Device control method, device and medium

Abstract

The application provides a device control method, device and medium. The method first acquires actual state data of a target device running according to a preset task target and simulation state data of a digital twin model of the target device running according to the preset task target, determines the difference between the target device and the digital twin model based on the actual state data and the simulation state data, adjusts the digital twin model based on the difference, simulates and predicts each control algorithm based on the preset task target and the adjusted digital twin model to obtain a simulation and prediction control result of each control algorithm, determines a target control algorithm from the multiple control algorithms according to the simulation and prediction control results, and controls the target device according to the target control algorithm to update the actual state data. The embodiments of the application can improve the device control precision.

Description

Technical Field

[0001] This application relates to the field of equipment control technology, and more specifically, to an equipment control method, equipment, and storage medium. Background Technology

[0002] With the rapid iteration of IoT, big data and artificial intelligence technologies, digital twin technology, with its core advantage of bidirectional mapping between "physical entity and virtual model", continues to expand in the depth and breadth of its application in the equipment field.

[0003] Equipment control systems built upon digital twin technology can break down information barriers between the physical world and virtual space, providing technical support for equipment operation status monitoring, fault prediction, and efficiency optimization. However, current mainstream digital twin systems still face numerous technical bottlenecks in the planning and control of transport equipment, making it difficult to meet the precise control requirements in complex and dynamic environments. Summary of the Invention

[0004] In view of this, this application provides a device control method, device, and storage medium to at least solve the problems existing in the related art.

[0005] Specifically, this application is implemented through the following technical solution: This application provides a device control method, including: Acquire the actual state data of the target device operating according to the preset task objectives, and the simulation state data of the digital twin model of the target device operating according to the preset task objectives; Based on the actual state data and the simulation state data, the differences between the target device and the digital twin model are determined, and the digital twin model is adjusted based on the differences to obtain the adjusted digital twin model. Based on the preset task objectives, the adjusted digital twin model is used to perform simulation prediction on multiple preset control algorithms to obtain the simulation prediction control results of each control algorithm; the simulation prediction control results include prediction control information and control evaluation results corresponding to the prediction control information. Based on the simulation predictive control results corresponding to the multiple control algorithms, a target control algorithm is determined from the multiple control algorithms, and the target device is controlled according to the target control algorithm so that the target device updates the actual state data.

[0006] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the device control methods described in the foregoing embodiments.

[0007] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the device control methods described in the foregoing embodiments.

[0008] This application also provides a computer program product, including a computer program that, when run by a processor, performs the steps of any of the possible device control methods described above.

[0009] The technical solutions provided by the embodiments of this application may include the following beneficial effects: In this embodiment, a closed-loop process of "data acquisition, model calibration, algorithm testing, equipment control, and data update" is formed, realizing an essential upgrade of equipment control from "passive monitoring of anomalies" to "active optimization and regulation". This enables continuous optimization of control effects in complex and ever-changing operating environments, thereby improving the reliability, adaptability, and overall operational efficiency of the target equipment.

[0010] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a device control method according to an exemplary embodiment of this application; Figure 2 This is a flowchart illustrating a device control method according to an exemplary embodiment of this application; Figure 3 This is a flowchart illustrating a difference estimation method as shown in an exemplary embodiment of this application; Figure 4 This is a flowchart illustrating a difference estimation method in an exemplary embodiment of this application; Figure 5 This is a hardware structure diagram of an electronic device illustrated in an exemplary embodiment of this application. Detailed Implementation

[0012] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0013] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0014] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0015] With the rapid iteration of IoT, big data, and AI technologies, digital twin technology, with its core advantage of bidirectional mapping between physical entities and virtual models, continues to expand in the depth and breadth of its application in the equipment field. Real-time closed-loop digital twin systems, by constructing high-fidelity virtual models, achieve dynamic mapping, simulation, and data feedback of the entire lifecycle operational status of physical equipment. This has become a key technology driving the intelligent and precise upgrading of transportation systems, demonstrating broad application prospects in the scheduling and management of various types of equipment.

[0016] Equipment management and control systems built upon digital twin technology can break down information barriers between the physical world and virtual space, providing technical support for equipment operation status monitoring, fault prediction, and efficiency optimization. However, current mainstream digital twin systems still face numerous technical bottlenecks in the planning and control of transport equipment, making it difficult to meet the precise control requirements under complex dynamic environments: First, real-time response capabilities are insufficient, with delays in synchronizing the virtual model with the physical equipment's state, resulting in weak adaptability to dynamic environments; second, coordination between control strategies and physical entities lags behind, with insufficient matching between model adjustments and actual equipment operation requirements, failing to achieve precise adaptation under dynamic operating conditions; third, the overall system closed-loop efficiency is limited, with most solutions only remaining at the level of status monitoring and prediction, making it difficult to form a complete closed loop of "monitoring-prediction-decision-control-feedback," thus hindering further improvements in the overall system's operational efficiency and safety performance.

[0017] In related technologies, some patents have applied digital twin technology to industrial control and monitoring scenarios, using digital twin technology to achieve state monitoring and data processing of specific physical objects. However, it usually has two major defects: First, the model adjustment mechanism is rigid, generally adopting the mode of "fitting physical laws + static calibration". It only corrects model parameters through fixed rules such as double consistency assimilation and self-calibration update, lacking a fine adjustment mechanism for the dynamic gap between the digital twin model and the physical equipment. In particular, it lacks targeted fine-tuning design for the black box digital twin model in the physical simulation engine, making it difficult for the virtual model to accurately reproduce the operating state and behavioral characteristics of physical equipment under complex dynamic working conditions. Second, the functional module design is imperfect. The module architecture of existing solutions is mostly centered around "data preprocessing-model prediction-anomaly judgment", which can only output monitoring data or early warning results. It does not have the function of testing, evaluating and intelligently selecting control algorithms, and cannot directly affect the control decision of physical equipment. It cannot complete the closed-loop management of the entire process of "simulation-decision-control-feedback". Essentially, it still belongs to the category of "passive monitoring" and is difficult to achieve active planning and precise control of physical equipment.

[0018] In summary, the current application of digital twin technology in the field of physical equipment is still limited to passive monitoring. It has not yet formed a real-time closed-loop framework that combines dynamic model adaptation capabilities with intelligent control decision-making capabilities, and cannot meet the intelligent needs of equipment scheduling and management in complex dynamic environments.

[0019] Based on the above research, this disclosure provides a device control method. The method first acquires actual state data of a target device operating according to a preset task objective, and simulated state data of a digital twin model of the target device operating according to the preset task objective. Second, based on the actual state data and the simulated state data, it determines the differences between the target device and the digital twin model, and adjusts the digital twin model based on these differences to obtain an adjusted digital twin model. Then, based on the preset task objective, it uses the adjusted digital twin model to perform simulation prediction on multiple preset control algorithms, obtaining simulation prediction control results for each control algorithm. The simulation prediction control results include prediction control information and control evaluation results corresponding to the prediction control information. Finally, based on the simulation prediction control results corresponding to the multiple control algorithms, it determines a target control algorithm from the multiple control algorithms, and controls the target device according to the target control algorithm to update the actual state data of the target device.

[0020] In this embodiment, a closed-loop process of "data acquisition, model calibration, algorithm testing, equipment control, and data update" is formed, realizing an essential upgrade of equipment control from "passive monitoring of anomalies" to "active optimization and regulation". This enables continuous optimization of control effects in complex and ever-changing operating environments, thereby improving the reliability, adaptability, and overall operational efficiency of the target equipment.

[0021] To facilitate understanding of this embodiment, a device control method disclosed in this disclosure will first be described in detail. The device control method provided in this disclosure is generally executed by an electronic device. The electronic device can be a physical entity device that needs to be controlled or monitored, including but not limited to transportation equipment, industrial equipment, medical equipment, etc.

[0022] In other embodiments, the electronic device can be a server, which can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing initial cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms. In still other embodiments, the electronic device can be a terminal device, which can be a terminal, handheld device, computing device, etc.

[0023] In other embodiments, the method can also be applied to an implementation environment consisting of electronic devices and servers, or an implementation environment consisting of terminal devices and servers. Furthermore, the device control method can also be implemented by a processor calling computer-readable instructions stored in memory.

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0025] Please see the appendix Figure 1 This is a flowchart illustrating a device control method according to an exemplary embodiment of this application. Figure 1 As shown, the device control method in this embodiment may include the following steps S101~S104: S101: Obtain the actual state data of the target device operating according to the preset task objective, and the simulation state data of the digital twin model operating according to the preset task objective.

[0026] The target equipment refers to the physical entity that needs to be controlled or monitored in operation. The types of target equipment include, but are not limited to, transportation equipment, industrial equipment, and medical equipment. For example, transportation equipment can refer to equipment with transportation functions, such as vehicles, drones, robots, and ships. Industrial equipment refers to various types of equipment applied in industrial production scenarios, including but not limited to machine tools, production line devices, industrial robots, pump and valve units, etc. Medical equipment refers to various types of equipment applied in medical scenarios, including but not limited to surgical robots, intelligent medical beds, and imaging equipment.

[0027] Preset task objectives can refer to the operational tasks or work indicators set for the target equipment. For example, for transport equipment, the corresponding preset task objective can be the fixed-line cruise in the automatic driving process. For machine tool equipment, the corresponding preset task objective can be the machining quality objective (such as accuracy, geometric tolerance, surface quality, etc.).

[0028] Actual state data refers to the real operating parameters collected by sensors when the target device is running in the physical space according to the preset task objectives, which are used to reflect the physical characteristics and operating status of the target device.

[0029] A digital twin model is a high-fidelity virtual digital model built based on the physical characteristics, structural parameters, and operating rules of a target device. It can simulate the operating behavior of the target device in a physical simulation environment to achieve monitoring, prediction, and optimization of the target device.

[0030] Simulation state data refers to the simulated operating parameters output by a digital twin model in a virtual environment when it runs according to the same preset task objectives as the target device. Its data dimensions correspond to the actual state data.

[0031] S102: Based on the actual state data and the simulation state data, determine the differences between the target device and the digital twin model, and adjust the digital twin model based on the differences to obtain the adjusted digital twin model.

[0032] In this step, the difference between the actual state data and the simulation state data is calculated to determine the source of error between the digital twin model and the target device (such as environmental factors, actuator failure, parameter perturbation, and insufficient fitting of nonlinear terms). Then, by correcting the model parameters and optimizing the simulation logic, an adjusted digital twin model that can more accurately reflect the current dynamic characteristics of the target device (physical entity) is obtained. In this way, the pain point of "static modeling and dynamic mismatch" of traditional digital twin models can be solved.

[0033] In some implementations, when adjusting the digital twin model based on the differences to obtain the adjusted digital twin model in step S102, the simulation step size of the digital twin model can be obtained first. Then, based on the simulation step size and the differences, the model adjustment amount can be determined, and the model parameters of the digital twin model can be adjusted based on the model adjustment amount to obtain the adjusted digital twin model.

[0034] Here, when a digital twin model runs in a physical simulation environment, it typically performs discrete-time integration at a fixed time step (i.e., simulation step size) to update its state. The simulation step size refers to the time interval between two adjacent simulation calculations.

[0035] The simulation step size can be set according to the simulation environment settings, real-time requirements, or numerical stability conditions. For example, the simulation step size can be 10 milliseconds, 1 millisecond, etc.

[0036] Furthermore, the product of the simulation step size and the difference can be determined, and this product can be used as the model adjustment amount, as shown in formula (1): (1) in, Adjustments to the model, For simulating step size, The difference between the target device and the digital twin model.

[0037] In this embodiment of the application, it is necessary to obtain the simulation step size currently used by the digital twin model so that when adjusting the model parameters later, the influence of the simulation step size on the discrete-time dynamics can be considered to ensure the accuracy and stability of the adjustment.

[0038] S103: Based on the preset task objective, the adjusted digital twin model is used to perform simulation prediction on multiple preset control algorithms respectively, and the simulation prediction control result of each control algorithm is obtained; the simulation prediction control result includes prediction control information and control evaluation result corresponding to the prediction control information.

[0039] Among them, multiple control algorithms include, but are not limited to, PID control algorithm, Model Predictive Control (MPC) algorithm, sliding mode control algorithm, adaptive control algorithm, etc.

[0040] In some implementations, the multiple control algorithms may also be algorithms corresponding to the device type of the target device. For example, if the target device is a vehicle, the multiple control algorithms may also include path planning algorithms (such as A-star algorithm, heuristic algorithms (e.g., Rapidly-exploring Random Tree Star, RRT), tabu search algorithm, Dijkstra algorithm, etc.) and trajectory optimization algorithms (such as polynomial trajectory optimization algorithm, Bézier curve planning algorithm, etc.).

[0041] Predictive control information refers to the control command data output by the control algorithm during the simulation prediction process of the digital twin model. Specifically, it can include control signals that the actuators of the target device can recognize (such as physical signal sequences such as current, voltage, speed, torque, and pulse width modulation waves). The control evaluation results are used to characterize the quality of the predictive control information. For example, the control evaluation results can include control accuracy, response speed, energy consumption level, stability, etc.

[0042] Specifically, based on the preset task objectives, in a physical simulation environment, the adjusted digital twin model can be used to perform simulation predictions on multiple preset control algorithms. For each control algorithm, the predicted control information (or control command sequence) generated during the entire simulation process and the state response of the digital twin model under this predicted control information are recorded. Based on this, the control evaluation result corresponding to the predicted control information is determined, and thus the simulation predictive control result of each control algorithm is obtained.

[0043] In this embodiment, two modes are supported: real-time computing (for online verification) and high-frequency non-real-time computing (for acceleration and batch simulation), which can efficiently simulate the control effects of various algorithms under long-term and multi-scenario conditions.

[0044] S104: Based on the simulation predictive control results corresponding to the plurality of control algorithms, a target control algorithm is determined from the plurality of control algorithms, and the target device is controlled according to the target control algorithm so that the target device updates the actual state data.

[0045] In this step, based on the control evaluation results of each control algorithm, the target control algorithm with the best adaptability is selected, and then the target control algorithm is applied to the actual control of the target device to drive the device to perform preset tasks. At the same time, the target device will generate new operating data under the action of the target control algorithm, that is, updated actual state data, which can be used as input for the next round of model calibration and algorithm testing.

[0046] In this embodiment, a closed-loop process of "data acquisition, model calibration, algorithm testing, equipment control, and data update" is formed, realizing an essential upgrade of equipment control from "passive monitoring of anomalies" to "active optimization and regulation". This enables continuous optimization of control effects in complex and ever-changing operating environments, thereby improving the reliability, adaptability, and overall operational efficiency of the target equipment.

[0047] Specifically, when determining the target control algorithm from the multiple control algorithms based on the simulation predictive control results corresponding to the multiple control algorithms, the simulation predictive control results can be evaluated using a pre-constructed multi-objective loss function. The multi-objective loss function is a function of the target device's state, error, and performance indicators as variables, as shown in formula (2): (2) in, The loss value. For the control error of the target equipment, Control error of the target device Convergence time, It is a monotonically increasing non-negative function. These are non-negative weight values. Except... In addition, the function can include performance variables, such as energy consumption, which are not limited here.

[0048] For example, in a trajectory tracking control scenario, It can be tracking and control error, and performance variables can also include various variables such as the distance between the target device and the obstacle, and the energy consumption of the target device.

[0049] In this way, after obtaining the loss value corresponding to each simulation predictive control result, the control algorithm corresponding to the simulation predictive control result with the minimum loss value can be determined as the target control algorithm, and the target control algorithm can be deployed to the target device (i.e., the physical entity system). At this point, the target control algorithm and the target device form a closed loop in the real world: the algorithm receives the actual state input from the target device as input, calculates the actual control information, and applies it to the target device, thereby achieving precise and optimized control of the target device.

[0050] In this embodiment, the adjusted digital twin model serves as a reliable virtual testbed. By defining a comprehensive and flexible loss function, the optimal algorithm for a specific control task can be objectively and automatically selected. This method streamlines the entire process from virtual testing to physical application, significantly improving the efficiency, safety, and final performance of control system design. It is particularly suitable for the intelligent upgrading of complex systems such as transportation equipment and industrial equipment.

[0051] Please see Figure 2This is a flowchart illustrating a device control method provided in an exemplary embodiment of this application. Figure 2 As shown, the difference between the target device and the digital twin model is first estimated. Based on the difference, the digital twin model is fine-tuned to obtain the adjusted digital twin model. In the physical simulation environment, the adjusted digital twin model is used to simulate and predict multiple control algorithms. Then, the target control algorithm is selected from multiple control algorithms based on the simulation prediction control results. Finally, the target device is controlled according to the target control algorithm so that the target device updates the actual state data and realizes closed-loop control.

[0052] The following is combined Figure 3 The difference estimation between the target device and the digital twin model is described in detail below. First, the physical model representing the target device, the digital twin model of the target device, and the mathematical theoretical model of the target device are pre-constructed in this application. The construction of each model is described below.

[0053] 1. Regarding the physical model, the physical model (black box model) can be represented by formula (3): (3) in, The derivative of the actual state is used to characterize the rate of change of the actual state of the target device over time. This is the actual state. For the actual state n dimensional vector, For actual control information, For information about actual control m dimensional vector, For the nonlinear black-box function of the target device. m and n It is a positive integer constant.

[0054] 2. For the digital twin model, the digital twin model (black box model) can be represented by formula (4): (4) in, The derivative of the simulation state is used to characterize the rate of change of the simulation state of the digital twin model over time. In simulation mode, Regarding the simulation state n dimensional vector, For simulation control information of digital twin systems, For information about simulation control m dimensional vector, This is a nonlinear black-box function for a digital twin model.

[0055] 3. Regarding the mathematical theoretical model, the mathematical theoretical model of the target device can be expressed by formula (5): (5) in, The state derivative of the target device. x The status of the target device. For the nonlinear dynamics term of the target device, it is only about x nonlinear functions, For the control gain term of the target device, for about x nonlinear functions, This refers to the control information for the target device.

[0056] here, It can be used to describe the inherent dynamic characteristics of a target when it is not subject to external control input, corresponding to the physical laws of the equipment itself, such as the mechanical friction resistance and aerodynamic resistance of the transport equipment, the component inertia and material damping of industrial equipment, etc., which are the internal characteristics of the target equipment. For example, if the throttle is not controlled during the operation of the transport equipment, the transport equipment will stop under the action of gravity and resistance. It can be used to characterize the coupling relationship between control information and state, and reflect the difference in the effectiveness of control information under different states. For example, the influence coefficient of rotor speed command on flight speed of UAV under different attitude angles, or how steering wheel angle (control information) affects the rate of change of heading angle (state derivative).

[0057] In this embodiment, formula (5) is the state model obtained by nonlinearly expressing the operating mechanism of the target device. It decomposes the state change of the target device into "dynamic term" and "gain term", which is a theoretical expression of the physical law of the target device. It is different from the black box expression of the physical model shown in formula (3). The black box expression of the physical model is the specific manifestation of the mathematical theoretical model under actual complex working conditions (incorporating information such as environmental interference and hardware failure).

[0058] Thus, after constructing the mathematical theoretical model shown in formula (5), the physical model shown in formula (3) can be decomposed and transformed according to the mathematical theoretical model, as shown in formula (6): (6) in, For mathematical theoretical models, This refers to the physical deviation term of the physical model relative to the mathematical theoretical model. The physical deviation term can refer to the deviation between the physical model and the mathematical theoretical model caused by actual environmental interference and / or hardware failure.

[0059] Similarly, based on the mathematical theoretical model, the digital twin model shown in formula (4) is decomposed and transformed as shown in formula (7): (7) in, For mathematical theoretical models, The twin deviation term is the twin deviation term of the digital twin model relative to the mathematical theoretical model. The twin deviation term can refer to the deviation of the digital twin model from the mathematical theoretical model caused by modeling errors and / or simulation errors.

[0060] Therefore, in this embodiment of the application, by determining the physical deviation term and the twin deviation term, the difference between the target device and the digital twin model is indirectly determined. For details, please refer to [link to relevant documentation]. Figure 3 This is a flowchart of a difference estimation method provided as an exemplary embodiment of this application.

[0061] like Figure 3 As shown, the process includes the following steps S1021~S1023: S1021: Based on the actual control information and the corresponding actual state data, the physical deviation term is estimated using the first neural network to obtain the physical deviation value of the physical model relative to the mathematical theoretical model.

[0062] As shown in formula (6), the physical deviation term It's about the actual state. and actual control information Therefore, based on the actual control information of the target device and the corresponding actual state data, the physical deviation term can be estimated using a neural network to obtain the physical deviation value.

[0063] In this application, neural networks have strong nonlinear mapping capabilities, which can be used to fit the unknown perturbation characteristics of the target device (or physical entity).

[0064] The expression for the first neuron network is shown in formula (8): (8) in, For physical deviations, This is the ideal weight matrix for the first neuron network. Let be the transpose of the ideal weight matrix of the first neuron network. For activation function, This is the first network bias variable.

[0065] Here, the first network bias variable refers to the fitting error between the fitted value output by the first neuron network and the physical bias term. It reflects the first neuron network's ability to fit the physical bias term. The smaller the value of the first network bias variable, the higher the fitting accuracy. The existence of the first network bias variable allows the model to retain a reasonable fitting error, avoiding overfitting of the neural network to the training data and improving its generalization ability to unknown perturbations. It should also be noted that the value of the first network bias variable will change during multiple iterations.

[0066] Due to the ideal weight matrix of the first neuron network Since the actual state data of the target device is unknown, this embodiment uses the real-time estimation deviation of the target device's actual state data to dynamically adjust the ideal weight matrix of the first neural network. Therefore, this application proposes to embed the above-mentioned first neural network into the first adaptive observer architecture as shown in formula (9), and to estimate the error term of the digital physical model online through the first adaptive observer and the first neural network.

[0067] (9) in, The derivative of the actual estimated state data, For real-time estimation of the physical deviation term, The actual state estimation error, This is the physical feedback correction value. m and n It is an exponent, and m ≠ n .

[0068] Specifically, regarding step S1021, when estimating the physical bias term using the first neural network, it can be achieved through the following steps: (A) Obtain the actual estimated state data of the current iteration cycle output by the first adaptive observer, and the ideal weight matrix of the first neuron network in the previous iteration cycle.

[0069] Specifically, in the implementation process, the first neural network is first embedded into the first adaptive observer. In each closed loop (or each iteration cycle), the first adaptive observer can output the actual estimated state data corresponding to the actual state data of the physical model, and at the same time read the ideal weight matrix of the first neural network in the previous iteration cycle.

[0070] (B) Based on the actual state data and the actual estimated state data, determine the actual state estimation error.

[0071] The actual state estimation error is shown in formula (10): (10) in, The actual state estimation error, This is actual state data. This is the actual estimated state data.

[0072] (C) Based on the actual control information and the corresponding actual state data, determine the activation function value of the first neural network.

[0073] Based on the actual control information of the current iteration cycle and the corresponding actual status data Calculate the activation function value of the first neural network. .

[0074] (D) Based on the actual state estimation error, the activation function value of the first neural network, and the ideal weight matrix of the first neural network in the previous iteration cycle, determine the ideal weight matrix of the first neural network in the current iteration cycle.

[0075] Specifically, a positive weight matrix adjustment term can be determined based on the actual state estimation error and the activation function value of the first neural network, and a weight adjustment magnitude constraint term can be determined based on the actual state estimation error and the ideal weight matrix of the first neural network in the previous iteration period. Based on the positive weight matrix adjustment term and the weight adjustment magnitude constraint term, the rate of change of the ideal weight matrix of the first neural network can be determined, and the rate of change of the ideal weight matrix of the first neural network can be integrated to obtain the ideal weight matrix of the first neural network in the current iteration period.

[0076] As shown in Equation (11), the ideal weight matrix of the first neuron network in the current iteration period is expressed as follows: (11) in, The derivative of the estimated ideal weight matrix of the first neural network is used to characterize the rate of change of the ideal weight matrix of the first neural network. k For the current iteration cycle, This is an estimate of the ideal weight matrix of the first neuron network in the previous iteration. The actual state estimation error, This is actual state data. This is the adjustment term for the first positive weight matrix. This is the first weight adjustment range constraint term.

[0077] Furthermore, after obtaining the rate of change of the ideal weight matrix of the first neural network as shown in formula (11), it can be integrated to obtain the ideal weight matrix of the first neural network in the current iteration period. .

[0078] It should be noted that, in the embodiments of this application, through and This discrete representation is used to distinguish the ideal weight matrix of the first neuron network in the current iteration cycle from that in the previous iteration cycle. However, in actual operation, the device control method proposed in this application, including the update process of the ideal weight matrix of the first neuron network, is continuous. Similarly, the expression and update of the ideal weight matrix of the second neuron network in the following text are similar to those of the ideal weight matrix of the first neuron network, and will not be elaborated on in the following text.

[0079] (E) Determine the twin bias estimate based on the ideal weight matrix of the first neuron network in the current iteration period and the activation function value of the first neuron network.

[0080] Here, after determining the ideal weight matrix of the first neuron network in the current iteration cycle, the ideal weight matrix of the first neuron network in the current iteration cycle and the activation function value of the first neuron network can be substituted into the expression of the first neuron network in formula (8) to obtain the twin bias value.

[0081] In this embodiment of the application, through the above-described online recursive process, the neural network weight matrix is ​​dynamically adjusted, thereby adjusting the error term. The estimates continuously approximate the true characteristics, thereby enabling adaptive fine-tuning of the physical model. Furthermore, combining the stable estimation framework of the adaptive observer with the powerful approximation capabilities of neural networks allows for real-time compensation for unmodeled dynamics and perturbations during system runtime, significantly improving the fidelity of the physical model.

[0082] Furthermore, after obtaining the physical deviation estimate, the actual estimated state data can be updated according to formula (9).

[0083] Specifically, this includes the following steps (a) to (c): (a) Determine the theoretical baseline value of the actual state based on the actual state data and the actual control information corresponding to the actual state data.

[0084] Specifically, the actual state data and corresponding actual control information are substituted into the mathematical theoretical model. In this way, the theoretical baseline value of the actual state can be determined.

[0085] (b) Determine the physical feedback correction value based on the actual state estimation error.

[0086] According to formula (9), the actual state estimation error is... Substitute into the physical feedback correction term The physical feedback correction value can then be obtained.

[0087] (c) Based on the physical deviation estimate, the physical feedback correction value, and the actual state theoretical benchmark value, determine the update rate of the actual estimated state, and integrate the update rate of the actual estimated state to obtain the actual estimated state data; the actual estimated state data is used as the actual state data for the next iteration.

[0088] In this way, the physical deviation estimate Physical feedback correction value and actual state theoretical benchmark value Substituting these values ​​into formula (7) will give the update rate (or derivative) of the actual estimated state. Furthermore, the update rate for the actual estimated state By performing integration, the actual estimated state data can be obtained. This actual estimated state data is used as the actual estimated state data for the next iteration cycle, thus achieving closed-loop update of the actual estimated state data.

[0089] S1022: Based on the simulation control information and the corresponding simulation state data, the twin deviation term is estimated using the neural network to obtain the twin deviation value of the digital twin model relative to the preset mathematical theoretical model.

[0090] As shown in formula (7), the twin bias term It's about the simulation state. and simulation control information Therefore, based on the simulation control information and corresponding simulation state data of the digital twin model, the twin bias term can be estimated using a neural network to obtain the twin bias value.

[0091] In this application, neural networks have strong nonlinear mapping capabilities, which can be used to fit the unknown perturbation characteristics of the target device (or physical entity).

[0092] To distinguish it from the first neural network in step S1021, this embodiment will describe the second neural network.

[0093] The expression for the second neural network is shown in formula (12): (12) in, For twin bias terms, This is the ideal weight matrix for the second neuron network. Let be the transpose of the ideal weight matrix of the second neuron network. For activation function, This is the second network bias variable.

[0094] Here, the second network bias variable refers to the fitting error between the fitted value output by the second neuron network and the twin bias term. It reflects the second neuron network's ability to fit the twin bias term. The smaller the value of the second network bias variable, the higher the fitting accuracy. The existence of the second network bias variable allows the model to retain a reasonable fitting error, avoiding overfitting of the neural network to the training data and improving its generalization ability to unknown perturbations. It should also be noted that the value of the second network bias variable will change during multiple iterations.

[0095] Due to the ideal weight matrix of the second neuron network Since the state is unknown, this embodiment uses the real-time estimation deviation of the state (actual state data of the target device or simulation state data of the digital twin model) to dynamically adjust the ideal weight matrix of the second neural network. Therefore, this application proposes to embed the above-mentioned second neural network into the second adaptive observer architecture as shown in formula (13), and to estimate the error term of the digital twin model online through the second adaptive observer and the second neural network.

[0096] (13) in, The derivative of the simulated estimated state data is used to characterize the update rate of the simulated estimated state. For real-time estimation of twin bias term, For simulation state estimation error, For twin feedback correction values, m and n Let be the power of the simulation state estimation error, and .

[0097] It should be noted that the second adaptive observer and the first adaptive observer mentioned above can refer to different observers or the same observer. If they are the same observer, then the observer can estimate the error terms of the digital twin model and the physical model in parallel online.

[0098] Specifically, regarding step S1022, when estimating the twin bias term using the second neural network, the following steps (1) to (5) can be used: (1) Obtain the simulation estimated state data of the current iteration cycle output by the second adaptive observer, and the ideal weight matrix of the second neuron network in the previous iteration cycle.

[0099] Specifically, in the implementation process, the second neural network is first embedded into the second adaptive observer. In each closed loop (or each iteration cycle), the second adaptive observer can output the simulation estimated state data corresponding to the simulation state data of the digital twin model, and at the same time read the ideal weight matrix of the second neural network in the previous iteration cycle.

[0100] (2) Based on the simulation state data and the simulation estimated state data, determine the simulation state estimation error.

[0101] The simulation state estimation error is shown in formula (14): (14) in, For simulation state estimation error, For simulation state data, This is for simulating and estimating state data.

[0102] (3) Based on the simulation control information and the corresponding simulation state data, determine the activation function value of the second neural network.

[0103] Based on the simulation control information of the current iteration cycle and the corresponding simulation state data Calculate the activation function value of the second neural network. .

[0104] (4) Based on the simulation state estimation error, the activation function value of the second neuron network and the ideal weight matrix of the second neuron network in the previous iteration cycle, determine the ideal weight matrix of the second neuron network in the current iteration cycle.

[0105] Specifically, based on the simulation state estimation error and the activation function value of the second neural network, a second positive weight matrix adjustment term can be determined. Based on the simulation state estimation error and the ideal weight matrix of the second neural network in the previous iteration period, a second weight adjustment magnitude constraint term can be determined. Based on the second positive weight matrix adjustment term and the second weight adjustment magnitude constraint term, the rate of change of the ideal weight matrix of the second neural network can be determined. The rate of change of the ideal weight matrix of the second neural network can be integrated to obtain the ideal weight matrix of the second neural network in the current iteration period.

[0106] As shown in Equation (15), the ideal weight matrix of the second neuron network in the current iteration cycle is expressed as follows: (15) in, The derivative of the estimated ideal weight matrix of the second neuron network is used to characterize the rate of change of the ideal weight matrix of the second neuron network. k For the current iteration cycle, For simulation state estimation error, For simulation state data, For simulation state estimation error norm, This is an estimate of the ideal weight matrix of the second neuron network in the previous iteration cycle. This is the adjustment term for the second positive weight matrix. This is the second weight adjustment range constraint term.

[0107] Furthermore, after obtaining the rate of change of the ideal weight matrix of the second neuron network as shown in formula (15), it can be integrated to obtain the ideal weight matrix of the second neuron network in the current iteration period. .

[0108] (5) Determine the twin bias value based on the ideal weight matrix of the second neuron network in the current iteration cycle and the activation function value of the second neuron network.

[0109] Here, after determining the ideal weight matrix of the second neuron network in the current iteration cycle, the ideal weight matrix of the second neuron network in the current iteration cycle and the activation function value of the second neuron network can be substituted into the expression of the second neuron network in formula (12) to obtain the twin bias value.

[0110] In this embodiment of the application, through the above-described online recursive process, the neural network weights are dynamically adjusted repeatedly, so that the error term... The estimates continuously approach the true characteristics, thereby enabling adaptive fine-tuning of the digital twin model.

[0111] Furthermore, combining the stable estimation framework of adaptive observers with the powerful approximation capabilities of neural networks enables real-time compensation for unmodeled dynamics and disturbances during system operation, significantly improving the fidelity of digital twin models.

[0112] Furthermore, after obtaining the twin bias estimate, the simulation estimated state data can be updated according to formula (13).

[0113] Specifically, this includes the following steps (a) to (c): (a) Based on the simulation state data and the simulation control information corresponding to the simulation state data, determine the theoretical benchmark value of the simulation state.

[0114] Specifically, the simulation state data and corresponding simulation control information are substituted into the mathematical theoretical model. In this way, the theoretical baseline value of the simulation state can be determined.

[0115] (b) Determine the twin feedback correction value based on the simulation state estimation error.

[0116] According to formula (7), the simulation state estimation error is substituted into the twin feedback correction term. The twin feedback correction value can be obtained from this.

[0117] (c) Based on the twin bias estimate, the twin feedback correction value, and the theoretical baseline value of the simulation state, determine the update rate of the simulation estimated state, and integrate the update rate of the simulation estimated state to obtain the simulation estimated state data; the simulation estimated state data is used as the simulation state data for the next iteration.

[0118] Thus, by substituting the twin bias estimate, twin feedback correction value, and simulation state theoretical baseline value into formula (13), the update rate (or derivative) of the simulation estimated state can be determined. Furthermore, the update rate of the simulated estimated state. By integrating, the simulation estimated state data can be obtained. This simulation estimated state data is used for the simulation estimated state data in the next iteration cycle, thus realizing the closed-loop update of the simulation estimated state data.

[0119] S1023: Determine the difference based on the twin deviation value and the physical deviation value.

[0120] Furthermore, the differences between the target device and the digital twin model are determined based on the difference between the twin deviation value and the physical deviation value.

[0121] For example, please see Figure 4 This is a flowchart illustrating a difference estimation method provided in an exemplary embodiment of this application. Figure 4 As shown, the target device and the digital twin model run in parallel. Using digital twin technology, the control of the physical world is mirrored and optimized in the virtual world. Both run according to the same preset task objective and execute the same target control algorithm. Here, the target control algorithm is the optimal control algorithm determined from multiple control algorithms based on the minimum loss value as described in step S104 above.

[0122] For the target device, based on the preset task objectives, the target control algorithm outputs actual control information to the target device, and the target device responds to the actual control information. Driven by this, its sensors measure actual state data. And based on actual status data and actual control information Calculate the physical deviation term between the target device and the mathematical theoretical model.

[0123] In addition, actual status data The data is also fed back to the target control algorithm, which compares the received actual state data with the expected actual state given by the preset task target, calculates the error, and adjusts and outputs new actual control information in real time based on this error, thereby changing the physical state of the target device. The sensor measures the new actual state data again, and so on, forming a dynamic, real-time physical world control closed loop.

[0124] For digital twin models, simulation control information is output through target control algorithms based on preset task objectives. Given a digital twin model, the digital twin model calculates simulation state data based on simulation control information. And based on simulation state data and simulation control information Calculate the twin deviation term between the digital twin model and the mathematical theoretical model.

[0125] In addition, simulation state data The received simulation state data is fed back to the target control algorithm, which compares it with the desired simulation state given by the preset task objective, calculates the error, and adjusts and outputs new simulation control information in real time based on this error, thereby changing the simulation state data of the digital twin model. This process is repeated continuously, forming a dynamic, real-time virtual control closed loop.

[0126] Corresponding to the above-described device control method, this disclosure also provides an electronic device, such as... Figure 5 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this disclosure. Figure 5 As shown, the electronic device 500 includes a processor 510, an internal bus 520, a memory 530, a network interface 540, and a non-volatile memory 550, and may also include other hardware required for its functions. One or more embodiments of this specification can be implemented in software, for example, the processor 510 reads the corresponding computer program from the non-volatile memory 550 into the memory 530 and then runs it. Of course, besides software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0127] The memory 530, also known as internal memory, is used to temporarily store the computational data in the processor 510, as well as the data exchanged with non-volatile memory 550 such as hard disk. The processor 510 exchanges data with non-volatile memory 550 through the memory 530.

[0128] In this embodiment, memory 530 is specifically used to store application code that executes the solution of this application, and its execution is controlled by processor 510. That is, when the electronic device is running, processor 510 communicates with network interface 540, memory 530 and non-volatile memory 550 through internal bus 520, so that processor 510 executes the application code stored in memory 530 and non-volatile memory 550, thereby executing the device control method described in the above method embodiment.

[0129] Processor 510 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware microservices. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0130] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device 500. In other embodiments of this application, the electronic device 500 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0131] This disclosure also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the device control method described in the above method embodiments. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0132] This disclosure also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the device control method in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0133] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0134] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0135] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0136] Computers suitable for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. Basic computer microservices include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0137] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0138] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0139] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and microservices in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program microservices and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0140] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0141] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A device control method, characterized in that, include: Acquire the actual state data of the target device operating according to the preset task objectives, and the simulation state data of the digital twin model of the target device operating according to the preset task objectives; Based on the actual state data and the simulation state data, the differences between the target device and the digital twin model are determined, and the digital twin model is adjusted based on the differences to obtain the adjusted digital twin model. Based on the preset task objectives, the adjusted digital twin model is used to perform simulation prediction on multiple preset control algorithms to obtain the simulation prediction control results of each control algorithm; the simulation prediction control results include prediction control information and control evaluation results corresponding to the prediction control information. Based on the simulation predictive control results corresponding to the multiple control algorithms, a target control algorithm is determined from the multiple control algorithms, and the target device is controlled according to the target control algorithm so that the target device updates the actual state data.

2. The method according to claim 1, characterized in that, The target device is characterized by a physical model, which includes a mathematical theoretical model and physical deviation terms. The digital twin model includes the mathematical theoretical model and twin deviation terms. The mathematical theoretical model is used to characterize the target device under ideal conditions. The actual state data is generated based on the actual control information of the target device, and the simulation state data is generated based on the simulation control information of the digital twin model; determining the differences between the target device and the digital twin model based on the actual state data and the simulation state data includes: Based on the actual control information and the corresponding actual state data, the physical deviation term is estimated using a first neural network to obtain the physical deviation value of the physical model relative to the mathematical theoretical model; and, Based on the simulation control information and the corresponding simulation state data, the twin bias term is estimated using a second neural network to obtain the twin bias value of the digital twin model relative to the preset mathematical theoretical model. The difference is determined based on the twin deviation value and the physical deviation value.

3. The method according to claim 2, characterized in that, The first neural network is embedded in the first adaptive observer; the step of estimating the physical deviation term using the first neural network based on the actual control information and the corresponding actual state data to obtain the physical deviation value of the physical model relative to the mathematical theoretical model includes: Obtain the actual estimated state data of the current iteration cycle output by the first adaptive observer, and the ideal weight matrix of the first neural network in the previous iteration cycle; Based on the actual state data and the actual estimated state data, the actual state estimation error is determined; Based on the actual control information and the corresponding actual state data, the activation function value of the first neural network is determined; Based on the actual state estimation error, the activation function value of the first neural network, and the ideal weight matrix of the first neural network in the previous iteration cycle, the ideal weight matrix of the first neural network in the current iteration cycle is determined. The twin bias estimate is determined based on the ideal weight matrix of the first neuron network in the current iteration cycle and the activation function value of the first neuron network.

4. The method according to claim 3, characterized in that, Determining the ideal weight matrix of the first neural network in the current iteration period based on the actual state estimation error, the activation function value of the first neural network, and the ideal weight matrix of the first neural network in the previous iteration period includes: Based on the actual state estimation error and the activation function value of the first neural network, the adjustment term of the first positive weight matrix is ​​determined; and, Based on the actual state estimation error and the ideal weight matrix of the first neuron network in the previous iteration cycle, the first weight adjustment magnitude constraint term is determined. Based on the first positive weight matrix adjustment term and the first weight adjustment magnitude constraint term, the rate of change of the ideal weight matrix of the first neural network is determined, and the rate of change of the ideal weight matrix of the first neural network is integrated to obtain the ideal weight matrix of the first neural network in the current iteration period.

5. The method according to claim 2, characterized in that, After obtaining the physical deviation value of the physical model relative to the mathematical theoretical model, the method further includes: Based on the actual state data and the actual control information corresponding to the actual state data, the theoretical benchmark value of the actual state is determined. Based on the actual state estimation error, determine the physical feedback correction value; Based on the physical deviation estimate, the physical feedback correction value, and the actual state theoretical benchmark value, the update rate of the actual estimated state is determined, and the update rate of the actual estimated state is integrated to obtain the actual estimated state data; the actual estimated state data is used as the actual state data for the next iteration.

6. The method according to claim 2, characterized in that, The second neural network is embedded in the second adaptive observer; the estimation of the twin bias term using the second neural network based on the simulation control information and the corresponding simulation state data to obtain the twin bias value of the digital twin model relative to the preset mathematical theoretical model includes: Obtain the simulation estimated state data of the current iteration cycle output by the second adaptive observer, and the ideal weight matrix of the second neural network in the previous iteration cycle; Based on the simulation state data and the simulation estimated state data, the simulation state estimation error is determined; Based on the simulation control information and the corresponding simulation state data, the activation function value of the second neural network is determined; Based on the simulation state estimation error, the activation function value of the second neural network, and the ideal weight matrix of the second neural network in the previous iteration cycle, the ideal weight matrix of the second neural network in the current iteration cycle is determined. The twin bias value is determined based on the ideal weight matrix of the second neuron network in the current iteration cycle and the activation function value of the second neuron network.

7. The method according to claim 6, characterized in that, The step of determining the ideal weight matrix of the second neural network in the current iteration period based on the simulation state estimation error, the activation function value of the second neural network, and the ideal weight matrix of the second neural network in the previous iteration period includes: Based on the simulation state estimation error and the activation function value of the second neural network, the adjustment term of the second positive weight matrix is ​​determined; Based on the simulation state estimation error and the ideal weight matrix of the second neuron network in the previous iteration cycle, the second weight adjustment magnitude constraint term is determined; Based on the second positive weight matrix adjustment term and the second weight adjustment magnitude constraint term, the rate of change of the ideal weight matrix of the second neural network is determined, and the rate of change of the ideal weight matrix of the second neural network is integrated to obtain the ideal weight matrix of the second neural network in the current iteration period.

8. The method according to claim 2, characterized in that, The method further includes: Based on the simulation state data and the simulation control information corresponding to the simulation state data, the theoretical baseline value of the simulation state is determined. Based on the twin deviation value, determine the twin feedback correction value; Based on the twin bias value, the twin feedback correction value, and the theoretical baseline value of the simulation state, the update rate of the simulation estimated state is determined, and the update rate of the simulation estimated state is integrated to obtain new simulation estimated state data; the new simulation estimated state data is used as the simulation state data for the next iteration.

9. The method according to claim 1, characterized in that, The adjustment of the digital twin model based on the differences to obtain the adjusted digital twin model includes: Obtain the simulation step size of the digital twin model; Based on the simulation step size and the difference, the model adjustment amount is determined, and the model parameters of the digital twin model are adjusted based on the model adjustment amount to obtain the adjusted digital twin model.

10. The method according to claim 1, characterized in that, The step of determining the target control algorithm from the plurality of control algorithms based on the simulation predictive control results corresponding to the plurality of control algorithms includes: For each control algorithm, the simulation predictive control result is used to determine the loss function value corresponding to the simulation predictive control result based on a preset multi-objective loss function. The control algorithm corresponding to the minimum loss function value is determined as the target control algorithm.

11. The method according to any one of claims 2-10, characterized in that, The target device includes a transport equipment, the actual control information includes control commands applied to the transport equipment, and the actual state data includes the position and kinematic information of the transport equipment.

12. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the device control method according to any one of claims 1-11.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the device control method according to any one of claims 1-11.

Images