods, which can be used to assist in rapid convergence under similar future scenarios. The optimization results are compared with the actual feedback error to correct the prediction bias of the surrogate model.

[0010] As a further aspect of the present invention: the periodic incremental training of the proxy model using newly added running data includes the following steps: Newly generated operational data is collected and stored periodically, including real-time energy consumption information, environmental parameter acquisition results, and equipment status monitoring values. Outlier detection and removal are performed on newly added data, and standardization is applied to improve data consistency. An incremental learning algorithm is used to gradually integrate new data into the existing agent model, updating the model parameters without losing historical knowledge; The predictive performance of the updated surrogate model was evaluated using cross-validation. Once the model met the set accuracy requirements, it was put into use in the next optimization cycle.

[0011] As a further aspect of the present invention: the incremental training of the proxy model includes: Time series analysis techniques are used to identify trends in logistics line usage patterns that change with seasons or time, and the learning strategy of the agent model is dynamically adjusted accordingly. A sliding window mechanism is set up to retain only the running data within a preset time range as model training samples, and discard outdated data to improve the model's response speed; Adjust the correction weights of the proxy model's output based on user feedback to make the model's predictions more closely resemble subjective evaluation metrics in real-world application scenarios; Establish a model version management mechanism to record the time of each model update, the source of training data, and the performance evaluation results.

[0012] As a further aspect of the present invention: the step of sending the optimized control parameters to the logistics line automation system and monitoring the system's operating status in real time includes the following steps: The optimal control parameters output by the multi-objective optimization and control module are converted into an instruction format that conforms to the standards of automated logistics line systems. The above instructions are sent to the corresponding logistics line equipment control system using a communication protocol; Real-time acquisition of operational data from each subsystem within the automated logistics line system, including docking coordinate errors, environmental parameters, and equipment operating status; The effectiveness of the current control strategy is evaluated based on real-time operational data, and the evaluation results are fed back to achieve closed-loop control.

[0013] As a further aspect of the present invention: the closed-loop optimization control includes: Before issuing control commands, simulation tools are used to rehearse the control strategy to be implemented, predict its potential impact, and adjust the control parameters based on the rehearsal results. A fault-tolerance mechanism is introduced to automatically switch to a backup control strategy when a communication failure or equipment malfunction is detected, ensuring system continuity and stability. By comparing the actual operating results with the expected goals, machine learning algorithms are used to analyze the causes of deviations, and the proxy model and algorithm parameters are dynamically adjusted and optimized accordingly. Detailed operation reports are generated regularly to record the implementation status of control strategies, equipment response time, and energy savings, which are used for subsequent performance audits and optimization decisions.

[0014] As a further aspect of the present invention: the method of predicting residuals based on a surrogate model to identify sensor anomalies or equipment failures, and switching to a backup control strategy when an anomaly occurs, includes the following steps: Real-time data collection of actual operation from various sensors and equipment within the logistics line is used to input the operational data into a trained surrogate model to generate predicted values. By comparing the predicted values ​​output by the surrogate model with the corresponding actual measured values, the residual between the two is calculated, and a threshold is set to determine whether there is a significant deviation. If the residual is detected to continuously exceed the set threshold, a multidimensional data analysis mechanism is activated to identify the anomaly type by combining historical data and statistical analysis methods, distinguishing between sensor drift, data transmission errors, or equipment failure. Once an anomaly is confirmed, the preset backup control strategy is automatically activated to maintain stable system operation, and alarm signals are triggered simultaneously to notify maintenance personnel to intervene. At the same time, the abnormal event is recorded for subsequent diagnosis and model correction.

[0015] The beneficial effects of this invention: (1) This invention uses a multi-source data acquisition module to perform real-time sensing and dynamic monitoring of the internal environmental parameters and equipment operating status of the logistics line, constructing a high spatiotemporal resolution data acquisition system. This system covers key environmental indicators such as temperature, humidity, and belt tension in the logistics line, as well as the operating data of major energy-consuming equipment such as motors in the logistics line. Based on this, the system uses the Gaussian process regression method to establish a nonlinear surrogate model, which can efficiently approximate the dynamic response behavior of complex logistics line systems and accurately predict the energy consumption performance and indoor environmental quality change trends under different control strategies. This modeling method not only has good nonlinear fitting ability but also quantifies the uncertainty of the prediction results, providing reliable confidence support for subsequent optimization. Furthermore, the system introduces a multi-objective optimization control mechanism based on the particle swarm optimization algorithm, which, with the support of the surrogate model, achieves collaborative optimization of multiple controllable variables such as motor setting current and start-stop time. By designing a multi-objective fitness function that integrates docking coordinate error, docking placement form error, and docking completion time error, and combining fuzzy logic to achieve dynamic adjustment of weights, the system can achieve adaptive trade-offs among multi-dimensional optimization objectives, ensuring that multiple errors can be integrated and considered, thereby achieving comprehensive optimization. In addition, the system integrates a local search mechanism, a historical memory bank, and a feedback correction strategy, which significantly improves the convergence speed and solution quality of the optimization process.

[0016] (2) This invention integrates online update and self-learning modules and constructs an intelligent control system with continuous evolution capabilities and high robustness. The online update mechanism relies on an incremental learning strategy, enabling the surrogate model to periodically integrate the latest operational data, achieving dynamic updates of model parameters without forgetting historical knowledge. Combining a sliding window mechanism and time series analysis methods, the system can effectively identify the evolutionary trends of logistics line usage patterns caused by external factors such as seasonal changes and time variations, and adjust the model training strategy and optimization objectives accordingly. This ensures that the surrogate model maintains high-precision prediction capabilities during long-term operation, significantly improving the system's adaptability and optimization stability. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a multi-source error fusion compensation control method for logistics line docking processes according to the present invention. Detailed Implementation

[0018] The following description is intended to disclose the invention so that those skilled in the art can implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Reference Figure 1 As shown, a multi-source error fusion compensation and control method for logistics line docking includes: A sensor network deployed within the logistics line collects environmental parameters, equipment operating status, and error data in real time. The environmental parameters include temperature, humidity, and belt tension in the logistics line. The collected data is cleaned and normalized, and a surrogate model is constructed to predict control strategy errors. The particle swarm optimization algorithm combined with a surrogate model is used to collaboratively optimize the operating parameters of motors in the logistics line; Periodically use newly added operational data to incrementally train the surrogate model, enabling it to adapt to changes in logistics line usage patterns; The optimized control parameters are sent to the logistics line automation system, and the system's operating status is monitored in real time to form a closed-loop optimization control. Based on the surrogate model, residuals are predicted to identify sensor anomalies or equipment failures, and a backup control strategy is switched when an anomaly occurs.

[0020] It's easy to understand that the logistics line mainly uses transport vehicles, robotic arms, and belts for transfer. Therefore, the transfer is mainly achieved through belts and motors. Motors are almost unaffected by the environment, but belts are affected by temperature and humidity, and their surface tension also affects the accuracy of the connection. Therefore, all of these factors need to be considered. In the multi-source data acquisition module, environmental parameters, equipment operating status, and error data are collected in real time through a sensor network deployed within the logistics line. The environmental parameters include temperature, humidity, and belt tension in the logistics line, specifically including: In a preferred embodiment of the present invention, the multi-source data acquisition module includes a wireless sensor network deployed in multiple functional areas within the logistics line. The sensor network comprises distributed temperature and humidity sensors, belt tension detection devices within the logistics line, and smart meters. The temperature and humidity sensors are uniformly distributed throughout the space of the logistics line to acquire real-time air temperature and relative humidity data for each area. Various sensors are installed within the logistics line to acquire the required data.

[0021] Furthermore, the sensor network employs a low-power communication protocol for networking and utilizes edge computing nodes to achieve local preliminary data fusion and timestamp synchronization, ensuring consistency of data from different sources across time dimensions. All collected data is uploaded to a cloud data platform via a gateway for subsequent preprocessing, modeling, and optimization modules. In addition, the sensor nodes possess self-diagnostic capabilities, proactively reporting status information when abnormal signals or equipment malfunctions are detected, providing early warning support for system maintenance.

[0022] This implementation method, by constructing a high spatiotemporal resolution data acquisition system, achieves comprehensive perception of logistics line environmental parameters, equipment operating status, and energy consumption characteristics, providing a high-quality data foundation for subsequent agent model training and multi-objective optimization and control.

[0023] In the data preprocessing and modeling module, the collected data is cleaned and normalized, and a surrogate model for predicting control strategy errors is constructed, specifically including: In a preferred embodiment of the present invention, the data preprocessing and modeling module includes data cleaning, normalization, and surrogate model construction processes executed sequentially. First, the system receives raw environmental parameters, equipment operating status, and error data from the multi-source data acquisition module, and performs noise reduction, missing value imputation, and outlier removal on these data to improve the data quality for subsequent modeling. Then, the cleaned historical data is normalized to ensure comparability of variables with different dimensions.

[0024] Furthermore, in the feature extraction stage, the system extracts input feature vectors closely related to the energy consumption optimization target of the logistics line from the processed data. These input features include, but are not limited to, current indoor temperature, relative humidity, belt tension in the logistics line, outdoor weather conditions, equipment operating time, and historical control strategy information. Based on these input features, a nonlinear mapping relationship is established using Gaussian process regression, where the input is the feature vector and the output is the corresponding predicted value, including the docking coordinate error, docking placement error, and docking completion time error per unit time.

[0025] Furthermore, to enhance the generalization ability of the surrogate model across different logistics line application scenarios, the system introduces a Bayesian optimization algorithm to automatically tune the hyperparameters of the Gaussian process regression kernel function, thereby obtaining the optimal model structure. The trained surrogate model is encapsulated into a standardized, callable interface, allowing the particle swarm optimization algorithm to quickly evaluate the multi-objective performance of candidate control strategies during each iteration.

[0026] In addition, the system includes an online update mechanism for the surrogate model. After each particle swarm optimization iteration, the system adds the newly generated control strategy and its corresponding actual operational feedback data to the training sample set. Subsequently, a sliding window mechanism is used to retain the valid operational data from the most recent N days, while removing outdated or abnormal samples to prevent the model from failing due to data drift. Based on this, the system uses an incremental learning strategy to dynamically update the Gaussian process regression model, ensuring that the surrogate model can adapt to changes in the logistics line's usage patterns.

[0027] Finally, during the optimization process, the system combines the confidence intervals output by the surrogate model to dynamically adjust the search strategy of the particle swarm optimization algorithm, achieving a balance between exploring the unknown control space and utilizing known efficient strategies, thereby improving the global optimization capability and optimization stability.

[0028] This implementation method, by introducing a surrogate modeling approach that combines Gaussian process regression with incremental learning, achieves efficient approximation and dynamic updating of responses to complex logistics line environments, significantly improving optimization efficiency and system adaptability.

[0029] In the multi-objective optimization and control module, a particle swarm optimization algorithm combined with a surrogate model is used to collaboratively optimize the operating parameters of the motors in the logistics line, specifically including: In a preferred embodiment of the present invention, the multi-objective optimization and control module includes an intelligent collaborative optimization mechanism based on a combination of particle swarm optimization algorithm and surrogate model, used for unified scheduling and dynamic optimization of the operating parameters of motors in the logistics line. First, the system constructs a multi-dimensional decision variable space containing motor set current and start-stop time, and defines it as the search space of the particle swarm optimization algorithm. Each particle represents a complete set of control strategies, covering the main controllable variables of the logistics line environment.

[0030] Based on this, the system quickly predicts the performance of the control strategy for each particle in actual operation by calling a pre-trained surrogate model, including the docking coordinate error, docking placement error, and belt tension in the logistics line per unit time. Subsequently, the system designs a multi-objective fitness function that comprehensively considers the docking coordinate error, docking placement error, and belt tension in the logistics line, and introduces a fuzzy logic control strategy to dynamically adjust the weight distribution among the objectives according to the current environmental state, thereby achieving adaptive trade-offs among multiple objectives.

[0031] Furthermore, during each particle swarm optimization iteration, the system introduces a local exploration mechanism to perform a refined search for control strategies within the neighborhood of the current optimal solution, thereby improving the accuracy and stability of the final solution. Simultaneously, the system establishes a memory of historical optimization results, recording the trends in logistics line load changes and environmental response characteristics over different time periods. This memory assists in rapid convergence under similar future scenarios, reducing redundant computational overhead.

[0032] Finally, the system compares and analyzes the control parameters output by each optimization with the actual execution feedback data, identifies the prediction error of the surrogate model, and feeds back the deviation information to correct the prediction capability of the surrogate model in reverse, forming a closed-loop enhanced optimization system.

[0033] This implementation method integrates particle swarm optimization algorithm with surrogate model prediction capabilities, and combines dynamic weight adjustment, local search, historical memory and feedback correction mechanisms to achieve efficient collaborative control of multi-equipment systems in the logistics line and control overall error.

[0034] Periodically use newly added operational data to incrementally train the surrogate model, enabling it to adapt to changes in logistics line usage patterns. Specifically, this includes: In a preferred embodiment of the present invention, the online update and self-learning module includes a dynamic update process for the agent model based on an incremental learning mechanism. This process periodically retrains the agent model using newly added operational data, enabling it to continuously adapt to changes in the logistics line's usage patterns. Specifically, the system periodically acquires newly generated operational data from a multi-source data acquisition module. This operational data includes, but is not limited to, real-time energy consumption information, environmental parameter acquisition results, and equipment status monitoring values, and stores the data in a local database.

[0035] Subsequently, the system performs preprocessing operations on the new data, including outlier detection and removal, missing value imputation, and standardization, to improve data quality and ensure consistency between the new samples and historical data in terms of dimensions and distribution. After preprocessing, the system uses an incremental learning algorithm to gradually integrate the new samples into the existing Gaussian process regression surrogate model, achieving online updates of model parameters without losing existing knowledge.

[0036] Furthermore, the system evaluates the predictive performance of the updated surrogate model using cross-validation, comparing its performance across multiple key metrics. Once the model's predictive accuracy meets the set threshold requirements, the system deploys the updated surrogate model to the next optimization cycle, supporting multi-objective optimization decisions that more closely reflect the current operating conditions.

[0037] Furthermore, the system incorporates time series analysis technology to identify trends in logistics line usage patterns that change seasonally or over time. Based on this, the system dynamically adjusts the learning strategy of the proxy model, such as adjusting the learning rate or increasing the weight of data within specific time periods. Simultaneously, a sliding window mechanism is implemented to retain only operational data within a preset time range as the training sample set, automatically discarding outdated data to prevent model drift due to historical noise interference, thereby improving model response speed and generalization ability.

[0038] To further enhance the applicability of the proxy model in practical application scenarios, the system adjusts the correction weights of the model output results based on user feedback, making the model predictions more aligned with unstructured indicators such as subjective current scoring or motor speed perception evaluation. Finally, the system establishes a model version management mechanism, recording the time node of each model update, the source of the training data used, key performance evaluation indicators, and the reason for model changes, facilitating subsequent tracking of model evolution paths, optimization of iteration efficiency, and fault tracing.

[0039] This implementation significantly improves the agent model's adaptability to changes in complex logistics line environments and its long-term predictive stability by constructing an online update system with incremental learning, time series awareness, user feedback fusion, and version management functions, thereby enhancing the self-learning and adaptive capabilities of the entire intelligent control system.

[0040] The optimized control parameters are sent to the logistics line automation system, and the system's operating status is monitored in real time to form a closed-loop optimization control, which specifically includes: In a preferred embodiment of the present invention, the control execution and feedback module includes core functions such as control command conversion, communication transmission, operation status monitoring and feedback optimization. It is used to send the optimal control parameters output by the multi-objective optimization and control module to the logistics line automation system and monitor the system operation status in real time to form a closed-loop optimization control system.

[0041] Specifically, the system first receives the optimal control parameters output from the multi-objective optimization and control module, including key variables such as motor set current and start / stop time, and converts them into instruction formats that conform to the standards of logistics line automation systems, such as data structures supported by BACnet or Modbus protocols. Subsequently, the system sends the standardized control instructions to the corresponding logistics line equipment control system via wired or wireless communication protocols.

[0042] During the execution of the control strategy, the system acquires real-time operational data from each subsystem within the automated logistics line, including docking coordinate errors, indoor temperature and humidity, belt tension, equipment start / stop status, and power consumption. Based on this operational data, the system evaluates the effectiveness of the current control strategy. The evaluation results are fed back as crucial input for the next round of optimization iterations, thus achieving closed-loop control.

[0043] Furthermore, before each control command is issued, the system uses simulation tools to rehearse the control strategy to be implemented, predicting the potential environmental response and energy consumption fluctuations. Based on the rehearsal results, the system dynamically adjusts the control parameters to avoid adverse effects. Simultaneously, the system incorporates a fault-tolerance mechanism. When abnormal situations such as communication interruptions, equipment failures, or response delays are detected, the system automatically switches to a preset backup control strategy to ensure the continuity and stability of system operation.

[0044] Furthermore, the system continuously compares the actual operating results with the expected targets, uses machine learning algorithms to analyze the sources of deviation, identifies whether there are model prediction errors, equipment response lags, or external interference factors, and dynamically adjusts the prediction logic of the proxy model and the search direction of the optimization algorithm accordingly. The system also has the ability to generate operation reports, regularly outputting detailed operation reports including control strategy execution status, equipment response time, energy savings, and user satisfaction ratings, for subsequent performance audits, system optimization, and energy efficiency management decision support.

[0045] This implementation method significantly improves the execution accuracy and closed-loop response capability of the logistics line energy consumption optimization system by constructing a control execution and feedback system with capabilities such as instruction conversion, communication control, real-time monitoring, simulation and pre-drilling, fault tolerance processing, and feedback optimization. It achieves efficient connection and continuous improvement from optimization decision-making to actual control.

[0046] Based on the surrogate model, residual identification of sensor anomalies or equipment failures is performed, and a backup control strategy is switched to when an anomaly occurs. Specifically, this includes: In a preferred embodiment of the present invention, the anomaly detection and fault-tolerant control module includes an anomaly identification mechanism based on surrogate model prediction residual analysis and a corresponding fault-tolerant response process, which is used to monitor the operating status of sensors and equipment inside the logistics line in real time, and automatically switch to a backup control strategy when an anomaly occurs to ensure stable system operation.

[0047] Specifically, the system acquires real-time operational data from various sensors and devices deployed within the logistics line through a multi-source data acquisition module. This operational data includes, but is not limited to, key parameters such as indoor temperature and humidity, belt tension in the logistics line, equipment start / stop status, and communication signal strength. This operational data is then input into a pre-trained Gaussian process regression surrogate model to generate predicted values ​​for the corresponding environmental conditions and equipment behavior.

[0048] Subsequently, the system compares the predicted values ​​output by the proxy model with the actual measured values ​​item by item, calculates the prediction residuals between the two, and determines whether there are any anomalies by setting a reasonable deviation threshold. If the calculated residuals continuously exceed the preset threshold, the system activates a multi-dimensional data analysis mechanism to further identify the anomaly type by combining historical operating data and statistical analysis methods, distinguishing possible causes of the anomaly, including sensor drift, data transmission errors, equipment response delays, or hardware failures.

[0049] Upon confirmation of an anomaly, the system automatically activates a backup control strategy matching the current operating conditions to maintain minimum stable motor operation and prevent overall control failure due to localized faults. Simultaneously, the system triggers an alarm signal and notifies maintenance personnel via the human-machine interface for on-site inspection and intervention. All abnormal events and their corresponding operational data are recorded in the database for subsequent diagnostic analysis and proxy model correction.

[0050] Furthermore, the system supports dynamically adjusting the residual threshold and detection sensitivity based on the frequency and type of anomalies, improving the adaptability and accuracy of anomaly identification. This mechanism not only enhances the system's robustness in complex operating environments but also provides strong support for the long-term reliable operation of the logistics line energy consumption optimization system.

[0051] The working principle of this invention: This invention includes a multi-source data acquisition module, a data preprocessing and modeling module, a multi-objective optimization and control module, an online update and self-learning module, a control execution and feedback module, and an anomaly detection and fault-tolerant control module. Through a sensor network deployed within the logistics line, the system can collect environmental parameters such as temperature, humidity, and belt tension, as well as motor operating data in the logistics line in real time. A surrogate model is constructed using Gaussian process regression to predict error performance under different control strategies. Based on this, combined with particle swarm optimization and a dynamic weight adjustment mechanism, the system achieves collaborative optimization of operating parameters for multiple devices. Simultaneously, the system possesses online incremental learning capabilities, continuously updating the surrogate model based on new operating data to adapt to changes in logistics line usage patterns. The optimized control parameters are sent to the logistics line automation system via a standardized communication protocol, and simulation pre-running, fault-tolerant switching, and feedback correction mechanisms are introduced during execution to form closed-loop control. Furthermore, the system performs anomaly detection based on the surrogate model's predicted residuals, automatically identifying sensor drift or equipment failure and switching to a backup control strategy to ensure stable system operation. This invention significantly improves the control of docking errors in logistics lines by constructing an intelligent control system that integrates perception, modeling, optimization, execution, and fault tolerance.

[0052] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for multi-source error fusion compensation and control during logistics line docking, characterized in that... ,include: A sensor network deployed within the logistics line collects environmental parameters, equipment operating status, and error data in real time. The environmental parameters include temperature, humidity, and belt tension in the logistics line. The collected data is cleaned and normalized, and a surrogate model is constructed to predict control strategy errors. The particle swarm optimization algorithm combined with a surrogate model is used to collaboratively optimize the operating parameters of motors in the logistics line; Periodically use newly added operational data to incrementally train the surrogate model, enabling it to adapt to changes in logistics line usage patterns; The optimized control parameters are sent to the logistics line automation system, and the system's operating status is monitored in real time to form a closed-loop optimization control. The system uses a proxy model to predict residuals to identify sensor anomalies or equipment failures, and switches to a backup control strategy when an anomaly occurs.

2. The multi-source error fusion compensation control method for logistics line docking process according to claim 1, characterized in that... The construction of the surrogate model for predicting control strategy errors specifically includes: Feature extraction is performed on the cleaned and normalized multi-source historical data to extract input feature vectors related to energy consumption, current, and motor speed; A nonlinear mapping relationship is established based on the Gaussian process regression method. The input feature vector is used as input, and the output is the corresponding docking coordinate error, docking placement shape error, and docking completion time error. The parameters of the Gaussian process kernel function are automatically tuned using the Bayesian optimization algorithm to improve the generalization ability of the surrogate model under different working conditions; The trained proxy model is encapsulated into a callable interface, which can be called in real time by the particle swarm optimization algorithm to evaluate the multi-objective performance of candidate solutions.

3. The multi-source error fusion compensation control method for logistics line docking process according to claim 2, characterized in that... The proxy model further includes: After each particle swarm optimization iteration, the newly generated control strategy and its corresponding actual operation feedback data are added to the training sample set; A sliding window mechanism is used to retain valid data from the most recent N days, while outdated or abnormal samples are removed to prevent model drift. An incremental learning strategy is used to update the Gaussian process regression model online, enabling the surrogate model to dynamically adapt to changes in logistics line usage patterns. By combining the confidence intervals predicted by the model, an exploration and utilization balance mechanism is introduced during the optimization process.

4. The multi-source error fusion compensation control method for logistics line docking process according to claim 3, characterized in that... The method of using particle swarm optimization algorithm combined with surrogate model to collaboratively optimize the operating parameters of motors in the logistics line specifically includes: A multi-dimensional decision variable space containing the motor set current and start / stop time is constructed as the search space for the particle swarm optimization algorithm; Based on the surrogate model, the docking coordinate error, docking placement shape error, and docking completion time error are quickly predicted for each particle under the corresponding control strategy. A multi-objective fitness function is designed, which comprehensively considers docking coordinate error, docking placement error, and docking completion time error, and dynamically adjusts the weights of each objective through fuzzy logic. The optimal combination of control parameters is iteratively searched using the particle swarm optimization algorithm under the guidance of the multi-objective fitness function, and the result is output to the control execution module.

5. The multi-source error fusion compensation control method for logistics line docking process according to claim 4, characterized in that... The collaborative optimization of the operating parameters of the motors in the logistics line includes: A local exploration mechanism is introduced in each optimization iteration to perform a fine-grained search for control strategies in the neighborhood of the current optimal solution, thereby improving the accuracy of the solution. A memory database is established by combining historical optimization results to record the logistics line load and environmental response patterns in different time periods, which can be used to assist in rapid convergence under similar future scenarios. The optimization results are compared with the actual feedback error to correct the prediction bias of the surrogate model.

6. The multi-source error fusion compensation control method for logistics line docking process according to claim 5, characterized in that... The periodic incremental training of the agent model using newly added running data includes the following steps: Newly generated operational data is collected and stored periodically, including real-time energy consumption information, environmental parameter acquisition results, and equipment status monitoring values. Outlier detection and removal are performed on newly added data, and standardization is applied to improve data consistency. An incremental learning algorithm is used to gradually integrate new data into the existing agent model, updating the model parameters without losing historical knowledge; The predictive performance of the updated surrogate model was evaluated using cross-validation. Once the model met the set accuracy requirements, it was put into use in the next optimization cycle.

7. The multi-source error fusion compensation control method for logistics line docking process according to claim 6, characterized in that... The incremental training of the proxy model includes: Time series analysis techniques are used to identify trends in logistics line usage patterns that change with seasons or time, and the learning strategy of the agent model is dynamically adjusted accordingly. A sliding window mechanism is set up to retain only the running data within a preset time range as model training samples, and discard outdated data to improve the model's response speed; Adjust the correction weights of the proxy model's output based on user feedback to make the model's predictions more closely resemble subjective evaluation metrics in real-world application scenarios; Establish a model version management mechanism to record the time of each model update, the source of training data, and the performance evaluation results.

8. The multi-source error fusion compensation control method for logistics line docking process according to claim 7, characterized in that... The process of sending the optimized control parameters to the logistics line automation system and monitoring the system's operating status in real time includes the following steps: The optimal control parameters output by the multi-objective optimization and control module are converted into an instruction format that conforms to the standards of automated logistics line systems. The above instructions are sent to the corresponding logistics line equipment control system using a communication protocol; Real-time acquisition of operational data from each subsystem within the automated logistics line system, including docking coordinate errors, environmental parameters, and equipment operating status; The effectiveness of the current control strategy is evaluated based on real-time operational data, and the evaluation results are fed back to achieve closed-loop control.

9. The multi-source error fusion compensation control method for logistics line docking process according to claim 8, characterized in that... The closed-loop optimization control includes: Before issuing control commands, simulation tools are used to rehearse the control strategy to be implemented, predict its potential impact, and adjust the control parameters based on the rehearsal results. A fault-tolerance mechanism is introduced to automatically switch to a backup control strategy when a communication failure or equipment malfunction is detected, ensuring system continuity and stability. By comparing the actual operating results with the expected goals, machine learning algorithms are used to analyze the causes of deviations, and the proxy model and algorithm parameters are dynamically adjusted and optimized accordingly. Detailed operation reports are generated regularly to record the implementation status of control strategies, equipment response time, and energy savings, which are used for subsequent performance audits and optimization decisions.

10. A multi-source error fusion compensation control method for logistics line docking process according to claim 9, characterized in that... The method of predicting residual identification of sensor anomalies or equipment failures based on a surrogate model and switching to a backup control strategy when an anomaly occurs includes the following steps: Real-time data collection of actual operation from various sensors and equipment within the logistics line is used to input the operational data into a trained surrogate model to generate predicted values. By comparing the predicted values ​​output by the surrogate model with the corresponding actual measured values, the residual between the two is calculated, and a threshold is set to determine whether there is a significant deviation. If the residual is detected to continuously exceed the set threshold, a multidimensional data analysis mechanism is activated to identify the anomaly type by combining historical data and statistical analysis methods, distinguishing between sensor drift, data transmission errors, or equipment failure. Once an anomaly is confirmed, the preset backup control strategy is automatically activated to maintain stable system operation, and alarm signals are triggered simultaneously to notify maintenance personnel to intervene. At the same time, the abnormal event is recorded for subsequent diagnosis and model correction.

Images