CN-121999466-A - Parking control method and system for railway contact net overhaul robot

CN121999466ACN 121999466 ACN121999466 ACN 121999466ACN-121999466-A

Abstract

The application discloses a parking control method and system for a railway contact net overhaul robot, and relates to the field of automatic control, wherein the method comprises the specific steps of collecting depth image data of a parking scene in real time; the method comprises the steps of inputting preprocessed data into a target detection model to obtain bounding box information of an insulator, calculating a depth estimation value, optimizing the depth estimation value through a space-time combined filtering algorithm to obtain a depth estimation optimized value, calculating an initial moving distance, obtaining original point cloud data taking a cantilever of a contact net as a main body after moving the maintenance robot based on the initial moving distance, and calculating three-dimensional pose of a key connecting piece of the contact net in a radar coordinate system, so that three-dimensional positions of the insulator are obtained, and obtaining final insulator comprehensive distance and pose information based on the initial moving distance and the three-dimensional positions of the insulator. The application ensures the stable, high-precision and automatic work of the contact net overhaul robot system in a complex dynamic environment.

Inventors

HUANG DEQING
CHEN LONG
XU JIAYI
MA LEI
QIN NA
SUN YONGKUI
ZHAO DUO
Tian Facun
LIN JIANFEI
XU JUN

Assignees

西南交通大学

Dates

Publication Date: 20260508
Application Date: 20260109

Claims (10)

1. A parking control method for a railway contact net overhauling robot is characterized by comprising the following specific steps: Collecting depth image data of a parking scene in real time, and preprocessing the depth image data; Inputting the preprocessed image data into a trained target detection model, identifying and positioning an insulator (4) in the preprocessed image data, acquiring boundary box information of the insulator (4), and calculating a depth estimation value based on the boundary box information; Optimizing the depth estimation value through a space-time combined filtering algorithm to obtain a depth estimation optimized value, and calculating the comprehensive distance between the overhauling robot body (3) and the contact network plane based on the depth estimation optimized value to serve as an initial moving distance; After the maintenance robot is moved based on the initial movement distance, acquiring original point cloud data taking a cantilever of the overhead line system as a main body; Processing the original point cloud data, and calculating the three-dimensional pose of a key connecting piece of the overhead line system (5) in a radar coordinate system; According to the three-dimensional pose of the key connecting piece in the radar coordinate system, based on the fixed geometric relationship among all the parts, acquiring the three-dimensional position of the insulator (4); obtaining final insulator comprehensive distance and pose information based on the initial movement distance and the three-dimensional position of the insulator (4); And calculating pose errors between the current position of the maintenance robot and the target maintenance point according to the final comprehensive distance and pose information of the insulator, and generating a control instruction to drive a slide rail compensation mechanism of the maintenance robot to perform displacement compensation so as to eliminate parking errors.
2. The parking control method for a railway catenary maintenance robot according to claim 1, wherein the depth estimation value obtaining step is: calculating the center point pixel coordinates of the bounding box based on the bounding box information of the insulator (4); selecting a local neighborhood from the preprocessed image data according to the pixel coordinates of the central point; and sampling and filtering the depth value of the local neighborhood to obtain the depth estimated value of the insulator (4).
3. A parking control method for a railway catenary maintenance robot according to claim 1, characterized in that an independent kalman filter instance is built and maintained for each insulator (4) in a spatio-temporal joint filtering algorithm, and the single processing cycle execution steps are: acquiring a depth estimation value of an insulator (4) in a current frame; Taking the depth estimation value as an observation value and inputting the observation value into the corresponding Kalman filter example; And executing a prediction step and an updating step in the Kalman filter example, and outputting a depth estimation optimization value of the current frame, wherein the prediction step performs state prediction based on the depth estimation optimization value of the last moment and a state space model, and the updating step corrects a prediction state by using the observed value.
4. The parking control method for the railway catenary maintenance robot according to claim 1, wherein the method for resolving the three-dimensional pose of the key connecting piece in the radar coordinate system comprises a point cloud registration algorithm, a geometric feature extraction algorithm or a deep learning segmentation algorithm.
5. The parking control method for a railway catenary maintenance robot according to claim 4, wherein the specific steps of the point cloud registration algorithm are as follows: Template point cloud data are obtained, the template point cloud data are processed based on normal distribution transformation to obtain a probability distribution model, and an initial transformation matrix is solved by maximizing likelihood probability of the original point cloud data under the probability distribution model; Taking the initial transformation matrix as a starting point, and performing iterative optimization on the original point cloud data and the template point cloud data based on an iterative nearest point algorithm to obtain an accurate transformation matrix; And determining the three-dimensional pose of the key connecting piece of the overhead line system (5) in a radar coordinate system according to the precise transformation matrix.
6. The parking control method for a railway catenary maintenance robot according to claim 5, wherein the step of obtaining the probability distribution model is: dividing the space of the template point cloud data into a plurality of voxel grids; Calculating a mean vector and a covariance matrix of each voxel grid based on three-dimensional coordinate vectors of all points in each voxel grid; and constructing the probability distribution model based on the mean vector and the covariance matrix of each voxel grid.
7. The parking control method for a railway catenary maintenance robot according to claim 4, wherein the geometric feature extraction algorithm comprises the following specific steps: and fitting a plurality of straight lines representing the structure from the original point cloud data by using a RANSAC algorithm, and determining the three-dimensional coordinates of the key connecting piece by calculating the intersection point of the plurality of straight lines.
8. The parking control method for a railway overhead line system maintenance robot according to claim 1, wherein the final insulator comprehensive distance and pose information obtaining step comprises the following steps: acquiring an initial moving distance and a confidence index of the three-dimensional position of the insulator (4); assigning a fusion weight based on the confidence indicator; And carrying out weighted fusion on the initial movement distance and the three-dimensional position of the insulator (4) according to the fusion weight to obtain final comprehensive distance and pose information of the insulator.
9. The parking control method for a railway catenary maintenance robot according to claim 1, wherein the displacement compensation amount is: ; in the formula, The distance between the radar and the intersection point of the radar perpendicular to the contact net plane; The distance from the radar to the standard position of the contact net after parking is set.
10. A parking control system for a railway catenary maintenance robot, comprising: the image acquisition and preprocessing module is used for acquiring depth image data of a parking scene in real time and preprocessing the depth image data; The depth estimation module is used for inputting the preprocessed image data into a trained target detection model, identifying and positioning an insulator (4) in the preprocessed image data, acquiring boundary box information of the insulator (4), and calculating a depth estimation value based on the boundary box information; The depth optimization module is used for optimizing the depth estimation value through a space-time combined filtering algorithm to obtain a depth estimation optimization value, and calculating the comprehensive distance between the overhauling robot body (3) and the contact network plane based on the depth estimation optimization value to serve as an initial moving distance; the point cloud data acquisition module is used for acquiring original point cloud data taking a cantilever of the overhead line system as a main body after the maintenance robot is moved based on the initial movement distance; The pose resolving module is used for processing the original point cloud data and resolving the three-dimensional pose of the key connecting piece of the overhead line system (5) in a radar coordinate system; The pose inference module is used for acquiring the three-dimensional position of the insulator (4) based on the fixed geometric relationship among all the parts according to the three-dimensional pose of the key connecting piece in the radar coordinate system; The fusion module is used for obtaining final comprehensive distance and pose information of the insulator based on the initial moving distance and the three-dimensional position of the insulator; and the error correction module is used for calculating the pose error between the current position of the maintenance robot and the target maintenance point according to the final comprehensive distance and pose information of the insulator, and generating a control instruction to drive the maintenance robot to perform displacement compensation so as to eliminate the parking error.

Description

Parking control method and system for railway contact net overhaul robot Technical Field The application relates to the field of automatic control, in particular to a parking control method and system for a railway contact net overhaul robot. Background At present, in railway contact net maintenance operation, the parking positioning of a maintenance vehicle at a target operation point mainly depends on manual visual and communication cooperation, the mode is low in efficiency and greatly influenced by personnel experience, and potential safety hazards are easily introduced due to visual limitation and communication delay at night, in rain and snow or in complex terrains. To improve the automation level, the prior art proposes a variety of parking assistance schemes including ultrasonic ranging, lidar ranging and vision-based ranging. In a complex structure with high reflection and strong shielding characteristics such as a contact net, the ultrasonic and laser radar technology is easy to cause ranging failure or precision reduction due to echo interference, while a vision-based method has a certain effect in a static environment, but in real dynamic scenes such as vehicle vibration, illumination change and local shielding, the stability and the robustness of the method are difficult to meet the actual engineering requirements of high-precision parking control. Therefore, how to realize an automatic parking control system with high precision, strong adaptability and high reliability in a complex railway environment is a problem to be solved by those skilled in the art. Disclosure of Invention In view of the above, the invention provides a parking control method and system for a railway overhead line system maintenance robot, which can effectively solve the problem of inaccurate parking positioning caused by environmental interference, mechanical clearance and single sensor limitation by fusing vision target assistance, radar accurate ranging and sliding rail dynamic compensation technology, improve the ranging reliability, parking precision and environmental adaptability of the system, and ensure the stable, high-precision and automatic work of the overhead line system maintenance robot system in a complex dynamic environment. In order to achieve the above object, the present application provides the following solutions: In a first aspect, the application provides a parking control method for a railway contact net overhaul robot, which comprises the following specific steps: Collecting depth image data of a parking scene in real time, and preprocessing the depth image data; Inputting the preprocessed image data into a trained target detection model, identifying and positioning insulators in the preprocessed image data, acquiring bounding box information of the insulators, and calculating a depth estimation value based on the bounding box information; optimizing the depth estimation value through a space-time combined filtering algorithm to obtain a depth estimation optimized value, and calculating the comprehensive distance between the overhauling robot body and the contact network plane based on the depth estimation optimized value to serve as an initial moving distance; After the maintenance robot is moved based on the initial movement distance, acquiring original point cloud data taking a cantilever of the overhead line system as a main body; Processing the original point cloud data, and calculating the three-dimensional pose of a key connecting piece of the overhead line system in a radar coordinate system; acquiring the three-dimensional position of the insulator based on the fixed geometric relationship among all the components according to the three-dimensional pose of the key connecting piece in the radar coordinate system; Acquiring final comprehensive distance and pose information of the insulator based on the initial moving distance and the three-dimensional position of the insulator; And calculating pose errors between the current position of the maintenance robot and the target maintenance point according to the final comprehensive distance and pose information of the insulator, and generating a control instruction to drive a slide rail compensation mechanism of the maintenance robot to perform displacement compensation so as to eliminate parking errors. Optionally, the step of obtaining the depth estimation value includes: calculating the center point pixel coordinates of the bounding box based on the bounding box information of the insulator; selecting a local neighborhood from the preprocessed image data according to the pixel coordinates of the central point; and sampling and filtering the depth value of the local neighborhood to obtain the depth estimated value of the insulator. Optionally, an independent kalman filter instance is built and maintained for each insulator in the space-time joint filtering algorithm, and the single processing cycle performs the steps of: Acquiring a depth est