An autonomous positioning method and system for a patrol robot

Abstract

The application discloses a kind of self-positioning method and system of inspection robot, applied to electric power inspection robot field, comprising obtaining environmental perception data and the running event of equipment, running event includes position change time;Extract feature information and establish the time corresponding relation with running event;According to position change time, define time window, determine the dynamic feature associated with running event in window;According to equipment type and event type, weight is assigned to dynamic feature;Based on weight, weighted processing is carried out to feature information and pose is solved.The self-positioning method of inspection robot provided by the application couples equipment running event with positioning process, accurately identifies feature distortion generated by equipment transient motion, and suppresses dynamic feature interference according to equipment type classification, effectively solves the problem that the positioning accuracy decreases due to the difficulty of accurately processing the transient motion of power special equipment in the prior art, and realizes high-precision autonomous positioning of inspection robot in complex dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of power inspection robot technology, and in particular to an autonomous positioning method and system for an inspection robot. Background Technology

[0002] With the continuous improvement of the automation level of power systems, substation inspection robots have become an important technical means to replace manual inspections. When performing autonomous navigation tasks, inspection robots typically use simultaneous localization and mapping (SLAM) technology. They perceive the surrounding environment through onboard sensors, estimate their own pose in real time, and build an environmental map, thereby achieving autonomous localization and navigation in the absence of external positioning facilities.

[0003] When existing inspection robots locate themselves in a power distribution substation environment, they typically incorporate all perceived features of the environment into their pose calculations. However, a power distribution substation environment contains numerous components whose operational status is directly related to the equipment, such as insulating rods during circuit breaker opening and closing, and rotating blades of transformer coolers. The movement of these components causes drastic changes in the perceived features of localized areas. Existing technologies struggle to accurately identify the transient movements of such equipment, and the feature changes generated during equipment movement cannot be effectively processed, resulting in a continuous decline in the positioning accuracy of the inspection robot, and in severe cases, potentially leading to positioning failure. Summary of the Invention

[0004] This invention provides an autonomous positioning method and system for inspection robots to solve the technical problem that the positioning accuracy of inspection robots is affected by the distortion of perception features caused by the inability to recognize the transient motion of the equipment during equipment operation. This enables inspection robots to maintain high-precision positioning during equipment operation and achieve autonomous positioning in complex dynamic environments.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide an autonomous positioning method for an inspection robot, comprising: Acquire environmental sensing data and operational events of at least one device, wherein the operational events include at least the displacement time; Extract feature information from the environmental perception data and establish a time correspondence between the feature information and the running events; A time window is defined based on the displacement time, and dynamic features associated with the running event are determined from the feature information within the time window according to the time correspondence. Weights are assigned to the dynamic features based on the device type and event type corresponding to the running event; The feature information is weighted based on the weights, and the real-time pose is obtained by solving the problem based on the weighted feature information.

[0006] As one preferred embodiment, the acquisition of environmental sensing data and operational events of at least one device includes: Laser point cloud data and image data are collected as the environmental perception data; The device receives a state change event transmitted through a one-way isolation device as the operation event.

[0007] As one preferred embodiment, the step of extracting feature information from the environmental perception data and establishing a time correspondence between the feature information and the operational events includes: Feature points are extracted from the environmental perception data, and trajectory sequences of each feature point are generated; The timestamps of the trajectory sequence are aligned with the shift times of the running events to establish the time correspondence.

[0008] As one preferred embodiment, the step of defining a time window based on the displacement time and determining the dynamic features associated with the running event from the feature information within the time window according to the time correspondence includes: A time window is defined based on the displacement time of the aforementioned running event; Calculate the instantaneous distortion of each feature point within the time window, and determine the peak distortion time based on the instantaneous distortion. Determine the phase difference between the peak distortion time and the displacement time. When the phase difference is within a preset range, determine that the feature point is the dynamic feature associated with the running event.

[0009] As one preferred embodiment, assigning weights to the dynamic features based on the device type and event type corresponding to the running event includes: When the device type is a switchgear and the event type is opening or closing, the weight of the dynamic feature is reset to zero during the first suppression period after the change time. When the device type is a rotating device and the event type is start or stop, during the second suppression period after the displacement time, the weight of the dynamic feature is reduced to the first weight value. When the device does not experience the operation event during the third recovery period, the weight of the feature information in the area where the device is located is restored to the second weight value.

[0010] Another embodiment of the present invention provides an autonomous positioning system for an inspection robot, comprising: The data acquisition module is used to acquire environmental sensing data and the operating events of at least one device, wherein the operating events include at least the displacement time. The time correspondence establishment module is used to extract feature information from the environmental perception data and establish a time correspondence between the feature information and the running events; The dynamic feature determination module is used to define a time window based on the displacement time, and determine the dynamic features associated with the running event from the feature information within the time window according to the time correspondence. The weight allocation module is used to allocate weights to the dynamic features according to the device type and event type corresponding to the running event; The pose solving module is used to perform weighted processing on the feature information based on the weights, and to solve the real-time pose based on the weighted feature information.

[0011] As one preferred embodiment, the data acquisition module includes: An environmental perception data acquisition unit is used to collect laser point cloud data and image data as the environmental perception data. The operation event acquisition unit is used to receive the status change event of the device transmitted through the one-way isolation device as the operation event.

[0012] As one preferred embodiment, the time correspondence establishment module includes: A trajectory sequence generation unit is used to extract feature points from the environmental perception data and generate a trajectory sequence for each feature point; The alignment processing unit is used to align the timestamps of the trajectory sequence with the shift times of the running events to establish the time correspondence.

[0013] As one preferred embodiment, the dynamic feature determination module includes: The time window delineation unit is used to delineate a time window based on the change time of the running event; The distortion peak time determination unit is used to calculate the instantaneous distortion of each feature point within the time window, and determine the distortion peak time based on the instantaneous distortion. The dynamic feature determination unit is used to determine the phase difference between the distortion peak time and the displacement time. When the phase difference is within a preset range, the feature point is determined to be the dynamic feature associated with the running event.

[0014] As one preferred embodiment, the weight allocation module includes: A complete suppression unit is used to reset the weight of the dynamic feature to zero during the first suppression period after the change time when the device type is a switching device and the event type is opening or closing. A partial weight reduction unit is used to reduce the weight of the dynamic feature to a first weight value during a second suppression period after the displacement time when the device type is a rotating device and the event type is start or stop. The long-term recovery unit is used to restore the weight of the feature information in the area where the device is located to the second weight value when the device does not experience the operating event during the third recovery period.

[0015] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: (1) This invention acquires environmental perception data and equipment operation events, establishes a correspondence between the displacement moments in the operation events and the feature information in the environmental perception data, and defines a time window based on the displacement moments. By determining the dynamic features within the window, the inspection robot can accurately identify the instantaneous feature changes that occur before and after the equipment's movement, avoiding the problem in the prior art where feature changes are ignored or misjudged due to the inability to predict equipment movements. In addition, this invention assigns different weights to dynamic features according to equipment type and event type, so that the instantaneous violent movement features of switching equipment are completely suppressed and the continuous movement features of rotating equipment are partially deweighted, solving the technical defect of the prior art that it is difficult to accurately identify the transient movement of power-specific equipment, thereby effectively improving the positioning accuracy of the inspection robot during equipment movement.

[0016] (2) This invention establishes a closed-loop mechanism of "event perception - feature suppression - pose calculation" by deeply coupling equipment operation events with synchronous positioning and map building processes, enabling the inspection robot to achieve autonomous positioning in complex dynamic environments. Compared with the processing methods of existing technologies, this invention can effectively adapt to the transient motion characteristics of power equipment, suppress the interference of feature distortion on positioning in scenarios where equipment operates frequently or multiple equipment operates simultaneously, avoid continuous decline in positioning accuracy, and significantly improve the positioning reliability of the inspection robot in complex environments of substations. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the autonomous positioning method of an inspection robot in one embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the calculation principle of the instantaneous distortion time difference between the device's operating events and visual feature points in one embodiment of the present invention. Figure 3 This is a diagram of a dynamic feature-based hierarchical scheduling strategy based on device type and event type in one embodiment of the present invention; Figure 4 This is a schematic diagram of the autonomous positioning system of the inspection robot in one embodiment of the present invention.

[0018] Figure label: Among them, 11 is the data acquisition module, 12 is the time correspondence establishment module, 13 is the dynamic feature determination module, 14 is the weight allocation module, and 15 is the pose solving module. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0021] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0023] One embodiment of the present invention provides an autonomous positioning method for an inspection robot. For details, please refer to [link / reference needed]. Figure 1 , Figure 1 The diagram shown is a flowchart illustrating the autonomous localization method of an inspection robot according to one embodiment of the present invention, which includes steps S1 to S5: S1: Acquire environmental sensing data and operational events of at least one device, where the operational events include at least the moment of displacement; The inspection robot acquires environmental perception data through onboard sensors and simultaneously obtains equipment operation events from the power production control area. This incorporates the precise timing of equipment actions as prior information into the localization process, enabling the robot to accurately determine the specific time of the equipment's actions and providing a time reference for subsequently actively identifying dynamic features within the corresponding time window.

[0024] Environmental perception data refers to the surrounding environment information collected by the inspection robot using airborne LiDAR and vision cameras, including LiDAR point cloud data (reflecting the three-dimensional spatial structure of the environment) and image data (reflecting the texture features of the environment). This data is used to extract feature points and construct trajectory sequences, which are the basis for subsequent pose calculations. Operational events refer to data records generated when the state of power equipment changes, specifically state change events. These events are captured by local state acquisition units deployed in the power production control area and include equipment identification, change type (such as opening, closing, starting, stopping), and change time.

[0025] Preferably, in one embodiment of the present invention, acquiring environmental sensing data and operational events of at least one device includes: Collect laser point cloud data and image data as environmental perception data; Receive the status change events of the equipment transmitted through the one-way isolation device as operation events.

[0026] Specifically, environmental perception data is collected by a 16-line LiDAR and a global shutter vision camera mounted on the inspection robot. The LiDAR collects point cloud data of the surrounding environment at a frequency of 10Hz, while the vision camera collects image data of the surrounding environment at a frequency of 30FPS. The LiDAR and vision camera are synchronized using the PTP precise time of the onboard inertial measurement unit (IMU), with the synchronization error controlled within ±10ms. The LiDAR point cloud data is used to construct a three-dimensional spatial structure, and the image data is used to extract feature point trajectories.

[0027] Status acquisition devices are deployed in the production control area of ​​the substation to monitor the switching status of equipment such as circuit breakers, disconnectors, grounding switches, and transformer coolers in real time. When a change in the switching status is detected, a status change event message is generated. The message includes the equipment identifier, the change type, and the millisecond-level change time. A status change event refers to an instantaneous action that changes the equipment status, such as circuit breaker opening or closing, or transformer cooler starting or stopping. Each status change corresponds to a precise millisecond-level time, called the change time.

[0028] The event message is unidirectionally transmitted to the robot wireless access gateway in the management information area via a dedicated power-specific forward physical isolation device. The inspection robot synchronizes the message from the gateway via wireless communication, and the actual change time is obtained after timestamp correction. The unidirectional isolation device uses a dedicated power-specific forward physical isolation device, which internally uses data diode technology to achieve unidirectional data transmission, physically blocking reverse data flow and ensuring that the production control area is not subject to external network attacks. The transmission delay is controlled within 200ms, meeting the robot's real-time requirements for change time.

[0029] S2: Extract feature information from environmental perception data and establish a time correspondence between feature information and runtime events; Feature distortion caused by equipment movement only occurs around the time of displacement. By establishing a time correspondence, the system can distinguish whether the feature distortion is caused by equipment movement or by the robot's own movement, thus avoiding the incorrect rejection of normal features or the omission of dynamic features.

[0030] Feature information refers to the set of features extracted from environmental perception data that can be used for localization, including feature points and their trajectory sequences, feature point coordinates, timestamps, and other attributes. In this embodiment, feature information is the foundational data for subsequent dynamic feature recognition and pose calculation. The time correspondence refers to the mapping relationship between the timestamps of the feature point trajectory sequences and the device displacement time. Specifically, for each displacement event, the system can know which feature points are being tracked before and after the event, and the current position and motion state of each feature point.

[0031] Preferably, in one embodiment of the present invention, extracting feature information from environmental perception data and establishing a time correspondence between feature information and runtime events includes: Feature points are extracted from environmental perception data, and trajectory sequences of each feature point are generated; Align the timestamps of the trajectory sequence with the shift times of the running events to establish a time correspondence.

[0032] Feature points refer to pixels in image data that have significant texture or geometric features, such as corner points and edge points, and can be stably tracked in consecutive frames; trajectory sequences refer to the sequence of position coordinates of the same feature point in consecutive time frames, represented as {(p1, t1), (p2, t2), ..., (p... n , t n )}, where p is the coordinate of the feature point in the image or three-dimensional space, t is the corresponding timestamp, and the trajectory sequence records the movement pattern of the feature point over time.

[0033] Specifically, the inspection robot extracts ORB feature points from image data using visual odometry, recording the pixel coordinates, 3D spatial coordinates, and timestamp of each feature point. For each feature point, the system continuously tracks its position changes in consecutive image frames, generating a trajectory sequence of the feature point within a time window T. In this embodiment, the time window T is 5 seconds.

[0034] In this embodiment, visual odometry estimates the camera's motion trajectory by comparing the positional changes of feature points in adjacent image frames; the ORB feature point extraction algorithm first detects key points in the image, and then calculates the orientation information and descriptor of each key point, so that feature points can still be stably matched under different viewing angles and lighting conditions.

[0035] The time correspondence is established using PTP (Precise Time Protocol) as a unified time reference. The timestamps in the trajectory sequences of each feature point are aligned with the change times of the running events to establish a time correspondence between feature information and running events. PTP is a network-based time synchronization protocol capable of achieving microsecond-level time synchronization accuracy. In this embodiment, the airborne IMU serves as the master clock, and the LiDAR, vision camera, and wireless access gateway are all synchronized with the IMU's PTP time. When the robot receives a state change event message, the system corrects it based on the original timestamp in the message and the synchronized local time to obtain the accurate change time.

[0036] S3: Determine the time window based on the displacement time, and determine the dynamic features associated with the running event from the feature information within the time window according to the time correspondence; The feature distortion caused by equipment operation has obvious transient characteristics. Taking circuit breaker tripping as an example, the entire process from the insulated rod going from rest to completing the tripping action lasts only tens of milliseconds. In this extremely short time, the image feature points in the area where the equipment is located will undergo drastic displacement, while the feature points before and after the action are in a relatively stable state. If a full analysis of all feature points in the entire time series is performed, not only will the computational load be large, but it is also easy to misclassify stable features before and after the equipment action as dynamic features, or to miss the distortion features at the moment of the action. By defining the time window based on the displacement moment, the system precisely limits the analysis range to the vicinity of the equipment action moment, which reduces the computational load and improves the accuracy of recognition.

[0037] Preferably, in one embodiment of the present invention, a time window is defined based on the displacement time, and dynamic features associated with the running event are determined from the feature information within the time window according to the time correspondence, including: The time window is defined based on the change time of the running event; Calculate the instantaneous distortion of each feature point within the time window, and determine the peak distortion time based on the instantaneous distortion. Determine the phase difference between the peak distortion time and the displacement time. When the phase difference is within a preset range, the feature point is determined to be a dynamic feature associated with the running event.

[0038] Among them, the instantaneous distortion variable is used to quantify the abnormal motion of feature points. In this embodiment, the pixel coordinate displacement magnitude is used, which reflects the positional change of the feature point between adjacent frames. The distortion peak moment refers to the moment when the feature point distortion index reaches its maximum value within the transient response window. This moment usually has a fixed time offset from the device action moment, and the offset depends on the device type and action type.

[0039] Specifically, the system defines the transient response window [t_event-δ1, t_event+δ2] based on the change time t_event. Here, δ1 is the pre-action stabilization period, and δ2 is the post-action stabilization period. In this embodiment, for fast-switching devices such as circuit breakers and disconnectors, δ1 is 100ms and δ2 is 300ms; for rotating devices such as transformer coolers, δ1 is 200ms and δ2 is 500ms.

[0040] Within the time window, the system traverses the trajectory sequence of all feature points and calculates the instantaneous distortion of each feature point. In this embodiment, the pixel coordinate displacement magnitude is used as the distortion index, that is, the Euclidean distance between the pixel coordinates of the feature point in two adjacent frames is calculated. Specifically, for each feature point, the system records its pixel coordinates (u_t, v_t) in the current frame and its pixel coordinates (u_{t+1}, v_{t+1}) in the next frame, and calculates the straight-line distance between the two coordinate points. This distance is the displacement magnitude of the feature point within this frame interval. The system calculates the displacement magnitude of each feature point frame by frame. When this value exceeds a preset threshold, the feature point is determined to be distorted.

[0041] For feature points exhibiting distortion, the system extracts the maximum point from the curve of its displacement modulus over time. The time corresponding to this maximum point is the peak distortion time, t_peak. The system calculates the difference Δt between the peak distortion time and the displacement time, which is t_peak minus t_event. When the absolute value of Δt is less than or equal to 30 milliseconds, the feature point is determined to be a dynamic feature point associated with the running event, and the feature point is associated with the device. If the absolute value of Δt is greater than 30 milliseconds, it is determined to be a static feature point, and the original weight is retained. The calculation principle of the time difference in this embodiment can be found in [link to relevant documentation]. Figure 2 , Figure 2 The diagram illustrates the principle of calculating the instantaneous distortion time difference between the device's operating events and visual feature points.

[0042] The preset range is set to 30 milliseconds because the operation of fast-switching devices such as circuit breakers and disconnectors typically completes within 20 to 50 milliseconds, with the peak distortion of feature points occurring 10 to 30 milliseconds after the action. Actual measurements show that during the tripping process of a 110kV GIS circuit breaker, the displacement modulus of the feature point reaches its peak 18 milliseconds after the action. Setting the preset range to 30 milliseconds covers the peak distortion time of most fast-switching devices, while also eliminating feature changes caused by robot movement or slow changes in lighting, ensuring the accuracy of dynamic feature recognition.

[0043] S4: Assign weights to dynamic features based on the device type and event type corresponding to the running event; Because the switching time of equipment is extremely short, feature points undergo drastic changes, and including them in the calculation would introduce significant errors. After rotating equipment starts, feature points exhibit periodic and continuous movement; completely removing them would lead to sparse features in that area. Similarly, after equipment has been inactive for a long time, previously suppressed feature points remain in a low-weight state, also causing feature sparsity. Therefore, in this embodiment, based on equipment type and event type, a three-level differentiated weight allocation strategy is adopted for dynamic features: Level A complete suppression, Level B partial weight reduction, and Level C long-term recovery. This achieves a balance between dynamic interference suppression and map feature integrity. For details, please refer to [link to relevant documentation]. Figure 3 , Figure 3 The diagram shows a dynamic feature-based hierarchical scheduling strategy based on device type and event type.

[0044] Preferably, in one embodiment of the present invention, weights are assigned to dynamic features based on the device type and event type corresponding to the running event, including: When the equipment type is a switchgear and the event type is opening or closing, the weight of the dynamic feature is reset to zero during the first suppression period after the change time. When the device type is a rotating device and the event type is start or stop, the weight of the dynamic feature is reduced to the first weight value during the second suppression period after the displacement time. If no operational events occur during the third recovery period, the weights of the feature information within the area where the device is located will be restored to the second weight value.

[0045] The first suppression period refers to the length of time after the switching equipment (such as a circuit breaker or disconnector) operates, during which the system completely suppresses the dynamic characteristics. In this embodiment, it is set to 100 to 300 milliseconds. This period is set because the opening or closing process is usually completed within tens of milliseconds, the peak of feature point distortion appears 10 to 30 milliseconds after the operation, and the entire distortion process basically ends within 100 to 300 milliseconds. Setting the first suppression period within this range can completely cover the distortion characteristics during the equipment operation without affecting the feature extraction of the regions before and after the operation.

[0046] The second suppression period refers to the length of time during which the system partially downweights the dynamic features after the rotating equipment (such as a transformer cooler or an on-load tap changer) starts or stops; in this embodiment, it is taken as 1 to 10 seconds. This period covers the continuous movement phase of the equipment, suppressing interference while preserving the feature density of the region.

[0047] The third recovery period refers to the time threshold for restoring feature weights after the equipment has been idle for a long time. In this embodiment, it is set to more than 24 hours. After the equipment has been idle for a long time, the features in this area have stabilized. The feature sparsity problem is solved by restoring the weights.

[0048] The first weight value refers to the weight to which the dynamic features are reduced when the B-level part is deweighted. In this embodiment, it is set to 0.1 to 0.3. This value allows the dynamic features to participate in the solution with a lower contribution, which reduces interference and avoids feature sparsity. The second weight value refers to the normal weight to which the feature information is restored when the C-level long-term recovery is performed. In this embodiment, it is set to 1, which is consistent with the static feature weight.

[0049] Specifically, the system obtains a list of dynamic feature points associated with the current running event from step S2, with each feature point labeled with a corresponding device ID. When the device type is a circuit breaker or disconnector, and the change type is opening or closing, the system starts a timer. Starting from the change time t_event, during the first suppression period, the weight values ​​of these feature points are set to 0. When performing feature matching and pose calculation at the SLAM front end, feature points with a weight of 0 are skipped directly and do not participate in any calculation. After the first suppression period ends, the system no longer performs special processing on these feature points, and newly extracted feature points in this area participate in the calculation normally.

[0050] When the equipment type is a transformer cooler or an on-load tap changer, and the change type is start or stop, the system starts a timer. Starting from the change time t_event, during the second suppression period, the weight values ​​of these feature points are set to the first weight value. During feature matching and pose calculation at the SLAM front end, these feature points still participate in the calculation, but their weight values ​​are reduced, decreasing their impact on the optimization results. Simultaneously, during the second suppression period, the system disables the extraction of new feature points in the area where the equipment is located; that is, no new ORB feature points are extracted within the image range of that area. After the second suppression period ends, the system restores the extraction of new feature points in that area, but the deweighted feature points continue to maintain low weights until level C recovery is triggered.

[0051] When a device does not experience any operational events during the third recovery period, the system restores the feature information weights of the area where the device is located to the second weight value. The system will pre-establish a database linking devices to map areas, recording the identifier, type, and the set of 3D voxels each device occupies in the global map. The system maintains a counter for each device, recording the interval since the last displacement event. When the interval for a device exceeds the third recovery period, long-term recovery is triggered. The system queries the map area corresponding to the device from the database based on the device identifier, retrieves all feature points within that area (including those previously zeroed or downweighted), restores their weight values ​​to the second weight value, and resets the map confidence score of that area to the initial value of 0.95. After recovery is complete, the system resets the device's interval timer to zero and restarts the timing.

[0052] S5: The feature information is weighted based on the weights, and the solution is obtained based on the weighted feature information to obtain the real-time pose.

[0053] By employing weighted processing, the system can reduce the impact of dynamic features on the solution results while retaining the dominant role of static features, thereby improving positioning accuracy. Specifically, based on the weight values ​​obtained in the preceding steps, a spatiotemporal dynamic feature mask matrix with the same size as the image data or laser point cloud is generated. Each pixel or voxel position in this matrix records the weight value of the corresponding feature point. For feature points with a weight of zero, the system directly filters them out of the feature set and does not participate in any subsequent calculations; for feature points with weights between 0 and 1, the system retains their weight values ​​and uses them for optimization in subsequent solutions.

[0054] At the SLAM front end, the system inputs the weighted feature set into the pose calculation module. Specifically, for visual SLAM, the system uses a weighted beam adjustment algorithm for pose calculation; for laser SLAM, the system uses a weighted iterative nearest-point algorithm for pose calculation.

[0055] The core of the weighted beam adjustment algorithm is constructing a weighted reprojection error function. Specifically, let the coordinates of a feature point in 3D space be P, its observed projection onto the image be p, and the camera pose be T. Then the reprojection error is e = p - π(T, P). Traditional beam adjustment treats the errors of all feature points as equally weighted, constructing the objective function as the sum of squared errors. Weighted beam adjustment, however, introduces weights w into the objective function, meaning the objective function is the weighted sum of the squared errors multiplied by the weights. The system uses an optimization algorithm to find the camera pose that minimizes the weighted objective function, thus obtaining the robot's real-time pose.

[0056] The principle of the weighted iterative nearest point algorithm is similar. When constructing the point cloud registration error, the system weights the error term according to the weight of the feature points, so that the static feature points with high weights contribute more to the pose estimation, and the dynamic feature points with low weights contribute less.

[0057] Another embodiment of the present invention provides an autonomous positioning system for an inspection robot. For details, please refer to [link / reference needed]. Figure 4 , Figure 4 The diagram shown illustrates an autonomous positioning system for an inspection robot according to one embodiment of the present invention, which includes: The data acquisition module is used to acquire environmental sensing data and the operating events of at least one device, the operating events including at least the displacement time; The time correspondence establishment module is used to extract feature information from environmental perception data and establish a time correspondence between feature information and running events; The dynamic feature determination module is used to define a time window based on the displacement time, and to determine the dynamic features associated with the running event from the feature information within the time window according to the time correspondence. The weight allocation module is used to assign weights to dynamic features based on the device type and event type corresponding to the running event; The pose solving module is used to perform weighted processing on feature information based on weights, and then solve for the real-time pose based on the weighted feature information.

[0058] Preferably, in one embodiment of the present invention, the data acquisition module includes: The environmental perception data acquisition unit is used to collect laser point cloud data and image data as environmental perception data. The operation event acquisition unit is used to receive the status change events of the equipment transmitted through the one-way isolation device as operation events.

[0059] Preferably, in one embodiment of the present invention, the time correspondence establishment module includes: The trajectory sequence generation unit is used to extract feature points from environmental perception data and generate trajectory sequences for each feature point. The alignment processing unit is used to align the timestamps of the trajectory sequence with the shift times of the running events to establish a time correspondence.

[0060] Preferably, in one embodiment of the present invention, the dynamic feature determination module includes: The time window delineation unit is used to delineate a time window based on the change time of the running event; The distortion peak time determination unit is used to calculate the instantaneous distortion of each feature point within the time window and determine the distortion peak time based on the instantaneous distortion. The dynamic feature determination unit is used to determine the phase difference between the peak distortion time and the displacement time. When the phase difference is within a preset range, the feature point is determined to be a dynamic feature associated with the running event.

[0061] Preferably, in one embodiment of the present invention, the weight allocation module includes: The complete suppression unit is used to reset the weight of the dynamic feature to zero during the first suppression period after the change time when the equipment type is a switchgear and the event type is opening or closing. Partial weight reduction unit, used to reduce the weight of dynamic features to the first weight value during the second suppression period after the displacement time when the equipment type is rotating equipment and the event type is start or stop; The long-term recovery unit is used to restore the weights of the feature information in the area where the device is located to the second weight value when no operating event occurs during the third recovery period.

[0062] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: (1) This invention establishes a correspondence between devices and map regions, and dynamically updates the feature weights of the regions where the devices are located based on device operation events. When the devices are inactive for a long time, the feature weights of the region are automatically restored, thereby realizing dynamic management and active correction of map confidence.

[0063] (2) This invention achieves low-latency and high-reliability acquisition of equipment operation events by transmitting operation events from the power production control area to the management information area through a one-way isolation device. Under the premise of meeting the power system safety protection regulations, it solves the technical problem of difficulty in data acquisition across safety zones and provides a data foundation for event-driven dynamic feature processing.

[0064] (3) The present invention adopts a hierarchical weight allocation strategy, and performs differentiated processing on dynamic features according to device type and event type, compared with the "one-size-fits-all" method of eliminating dynamic features in the prior art. The present invention retains some deweighted features to participate in the localization calculation while suppressing the interference of dynamic features, avoiding the map sparsity problem caused by excessive feature elimination, and achieving a balance between dynamic feature suppression and map integrity.

[0065] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An autonomous positioning method for an inspection robot, characterized in that, include: Acquire environmental sensing data and operational events of at least one device, wherein the operational events include at least the displacement time; Extract feature information from the environmental perception data and establish a time correspondence between the feature information and the running events; A time window is defined based on the displacement time, and dynamic features associated with the running event are determined from the feature information within the time window according to the time correspondence. Weights are assigned to the dynamic features based on the device type and event type corresponding to the running event; The feature information is weighted based on the weights, and the real-time pose is obtained by solving the problem based on the weighted feature information. The step of extracting feature information from the environmental perception data and establishing a time correspondence between the feature information and the operational events includes: Feature points are extracted from the environmental perception data, and a trajectory sequence of each feature point is generated; Align the timestamps of the trajectory sequence with the shift times of the running events to establish the time correspondence; The step of defining a time window based on the displacement time and determining the dynamic features associated with the running event from the feature information within the time window according to the time correspondence includes: A time window is defined based on the displacement time of the aforementioned running event; Calculate the instantaneous distortion of each feature point within the time window, and determine the peak distortion time based on the instantaneous distortion. Determine the phase difference between the peak distortion time and the displacement time. When the phase difference is within a preset range, determine that the feature point is the dynamic feature associated with the running event. The step of assigning weights to the dynamic features based on the device type and event type corresponding to the running event includes: When the device type is a switchgear and the event type is opening or closing, the weight of the dynamic feature is reset to zero during the first suppression period after the change time. When the device type is a rotating device and the event type is start or stop, during the second suppression period after the displacement time, the weight of the dynamic feature is reduced to the first weight value. When the device does not experience the operational event during the third recovery period, the weight of the feature information in the area where the device is located is restored to the second weight value.

2. The autonomous positioning method for an inspection robot as described in claim 1, characterized in that, The acquisition of environmental sensing data and operational events of at least one device includes: Laser point cloud data and image data are collected as the environmental perception data; The device receives a state change event transmitted through a one-way isolation device as the operation event.

3. An autonomous positioning system for an inspection robot, characterized in that, include: The data acquisition module is used to acquire environmental sensing data and the operating events of at least one device, wherein the operating events include at least the displacement time. The time correspondence establishment module is used to extract feature information from the environmental perception data and establish a time correspondence between the feature information and the running events; The dynamic feature determination module is used to define a time window based on the displacement time, and determine the dynamic features associated with the running event from the feature information within the time window according to the time correspondence. The weight allocation module is used to allocate weights to the dynamic features according to the device type and event type corresponding to the running event; The pose solving module is used to perform weighted processing on the feature information based on the weights, and to solve the real-time pose based on the weighted feature information. The time correspondence establishment module includes: A trajectory sequence generation unit is used to extract feature points from the environmental perception data and generate a trajectory sequence for each feature point; The alignment processing unit is used to align the timestamps of the trajectory sequence with the shift times of the running events to establish the time correspondence. The dynamic feature determination module includes: The time window delineation unit is used to delineate a time window based on the change time of the running event; The distortion peak time determination unit is used to calculate the instantaneous distortion of each feature point within the time window, and determine the distortion peak time based on the instantaneous distortion. A dynamic feature determination unit is used to determine the phase difference between the distortion peak time and the displacement time. When the phase difference is within a preset range, the feature point is determined to be the dynamic feature associated with the running event. The weight allocation module includes A complete suppression unit is used to reset the weight of the dynamic feature to zero during the first suppression period after the change time when the device type is a switching device and the event type is opening or closing. A partial weight reduction unit is used to reduce the weight of the dynamic feature to a first weight value during a second suppression period after the displacement time when the device type is a rotating device and the event type is start or stop. The long-term recovery unit is used to restore the weight of the feature information in the area where the device is located to the second weight value when the device does not experience the operating event during the third recovery period.

4. The autonomous positioning system for an inspection robot as described in claim 3, characterized in that, The data acquisition module includes: An environmental perception data acquisition unit is used to collect laser point cloud data and image data as the environmental perception data. The operation event acquisition unit is used to receive the status change event of the device transmitted through the one-way isolation device as the operation event.

Images