Method for automatically detecting electric energy meter through robot and related product

Abstract

The invention discloses a method for automatically detecting an electric energy meter through a robot and a system for automatically detecting the electric energy meter. The method comprises the steps that the robot is controlled to grab the detected electric energy meter from a storage box; controlling the robot to hang the detected electric energy meter on the meter hanging interface; controlling the robot to press a verification start button so as to verify the to-be-verified electric energy meter through the verification table; and controlling the robot to sort the detected electric energy meters according to the verification result. The robot is controlled to complete the whole process from feeding, meter hanging, verification triggering to discharging and sorting in the electric energy meter verification process, so that unattended operation of the whole process of electric energy meter verification is achieved, the electric energy meter verification work efficiency is improved, and verification precision and result reliability are guaranteed.

Description

Technical Field

[0001] This application generally relates to the field of power system technology. More specifically, this application relates to a method and related products for automating the calibration of electricity meters using robots. Background Technology

[0002] The automated verification of electricity meters mainly relies on fixed automated production lines. Typical configurations include: loading / unloading robots, dedicated meter-hanging robotic arms moving along guide rails, and independent verification initiation and sorting units. While such systems can replace manual labor to some extent, they have significant limitations: each functional module must be strictly aligned and installed at a fixed workstation; the production line layout is rigid and lacks scalability, making it difficult to adapt to changes in the layout of different meter models or verification stations; this mode cannot operate outside of preset tracks and fixed bases, and is essentially still "positional automation" rather than "task automation."

[0003] In recent years, mobile collaborative robots have gained attention in flexible manufacturing scenarios due to their integrated "mobility + operation" capabilities. However, in the field of electricity meter calibration, no solution has yet been found that can effectively integrate the omnidirectional accessibility of a mobile chassis with the precision operation capabilities of a dual-arm robot to complete a fully closed-loop unmanned calibration task in a single-machine configuration, from retrieving meters from the turnover box, moving them to the calibration table, high-precision mounting, button triggering, to classification and playback. Especially in key stages such as meter alignment and button press confirmation, there is a lack of reliable execution strategies combining 3D vision and force feedback, resulting in insufficient system robustness and difficulty in long-term stable operation in real industrial environments. Furthermore, existing automation solutions generally lack anomaly detection and autonomous recovery mechanisms for the entire process. If any deviation occurs in the grasping, mounting, or pressing stages (such as missing the meter, misalignment, or button unresponsiveness), the system often shuts down and alarms, requiring manual intervention to continue operation. This severely restricts the continuous operation capability and true "unmanned" level of the calibration system.

[0004] In view of this, this application proposes a method for calibrating electricity meters using robots and a system for automating electricity meter calibration, so as to improve the flexibility and efficiency of electricity meter calibration work. Summary of the Invention

[0005] In order to at least solve one or more of the technical problems mentioned above, this application proposes a method for automating the verification of electricity meters using robots and a system for automating the verification of electricity meters in several aspects.

[0006] In a first aspect, this application provides a method for automating the verification of electricity meters using a robot, comprising: controlling the robot to pick up an electricity meter to be verified from a storage box; controlling the robot to hang the electricity meter to be verified on a meter mounting interface; controlling the robot to press a verification start button to verify the electricity meter to be verified via the verification table; and controlling the robot to sort the electricity meters to be verified based on the verification results.

[0007] In a second aspect, this application provides a system for automated verification of electricity meters, comprising: a verification table; and a robot configured to: pick up an electricity meter to be tested from a storage box; hang the electricity meter to be tested on a meter mounting interface; press a verification start button to facilitate verification of the electricity meter to be tested via the verification table; and sort the electricity meters to be tested according to the verification results.

[0008] By utilizing the method and system for automated meter verification via robots as described above, this application enables robots to complete the entire meter verification process, from loading, meter mounting, verification triggering to unloading and sorting. This facilitates unattended operation of the entire meter verification process, thereby improving the efficiency of meter verification and ensuring verification accuracy and reliability. Due to the flexible mobility of robots, configuring robots with autonomous movement capabilities can enhance the flexibility of the automated meter verification system. Attached Figure Description

[0009] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein:

[0010] Figure 1 An exemplary schematic diagram of an electricity meter calibration bench in some embodiments of this application is shown.

[0011] Figure 2 An exemplary schematic diagram of an automated energy meter calibration system according to some embodiments of this application is shown.

[0012] Figure 3 An exemplary flowchart of a method for automating the verification of electricity meters using a robot, according to some embodiments of this application, is shown. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] It should be understood that the terms "comprising" and "including" used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0015] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0016] As used in this specification and claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0017] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise expressly specified. "Several" means one or more, unless otherwise expressly specified.

[0018] The specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0019] The electricity meter verification system simulates actual electricity usage conditions on a verification platform to automatically test, calibrate, and evaluate the performance of the electricity meter under test, thereby determining whether the metering performance of the electricity meter under test meets the requirements.

[0020] Figure 1 An exemplary schematic diagram of an electricity meter calibration bench according to some embodiments of this application is shown. For example... Figure 1 As shown, in some embodiments, the electricity meter calibration platform includes a host computer, a slave computer, a calibration device, and a data acquisition system. The host computer runs host computer software, serves as the human-computer interaction interface, and is responsible for issuing calibration tasks, performing data analysis, data storage, querying, and generating calibration reports. The slave computer is equipped with a slave controller, which receives instructions from the host computer, controls the calibration device to perform specific calibration operations, and coordinates the linkage of various hardware components. The calibration device simulates actual electricity consumption scenarios, providing stable and standard calibration conditions for the electricity meter being tested.

[0021] like Figure 1 As further illustrated, the calibration device is equipped with a power supply, a load, a standard meter, and a meter under test. The standard meter serves as the calibration reference, possessing extremely high measurement accuracy. Its measurement data is synchronously transmitted to the host computer along with the data from the meter under test to calculate the error value of the meter under test, ensuring the accuracy of the calibration results. The meter under test, i.e., the electricity meter to be tested, needs to be installed in the calibration device according to the calibration bench requirements, operating in a simulated power environment, and outputting its own measurement data for subsequent comparative analysis. In some embodiments, the standard meter and the meter under test can both be single-phase or three-phase electricity meters. The power supply provides a stable and reliable power input to the entire calibration device, simulating different voltage levels and power supply stability scenarios according to calibration needs, ensuring that the calibration covers diverse actual operating conditions. The load is used to simulate the electrical load in actual use of the electricity meter. By adjusting the load parameters, different power factors and different load sizes can be simulated, comprehensively verifying the measurement accuracy of the meter under test under complex operating conditions.

[0022] continue Figure 1 The data acquisition system is used to collect key data during the verification process in real time and transmit it to the host computer. In some embodiments, the electricity meter verification station can verify the metering performance of various electrical parameters of the meter under test, such as voltage, current, and cumulative energy value. The host computer software sends specific verification tasks to the lower-level controller, which starts the verification device, simulates the actual power consumption environment, and begins the verification. The data acquisition system collects voltage, current, and other data during the verification process in real time and transmits them synchronously to the host computer. The host computer software processes and analyzes the data, displays and stores it, and generates a verification report.

[0023] Figure 2 Exemplary schematic diagrams of automated energy meter calibration systems according to some embodiments of this application are shown. For example... Figure 2As shown, in some embodiments, multiple forms to be tested are placed in a loading and turnover box. A robotic arm can pick up one form to be tested from the loading and turnover box and place it on the form mounting interface of the calibration device. This loading and mounting operation is used to mount the form to be tested onto the calibration device. The calibration device is equipped with a calibration start button. After pressing the calibration start button, the calibration device executes various calibration items according to the program. After completing the calibration program, the robotic arm removes the calibrated form from the form mounting interface, and the forms are sorted into a qualified or unqualified temporary storage area based on the calibration results.

[0024] exist Figure 2 The automated electricity meter calibration system shown primarily relies on fixed automated production lines for automated calibration. Typical configurations include: loading / unloading robots, dedicated meter-hanging robotic arms moving along guide rails, and independent calibration start-up and sorting units. While such systems can replace manual labor to some extent, they have significant limitations: each functional module must be strictly aligned and installed at a fixed workstation; the production line layout is rigid and lacks scalability, making it difficult to adapt to changes in the layout of different meter models or calibration stations. This mode cannot operate independently of preset tracks and fixed bases, essentially remaining a form of "positional automation" rather than "task automation."

[0025] In recent years, mobile collaborative robots have gained attention in flexible manufacturing scenarios due to their integrated "mobility + operation" capabilities. However, in the field of electricity meter calibration, no solution has yet been found that can effectively integrate the omnidirectional accessibility of a mobile chassis with the precision operation capabilities of a dual-arm robot to complete a fully closed-loop unmanned calibration task in a single-machine configuration, from retrieving meters from the turnover box, moving them to the calibration table, high-precision mounting, button triggering, to classification and playback. Especially in key stages such as meter alignment and button press confirmation, there is a lack of reliable execution strategies combining 3D vision and force feedback, resulting in insufficient system robustness and difficulty in long-term stable operation in real industrial environments. Furthermore, existing automation solutions generally lack anomaly detection and autonomous recovery mechanisms for the entire process. If any deviation occurs in the grasping, mounting, or pressing stages (such as missing the meter, misalignment, or button unresponsiveness), the system often shuts down and alarms, requiring manual intervention to continue operation. This severely restricts the continuous operation capability and true "unmanned" level of the calibration system. In view of this, this application proposes a method for calibrating electricity meters using robots and an automated electricity meter calibration system, so as to improve the flexibility and efficiency of electricity meter calibration work.

[0026] Figure 3 An exemplary flowchart of a method 300 for automating the verification of electricity meters using a robot, according to some embodiments of this application, is shown. Figure 3As shown, in some embodiments, method 300 includes: S301, controlling a robot to pick up the energy meter under test from a storage box; S302, controlling the robot to hang the energy meter under test on the meter mounting interface; S303, controlling the robot to press the calibration start button to calibrate the energy meter under test through the calibration table; and S304, controlling the robot to sort the energy meters under test according to the calibration results.

[0027] In some embodiments, the robot can navigate to a storage box (or loading / turnaround box) containing the energy meters to be tested and pick them up from the box. These energy meters may be single-phase or three-phase. The robot picks up the energy meter and hangs it on the meter mounting interface of the meter calibration platform. After the robot presses the calibration start button, the calibration platform calibrates the energy meter. After calibration, the robot can sort the tested energy meters based on the calibration results.

[0028] This application utilizes robot control to enable robots to complete the entire process of electricity meter verification, from loading, meter mounting, verification triggering to unloading and sorting. This facilitates unattended operation of the entire electricity meter verification process, thereby improving the efficiency of electricity meter verification and ensuring verification accuracy and reliability. Due to the flexible mobility of robots, configuring robots with autonomous movement capabilities can enhance the flexibility of the automated electricity meter verification system.

[0029] In some embodiments, controlling a robot to pick up an energy meter under test from a storage box includes: controlling the robot to move toward the storage box; acquiring an image of the storage box using a first camera mounted on the robot's left robotic arm to obtain a first image; acquiring a three-dimensional model related to the energy meter under test; calculating the pose of the energy meter under test in the storage box based on the first image and the three-dimensional model to obtain a first pose; and controlling the left robotic arm to pick up the energy meter under test based on the first pose.

[0030] In some embodiments, a first camera is mounted at the end of the robot's left robotic arm. This first camera is a depth camera, capable of simultaneously acquiring color visual information and spatial depth information of the shooting area, and directly outputting RGB-D fused data. In these embodiments, the robot navigates to the storage box containing the tested energy meter and uses the first camera to capture a first image of the storage box. The first image is RGB-D fused data including color visual information and spatial depth information.

[0031] In some embodiments, the energy meter under test may include multiple models, and a 3D model can be built for each model and stored in a model library. For example, the 3D model stored in the model library can be a 3D OBJ model with the filename extension .obj. The 3D model of the energy meter retains its key geometric features. In these embodiments, a 3D model related to the energy meter under test can be obtained from the model library based on the model of the energy meter under test; for example, a 3D model of the same model as the energy meter under test can be obtained from the model library.

[0032] In some embodiments, the first pose of the tested energy meter in the storage box is calculated based on the first image and the three-dimensional model. The first pose can be a 6D (Dimension) pose, including the coordinates of the X, Y, and Z axes based on a specified coordinate system and the rotation angles around the X, Y, and Z axes. The first pose records the position and attitude information of the tested energy meter in three-dimensional space, so that the left robotic arm can be controlled to accurately grasp the tested energy meter based on the first pose.

[0033] In some embodiments, the robot navigates to the storage box containing the energy meter under test, collects point cloud data of the box using a first camera, and loads the corresponding complete 3D OBJ model from the model library according to the type of the energy meter under test. Based on the 3D OBJ model, a 6D pose estimation algorithm is used to calculate the precise position and pose of the energy meter under test in the coordinate system of the first camera, and the result is converted into the first pose in the coordinate system of the left robotic arm using a hand-eye calibration algorithm. The left robotic arm plans a collision-free grasping trajectory based on the first pose and drives the end effector to complete a stable grasp. The 6D pose estimation algorithm can be an Iterative Closest Point (ICP) algorithm or a pose regression network based on deep learning.

[0034] In some embodiments, the robot navigates to the storage box and collects point cloud data inside the box using a first camera mounted on the end of its left arm, identifying the energy meter at the center of its field of view as the current energy meter to be inspected. Based on the type of the current energy meter, a corresponding complete 3D OBJ model is loaded from the model library. A FoundationPose-based 6D pose estimation algorithm is used to calculate the position and orientation of the energy meter in the first camera coordinate system. The position and orientation in the first camera coordinate system are then transformed to the left robotic arm base coordinate system using a hand-eye calibration matrix to obtain the first pose, which serves as the target pose for grasping. In some embodiments, the FoundationPose algorithm is used to extract the first pose of the energy meter in the storage box based on the first image and the 3D model. Specifically, the FoundationPose algorithm can be implemented as a two-stage pose estimation method that integrates geometric optimization and neural implicit field representation. The first stage is coarse localization, using a deep learning network to predict the initial 6D pose of the energy meter and the correspondence of surface key points from the first image. The second stage is fine registration. Based on the 3D model associated with the tested electricity meter, a signed distance field (SDF) is constructed. The geometric residual between the observed point cloud and the model SDF is minimized through iterative optimization, resulting in a high-precision, sub-millimeter-level final pose estimation. This algorithm is robust to changes in illumination, partial occlusion, and surface reflections, making it particularly suitable for targets with regular geometric structures but with highlights or missing textures, such as electricity meters. It supports arbitrary 3D OBJ model inputs, requires no training of a specific detector, and is ideal for verification scenarios involving rapid switching between multiple electricity meter models.

[0035] In some embodiments, controlling a robot to mount the tested energy meter onto the meter mounting interface includes: controlling the robot to move towards a calibration platform; acquiring an image of the calibration platform using a first camera to obtain a second image; identifying preset calibration platform markers from the second image to obtain marker identification results; calculating the pose of the meter mounting interface based on the marker identification results to obtain a second pose; and controlling the left robotic arm to mount the tested energy meter onto the meter mounting interface based on the second pose.

[0036] In some embodiments, the robot navigates to the calibration station and acquires a second image of the calibration station area using a first camera. This second image can be RGB-D fused data. Then, an object detection algorithm, such as a deep learning object detection algorithm, is used to identify multiple pre-set high-contrast visual markers on the calibration station to obtain marker recognition results. These visual markers can be circular marks. In some embodiments, based on the position of each marker in the image, its corresponding local point cloud is extracted, and the three-dimensional center point coordinates of each marker area are calculated. The geometric center of all marker center points is then used as the reference origin of the meter mounting interface. The normal vector of this plane is calculated using a plane fitting method, thereby constructing the second pose of the meter mounting interface in the camera coordinate system. This second pose can be a 6D pose. Based on the second pose of the meter mounting interface in the camera coordinate system, the left robotic arm can be controlled to accurately mount the tested energy meter onto the meter mounting interface.

[0037] In some embodiments, controlling the left robotic arm to mount the energy meter under test onto the meter mounting interface according to the second pose includes: converting the second pose into a pose in the left robotic arm coordinate system to obtain a third pose; setting a pre-approach target located above the meter mounting interface according to the third pose; controlling the left robotic arm to move towards the pre-approach target; and controlling the energy meter under test to sink so as to facilitate mounting the energy meter under test onto the meter mounting interface.

[0038] In some embodiments, a hand-eye calibration algorithm is used to convert the second pose of the meter mounting interface in the camera coordinate system into a third pose in the left robotic arm coordinate system. The left robotic arm, carrying the energy meter, plans its approach to a pre-approach target based on the third pose. The pre-approach target is set approximately 1-2 cm above the meter mounting interface. Once the end effector reaches the pre-approach target, the robot control system switches to a flexible admittance control mode. In this mode, the end effector allows the meter to sink naturally under its own weight and with slight guidance in the vertical direction (Z-axis, i.e., the direction of gravity). The robot control system monitors the Z-axis force value (i.e., the support reaction force) of the six-dimensional force sensor in real time. If the Z-axis force value is detected to be consistently stable within the preset support force range (e.g., 3-8N, corresponding to the meter's own weight and slight pre-pressure), and the force value fluctuation is less than the threshold (e.g., 0.5N) and maintained for more than 500ms, it is determined that the meter has been stably placed on the meter mounting interface; at the same time, visual verification is used: the distance between the bottom of the meter and the top of the meter mounting interface bracket is confirmed to be less than the tolerance (e.g., <1mm) through the first camera; after the above force-visual dual conditions are met, the left arm releases the gripper, completing the stress-free and highly reliable mounting operation.

[0039] In some embodiments, the meter placement operation is performed by the left robotic arm, the core objective of which is to place the meter smoothly and stress-free on the two extended supports of the calibration table. The entire process is divided into two stages: high-precision pose estimation and flexible placement control. Details are as follows: (1) 6D pose estimation of the hanging table interface.

[0040] The robot navigates and stops approximately 0.5m in front of the calibration platform, ensuring that the first camera fully covers the calibration platform area. Four green circular high-contrast markers, each 16mm in diameter, are pre-set on the calibration platform, symmetrically distributed in a rectangular pattern around the outer perimeter of the dial indicator bracket.

[0041] First, the YOLOv11-seg instance segmentation model is run to detect and segment all circular marker regions in the RGB image. For each detected marker, its corresponding local point cloud in the depth map is extracted, and outliers are removed while retaining valid data. Then, the centroid method is used to calculate the 3D center point of each marker region. , where i = 1, 2, 3, 4.

[0042] Secondly, calculate the geometric center of the four center points as the reference origin for the wall clock:

[0043] Next, using the four center points as input, the Random Sample Consensus (RANSAC) plane fitting algorithm is employed to fit the optimal supporting plane, and its unit normal vector is output. Therefore, the second pose of the hanging table interface in the left camera coordinate system is constructed: position O, orientation n.

[0044] The second pose is the pose under the first camera. It is transformed to the left robotic arm base coordinate system through a pre-calibrated hand-eye matrix to obtain the third pose, and the target of subsequent motion planning is set according to the third pose.

[0045] This design avoids relying directly on the point cloud of the bracket itself, and instead uses highly robust marker points to indirectly reconstruct the hanging plane, significantly improving positioning stability and avoiding the influence of reflections and obstructions.

[0046] (2) Flexible placement and success determination.

[0047] The left robotic arm carries the already grasped energy meter under test and plans its pre-approach to the target based on the third pose. The position of the pre-approach target is set 15mm above O. The attitude of the pre-approach target is set as follows: the Z-axis of the end effector is aligned with the normal vector n, and the overall attitude is slightly deflected upward by 2°, so as to prevent the side wall of the energy meter under test from touching the support column in advance.

[0048] Once the left robotic arm's end effector reaches the pre-approach target, the robot control system switches to admittance control mode, and the spare energy meter slowly sinks under its own weight. The six-dimensional force sensor outputs the Z-axis force value in real time. The robot control system continuously monitors force and visual conditions. When both conditions are simultaneously met, the robot control system determines that the mounting is successful, releases the gripper, raises the left arm 50mm to disengage, and completes the gauge mounting operation. The force conditions are: The force fluctuation amplitude is less than or equal to 0.5N and the duration is ≥500ms. The visual condition is: the distance d between the bottom of the tested energy meter and the top of the support is ≤1mm. In some embodiments, a third camera set on the robot's head collects point cloud data of the bottom of the tested energy meter and the top of the support, and d is calculated based on the distance from the point to the plane. In some embodiments, if either the force condition or the visual condition is not met, the recovery process of the meter hanging sub-state machine is triggered: retreat to the safe position, re-collect data and fine-tune the initial approach height, and retry a maximum of 3 times.

[0049] In some embodiments, controlling the robot to press the calibration start button includes: acquiring an image of the calibration table using a second camera mounted on the robot's right robotic arm to obtain a third image; determining the pose of the calibration start button based on the third image to obtain a fourth pose; controlling the right robotic arm to move toward the calibration start button based on the fourth pose; and controlling the right robotic arm to apply pressure to the calibration start button to trigger the calibration start button.

[0050] In some embodiments, after the calibration table is mounted, the right robotic arm performs the operation of pressing the calibration start button. The right robotic arm is equipped with a second camera, which is a depth camera capable of simultaneously acquiring color visual information and spatial depth information of the shooting area, directly outputting RGB-D fused data. The right robotic arm acquires a third image of the calibration table through the second camera. The third image is RGB-D fused data including color visual information and spatial depth information. A 2D object detection model can be used to locate the calibration start button area in the third image. Combined with the corresponding point cloud data, the button surface plane is fitted, and its center point coordinates and normal vector are calculated to obtain the 6D pose of the calibration start button in the second camera coordinate system. Through hand-eye calibration conversion, the 6D pose of the calibration start button in the second camera coordinate system can be converted to a fourth pose in the right robotic arm coordinate system. The right robotic arm approaches the calibration start button according to the fourth pose and, upon contact, activates force closed-loop control: continuously monitoring the force value in the Z-axis direction. When the force value reaches a preset threshold... And the duration exceeds When the test starts, it is determined that the test start button has been effectively triggered, and the test start is completed.

[0051] In some embodiments, after the energy meter under test is successfully mounted on the calibration table, the robot control system triggers the calibration start process, and the right robotic arm performs the calibration start button pressing operation. This process requires precise alignment of the button position, application of force along the normal direction, and reliable determination of whether the triggering is effective. Specifically, it may include the following two stages: (1) Button 6D Pose Estimation. A second camera is mounted on the end of the right robotic arm. The second camera acquires images at a distance of approximately 30 cm from the area of ​​the test start button to obtain a third image. First, a lightweight YOLOv11s object detection model is run. This model has been trained on a dataset of button images under various lighting and viewing conditions to locate the 2D bounding box of the physical button in the third image. After obtaining the 2D region, the point cloud data in the corresponding depth map is extracted and filtered. The filtering process includes removing outliers and filling holes. Subsequently, plane fitting is performed on the extracted local point cloud: the RANSAC algorithm is used to fit the best plane to obtain the unit normal vector of the test start button surface. The geometric center of all points on the plane is calculated as the center point of the verification start button. Therefore, the 6D pose of the start button in the second camera coordinate system is constructed: the position is... Orientation is By using a pre-calibrated hand-eye matrix, the 6D pose of the start button in the second camera coordinate system can be transformed to the right robotic arm base coordinate system to obtain the fourth pose in the right robotic arm coordinate system, which serves as the pressing target. Thus, by employing a "2D detection + 3D fitting" strategy, both detection speed and pose accuracy are balanced, avoiding the computational overhead of directly using dense point cloud matching, making it suitable for real-time operation scenarios.

[0052] (2) Force closed-loop pressing and trigger determination. An integrated pressing head and a six-dimensional force sensor are installed at the end of the right robotic arm. A force loop along the normal direction is planned based on the fourth pose. The robot approaches the test start button at a low speed of 5 m / s along a straight approach trajectory. Once the pressing head contacts the surface of the test start button, the robot control system switches to force closed-loop control mode, aiming to maintain the force value along the Z-axis (i.e., the pressing direction). Stabilize at the target force value. In some embodiments, the minimum force required to trigger the test start button is 4.5N; the target force value is set to allow for a safety margin. The force is set to 5.5N; the force control bandwidth is set to 100Hz to ensure a fast response. In these embodiments, the robot control system continuously monitors the actual force value. To eliminate the possibility of momentary impact or accidental touch, and to help determine if there is a jam, when If the state persists for more than 500ms and there are no abnormal sudden changes in the joint current, it is determined that the verification start button has been effectively triggered. The right robotic arm maintains the pressed state for 200ms to ensure signal stability, and then retracts 50mm to complete the pressing action. If the judgment condition is not met within 5 seconds, the recovery process of the pressing sub-state machine is triggered: the pressing point is fine-tuned and the target force value is increased through a ±2mm grid search. The recovery process is attempted a maximum of three times. If it still fails, "Button pressing failure" is reported and manual intervention is requested.

[0053] In some embodiments, controlling the robot to sort the tested electricity meters according to the inspection results includes: in response to the tested electricity meter being a qualified product, acquiring an image of the meter hanging area using a first camera to obtain a fourth image; extracting the pose of the tested electricity meter on the meter hanging interface based on the fourth image and a three-dimensional model to obtain a fifth pose; controlling the left robotic arm to grasp the tested electricity meter based on the fifth pose; and controlling the robot to place the tested electricity meter into a qualified product temporary storage box.

[0054] In some embodiments, the robot control system receives communication signals from the calibration station and obtains the calibration results. If the meter is qualified, the left robotic arm places the tested energy meter into a qualified product storage box (or qualified product turnover box). Specifically, if the tested energy meter is qualified, the following steps S1-S7 are performed: S1. The left robotic arm moves to the front of the watch mount and collects point cloud data of the watch mount area using the first camera.

[0055] S2. Based on the type of the energy meter under test, load the corresponding 3D OBJ model, use the 6D pose estimation algorithm to calculate the precise position and attitude of the energy meter under test on the meter mounting interface, and then transform the position and attitude of the energy meter under test on the meter mounting interface into the coordinate system of the left robotic arm to obtain the fifth pose.

[0056] S3. The left robotic arm plans a collision-free grasping trajectory based on the fifth pose and drives the end gripper to complete a stable grasp; S4. The robot navigates to the qualified product storage box and scans the existing power meter layout inside the box using a third camera mounted on the robot's head. The third camera is a depth camera, capable of simultaneously acquiring color visual information and spatial depth information of the shooting area, and directly outputting RGB-D fused data.

[0057] S5. Based on the vacancy detection algorithm, an unoccupied vacancy table region is identified.

[0058] S6. Calculate the coordinates of the three-dimensional center point of the empty surface area, and obtain its surface normal vector through plane fitting to construct the 6D pose of the target.

[0059] S7. The left robotic arm carries the tested energy meter along the optimized path to approach the 6D pose of the target, and switches to flexible control mode at a distance of 1-2cm from the target surface to achieve stable placement based on force feedback.

[0060] In some embodiments, if the tested electricity meter is qualified, the robot navigates to the meter mounting position, and the left robotic arm uses the same 6D pose estimation algorithm as in the meter mounting stage to complete the meter grasping. After grasping, the robot autonomously navigates to a stop approximately 0.3m in front of the qualified product storage box. The interior of the qualified product storage box is divided into regularly arranged meter positions; exemplarily, the meter positions inside the qualified product storage box adopt a 4×3 layout. A third camera positioned on the robot's head captures an overhead view of the box, obtaining an RGB image of the electricity meter placement. A lightweight YOLOv11s object detection model is run, trained on a qualified product storage box image dataset containing various lighting and viewing conditions, to locate the empty meter positions in the storage box within the RGB image. Then, the left robotic arm moves to the empty meter position, acquires the point cloud data of the current meter position, and performs filtering processing, including outlier removal and hole filling. Subsequently, plane fitting is performed on this local point cloud: the RANSAC algorithm is used to fit the optimal plane to obtain the unit normal vector of the empty meter position surface. And calculate the geometric center of all points on the plane as the center position of the empty table. .Depend on and The defined pose, after being converted to the coordinate system of the left robotic arm via hand-eye calibration, is used to plan the placement trajectory. The left robotic arm, carrying the energy meter under test, approaches the target position and switches to flexible admittance control mode at a distance of 15mm from the placement surface, allowing the energy meter to fall naturally under its own weight. When the six-dimensional force sensor detects that the Z-axis support force is stable within the weight range of the energy meter under test and remains stable for more than 500ms, the placement is considered successful, and the gripper is released.

[0061] In some embodiments, controlling the robot to sort the tested electricity meters according to the inspection results includes: in response to the tested electricity meter being a defective product, acquiring an image of the meter hanging area using a first camera to obtain a fifth image; extracting the pose of the tested electricity meter on the meter hanging interface based on the fifth image and a three-dimensional model to obtain a sixth pose; controlling the left robotic arm to grasp the tested electricity meter based on the sixth pose; and controlling the robot to place the tested electricity meter into the defective product isolation area.

[0062] In some embodiments, if the tested electricity meter is a defective product, the following operations are performed: the pose of the tested electricity meter on the meter mounting position is estimated based on the three-dimensional model of the tested electricity meter, and the tested electricity meter is picked up; the robot moves to the defective product isolation area preset on the workbench, the defective product isolation area is equipped with fixed placement marks, and the robot control system directly calls the pre-calibrated placement pose to place the tested electricity meter in the designated position with a standard pose.

[0063] In some embodiments, if the energy meter under test is defective, the left robotic arm grasps the energy meter in the meter mounting position. Subsequently, the robot navigates to a pre-defined defective product isolation area on the right side of the calibration table. This isolation area is a fixed location and does not require real-time sensing. The robot directly invokes a pre-calibrated fixed placement posture, and the left robotic arm smoothly places the energy meter under test in this position in a standard posture. Once completed, the gripper releases, and the task ends. It is understandable that when the number of defective products is usually small and the placement area is fixed, using a pre-calibrated posture can improve efficiency; however, when qualified products are densely packed, it is necessary to dynamically identify empty spaces to avoid collisions.

[0064] In some embodiments, the verification process is controlled by a master state machine and multiple sub-state machines; the master state machine sequentially activates the grasping sub-state machine, the table-mounting sub-state machine, the pressing sub-state machine, and the sorting sub-state machine. In these embodiments, the robot controller employs a hierarchical finite state machine (HFSM) architecture to schedule and monitor the entire verification process. The HFSM consists of a top-level master state machine and multiple sub-task state machines. The top-level master state machine is responsible for macroscopic process advancement, sequentially activating the following sub-task state machines: grasping sub-state machine → table-mounting sub-state machine → pressing sub-state machine → sorting sub-state machine.

[0065] In some embodiments, the robot controller runs an event-driven hierarchical finite state scheduling framework to manage the entire process from grasping to sorting. The scheduling framework consists of a top-level master state machine and grasping sub-task state machines, hanging table sub-task state machines, pressing sub-task state machines, and sorting sub-task state machines. Each sub-state machine is independent of the others and executes sequentially.

[0066] In some embodiments, the top-level master state machine is initially in a "standby" state. Upon receiving a verification task instruction, the sub-state machines are activated sequentially. First, the grasping sub-state machine is started. If the grasping sub-state machine returns "success," the table hanging sub-state machine is activated; and so on, until the sorting sub-state machine completes. If any sub-state machine returns "failure," the top-level master state machine immediately pauses the process, triggers an alarm, and waits for manual confirmation.

[0067] In some embodiments, all state transitions are driven by discrete events, such as a "grip complete" event from the gripper closure sensor, a "force value met" event from the six-dimensional force sensor, a "retry exceeded" event triggered by an internal counter, and a "communication timeout" event caused by the loss of the calibration equipment heartbeat signal.

[0068] In some embodiments, the sub-state machines flow sequentially in a one-way direction. The top-level state machine activates the next subtask only when the current sub-state machine exits in a "successful" state. All state transitions are event-driven to ensure that the process is controllable and traceable. Events that drive state transitions include "grabbing completed", "force value met", "visual verification passed", and "retry exceeded".

[0069] In some embodiments, the sub-state machine includes: an execution state configured to execute a main operation related to the current sub-state machine; a verification state configured to verify whether the main operation has achieved the expected goal; and a recovery state configured to execute a recovery operation related to the current sub-state machine if the main operation has not achieved the prefetched goal.

[0070] In some embodiments, each subtask state machine is an independent state module, which contains: Execution phase. This phase involves executing the main operations, such as fetching and table mounting. Verification state. Multimodal sensors are used to determine whether the main operation was successful.

[0071] Recovery state. If verification fails, a recovery operation is performed according to a preset strategy, and the process returns to the execution state to retry. Recovery operations include repositioning, fine-tuning pose, and changing the gripping point.

[0072] Failure exit. When the number of retries exceeds the threshold N, the current sub-state machine is exited, the fault is reported to the top layer, and a manual intervention request is triggered. The value of N can be 3, 4, or 5.

[0073] Taking the state machine of the hanging meter as an example, it contains the following four states: Execution mode. The left robotic arm moves to the pre-approach target according to the pose of the dial indicator interface in the left robotic arm coordinate system, switches to flexible admittance control mode, and performs the sinking and placement action.

[0074] Verification State. After the main action is completed, force verification and visual verification are performed in parallel. Force verification is: checking whether the Z-axis support force is within the range of [3,8]N and remains stable for more than 500ms. Visual verification is: measuring whether the distance between the bottom surface of the tested electricity meter and the bracket is less than 1mm using the first camera. If both visual and force verifications pass, the "meter hanging successful" event is triggered, the meter hanging sub-state machine exits and returns "success". If any verification fails, the system enters the recovery state. Among these, a sudden increase in force value caused by dry sidewalls will cause force verification to fail, and visual display suspension will cause visual verification to fail.

[0075] Recovery state. Control the left robotic arm to retract along the Z-axis, re-acquire the point cloud of the calibration table area, rerun the marker point detection and pose reconstruction, fine-tune the initial approach height ±2mm or offset angle ±1°, reset the counter and return to the execution state to retry.

[0076] Failure exit. The sub-state machine maintains a retry counter. If the number of retries is greater than or equal to 3, a "table failure" event is triggered, a fault code is reported to the top-level state machine, and a manual intervention request is sent to the monitoring terminal.

[0077] In some embodiments, the table sub-state machine is implemented as a ROS2 node, with each state being an independent callback function. Events are delivered through a topic publish / subscribe mechanism to ensure real-time performance and module decoupling.

[0078] In some embodiments, the recovery strategy of the grasping sub-state machine includes: if the gripper force control verification fails after grasping, such as the meter not being successfully grasped, adjusting the navigation position, recalculating the target attitude, resetting the counter, and returning to the execution state to retry, with a maximum of 3 retries.

[0079] In some embodiments, the recovery strategy of the pressing sub-state machine includes: if the calibration table start button is not triggered, searching for a new pressing point within a ±2mm grid around the original detection position, or gradually increasing the target force value, wherein each increase is 0.5N and the upper limit of the target force value is 7N; and retrying a maximum of 3 times.

[0080] In some embodiments, the recovery strategy of the sorting sub-state machine includes: if the force feedback is abnormal when a qualified product is placed, the left robotic arm is moved to the empty position, the attitude is re-estimated, and the placement process is executed; the process is repeated up to three times.

[0081] Understandably, this application uses a single mobile dual-arm robot to replace the traditional fixed multi-machine production line. It eliminates the need for pre-set tracks or specialized tooling, allowing for rapid adaptation to different models of electricity meters and calibration station layouts. This achieves "task autonomy" rather than "position autonomy," resulting in high flexibility and strong adaptability. Through dual pose estimation using "3D model matching + marker point positioning," combined with end-effector flexible control, the success rate of meter placement is effectively improved, providing high-precision meter placement operation. By integrating 2D detection, 3D normal estimation, and force feedback closed-loop, the system ensures the calibration station's start button is effectively triggered, preventing inattentive testing due to false presses, thus providing verifiable button press status. A built-in state-driven anomaly self-recovery mechanism enables autonomous retrying of operational deviations such as missed grabs, misalignment, or failed presses, significantly reducing the frequency of manual intervention and achieving fully unattended operation. Furthermore, this application completes all processes with a single machine, simplifying the production line structure and reducing initial investment and maintenance complexity, resulting in high system integration and low deployment costs.

[0082] Corresponding to the aforementioned application function implementation method embodiments, this application also provides a system for automated verification of electricity meters, including: a verification table; and a robot configured to: pick up the electricity meter to be tested from a storage box; hang the electricity meter to be tested on the meter mounting interface; press the verification start button to facilitate the verification of the electricity meter to be tested through the verification table; and sort the electricity meters to be tested according to the verification results.

[0083] In summary, the specific functions of the system for automatically verifying energy meters provided in the embodiments of this specification can be explained in comparison with the aforementioned embodiments in this specification, and can achieve the technical effects of the aforementioned embodiments. Therefore, they will not be repeated here.

[0084] While numerous embodiments of this application have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of this application. It should be understood that various alternatives to the embodiments of this application described herein may be employed in the practice of this application. The appended claims are intended to define the scope of protection of this application and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for automating the calibration of electricity meters using robots, characterized in that, include: Control the robot to pick up the tested electricity meter from the storage box; The robot is controlled to mount the tested electricity meter onto the meter mounting interface. The robot is controlled to press the calibration start button so that the energy meter under test can be calibrated through the calibration platform. as well as Based on the test results, the robot is controlled to sort the tested electricity meters.

2. The method according to claim 1, characterized in that, Controlling the robot to retrieve the tested electricity meter from the storage box includes: Control the robot to move towards the storage box; The storage box is imaged by a first camera mounted on the left robotic arm of the robot to obtain a first image; Obtain a three-dimensional model related to the tested energy meter; Based on the first image and the three-dimensional model, the pose of the tested energy meter in the storage box is calculated to obtain the first pose; Based on the first pose, the left robotic arm is controlled to grasp the energy meter being tested.

3. The method according to claim 2, characterized in that, Controlling the robot to mount the tested electricity meter onto the meter mounting interface includes: Control the robot to move towards the calibration table; The first camera is used to capture images of the calibration table to obtain a second image; Identify the preset verification table markers from the second image to obtain the marker identification results; The pose of the hanging table interface is calculated based on the identification results of the marker points to obtain the second pose; Based on the second pose, the left robotic arm is controlled to mount the energy meter to be tested onto the meter mounting interface.

4. The method according to claim 3, characterized in that, Based on the second pose, controlling the left robotic arm to mount the tested energy meter onto the meter mounting interface includes: The second pose is converted into a pose in the left robotic arm coordinate system to obtain the third pose; Based on the third pose, a pre-approach target is set above the watch interface; Control the left robotic arm to move toward the pre-approaching target; and Control the lowering of the tested energy meter so that it can be mounted on the meter mounting interface.

5. The method according to claim 1, characterized in that, Controlling the robot to press the calibration start button includes: A third image is obtained by acquiring images of the calibration table using a second camera mounted on the robot's right robotic arm. The pose of the detection start button is determined based on the third image to obtain the fourth pose; Based on the fourth pose, control the right robotic arm to move toward the calibration start button; Control the right robotic arm to apply pressure to the calibration start button in order to trigger the calibration start button.

6. The method according to claim 2, characterized in that, Based on the inspection results, controlling the robot to sort the inspected energy meters includes: In response to the test electricity meter being deemed a qualified product, The first camera captures images of the watch area to obtain the fourth image; Based on the fourth image and the three-dimensional model, the pose of the tested energy meter on the meter mounting interface is extracted to obtain the fifth pose; Based on the fifth pose, control the left robotic arm to grasp the tested energy meter; and The robot is controlled to place the tested electricity meter into a qualified product storage box.

7. The method according to claim 2, characterized in that, Based on the inspection results, controlling the robot to sort the inspected energy meters includes: In response to the fact that the tested electricity meter was a defective product, The first camera captures images of the watch area to obtain the fifth image; Based on the fifth image and the three-dimensional model, the pose of the tested energy meter on the meter mounting interface is extracted to obtain the sixth pose; Based on the sixth pose, control the left robotic arm to grasp the tested energy meter; and The robot is controlled to place the tested electricity meter into the defective product isolation area.

8. The method according to any one of claims 1-7, characterized in that, include: The verification process is controlled by a main state machine and multiple sub-state machines. The main state machine sequentially activates the grab sub-state machine, the table mounting sub-state machine, the pressing sub-state machine, and the sorting sub-state machine.

9. The method according to claim 8, characterized in that, The sub-state machine includes: The execution state is configured to execute the main operation related to the current sub-state machine. The verification state is configured to verify whether the main operation has achieved the expected goal. The recovery state is configured to perform a recovery operation related to the current sub-state machine if the main operation fails to achieve the expected goal.

10. A system for automatically verifying electricity meters, characterized in that, include: Inspection station; as well as The robot is configured as follows: Retrieve the tested electricity meter from the storage box; Mount the electricity meter to be tested onto the meter mounting interface; Press the calibration start button to allow the energy meter under test to be calibrated on the calibration bench; Based on the test results, the tested electricity meters are sorted.

Images