CN-121978918-A - Complex system control method, device and equipment based on diffusion model

CN121978918ACN 121978918 ACN121978918 ACN 121978918ACN-121978918-A

Abstract

The application provides a complex system control method, device and equipment based on a diffusion model, and relates to the technical field of control systems. The method comprises the steps of obtaining the current system state, the target system state and initial system noise information sampled in a preset time period of a complex system, inputting the current system state, the target system state and the initial system noise information into a denoising network in a diffusion model, denoising the initial system noise information through the denoising network to obtain a plurality of target prediction system states of the complex system in the preset time period, inputting a prediction system state sequence formed by the plurality of target prediction system states into an inverse dynamic model, estimating two adjacent target prediction system states in the prediction system state sequence through the inverse dynamic model to generate a corresponding control signal sequence, and carrying out limited time control on the complex system based on the control signal sequence. By adopting the technical scheme provided by the application, the control effect of the complex system can be improved.

Inventors

DING JINGTAO
LI YONG
CHEN HONGYI

Assignees

清华大学

Dates

Publication Date: 20260505
Application Date: 20260106

Claims (10)

1. A complex system control method based on a diffusion model, comprising: acquiring the current system state, the target system state and initial system noise information sampled in a preset time period of a complex system; the current system state, the target system state and the initial system noise information are all input to a denoising network in a diffusion model, and the denoising network is used for denoising the initial system noise information to obtain a plurality of target prediction system states of the complex system in the preset time period; Inputting a predicted system state sequence formed by the plurality of target predicted system states into an inverse dynamic model, and estimating the states of two adjacent target predicted system states in the predicted system state sequence through the inverse dynamic model to generate a corresponding control signal sequence; and performing limited time control on the complex system based on the control signal sequence.
2. The method of claim 1, wherein the inputting the current system state, the target system state, and the initial system noise information into a denoising network of a diffusion model, denoising the initial system noise information through the denoising network, and obtaining a plurality of target prediction system states of the complex system within the preset time period, comprises: Inputting the current system state, the target system state and the initial system noise information into the denoising network, and denoising the initial system noise information through the denoising network to obtain denoised system noise information; Under the condition that the denoising times do not reach the preset denoising times, re-determining the denoised system noise information as new initial system noise information, inputting the new initial system noise information, the current system state and the target system state into the diffusion model again, and denoising the new initial system noise information through the diffusion model until the denoising times reach the preset denoising times; and under the condition that the denoising times reach the preset denoising times, determining a plurality of prediction system states obtained when the denoising times reach the preset denoising times as the plurality of target prediction system states.
3. The method according to claim 2, wherein the denoising network includes a dual U-shaped network module and a residual connection module, the inputting the current system state, the target system state and the initial system noise information into the denoising network, denoising the initial system noise information through the denoising network, and obtaining denoised system noise information includes: Inputting the current system state, the target system state and the initial system noise information into the double-U-shaped network module, and denoising the initial system noise information through the double-U-shaped network module to obtain first system noise information and second system noise information; And inputting the first system noise information and the second system noise information into the residual error connection module, and determining the sum of the first system noise information and the second system noise information as the denoised system noise information through the residual error connection module.
4. The method of claim 3, wherein the dual U-network module includes a first U-network element and a second U-network element, the inputting the current system state, the target system state, and the initial system noise information into the dual U-network module, denoising the initial system noise information by the dual U-network module, and obtaining the first system noise information and the second system noise information includes: Inputting the current system state, the target system state and the initial system noise information into the first U-shaped network unit, and denoising the initial system noise information through the first U-shaped network unit to obtain a first system noise characteristic and the first system noise information; And inputting the current system state, the target system state, the initial system noise information and the first system noise characteristics into the second U-shaped network unit, and obtaining the second system noise information through the second U-shaped network unit.
5. The method of claim 4, wherein the first U-shaped network element comprises a first one-dimensional U-shaped network and a first-order expansion system estimation network, wherein the inputting the current system state, the target system state and the initial system noise information into the first U-shaped network element, denoising the initial system noise information through the first U-shaped network element to obtain a first system noise characteristic and the first system noise information comprises: Inputting the current system state, the target system state and the initial system noise information into the first one-dimensional U-shaped network, and denoising the initial system noise information through the first one-dimensional U-shaped network to obtain the first system noise characteristics; And inputting the first system noise characteristics into the first-order expansion system estimation network, and carrying out first-order coefficient estimation on the first system noise characteristics through the first-order expansion system estimation network to obtain the first system noise information.
6. The method of claim 4, wherein the second U-shaped network element comprises a second one-dimensional U-shaped network and a second order expansion system estimation network, wherein the inputting the current system state, the target system state, the initial system noise information, and the first system noise characteristic into the second U-shaped network element, the second system noise information being obtained by the second U-shaped network element comprises: The current system state, the target system state, the initial system noise information and the first system noise characteristics are input into the second one-dimensional U-shaped network, and the second system noise characteristics are obtained through the second one-dimensional U-shaped network; And inputting the second system noise characteristics into the second-order expansion system estimation network, and carrying out first-order coefficient estimation on the second system noise characteristics through the second-order expansion system estimation network to obtain the second system noise information.
7. A diffusion model-based complex system control device, comprising: the acquisition unit is used for acquiring the current system state, the target system state and the initial system noise information sampled in a preset time period of the complex system; The denoising unit is used for inputting the current system state, the target system state and the initial system noise information into a denoising network in a diffusion model, denoising the initial system noise information through the denoising network, and obtaining a plurality of target prediction system states of the complex system in the preset time period; The generation unit is used for inputting a predicted system state sequence formed by the plurality of target predicted system states into an inverse dynamic model, and estimating the states of two adjacent target predicted systems in the predicted system state sequence through the inverse dynamic model to generate a corresponding control signal sequence; And the control unit is used for carrying out limited time control on the complex system based on the control signal sequence.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the processor implements the diffusion model based complex system control method according to any one of claims 1 to 6 when executing the computer program.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the diffusion model based complex system control method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the diffusion model-based complex system control method of any one of claims 1 to 6.

Description

Complex system control method, device and equipment based on diffusion model Technical Field The present application relates to the field of control systems, and in particular, to a method, an apparatus, and a device for controlling a complex system based on a diffusion model. Background A complex system is composed of a plurality of interactive components, exhibiting nonlinear dynamics and emerging behavior. The complex system is controlled for a limited time, and the purpose of the complex system is to control the complex system to reach a target system state in a limited time. In the related prior art, the method mainly depends on an accurate system dynamics model to realize the finite time control of a complex system. However, in practical applications, in view of the fact that it is generally difficult to obtain accurate system equations and parameters by using a system dynamics model, especially for a complex system scene with nonlinear dynamics, the uncertainty of the system dynamics model seriously affects the control effect, so that the control effect is poor. Disclosure of Invention The application provides a complex system control method, device and equipment based on a diffusion model, which are used for solving the defect of poor control effect caused by uncertainty of a system dynamics model in the prior art, thereby improving the control effect of a complex system. The application provides a complex system control method based on a diffusion model, which comprises the following steps: acquiring the current system state, the target system state and initial system noise information sampled in a preset time period of a complex system; the current system state, the target system state and the initial system noise information are all input to a denoising network in a diffusion model, and the denoising network is used for denoising the initial system noise information to obtain a plurality of target prediction system states of the complex system in the preset time period; Inputting a predicted system state sequence formed by the plurality of target predicted system states into an inverse dynamic model, and estimating the states of two adjacent target predicted system states in the predicted system state sequence through the inverse dynamic model to generate a corresponding control signal sequence; and performing limited time control on the complex system based on the control signal sequence. According to the complex system control method based on the diffusion model provided by the application, the current system state, the target system state and the initial system noise information are all input into a denoising network of the diffusion model, the initial system noise information is denoised through the denoising network, and a plurality of target prediction system states of the complex system in the preset time period are obtained, and the complex system control method comprises the following steps: Inputting the current system state, the target system state and the initial system noise information into the denoising network, and denoising the initial system noise information through the denoising network to obtain denoised system noise information; Under the condition that the denoising times do not reach the preset denoising times, re-determining the denoised system noise information as new initial system noise information, inputting the new initial system noise information, the current system state and the target system state into the diffusion model again, and denoising the new initial system noise information through the diffusion model until the denoising times reach the preset denoising times; and under the condition that the denoising times reach the preset denoising times, determining a plurality of prediction system states obtained when the denoising times reach the preset denoising times as the plurality of target prediction system states. According to the complex system control method based on the diffusion model provided by the application, the denoising network comprises a double-U-shaped network module and a residual error connection module, the current system state, the target system state and the initial system noise information are all input into the denoising network, the denoising network is used for denoising the initial system noise information to obtain denoised system noise information, and the method comprises the following steps: Inputting the current system state, the target system state and the initial system noise information into the double-U-shaped network module, and denoising the initial system noise information through the double-U-shaped network module to obtain first system noise information and second system noise information; And inputting the first system noise information and the second system noise information into the residual error connection module, and determining the sum of the first system noise information and the second system noise information as