CN-121979268-A - Autonomous navigation method and system for power station inspection robot

CN121979268ACN 121979268 ACN121979268 ACN 121979268ACN-121979268-A

Abstract

The invention relates to the technical field of high-precision positioning, and discloses an autonomous navigation method and an autonomous navigation system for a power station inspection robot, wherein the method comprises the steps of acquiring original point cloud of a power station, a visual image, environment constraint data and a reference map; the method comprises the steps of carrying out geometric consistency processing on an original point cloud to remove dynamic interference and generate a static environment map, fusing visual data to correct the pose of a robot if a deviation threshold value is exceeded, fusing the point cloud and the visual data to identify the position of the dynamic interference, planning a plurality of candidate tracks by combining the pose and the kinematic constraint, screening an optimal track to generate an obstacle avoidance track sequence, adjusting the track according to real-time sensor data and generating a navigation instruction, determining motion control parameters, executing smooth movement if the motion control parameters meet the environmental constraint, otherwise executing the smooth movement after the correction, and completing the inspection task. The method can realize high-precision autonomous navigation of the power station inspection robot in a dynamic complex environment, and meets the dual requirements of power station operation and maintenance on navigation safety and high efficiency.

Inventors

HE FUCHANG
ZHANG TIANXING
LU ZEHUA
LUO JINGXIANG
Que Kaihua

Assignees

紫金龙净清洁能源有限公司
西藏麻米紫金龙净清洁能源有限公司

Dates

Publication Date: 20260505
Application Date: 20260305

Claims (10)

1. The autonomous navigation method of the power station inspection robot is characterized by comprising the following steps of: acquiring original point cloud data, visual image data, power station environment constraint data and a power station reference map of the current environment of a power station; Performing geometric consistency processing on the original point cloud data, and removing dynamic interference components in the original point cloud data to obtain a static environment map; calculating a position deviation value of a corresponding area of the static environment map and the power station reference map, and if the position deviation value exceeds a preset deviation threshold value, correcting the pose of the robot according to the visual image data to obtain the current pose of the robot with high precision; performing fusion processing on the original point cloud data and the visual image data, and identifying a dynamic object newly added or continuously existing in a real-time environment to obtain a dynamic interference position; Planning and generating candidate tracks of a plurality of robots according to the dynamic interference positions and the current gesture of the robots and combining the pre-acquired robot kinematics constraint rules; according to a pre-acquired track optimization standard, selecting an optimal track from the candidate tracks, and sorting the optimal track with a space coordinate according to a time stamp to generate an obstacle avoidance track sequence; based on the obstacle avoidance track sequence, acquiring real-time sensor data of the robot, analyzing a dynamic object movement trend according to the real-time sensor data, adjusting the obstacle avoidance track sequence, and generating a navigation instruction of the robot according to the adjusted obstacle avoidance track sequence; And determining motion control parameters according to the current gesture of the robot and the navigation instruction, if the motion control parameters accord with the power station environment constraint data, executing smooth movement, if the motion control parameters do not accord with the power station environment constraint data, correcting the motion control parameters to accord with the power station environment constraint data, and then executing the smooth movement to complete the inspection task.
2. The autonomous navigation method of the inspection robot of the power station according to claim 1, wherein the obtaining the original point cloud data, the visual image data, the power station environment constraint data and the power station reference map of the current environment of the power station comprises: starting a laser radar to scan the current environment of the power station, and collecting three-dimensional original point cloud data of the power station; synchronously starting a visual sensor, shooting a current scene image of the power station, and obtaining visual image data; collecting equipment layout, channel width, safety distance and forbidden area data in a power station, and integrating to form power station environment constraint data; A pre-stored station reference map is retrieved, the station reference map containing accurate coordinate information of the station static equipment.
3. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the performing geometric consistency processing on the original point cloud data, removing dynamic interference components therein, and obtaining a static environment map includes: performing inter-frame feature matching on the original point cloud data to construct an inter-frame corresponding point set; Solving a space transformation relation according to the inter-frame corresponding point set, and calculating a pose transformation matrix; aligning the original point cloud data of the continuous frames according to the pose transformation matrix, and carrying out point cloud fusion on the aligned point cloud data to form a temporary global map; extracting geometric consistency constraint conditions from the temporary global map, identifying point cloud clusters which do not accord with the constraint, and marking the point cloud clusters as potential dynamic areas; voxel clustering is carried out on the potential dynamic areas, and the spatial distribution characteristic of each cluster is calculated; And comparing the spatial distribution characteristics with a preset volume threshold value and a preset density threshold value, if the spatial distribution characteristics exceed any one of the volume threshold value and the density threshold value, removing the corresponding point cloud cluster, retaining static point cloud data conforming to constraint, and generating a static environment map.
4. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the calculating the position deviation value of the region corresponding to the static environment map and the power station reference map, if the position deviation value exceeds a preset deviation threshold, correcting the pose of the robot according to the visual image data to obtain the current pose of the robot with high precision comprises: Comparing the coordinates of the corresponding areas of the static environment map and the power station reference map, and calculating a position deviation value; If the position deviation value exceeds a preset deviation threshold, extracting characteristic points in the visual image data, calculating an optical flow field of the characteristic points between continuous frames, and constructing an initial visual pose of the robot according to the motion association of the optical flow field and the characteristic points; Initializing a particle set based on the initial visual pose, wherein the initialized particle weights are equal values; projecting particles in the particle set to an image to generate projection feature points, matching the projection feature points with actual feature points of the image, calculating likelihood values of particle observation according to matching errors, and updating particle weights according to the likelihood values; Screening high-weight particles to form a high-weight particle set, and correcting the pose deviation of the particles according to the geometric constraint of the static environment map to obtain corrected pose of the particles; And calculating the weighted average pose of the high-weight particle set according to the corrected particle pose to obtain the current pose of the robot.
5. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the fusing processing is performed on the original point cloud data and the visual image data, and the dynamic object newly added or continuously existing in the real-time environment is identified to obtain a dynamic interference position, and the method comprises the following steps: Projecting the original point cloud data to an image plane of the visual image data, and performing space matching on the original point cloud data and the contour of the dynamic object which is preliminarily segmented in the image to obtain a dynamic interference candidate point set consistent with the space matching; calculating displacement vectors of the original point cloud data between continuous frames according to the dynamic interference candidate point set, and simultaneously extracting a dynamic target optical flow field of the visual image data; the true motion track of the dynamic object in the real-time environment is confirmed through mutual verification of the displacement vector and the dynamic target optical flow field; And marking the accurate position of the dynamic object according to the real motion trail to obtain a dynamic interference position.
6. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the planning and generating candidate trajectories of a plurality of robots according to the dynamic interference position and the current gesture of the robots and by combining with a pre-acquired robot-combined kinematic constraint rule comprises: taking the current gesture of the robot as a center, intercepting a local area of the static environment map and performing rasterization processing to obtain a local grid map; overlapping the predicted motion area of the dynamic interference position to the local grid map to generate a cost map for dynamic obstacle avoidance; And generating a plurality of candidate tracks in the cost map based on a pre-acquired robot kinematics constraint rule, wherein the constraint rule comprises a maximum linear speed and a maximum angular speed limit.
7. The autonomous navigation method of the power station inspection robot according to claim 6, wherein the selecting the optimal track from the candidate tracks according to the pre-acquired track optimization criteria and sorting the optimal track with the space coordinates according to the time stamp, and generating the obstacle avoidance track sequence comprises: Removing the candidate tracks overlapped with the obstacle area in the cost map, and calculating the comprehensive score of the rest candidate tracks by comparing with a pre-acquired track optimization standard, wherein the track optimization standard comprises path length, smoothness, safety distance and avoidance margin with dynamic interference; And selecting the candidate track with the highest comprehensive score, sorting according to the time stamp and the space coordinate, and generating an obstacle avoidance track sequence.
8. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the acquiring real-time sensor data of the robot based on the obstacle avoidance track sequence, analyzing dynamic object movement trend according to the real-time sensor data, adjusting the obstacle avoidance track sequence, and generating a navigation instruction of the robot according to the adjusted obstacle avoidance track sequence comprises: Acquiring real-time sensor data of a robot, mapping the real-time sensor data to a coordinate system of the obstacle avoidance track sequence, calculating a displacement vector and a motion speed of the dynamic object, and deducing a predicted occupation area in a future preset time; if the overlapping degree of the predicted occupied area and the obstacle avoidance track sequence exceeds a preset overlapping threshold value, correcting the obstacle avoidance track sequence to generate a smooth correction path; And generating a navigation instruction comprising a linear speed instruction and an angular speed instruction according to the corrected path.
9. The autonomous navigation method of the power station inspection robot according to claim 1, wherein the determining a motion control parameter according to the current gesture of the robot and the navigation instruction, if the motion control parameter accords with the power station environment constraint data, executing smooth movement, if not, correcting to accord, executing smooth movement, and completing the inspection task comprises: analyzing the navigation instruction to obtain discrete path point coordinates, and mapping the discrete path point coordinates to a robot body coordinate system; Generating a deviation vector of track tracking according to the current gesture of the robot and the discrete path point coordinates in the body coordinate system; According to the deviation vector, calculating the expected linear speed and the angular speed, and then carrying out boundary clamping to obtain a motion control parameter; if the motion control parameters accord with the power station environment constraint data, planning a speed curve to generate a smooth movement track sequence, and driving the robot to move to a preset inspection point; If not, correcting the motion control parameters based on the power station environment constraint data until the motion control parameters meet constraint requirements, and executing smooth movement; and acquiring an image of the current equipment, and if the acquired image meets the preset task requirement, confirming that the inspection task is completed.
10. An autonomous navigation system of a power station inspection robot, comprising: the data acquisition module is used for acquiring original point cloud data, visual image data, power station environment constraint data and a power station reference map of the current environment of the power station; the static map construction module is used for carrying out geometric consistency processing on the original point cloud data, removing dynamic interference components in the original point cloud data and obtaining a static environment map; The pose correction module is used for calculating the position deviation value of the corresponding area of the static environment map and the power station reference map, and correcting the pose of the robot according to the visual image data if the position deviation value exceeds a preset deviation threshold value, so as to obtain the current pose of the robot with high precision; The dynamic interference identification module is used for carrying out fusion processing on the original point cloud data and the visual image data, and identifying newly-added or continuously-existing dynamic objects in a real-time environment to obtain dynamic interference positions; The track generation module is used for planning and generating candidate tracks of a plurality of robots according to the dynamic interference position and the current gesture of the robots and combining with a pre-acquired robot kinematics constraint rule; The instruction optimization module is used for screening an optimal track from the candidate tracks according to a track optimization standard acquired in advance, and sorting the optimal track with the space coordinates according to time stamps to generate an obstacle avoidance track sequence; The instruction generation module is used for acquiring real-time sensor data of the robot based on the obstacle avoidance track sequence, analyzing the moving trend of the dynamic object according to the real-time sensor data, adjusting the obstacle avoidance track sequence, and generating a navigation instruction of the robot according to the adjusted obstacle avoidance track sequence; And the inspection execution module is used for determining motion control parameters according to the current gesture of the robot and the navigation instruction, executing smooth movement if the motion control parameters accord with the power station environment constraint data, and executing the smooth movement after correcting to accord with the motion control parameters if the motion control parameters do not accord with the power station environment constraint data, so as to complete the inspection task.

Description

Autonomous navigation method and system for power station inspection robot Technical Field The invention relates to the technical field of high-precision positioning, in particular to an autonomous navigation method and an autonomous navigation system for a power station inspection robot. Background At present, in the technical field of high-precision positioning, along with the continuous expansion of the scale of a power station and the increasing complexity of an operation environment, a power station inspection robot is used as core equipment for equipment operation and maintenance, and the autonomous navigation capability of the power station inspection robot is directly related to the safety and the high efficiency of inspection work. The existing autonomous navigation method of the power station inspection robot in the industry mainly depends on a single sensor positioning or static obstacle avoidance strategy, for example, an environment map is built only through a laser radar SLAM, or dynamic interference is ignored by adopting a fixed path planning, or the geometric consistency processing is not carried out by simply fusing sensor data. However, this approach presents significant drawbacks in complex operating environments. The single sensor is easy to be interfered by dynamic objects, so that deviation and positioning drift occur in map construction, static equipment cannot be accurately distinguished from dynamic personnel and vehicles, obstacle avoidance misjudgment is easy to be caused, pre-judgment and track real-time adjustment on the moving trend of the dynamic objects are not available, and particularly in power station scenes with dense equipment and frequent personnel flow, robots are easy to collide and frequently pause, and smooth and efficient navigation is difficult to realize. In conclusion, the prior art is difficult to realize high-precision autonomous navigation of the power station inspection robot in a dynamic complex environment, and cannot meet the dual requirements of power station operation and maintenance on navigation safety and high efficiency. Disclosure of Invention The invention provides an autonomous navigation method and an autonomous navigation system for a power station inspection robot, which are used for realizing high-precision autonomous navigation of the power station inspection robot in a dynamic complex environment and meeting the dual requirements of power station operation and maintenance on navigation safety and high efficiency. In order to solve the technical problems, the invention provides an autonomous navigation method of a power station inspection robot, comprising the following steps: acquiring original point cloud data, visual image data, power station environment constraint data and a power station reference map of the current environment of a power station; Performing geometric consistency processing on the original point cloud data, and removing dynamic interference components in the original point cloud data to obtain a static environment map; calculating a position deviation value of a corresponding area of the static environment map and the power station reference map, and if the position deviation value exceeds a preset deviation threshold value, correcting the pose of the robot according to the visual image data to obtain the current pose of the robot with high precision; performing fusion processing on the original point cloud data and the visual image data, and identifying a dynamic object newly added or continuously existing in a real-time environment to obtain a dynamic interference position; Planning and generating candidate tracks of a plurality of robots according to the dynamic interference positions and the current gesture of the robots and combining the pre-acquired robot kinematics constraint rules; according to a pre-acquired track optimization standard, selecting an optimal track from the candidate tracks, and sorting the optimal track with a space coordinate according to a time stamp to generate an obstacle avoidance track sequence; based on the obstacle avoidance track sequence, acquiring real-time sensor data of the robot, analyzing a dynamic object movement trend according to the real-time sensor data, adjusting the obstacle avoidance track sequence, and generating a navigation instruction of the robot according to the adjusted obstacle avoidance track sequence; And determining motion control parameters according to the current gesture of the robot and the navigation instruction, if the motion control parameters accord with the power station environment constraint data, executing smooth movement, if the motion control parameters do not accord with the power station environment constraint data, correcting the motion control parameters to accord with the power station environment constraint data, and then executing the smooth movement to complete the inspection task. In a second aspect, the present invention provides an aut