A data-driven cloud-edge collaborative control method and system based on a disturbance observer

Abstract

The application discloses a data-driven cloud-edge collaborative control method and system based on a disturbance observer, wherein total disturbance of a control process in a cloud control system is estimated in an edge system, and a cloud-edge composite control variable is formed after the cloud control variable is compensated by the total disturbance; the cloud-edge composite control variable with precise anti-disturbance is used to control the controlled object, the disturbance caused by cloud computing errors, cloud-edge communication time delays and other factors existing in the prior art is eliminated, the utilization rate of the computing resources of the edge system is improved, and the control quality of the data-driven predictive cloud control system is improved.

Description

Technical Field

[0001] This invention belongs to the field of automatic control technology, specifically relating to a data-driven cloud-edge collaborative control method and system based on a disturbance observer. Background Technology

[0002] The development of information technology has spurred numerous scientific and technological achievements, including well-known ones such as cloud computing, the Internet of Things (IoT), networked control, and cyber-physical systems. Traditional control systems suffer from weak processing capabilities and low intelligence, while reality demands more powerful and intelligent functions, along with improved information interaction capabilities. Currently, the complexity and scale of control systems are increasing, placing higher demands on storage and computing power. Based on the development of theories and technologies such as cloud computing, IoT, and networked control, the concept of cloud control has been proposed. However, how to achieve mutual collaboration between cloud platforms, edge systems, and physical systems (referred to as "cloud-edge collaboration") to optimize the overall manufacturing system and efficiently, safely, and with high quality complete all activities and tasks throughout the manufacturing lifecycle is a major challenge facing intelligent manufacturing systems. In recent years, the cloud-edge collaboration problem in intelligent manufacturing systems has attracted attention from both industry and academia.

[0003] Currently, data-driven predictive cloud control has been proposed. Combining the advantages of cloud computing and networked control, it leverages the Internet of Things (IoT) and cyber-physical systems (CPS) to establish complex and powerful intelligent control systems. This intelligent control system relies solely on historical input-output data and does not require a pre-established mathematical model of the controlled object to generate optimal predictive control sequences. As a control algorithm, data-driven predictive control is applicable to controlled objects in various fields, including autonomous vehicles, drones, air conditioning, and power systems. Taking autonomous vehicle control as an example, the vehicle's position and speed are used as output signals, and acceleration is used as the control input signal to control the vehicle. For other controlled objects, based on their physical characteristics, specific control input signals can be applied to generate corresponding output signals, achieving the desired control effect.

[0004] However, existing data-driven predictive cloud control systems lack collaborative interaction with edge systems. Especially when computational errors exist in the cloud and communication latency exists between the cloud and the edge, the control quality of edge nodes will be significantly impacted. Summary of the Invention

[0005] In view of this, the present invention provides a data-driven cloud-edge collaborative control method and system based on a disturbance observer, which realizes collaborative control of the physical system by the cloud platform and the edge system.

[0006] This invention provides a data-driven cloud-edge collaborative control method based on a disturbance observer, comprising the following steps:

[0007] Step 1: At sampling time k, the edge controlled node simultaneously sends the input data of the controlled object and the output data it generates to the cloud control platform;

[0008] Step 2: The cloud control platform generates cloud control variables based on the received input and output data of the controlled object;

[0009] Step 3: The cloud control platform sends the cloud parameters and cloud control variables to the edge controlled nodes;

[0010] Step 4: When the sampling time k is not greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node sends the cloud control variable as new input data to the controlled object; when the sampling time k is greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node performs compensation processing on the cloud parameters, cloud control variable, input data and output data to obtain the cloud-edge composite control variable, and then sends the cloud-edge composite control variable as new input data to the controlled object.

[0011] Step 5: If k is less than the threshold, increment k by 1 and execute step 1; if k is not less than the threshold, end the process.

[0012] Furthermore, in step 2, the cloud control platform generates cloud control variables based on the received input and output data of the controlled object, including the following steps:

[0013] Step 2.1: Calculate the control rate parameter L of the cloud controller using formula (1). w and L u :

[0014]

[0015] Among them, V p Let the intermediate variable be denoted as W p Let W be the intermediate variable. p =[U p ,Y p ], Y p U is the Hankel matrix of the output data of the controlled object over past time periods. p Input data for the controlled object over past time periods. For V p Moore-Penrose generalized inverse matrix;

[0016] Step 2.2: Calculate the predictive control sequence u using formula (2). f (k):

[0017]

[0018] in, For the control rate parameter L u The transpose of .

[0019] Step 2.3: The predicted control sequence u f The first term of (k) is used as the cloud control variable u. cloud (k).

[0020] Furthermore, in step 4, when the sampling time k is greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node performs compensation processing on the cloud parameters, cloud control variables, input data, and output data to obtain the cloud-edge composite control variable, including the following steps:

[0021] Step 4.1: Estimate the total control disturbance using a disturbance observer based on cloud parameters. As shown in formula (3):

[0022]

[0023] Where P(k) is an auxiliary variable, L is the observer gain coefficient, and a i For cloud parameter L w The vector in the first row, b i For cloud parameter L w The vector in the first row, b is the cloud parameter L. u The first block line, where i is the parameter number and satisfies 1≤i≤N;

[0024] Step 4.2: Based on the total control disturbance and the cloud control variable, calculate the cloud-edge composite control variable u(k) using formula (4):

[0025]

[0026] Among them, u com (k) is the compensation control quantity for controlling the total disturbance.

[0027] The present invention provides a data-driven cloud-edge collaborative control system based on a disturbance observer, comprising a cloud control unit, a control scheduling unit, a disturbance observation unit, a total disturbance compensation unit, and a controlled object;

[0028] The cloud control unit is deployed in the cloud platform and is used to calculate cloud control variables and cloud parameters based on the input and output data of the controlled object. The cloud control variables are used to control the controlled object, and the cloud parameters serve as the input of the disturbance observation unit.

[0029] The control scheduling unit is deployed in the edge system and is used to set the sampling time. At the sampling time, the input and output data of the controlled object are sent to the cloud control unit. When the sampling time is less than or equal to the maximum number of rows in the Hankel matrix of the input data, the cloud control variable output by the cloud control unit is sent to the controlled object as input data. When the sampling time is greater than the maximum number of rows in the Hankel matrix of the input data, the cloud parameters output by the cloud control unit are input to the disturbance observation unit, and the cloud control variable is input to the total disturbance compensation unit. The cloud-edge composite control variable obtained from the total disturbance compensation unit is sent to the controlled object as input data.

[0030] The disturbance observation unit is deployed in the edge system and is used to calculate the total control disturbance based on the cloud parameters output by the cloud control unit, and send the total control disturbance to the total disturbance compensation unit.

[0031] The total disturbance compensation unit is deployed in the edge system and is used to compensate the cloud control variables output by the cloud control unit by using the total control disturbance output by the disturbance observation unit to obtain the cloud-edge composite control variables, and then send the cloud-edge composite control variables to the control scheduling unit.

[0032] The controlled object is deployed in the physical system and is used to perform operations and generate output data based on the input data output by the control and scheduling unit.

[0033] Beneficial effects:

[0034] This invention estimates the total disturbance in the control process of the cloud control system in the edge system, and then uses the total disturbance to compensate the cloud control variables to form a cloud-edge composite control quantity. The control of the controlled object is achieved by using the precise disturbance-resistant cloud-edge composite control quantity, which eliminates the disturbances caused by cloud computing errors, cloud-edge communication delays and other factors, improves the utilization rate of edge system computing resources, and enhances the control quality of the data-driven predictive cloud control system. Attached Figure Description

[0035] Figure 1 The present invention provides a structural diagram of a data-driven cloud-edge collaborative control system based on a disturbance observer. Detailed Implementation

[0036] The following examples illustrate the invention in detail.

[0037] To better understand the technical solution provided by this invention, the following is a detailed explanation of the relevant concepts involved in this invention. This invention employs a Hankel matrix, the form of which is shown in the following equation:

[0038]

[0039]

[0040] Where U represents the input data of the controlled object, and the subscripts p and f represent the past time and future time, respectively, i.e., U p For the input data of the controlled object at past times, U f The input data represents the future timeframe of the controlled object; N and j are scale parameters, representing the rows and columns of the Hankel matrix, respectively. In this invention, the controlled object includes, but is not limited to, various systems such as unmanned vehicles, drones, air conditioners, and power systems.

[0041] Define intermediate variable W p Represented as W p =[U p ,Y p ], where Y p The Hankel matrix represents the output data of the controlled object at past time points.

[0042] definition W p Y represents f The line space along U f The line space is projected towards W p The projection of the row space, where Y is the line space. f The Hankel matrix represents the output data of the controlled object at future moments.

[0043] The column vectors and intermediate variables of the input-output matrix of the controlled object at sampling time k are defined as follows:

[0044]

[0045]

[0046]

[0047] This invention provides a data-driven cloud-edge collaborative control method and system based on a disturbance observer. The basic idea is to calculate the control quantity in the cloud and observe and compensate for the total disturbance in the edge system. Through data interaction between the cloud control platform, the edge controlled nodes, and the controlled object, the cloud control platform controls the controlled object. The edge controlled nodes are used to compensate for the cloud control variables and cloud parameters output by the cloud control platform and the output data generated by the controlled object, and then send them as input data to the controlled object. The output data and input data of the controlled object are then fed back to the cloud control platform, thereby realizing collaborative control between the cloud platform, the edge system, and the physical system.

[0048] This invention provides a data-driven cloud-edge collaborative control method based on a disturbance observer, the specific steps of which are as follows:

[0049] Step 1: At sampling time k, the edge controlled node simultaneously sends the input data of the controlled object and the output data it generates to the cloud control platform.

[0050] In addition, to facilitate subsequent data processing, the edge controlled nodes also store the input data of the controlled object and the output data generated therefrom in the edge system's data cache.

[0051] Step 2: The cloud control platform generates cloud control variables based on the received input and output data of the controlled object, specifically including the following steps:

[0052] Step 2.1: Calculate the control rate parameter L of the cloud controller using formula (1). w and L u :

[0053]

[0054] Among them, V p Let the intermediate variable be denoted as W p Let W be the intermediate variable. p =[U p ,Y p ], Y p U is the Hankel matrix of the output data of the controlled object over past time periods. p Input data for the controlled object over past time periods. For V p The Moore-Penrose generalized inverse matrix.

[0055] Step 2.2: Calculate the predictive control sequence u using formula (2). f (k):

[0056]

[0057] in, For the control rate parameter L u The transpose of .

[0058] Step 2.3: The predicted control sequence u f The first term of (k) is used as the cloud control variable u. cloud (k).

[0059] Step 3: The cloud control platform sends the cloud parameters and cloud control variables to the edge controlled nodes.

[0060] Step 4: The edge controlled node determines the new input data of the controlled object according to the situation. When the sampling time k is less than or equal to the maximum number of rows in the input data Hankel matrix, the cloud control variable is used as the new input data of the controlled object. When the sampling time k is greater than the maximum number of rows in the input data Hankel matrix, the received cloud parameters and cloud control variables, as well as the input and output data of the controlled object saved in Step 1, are compensated to obtain the cloud-edge composite control variable. The cloud-edge composite control variable is used as the new input data of the controlled object, and then the new input data is sent to the controlled object.

[0061] The processing of edge-controlled nodes includes the following steps:

[0062] Step 4.1: If k is less than or equal to N, then the cloud control variable u... cloud (k) is sent as input data to the controlled object, and step 5 is executed; if k is greater than N, then step 4.2 is executed. Wherein, N is the maximum number of rows in the Hankel matrix of the input data of the controlled object.

[0063] Step 4.2: Estimate the total control disturbance using a disturbance observer based on cloud parameters. The disturbance observer is shown in equation (3):

[0064]

[0065] Where P(k) is an auxiliary variable, L is the observer gain coefficient, and a i For cloud parameter L w The vector in the first row, b i For cloud parameter L w The vector in the first row, b is the cloud parameter L. u The first block row, i is the parameter number and satisfies 1≤i≤N. In this invention, a block row refers to a data block composed of multiple rows of data.

[0066] Step 4.3: Based on the total control disturbance and the cloud control variable, calculate the cloud-edge composite control variable u(k) using formula (4):

[0067]

[0068] Among them, u com (k) is the compensation control quantity for controlling the total disturbance.

[0069] Step 4.4: The cloud-edge composite control variable u(k) is sent as input data to the controlled object.

[0070] Step 5: If k is less than the threshold, increment k by 1 and execute Step 1; if k is not less than the threshold, end the process. The threshold value is set according to the actual control needs of the controlled object; a larger threshold results in a longer control time.

[0071] This invention provides a data-driven cloud-edge collaborative control system based on a disturbance observer, including a cloud control unit, a control scheduling unit, a disturbance observation unit, a total disturbance compensation unit, and a controlled object.

[0072] The cloud control unit, deployed in the cloud platform, is used to calculate cloud control variables and cloud parameters based on the input and output data of the controlled object. The cloud control variables are used to control the controlled object, and the cloud parameters serve as the input to the disturbance observation unit.

[0073] The control and scheduling unit, deployed in the edge system, is used to set the sampling time. At the sampling time, the input and output data of the controlled object are sent to the cloud control unit. When the sampling time is less than or equal to the maximum number of rows in the Hankel matrix of the input data, the cloud control variables output by the cloud control unit are sent to the controlled object as input data. When the sampling time is greater than the maximum number of rows in the Hankel matrix of the input data, the cloud parameters output by the cloud control unit are input to the disturbance observation unit, and the cloud control variables are input to the total disturbance compensation unit. The cloud-edge composite control variables obtained from the total disturbance compensation unit are sent to the controlled object as input data.

[0074] The disturbance observation unit, deployed in the edge system, is used to calculate the total control disturbance based on the cloud parameters output by the cloud control unit and send the total control disturbance to the total disturbance compensation unit.

[0075] The total disturbance compensation unit, deployed in the edge system, is used to compensate the cloud control variables output by the cloud control unit using the total control disturbance output by the disturbance observation unit to obtain the cloud-edge composite control variables, and then send the cloud-edge composite control variables to the control scheduling unit.

[0076] The controlled object is deployed in the physical system to perform operations and generate output data based on the input data output by the control and scheduling unit.

[0077] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data-driven cloud-edge collaborative control method based on a disturbance observer, characterized in that, Includes the following steps: Step 1: At sampling time k, the edge controlled node simultaneously sends the input data of the controlled object and the output data it generates to the cloud control platform; Step 2: The cloud control platform generates cloud control variables based on the received input and output data of the controlled object; Step 3: The cloud control platform sends the cloud parameters and cloud control variables to the edge controlled nodes; Step 4: When the sampling time k is not greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node sends the cloud control variable as new input data to the controlled object; when the sampling time k is greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node performs compensation processing on the cloud parameters, cloud control variable, input data and output data to obtain the cloud-edge composite control variable, and then sends the cloud-edge composite control variable as new input data to the controlled object. Step 5: If k is less than the threshold, increment k by 1 and execute Step 1; if k is not less than the threshold, end the process. In step 4, when the sampling time k is greater than the maximum number of rows in the input data Hankel matrix, the edge controlled node performs compensation processing on the cloud parameters, cloud control variables, input data, and output data to obtain the cloud-edge composite control variable, including the following steps: Step 4.1: Estimate the total control disturbance using a disturbance observer based on cloud parameters. As shown in formula (3): Where P(k) is an auxiliary variable, L is the observer gain coefficient, and a i For cloud parameter L w The vector in the first row, b i For cloud parameter L w The vector in the first row, b is the cloud parameter L. u The first block line, where i is the parameter number and satisfies 1 ≤ i ≤ N, N is the scale parameter, u cloud (k) represents the cloud control variable; Step 4.2: Based on the total control disturbance and the cloud control variable, calculate the cloud-edge composite control variable u(k) using formula (4): Among them, u com (k) is the compensation control quantity for controlling the total disturbance.

2. The data-driven cloud-edge collaborative control method according to claim 1, characterized in that, In step 2, the cloud control platform generates cloud control variables based on the received input and output data of the controlled object, including the following steps: Step 2.1: Calculate the control rate parameter L of the cloud controller using formula (1). w and L u : Among them, V p Let the intermediate variable be denoted as W p Let W be the intermediate variable. p =[U p ,Y p ], Y p U is the Hankel matrix of the output data of the controlled object over past time periods. p Input data for the controlled object over past time periods. For V p The Moore-Penrose generalized inverse matrix, Y f U is the Hankel matrix of the output data of the controlled object at future times. f Input data for the controlled object at future moments; Step 2.2: Calculate the predictive control sequence u using formula (2). f (k): in, For the control rate parameter L u Transpose of; Step 2.3: The predicted control sequence u f The first term of (k) is used as the cloud control variable u. cloud (k).

3. A data-driven cloud-edge collaborative control system based on a disturbance observer, implementing the data-driven cloud-edge collaborative control method of claim 1, characterized in that, It includes a cloud-based control unit, a control and scheduling unit, a disturbance observation unit, a total disturbance compensation unit, and the controlled object; The cloud control unit is deployed in the cloud platform and is used to calculate cloud control variables and cloud parameters based on the input and output data of the controlled object. The cloud control variables are used to control the controlled object, and the cloud parameters serve as the input of the disturbance observation unit. The control scheduling unit is deployed in the edge system and is used to set the sampling time. At the sampling time, the input and output data of the controlled object are sent to the cloud control unit. When the sampling time is less than or equal to the maximum number of rows in the Hankel matrix of the input data, the cloud control variable output by the cloud control unit is sent to the controlled object as input data. When the sampling time is greater than the maximum number of rows in the Hankel matrix of the input data, the cloud parameters output by the cloud control unit are input to the disturbance observation unit, and the cloud control variable is input to the total disturbance compensation unit. The cloud-edge composite control variable obtained from the total disturbance compensation unit is sent to the controlled object as input data. The disturbance observation unit is deployed in the edge system and is used to calculate the total control disturbance based on the cloud parameters output by the cloud control unit and send the total control disturbance to the total disturbance compensation unit. The total disturbance compensation unit is deployed in the edge system and is used to compensate the cloud control variables output by the cloud control unit by using the total control disturbance output by the disturbance observation unit to obtain the cloud-edge composite control variables, and then send the cloud-edge composite control variables to the control scheduling unit. The controlled object is deployed in the physical system and is used to perform operations and generate output data based on the input data output by the control and scheduling unit.

Images