CN-122008890-A - Magnetic levitation system multipoint collaborative reinforcement learning compensation control method, system, equipment and medium based on UAV-InSAR

CN122008890ACN 122008890 ACN122008890 ACN 122008890ACN-122008890-A

Abstract

The invention provides a magnetic levitation system multipoint collaborative reinforcement learning compensation control method, system, equipment and medium based on UAV-InSAR, wherein the method comprises the steps of carrying out high-precision mapping on a magnetic levitation line by using an unmanned aerial vehicle to carry a lightweight X-band interference synthetic aperture radar system, and constructing a track irregularity power spectrum model; the method comprises the steps of constructing a layered control system, generating a basic electromagnetic force control law by an amplitude saturation controller of a lower layer according to the state quantity of a current suspension point, interacting an environment formed by an upper layer based on reinforcement learning, a suspension frame, a track and the amplitude saturation controller as an intelligent body, training by using a standardized irregularity function obtained from a track irregularity power spectrum model, outputting compensation control quantity of each suspension point, superposing the basic electromagnetic force control law and the compensation control quantity, generating a total electromagnetic force control law of each suspension point, and carrying out active cooperative compensation and global stable suspension control on the suspension system. The invention obviously improves the comprehensive control performance of the suspension system.

Inventors

JI WEN
SUN YOUGANG
XU JUNQI
ZHANG ZHIQIANG
LIANG XIN
ZHAO XIAONING
ZHONG ZAIMIN

Assignees

同济大学
中车青岛四方机车车辆股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A magnetic levitation system multipoint collaborative reinforcement learning compensation control method based on UAV-InSAR is characterized by comprising the following steps: Carrying out high-precision mapping on a magnetic levitation line by using an unmanned aerial vehicle carried light X-band interference synthetic aperture radar system, and constructing a track irregularity power spectrum model based on interference phase inversion; constructing a layered control system, wherein the lower layer is an amplitude saturation controller arranged on each suspension point, and the upper layer is a compensator based on reinforcement learning; the amplitude saturation controller generates a basic electromagnetic force control law according to the state quantity of the current suspension point; The compensator based on reinforcement learning is used as an intelligent body, interacts with an environment formed by the suspension frame, the track and the amplitude saturation controller, trains by using a standardized irregularity function obtained from the track irregularity power spectrum model, and outputs compensation control quantity for each suspension point; and superposing the basic electromagnetic force control law and the compensation control quantity to generate a total electromagnetic force control law of each suspension point, and carrying out active cooperative compensation and global stable suspension control on the suspension system.
2. The method of claim 1, wherein the high-precision mapping of the magnetic levitation line by using the unmanned aerial vehicle-mounted lightweight X-band interferometric synthetic aperture radar system, and constructing the track irregularity power spectrum model based on interferometric phase inversion, comprises: Carrying out high-precision mapping on a magnetic levitation line by using an unmanned aerial vehicle carried light X-band interference synthetic aperture radar system; Determining an interference phase according to the geometrical relationship between a main track and a slave track in the mapping process; inversion is carried out through the interference phase, so that elevation disturbance in the track normal direction is obtained; and establishing a track irregularity power spectrum model based on the elevation disturbance field.
3. The method of claim 2, wherein the elevation perturbation expression for the track normal direction is: Wherein, the As the coordinates of the azimuth direction, As a distance-wise coordinate, the coordinate, For the elevation change obtained by the interferometric phase inversion, Is the angle of incidence.
4. The method of claim 3, wherein the expression of the track irregularity power spectrum model is: Wherein, the Is the number of waves to be used, In units of imaginary numbers, As the coordinates of the azimuth direction, L is the observation length; the normal Gao Chengrao dynamic field of the standard track is shown.
5. The method of claim 4, wherein the expression of the normalized irregularity function is: Wherein, the Is a random phase, satisfies , Is the first The number of discrete wavenumber points, Is the first Interval of the wavenumber intervals.
6. The method of claim 1, wherein the expression of the total electromagnetic force control law is: Wherein, the Is a basic electromagnetic force control law; the compensation control quantity for each suspension point is obtained by the learning of the reinforcement learning agent.
7. The method of claim 6, wherein the expression of the basic electromagnetic force control law is: Wherein, the , Is a gain coefficient; for a mass equivalent to a single suspension point, Is the first The error of the individual floating points is calculated, Is the first The derivative of the error for each of the floating points, As a function of the arc-tangent, Gravitational acceleration.
8. A magnetic levitation system multipoint collaborative reinforcement learning compensation control system based on UAV-InSAR is characterized by comprising: The track irregularity power spectrum model construction module is used for carrying out high-precision mapping on a magnetic levitation line by utilizing an unmanned aerial vehicle carried light X-band interference synthetic aperture radar system and constructing a track irregularity power spectrum model based on interference phase inversion; the hierarchical control system construction module is used for constructing a hierarchical control system, wherein the lower layer is an amplitude saturation controller arranged on each suspension point, and the upper layer is a compensator based on reinforcement learning; the total electromagnetic force control law calculation module is used for generating a basic electromagnetic force control law according to the state quantity of the current suspension points by the amplitude saturation controller, interacting with an environment formed by the suspension frame, the track and the amplitude saturation controller by using the reinforcement learning-based compensator as an intelligent body, training by using a standardized irregularity function obtained from the track irregularity power spectrum model, outputting compensation control quantity of each suspension point, superposing the basic electromagnetic force control law and the compensation control quantity, generating the total electromagnetic force control law of each suspension point, and carrying out active cooperative compensation and global stable suspension control on a suspension system.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the UAV-InSAR based magnetic levitation system multipoint collaborative reinforcement learning compensation control method of any one of claims 1 to 7.
10. A computer storage medium, wherein a computer program is stored on the computer storage medium, and when executed by a processor, the computer program realizes the steps in the UAV-InSAR-based magnetic levitation system multipoint collaborative reinforcement learning compensation control method according to any one of claims 1 to 7.

Description

Magnetic levitation system multipoint collaborative reinforcement learning compensation control method, system, equipment and medium based on UAV-InSAR Technical Field The invention relates to the technical field of high-speed magnetic levitation suspension, in particular to a magnetic levitation system multipoint collaborative reinforcement learning compensation control method, system, equipment and medium based on UAV-InSAR. Background The high-speed magnetic levitation train realizes non-contact high-speed operation by means of electromagnetic force, has the advantages of speed and environmental friendliness, and has irreplaceable strategic significance for constructing an efficient and convenient urban mass rapid traffic network and promoting the collaborative development of areas. The levitation system is a core subsystem of the magnetic levitation train, and can maintain the levitation air gap to be stable near a rated value by actively controlling and adjusting the electromagnetic force in real time, and the performance of the levitation system directly determines the running safety and stability of the train. However, suspension systems are inherently complex systems with strong nonlinearities, open loop instability, and are subject to complex and varying external disturbances, a serious set of constraints during operation. In particular, the complex superposition disturbance such as the multipoint coupling effect and track irregularity, the electromagnet output saturation constraint, the inherent hysteresis characteristic of the chopper and the like become key points for restricting the improvement of the performance of the suspension control system. The geometric irregularity of the magnetic levitation line is a key physical quantity affecting the dynamic response, levitation control robustness and running stability of the vehicle, and the high-precision, continuous and spatial distribution representation of the magnetic levitation line has important significance for the design of modern control strategies. The traditional point measurement method is limited by low coverage efficiency and dependence on-site operation and maintenance despite higher precision, and is difficult to meet the requirements of high space-time resolution and large coverage in real-time monitoring of a long-line magnetic levitation system. In recent years, unmanned aerial vehicles have gained widespread attention as a highly flexible, low-cost mobile sensing platform in track and rail infrastructure monitoring. UAV-InSAR research shows that under the condition of high-precision deformation monitoring, unmanned aerial vehicle carrying related system interference imaging has the potential of continuously acquiring local orbit disturbance and inverting with high resolution. The inversion result of the high-precision track irregularity spectrum not only can establish a real external disturbance model for the magnetic levitation system, but also provides an input data source for training and verifying an intelligent control algorithm. The irregularity modeling process based on the UAV-InSAR constitutes a key link of environment modeling in the reinforcement learning control framework, and completes an important task of external disturbance recognition. However, the method of directly applying reinforcement learning cannot theoretically ensure the stability of the suspension system, and in practical application, instability accidents such as rail collision and the like are difficult to avoid. Disclosure of Invention In view of the above, the present invention provides a magnetic levitation system multipoint collaborative reinforcement learning compensation control method, system, device and medium based on UAV-InSAR to solve the above problems. The invention provides a magnetic levitation system multipoint collaborative reinforcement learning compensation control method based on UAV-InSAR, which comprises the steps of carrying a lightweight X-band interference synthetic aperture radar system on an unmanned aerial vehicle to conduct high-precision mapping on a magnetic levitation line, constructing a track irregularity power spectrum model based on interference phase inversion, constructing a layered control system, wherein the lower layer is an amplitude saturation controller arranged on each levitation point, the upper layer is a reinforcement learning-based compensator, the amplitude saturation controller generates a basic electromagnetic force control law according to the state quantity of the current levitation point, the reinforcement learning-based compensator is used as an intelligent body, interacts with an environment formed by a levitation frame, a track and the amplitude saturation controller, trains by utilizing a standardized irregularity function obtained from the track irregularity power spectrum model, outputs compensation control quantity of each levitation point, superimposes the basic electromagnetic force cont