Mine heavy-load battery replacing robot, battery replacing manipulator and control method thereof

Abstract

The application provides a mine heavy-load battery replacing robot, a battery replacing manipulator and a control method thereof. The battery replacing manipulator comprises a frame, a plurality of lifting units, a visual positioning unit, a locking assembly, a plurality of force sensors and a control unit. The control unit is configured to calculate the pose deviation of the frame and a target object according to image data, and output a pose adjustment signal in real time. When the locking signals of all the locking assemblies are received and the detection data of all the force sensors meet the preset threshold condition, a lifting permission signal is output. The battery replacing manipulator provided by the application realizes accurate alignment and full-range guidance of the lifting device and the battery pack through a multi-point visual positioning system, realizes reliable hooking through a plurality of independently driven lifting units, and forms a double-dimensional series interlocking logic through mechanical locking confirmation of the locking assembly and bearing verification of the plurality of force sensors. The lifting is allowed only when all the locking is in place and all the lifting units effectively bear the force without bias.

Description

Technical Field

[0001] This invention relates to the field of vehicle battery swapping robot technology, specifically to a heavy-duty battery swapping robot for mining operations, a battery swapping manipulator, and a control method thereof. Background Technology

[0002] With the advancement of electrification in mining equipment, the demand for battery swapping for heavy-duty mining machinery continues to grow. The confined space, high dust concentration, and poor lighting conditions in mines, coupled with the fact that power battery packs typically weigh several tons, place extremely high demands on the alignment accuracy, operational efficiency, and safety of battery swapping operations. For pickup trucks with top-mounted batteries, a robotic arm is needed to grasp and lift / insert the battery pack in the trunk area for swapping. For example, Chinese Patent 202210997035.1 discloses a vision-based grasping system and method for swapping batteries in heavy-duty trucks, controlling a gantry crane robotic arm to move to the corresponding grasping coordinate point to replace the battery in the electric heavy-duty truck.

[0003] The existing technologies mentioned above rely solely on visual recognition for positioning, which cannot identify hidden failure conditions such as false locking of a single claw, false hanging, and unilateral load, posing a safety hazard of heavy-duty battery packs falling off during lifting. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned technical deficiencies and propose a battery swapping robot, a battery swapping robot and its control method, to solve the technical problem of how to avoid hidden failure conditions such as single-claw false locking, false engagement, and unilateral off-center loading in the prior art.

[0005] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a battery swapping robot for grasping a battery pack having a plurality of lifting lugs, comprising: The frame has a connecting part at its top for connecting the lifting mechanism; Multiple lifting units are respectively located at the corresponding corners of the frame and correspond one-to-one with the lifting lugs of the battery pack, for attaching, gripping or detaching from the corresponding lifting lugs. A visual positioning unit includes multiple cameras respectively located at corresponding corners of the frame and facing downwards, used to acquire target positioning feature images and output image data; A locking assembly, which is installed on the frame in a one-to-one correspondence with the lifting unit, is used to lock the lifting unit when it is attached to the corresponding lifting lug and output a locking signal. Multiple force sensors are installed one-to-one on the axial force path of the lifting unit to detect axial force and output detection data; The control unit is signal-connected to the lifting unit, visual positioning unit, locking components, and force sensors, and is configured to calculate the pose deviation between the frame and the target object based on image data and output a pose adjustment signal in real time; and outputs a lifting permission signal when it receives locking signals from all the locking components and when the detection data from all the force sensors meet the preset threshold conditions.

[0006] In some embodiments, the lifting unit includes a main shaft, a lifting claw, and a drive mechanism. The main shaft is vertically rotatably connected to the frame. The lifting claw is fixed to the lower end of the main shaft. The drive mechanism is drively connected to the main shaft and drives the main shaft to rotate, so that the lifting claw has an insertion / removal position for inserting and removing from the corresponding lifting lug, and a lifting position for being hooked onto the inside of the lifting lug after rotation.

[0007] In some embodiments, the drive mechanism includes an explosion-proof servo motor and a worm gear reducer. The explosion-proof servo motor is fixedly mounted on the frame, and its output shaft is connected to the input end of the worm gear reducer. The output end of the worm gear reducer is connected to the main shaft.

[0008] In some embodiments, the locking assembly includes a wheel and a telescopic locking pin. The wheel is fixed to the main shaft and has a limit groove. The telescopic locking pin has a retractable pin end. When the main shaft rotates to the lifting position, the pin end extends and inserts into the limit groove, preventing the main shaft from rotating. When the main shaft rotates to the insertion / removal position, the pin end retracts and disengages from the limit groove. A position sensor is provided in the limit groove to detect whether the pin end is inserted into the limit groove and output a locking signal.

[0009] In some embodiments, the frame is rectangular, and the number of cameras is four, with each of the four cameras mounted on the outer edges of the four corners of the frame. The control unit is configured to receive four channels of image data output by the four cameras when the frame is in a high position, and to calculate the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation of the frame relative to the target object in the horizontal plane. Furthermore, the control unit is configured to, during the descent of the frame, when the hoisting lug or positioning feature moves out of the field of view of any of the cameras, to stop the identification of the hoisting lug and start the identification of the edge features of the target object, and to continuously output the real-time position deviation of the frame relative to the target object based on the identified features.

[0010] In some embodiments, the control unit calculates the pose deviation of the frame relative to the target object, including: When the frame is in a high position, the positioning reference points of the hoisting lugs in the images captured by the four cameras are identified respectively, and the pixel coordinates of the four hoisting lugs at the four corners of the battery pack are obtained. Based on the pre-stored standard physical coordinates of the hoisting lugs at the four corners of the battery pack, and combined with the internal and external parameters of the camera, the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation around the vertical axis of the frame relative to the battery pack in the horizontal plane are obtained through four-point coordinate registration and perspective projection conversion. After the frame is lowered until the hoisting lugs are out of the camera's field of view, the vertical corner edges of the battery pack in the images captured by the four cameras are identified respectively, and the straight line features of the four corner edges are fitted. Based on the standard relative positions of the four corner edges, combined with the camera's internal and external parameters, and according to the image normal offset of each corner edge, the X-axis translational deviation and Y-axis translational deviation of the frame relative to the battery pack in the horizontal plane are calculated.

[0011] In some embodiments, the visual positioning unit further includes a global camera, which is mounted at the bottom center of the frame, for acquiring positioning feature images of the battery pack and outputting image data.

[0012] In some embodiments, the control unit is further configured to calculate the coarse pose deviation between the frame and the target based on the image data output by the global camera, in order to assist in guiding the four corner cameras to lock the hoisting ear hole area.

[0013] Secondly, the present invention also provides a heavy-duty battery swapping robot for mines, comprising: The traveling mechanism is used to translate along a preset path in the horizontal plane. A lifting mechanism, mounted on the traveling mechanism, is used to output vertical lifting drive and rotation drive around a vertical axis; and The battery swapping robot described in any of the above embodiments is connected at the top of its frame to the lifting end of the hoisting mechanism; The overall controller is connected to the walking mechanism, the lifting mechanism, and the battery swapping robot via signals. Based on the positional deviation reported by the control unit of the battery swapping robot, it controls the walking mechanism and the lifting mechanism to adjust their positions to complete alignment, and controls the lifting mechanism to drive the battery swapping robot to descend. It only controls the lifting mechanism to perform the lifting action when it receives the lifting permission signal output by the control unit of the battery swapping robot.

[0014] Thirdly, the present invention also provides a control method for a heavy-duty battery swapping robot in a mine as described above, comprising the following steps: S1. Collect target positioning feature images through multiple cameras at each corner of the frame, identify the positioning reference points of the lifting lugs at each corner of the battery pack and obtain pixel coordinates. Based on the pre-stored standard physical coordinates, and combined with the internal and external parameters of the cameras, calculate the X-direction translation deviation, Y-direction translation deviation and yaw angle deviation of the frame relative to the battery pack in the horizontal plane. S2. Based on the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation, drive the traveling mechanism to adjust the position of the lifting mechanism and frame in the horizontal plane, and drive the lifting mechanism to adjust the yaw angle of the frame around the vertical axis so that each lifting unit is vertically aligned with the corresponding lifting lug. S3. Drive the lifting mechanism to lower the frame. During the descent, continuously collect target positioning feature images and calculate and obtain the pose deviation in real time. Adjust the frame pose in real time according to the pose deviation so that each lifting unit is vertically aligned with the corresponding lifting lug during the descent. S31. During the descent, when the hoisting lug is within the camera's field of view, the hoisting lug of the battery pack is used as the target positioning feature, and the pose deviation is calculated by four-point coordinate registration and perspective projection conversion. S32. After descending to the preset transition zone height, simultaneously perform hoisting ear hole recognition and corner edge recognition. Based on the confidence levels of the hoisting ear hole features and the corner edge features, output continuous pose deviations through weighted fusion. S33. After the mounting ear hole moves out of the camera's field of view, the vertical angle edge of the battery pack is used as the target positioning feature. The real-time pose deviation is calculated by fitting the straight line of the angle edge and continuously fine-tuning the frame pose. S4. After each lifting unit is inserted into the corresponding lifting lug, each lifting unit is driven to perform hooking and grabbing in the lifting lug. After hooking and grabbing, each lifting unit is locked by the locking structure, and the locking position signal is detected and output. S5. The lifting mechanism performs a pre-lifting action, supports the frame and attaches the grabbed battery, and detects the axial load of each lifting unit; S6. If all locking signals are received and the axial load values ​​of all lifting units meet the preset conditions, output a lifting permission signal and execute the lifting action through the lifting mechanism.

[0015] Compared with existing technologies, the battery swapping robot provided by this invention achieves precise alignment and full-process descent guidance of the lifting device and battery pack through a multi-point vision positioning system. It also achieves reliable attachment with multiple independently driven lifting units. Furthermore, it forms a two-dimensional series interlocking logic through mechanical locking confirmation of the locking components and load verification of multiple force sensors. Lifting is only allowed when all locking is in place and all lifting units are effectively bearing force and there is no off-center load. This significantly improves the alignment accuracy and safety redundancy of heavy-duty battery swapping operations in mines, effectively avoiding safety hazards such as false locking, false attachment, and off-center overturning. It is suitable for the high dust, heavy load, and high safety level working environment of mines. Attached Figure Description

[0016] Figure 1 This is a three-dimensional structural diagram of the battery swapping manipulator and heavy-duty battery swapping robot in the mine provided in the embodiments of the present invention; Figure 2 This is a structural diagram of the visual positioning unit of the battery swapping robot provided in an embodiment of the present invention; Figure 3 This is a structural diagram of the lifting unit of the battery swapping robot provided in an embodiment of the present invention; Figure 4 This is a control block diagram of the heavy-duty battery swapping robot in the mine provided in an embodiment of the present invention; Figure 5 This is a flowchart of the control method provided in an embodiment of the present invention.

[0017] Explanation of reference numerals in the attached figures: 100. Battery swapping robotic arm; 1. Frame; 11. Connecting parts; 2. Lifting unit; 21. Main spindle; 22. Lifting claw; 23. Drive mechanism; 231. Explosion-proof servo motor; 232. Worm gear reducer; 3. Visual positioning unit; 31. Camera; 32. Global camera; 4. Locking assembly; 41. Wheel; 42. Telescopic lock stop pin; 43. Limit groove; 44. Position sensor; 5. Force sensor; 6. Control unit; 200. Walking mechanism; 300. Lifting mechanism; 400. Battery pack; 401. Lifting lug; 402. Corner edges; 500. Complete machine controller. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] To address the technical challenges of avoiding hidden failures such as false locking, incomplete attachment, and unilateral overloading, this invention provides a battery swapping robot. It achieves precise alignment and full-process descent guidance between the lifting device and the battery pack through a multi-point vision positioning system. Reliable attachment is achieved through multiple independently driven lifting units. A two-dimensional series interlocking logic is formed by mechanical locking confirmation from the locking components and load-bearing verification from multiple force sensors. Lifting is only permitted when all components are fully locked in place, all lifting units are effectively under load, and there is no unilateral overloading. This significantly improves the alignment accuracy and safety redundancy of heavy-duty battery swapping operations in mines, effectively avoiding safety hazards such as false locking, incomplete attachment, and overturning due to unilateral overloading. It is suitable for the high-dust, heavy-load, and high-safety-level working environments of mines.

[0020] It should be noted that the battery swapping robot described in this invention is used in, but not limited to, heavy-duty pickup trucks and new energy vehicles in mines. For ease of explanation, this invention will only use the application of the battery swapping robot in heavy-duty pickup trucks and new energy vehicles in mines as an example. The principle of the battery swapping robot in other types of equipment is essentially the same as that in heavy-duty pickup trucks and new energy vehicles in mines, and will not be described in detail here.

[0021] Please see Figure 1-3 This invention provides a battery swapping robot for grasping a rectangular heavy-duty battery pack 400 with four lifting lugs 401. The four lifting lugs 401 are respectively arranged at the four corners of the top surface of the battery pack 400, that is, the top of the battery pack 400 has four lifting lugs, and the lifting lugs have lifting lugs 401 that match the shape of the lifting claw 22. The battery swapping robot 100 includes a frame 1, four lifting units 2, a vision positioning unit 3, four sets of locking components 4, four force sensors 5, and a control unit 6.

[0022] In this embodiment, frame 1 is a rectangular high-strength steel structure frame, welded from steel profiles, meeting the requirements for heavy load bearing of several tons. A connecting part 11 is provided at the top of frame 1, which is fixedly connected to the lifting end of the lifting mechanism 300 via high-strength bolts, used to transfer the overall lifting load. Four lifting units 2 are built into the four corners of frame 1, corresponding one-to-one with the four lifting lugs 401 of the battery pack 400, used for attaching to, gripping, or detaching from the corresponding lifting lugs 401. The visual positioning unit 3 includes four downward-facing cameras 31 respectively located at the corresponding corners of frame 1, used to acquire target positioning feature images and output image data. When gripping the battery pack 400, the target positioning features are the lifting lugs 401 and the corner edges 402 of the battery pack. Locking components 4 are installed on the frame 1 in a one-to-one correspondence with the lifting units 2. They are used to lock the lifting units 2 when they are hooked onto the corresponding lifting lugs 401 and output a locking signal, preventing disengagement during hooking and improving the safety of grabbing the battery pack 400. Multiple force sensors 5 are correspondingly located on the axial force path of the lifting units 2 to detect axial force and output detection data. Grabbing and lifting can only be performed when the force on each lifting unit 2 meets the preset threshold conditions, avoiding safety hazards such as false hooking and uneven loading. The control unit 6 is signal-connected to the lifting unit 2, the visual positioning unit 3, the locking assembly 4, and the force sensor 5. It is configured to calculate the positional deviation between the frame 1 and the target object based on image data, output positional adjustment signals in real time, and perform visual positioning to ensure that the lifting unit 2 is aligned with the corresponding lifting ear hole 401. The more accurate the visual alignment, the more correct the posture of insertion into the ear hole, which can avoid locking jamming or incomplete locking due to alignment deviation, thus reducing the probability of locking failure from the source. Furthermore, when the locking position signals of all the locking assemblies 4 are received and the detection data of all the force sensors 5 meet the preset threshold conditions, the lifting permission signal is output, which can provide a safe battery swapping grabbing action.

[0023] Understandably, control unit 6 can be selected from a mining explosion-proof programmable logic controller (PLC), a mining embedded industrial control computer, or a mining explosion-proof and intrinsically safe controller. All three are mature solutions for mining environments, and the specific choice depends on computing requirements and cost considerations.

[0024] In one embodiment, please refer to Figure 1 and Figure 3Four sets of lifting units 2 are respectively assembled at the four corners of the frame 1, corresponding one-to-one with the lifting lugs 401 of the battery pack 400. Each set of lifting units 2 includes a main shaft 21, a lifting claw 22, and a drive mechanism 23. The main shaft 21 is a vertically set stepped shaft, which is vertically rotatably connected to the frame 1 through a bearing seat and can rotate around its own vertical axis. The lifting claw 22 is fixedly installed at the lower end of the main shaft 21 and is L-shaped. When the lower end of the lifting claw 22 is aligned with the elongated hole of the lifting lug 401, the lifting claw 22 can be vertically inserted into the lifting lug 401, i.e., the insertion and removal position; when the main shaft 21 drives the lifting claw 22 to rotate 90°, the length direction of the lifting claw 22 is perpendicular to the elongated hole direction of the lifting lug 401, and the claw head of the lifting claw 22 can abut against the inner top wall bearing end face of the inner side of the lifting lug 401, forming a stable lifting force state, i.e., the lifting position state.

[0025] Understandably, the main shaft 21 is rotatably supported on the frame 1 by bearings. The specific type and installation method of the bearings are set according to the actual load requirements and assembly conditions using conventional bearing configuration methods in the prior art, which will not be elaborated in this embodiment. Because the lower end of the main shaft 21 bears the axial tensile force and radial off-center load during lifting, a pair of tapered roller bearings installed back-to-back can be used to provide sufficient axial and radial load-bearing capacity and anti-overturning rigidity. The upper end of the main shaft 21 adopts a non-locking installation method, allowing the main shaft to undergo slight axial displacement due to force deformation or temperature difference, avoiding over-positioning. The bearing configuration is not a limitation of this invention. Those skilled in the art can select different types such as tapered roller bearings, self-aligning roller bearings, deep groove ball bearings, or angular contact ball bearings, or adopt different installation methods such as face-to-face, back-to-back, or tandem, according to the actual load, speed, and installation space conditions, as long as the rotational support and load transfer functions of the main shaft 21 can be achieved.

[0026] Furthermore, to provide a first-order rotational locking mechanism for moving the battery pack, the drive mechanism 23 includes an explosion-proof servo motor 231 and a worm gear reducer 232. The explosion-proof servo motor 231 is fixedly mounted on the upper surface of the frame 1 via a motor mount, meeting the explosion-proof electrical requirements of underground mines. The output shaft of the explosion-proof servo motor 231 is coaxially connected to the input end of the worm gear reducer 232, and the output end of the worm gear reducer 232 is coaxially fixedly connected to the upper end of the main shaft 21. The rotational power output by the explosion-proof servo motor 231 is reduced and amplified by the worm gear reducer 232, driving the main shaft 21 to rotate around its own axis, realizing the switching of the lifting claw 22 between the insertion / removal position and the lifting position. At the same time, the worm gear reducer 232 has a reverse self-locking characteristic, which can prevent the main shaft 21 from rotating under load in the lifting state, thus forming the first-order rotational locking mechanism.

[0027] Furthermore, to provide a second rotary locking mechanism for moving the battery pack, the locking assembly 4 includes a wheel 41 and a telescopic lock stop pin 42. The wheel 41 is fixed to the main shaft 21 and rotates synchronously with the main shaft 21. Two limiting grooves 43 are formed on the circumferential end face of the wheel 41, with the central angles of the two limiting grooves 43 differing by 90°, corresponding to the insertion / removal position and the lifting position, respectively. The telescopic lock stop pin 42 is a hydraulically, pneumatically, electrically, or electromagnetically driven telescopic pin structure, preferably a structure with a pneumatic push rod connected to a pin shaft. It is fixedly installed on the frame 1 by a bracket, with its insertion end facing the circumferential surface of the wheel 41, and can extend and retract radially along the wheel 41. When the main shaft 21 drives the lifting claw 22 to the lifting position, the pin end of the telescopic lock stop pin 42 extends and inserts into the limiting groove 43 corresponding to the lifting position, mechanically preventing the main shaft 21 from rotating. This, together with the reverse self-locking characteristic of the worm gear reducer 232, forms a double locking structure. When the main shaft 21 needs to rotate back to the insertion / removal position, the pin end of the telescopic lock stop pin 42 retracts, disengages from the limiting groove 43, and releases the lock. Each limiting groove 43 has a position sensor 44 embedded in its bottom. The position sensor 44 is a proximity sensor used to detect whether the pin end is fully inserted into the limiting groove 43 and outputs a locking signal to the control unit 6.

[0028] Understandably, hydraulic, pneumatic, electric, or electromagnetically driven telescopic pin structures can all form effective telescopic movements. To meet the needs of mine use, equipment with explosion-proof standards can be used.

[0029] Furthermore, the force sensor 5 is a spoke-type annular pressure force sensor with a through hole in its center for the main shaft 21 to pass through. The sensor body is fixedly mounted on the frame 1 by a bracket, leaving a gap between it and the main shaft 21, and does not rotate with the main shaft 21. Specifically, the main shaft 21 has a shoulder, and a bearing is fitted on the shoulder. The bearing is installed in a bearing seat on the frame, and the annular force sensor 5 is sandwiched between the upper end face of the bearing seat and the lower crossbeam of the frame 1. In the lifting state, the axial pulling force of the lifting claw 22 is transmitted sequentially through the main shaft shoulder, bearing, bearing seat, and force sensor 5 to the lower crossbeam of the frame 1. This layout ensures that all axial forces must pass through the force sensor 5 before being transmitted to the frame 1, achieving accurate measurement of the actual load.

[0030] Understandably, the four force sensors 5 detect the axial load value of the corresponding lifting unit 2 and output digital detection data to the control unit 6. The logic for the control unit 6 to determine whether the detection values ​​of all force sensors 5 meet the preset conditions is that the detection values ​​of each force sensor 5 are greater than the preset load threshold, confirming that all lifting claws 22 are effectively bearing load; the deviations of the detection values ​​of each force sensor 5 from the average value are all within the preset range, confirming that the battery pack is lifted horizontally without off-center load. Only when the above two conditions are met simultaneously and all locking signals are valid, the control unit 6 outputs a lifting permission signal. After the lifting mechanism 300 tightens slightly upwards, the control unit 6 first checks whether the force on all four claws has reached the lower limit threshold. Only after confirming that all are effectively engaged does it proceed to the off-center load check; if any claw fails to reach the lower limit threshold, a false engagement alarm is triggered. Only when both the no-load pre-tightening check and the off-center load check pass, the control unit 6 outputs a lifting permission signal.

[0031] In one embodiment, please refer to Figure 1 and Figure 2 The frame 1 is rectangular, and there are four cameras 31, which are respectively installed on the outer edges of the four corners of the frame 1. All four cameras 31 are explosion-proof industrial cameras, suitable for dusty and low-light environments in mines. The camera lenses are equipped with dustproof masks and active supplementary lights. The four corner cameras 31 are respectively installed on the outer edges of the four corners of the frame 1, with their lenses facing downwards and adjustable according to actual needs, and have a certain tilt angle to achieve the required capture of the vertical angle edge 402 and ear hole features. The fields of view of each camera partially overlap. At a high position, the corner cameras 31 can completely capture the image of the corresponding hoisting ear hole; during descent, they can capture the image of the vertical angle edge 402 of the battery pack.

[0032] It should be noted that the camera 31 can not only identify the outer edges of the four corners of the frame 1, but also identify the body contour features and the corresponding installation area features of the battery pack 400 when installing the battery pack 400 onto the new energy pickup truck, so as to realize the requirement of one camera 31 to simultaneously complete the grabbing, disassembly and installation.

[0033] Furthermore, the visual positioning unit 3 also includes a global camera 32, which is also an explosion-proof industrial camera, adapted to dusty and low-light environments in mines. The camera lens is equipped with a dustproof lens and an active fill light. The global camera 32 is installed in the center of the bottom of the frame 1, with the lens facing downwards. Its field of view can cover the entire top contour of the battery pack, used to acquire the overall contour image of the battery pack. In addition, the top of the battery pack has a rectangular groove or positioning mark that can be recognized by the global camera 32. If there is a rectangular groove in the center of the top surface of the battery pack 400, its contour has significant straight edge features. The control unit 6 extracts the straight edges in the image through an edge detection algorithm, uses a polygon approximation algorithm to filter out the rectangular contour, and obtains the coordinates of the four corner points and the geometric center coordinates of the rectangular groove. If there is a visual target or reflective label on the top surface of the battery pack 400, the control unit 6 identifies the center position of the label through template matching or feature point extraction algorithms. If there is no dedicated positioning feature on the top surface of the battery pack 400, the control unit 6 directly extracts the outer rectangle of the overall contour of the battery pack 400 and calculates its center coordinates and deflection angle. In this embodiment, a rectangular groove contour is preferred as a coarse positioning feature because it has clear edges, strong anti-interference ability, and is not easily affected by the significant influence of the surface of the battery pack.

[0034] In one embodiment, please refer to Figure 1 and Figure 4 The control unit 6 is an embedded explosion-proof controller installed inside the frame 1. It is connected to the explosion-proof servo motor 231 of the lifting unit 2, each camera of the visual positioning unit 3, the telescopic lock stop pin 42 of the locking assembly 4, the position sensor 44, and the force sensor 5 via explosion-proof signal lines. The control unit 6 is configured to receive four channels of image data output by the four cameras 31 when the frame 1 is in a high position, and calculate the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation of the frame 1 relative to the target object in the horizontal plane. Furthermore, the control unit is configured to stop the recognition of the lifting ear hole 401 and start the recognition of the edge features of the target object when the lifting ear hole 401 or the positioning feature moves out of the field of view of any of the cameras 31 during the descent of the frame 1, and continuously output the real-time position deviation of the frame 1 relative to the target object based on the recognized features.

[0035] Specifically, when the frame 1 is in a high position, the positioning reference points of the hoisting lugs 401 in the images captured by the four cameras 31 are identified respectively, and the pixel coordinates of the four hoisting lugs 401 at the four corners of the battery pack 400 are obtained; based on the pre-stored standard physical coordinates of the hoisting lugs 401 at the four corners of the battery pack 400, and combined with the internal and external parameters of the camera 31, the X-direction translation deviation, Y-direction translation deviation, and yaw angle deviation around the vertical axis of the frame 1 relative to the battery pack 400 in the horizontal plane are obtained through four-point coordinate registration and perspective projection conversion.

[0036] Understandably, the positioning reference point for the lifting ear hole 401 is the centroid of its contour, which is the geometric center of the contour obtained by extracting the image contour and calculating it using the zero-order moment. Specifically, the calculation method involves averaging the coordinates of all pixels of the extracted closed contour of the lifting ear hole to obtain the centroid coordinates, which serve as the positioning reference point for the lifting ear hole. The pixel coordinates of this reference point are the input feature points for subsequent four-point coordinate registration and perspective projection conversion. Regardless of whether the shape of the lifting ear hole 401 is circular, strip-shaped, or irregular, this centroid can be uniquely determined and is stable and repeatable. Even if there are manufacturing deviations or wear deformations at the edge of the ear hole, the change in the centroid position is within a controllable range and will not affect the positioning accuracy.

[0037] Furthermore, after the frame 1 descends until the hoisting ear hole 401 moves out of the camera's field of view, the vertical angular edges 402 of the battery pack 400 in the images captured by the four cameras 31 are identified respectively, and the straight line features of the four angular edges 402 are fitted. Based on the standard relative position of the four angular edges 402, combined with the internal and external parameters of the camera 31, the X-direction translational deviation and Y-direction translational deviation of the frame relative to the battery pack in the horizontal plane are calculated according to the image normal offset of each angular edge 402.

[0038] It should be noted that the camera's imaging model adopts the general pinhole imaging model. This model describes the mathematical mapping relationship between spatial points in the three-dimensional world coordinate system and their projection points on the camera's image plane, and is characterized by the camera's intrinsic and extrinsic parameter matrices. The intrinsic parameter matrix contains the camera's equivalent focal length and principal point coordinates, used to describe the projection relationship of three-dimensional spatial points in the camera coordinate system; the extrinsic parameter matrix contains the camera's rotation matrix and translation vector relative to the frame coordinate system, used to describe the camera's mounting pose on the frame.

[0039] Furthermore, the control unit 6 is also configured to calculate the coarse pose deviation between the frame 1 and the target object based on the image data output by the global camera 32, in order to assist in guiding the four corner cameras 31 to lock the area of ​​the hoisting ear hole 401.

[0040] Furthermore, to better understand the present invention, the visual positioning and pose calculation method built into the control unit 6 will be described in detail below. The pose calculation method built into the control unit 6 is divided into three stages: high-level coarse alignment stage, transition fusion stage, and low-level fine alignment stage.

[0041] In this embodiment, during the high-position coarse alignment stage, when the frame 1 is in a high position (exemplarily, the height from the top surface of the battery pack is ≥500mm), all four corner cameras 31 can completely capture the images of the corresponding corner mounting ear holes. At this time, the four-point coordinate registration method is used to calculate the horizontal pose deviation.

[0042] Specifically, the four-point coordinate registration method is used to solve the horizontal pose deviation. First, image preprocessing and feature extraction are performed. Each camera image is sequentially processed by grayscale conversion, Gaussian filtering, and adaptive binarization. The contour extraction algorithm is used to identify the contour of the hoisting ear hole in the image, and the pixel coordinates of the positioning reference point of the hoisting ear hole are calculated using the zero-order moment. ,in The corresponding lifting lugs are located at the four corners.

[0043] Simultaneously calculate the confidence level of the ear canal features. It is calculated by weighting contour completeness, contour area matching degree, and aspect ratio matching degree: ; in The score is given for the completeness of the outline (values ​​range from 0 to 1, with higher scores for better outline closure). The matching score (0-1) between the detected area and the standard ear canal area is used. The score (0-1) is given to the degree of matching between the aspect ratio of the outline and the standard aspect ratio of the ear canal.

[0044] Next, pixel coordinates are converted to physical coordinates. The current height of the frame is obtained from the height encoder of the lifting mechanism or the laser rangefinder, serving as the known depth value. Combining the pinhole imaging model with the camera's intrinsic and extrinsic parameter matrices (the camera's mounting pose relative to the frame center), the pixel coordinates are... Convert to 3D coordinates in frame coordinate system ,in This is a known constant, representing the depth value corresponding to the current frame height. The planar physical coordinates can be obtained by using the pinhole imaging coordinate transformation relationship. .

[0045] Then, pose calculation and residual verification are performed, and the standard physical coordinates of the positioning reference points of the four lifting lugs of the battery pack are pre-stored. (World coordinate system, with the center of the top surface of the battery pack as the origin). Rigid body transformation in the horizontal plane includes three degrees of freedom: X-axis translational deviation. Y-axis translational deviation Yaw angle deviation The transformation relationship is as follows: ; An overdetermined system of equations was constructed using the coordinates of the four ear holes, and the optimal solution was obtained by using the least squares method. , , This refers to the horizontal pose deviation of the frame relative to the battery pack.

[0046] After the solution is completed, a reprojection residual check is performed. The obtained pose parameters are substituted into the transformation formula to calculate the reprojection error (physical coordinate deviation value) of each set of coordinates. If the maximum residual exceeds a preset threshold (e.g., 5mm), the feature point with the largest error is removed, and the remaining three points are used to re-perform the three-point registration solution; if there are fewer than 3 valid feature points, the pose solution is deemed invalid, and a new photo re-positioning is triggered.

[0047] Furthermore, during the transition fusion phase (for example, when the frame descends to a distance of 200-500mm from the top surface of the battery pack), a smooth feature switching is performed. The hanging ear hole gradually moves out of the field of view of the corner camera, the confidence of the ear hole feature decreases, and at the same time, the corner edges and features gradually become clear. At this time, the transition zone is entered, and a weighted fusion algorithm is used to achieve smooth feature switching and avoid abrupt changes in pose output.

[0048] Specifically, for corner and edge feature extraction and geometric constraint screening, probabilistic Hough transform (PPHT) is used to detect straight lines in each camera image, and lines that are close to the vertical direction are selected as candidate corners and edges. All candidate corners and edges must simultaneously meet the following geometric constraints to be included in the valid corner and edge set: The deviation between the tilt angle of the candidate corner and edge and the pre-stored standard corner and edge tilt angle is less than a preset threshold. The pre-stored standard corner and edge tilt angle is the calibration tilt angle of the i-th corner and edge in the corresponding camera image under standard alignment posture, which is measured and pre-stored by the system during the initial installation or calibration by manual alignment; The horizontal distance between the candidate corner and edge and the candidate corner and edge detected by adjacent cameras in the image matches the nominal width of the battery pack, with a relative error of less than 10%.

[0049] Understandably, the process begins with multi-camera joint calibration to obtain the extrinsic parameters of each camera relative to the same world coordinate system. The pixel coordinates of the corner edges detected by each camera are then uniformly transformed to the same world coordinate system, yielding the three-dimensional physical coordinates of each corner edge. In three-dimensional space, points on two corner edges each have a definite spatial position, and the difference in their horizontal coordinates is the spacing. This spacing has the same physical unit as the nominal width of the battery pack in the world coordinate system, allowing for direct comparison. Based on this, the horizontal physical distance between two corner edges located on the same side of the battery pack is calculated. This measured spacing is then compared with the pre-stored nominal width of the battery pack. If the relative error is less than 10%, the candidate corner edges are deemed to have passed the spacing matching constraint and can be included in the set of valid corner edges. If the error exceeds the threshold, it indicates that at least one corner edge is a background interference line and is discarded. The aforementioned multi-camera joint calibration and coordinate transformation methods are existing technologies in the field of machine vision, and can be implemented by those skilled in the art based on well-known principles.

[0050] It should be noted that the preset threshold for the deviation between the tilt angle of the candidate corner edge and the pre-stored standard corner edge tilt angle, as well as the relative error value between the horizontal distance between the candidate corner edge and the candidate corner edge detected by adjacent cameras in the image and the nominal width of the battery pack, can all be adjusted according to the actual situation. In the image coordinate system, the tilt angle of the corner edge is defined as the angle between the straight line of the corner edge and the horizontal axis of the image. The standard corner edge tilt angle is measured and pre-stored under standard alignment posture during the initial installation or calibration of the system, serving as the reference direction for the corner edge.

[0051] Only candidate corner edges that simultaneously satisfy the constraints are included in the set of valid corner edges. For each valid corner edge, its confidence level is calculated using its line length, edge gradient magnitude, and edge continuity. : ; in The normalized score for the line length is (0-1). The edge gradient magnitude is divided into (0-1). The score is based on the continuity of the line (0-1).

[0052] When there are multiple valid corner edges in a single image, the one with the highest confidence is selected as the final valid corner edge of that path; when the confidence of a single camera corner edge is less than 0.5, the corner edge result of that path is not included in the fusion calculation.

[0053] Control unit 6 calculates the average confidence of all valid corner edges and the average confidence of the four ear hole features. If the number of valid corner edges is zero (i.e., none of the cameras detect any valid corner edges), control unit 6 maintains the pose output of the previous frame, while the control frame pauses its descent and fine-tunes the horizontal field of view, continuing the descent only after the corner edges are recaptured. If no valid corner edges are detected for five consecutive frames, an alarm is directly output to prompt manual intervention. If the number of valid corner edges is greater than or equal to one, control unit 6 dynamically assigns fusion weights to the two types of pose results based on the relative magnitudes of the average ear hole confidence and the average corner edge confidence. The higher the ear hole confidence, the greater the ear hole pose weight; the higher the corner edge confidence, the greater the corner edge pose weight. The sum of the two weights is normalized. Control unit 6 directly weights and fuses the ear hole pose result with the local corner edge pose result to obtain the final fused pose deviation, including X-axis translation deviation, Y-axis translation deviation, and yaw angle deviation. After the switching is completed, the control unit 6 performs a moving average filter on the pose deviation results of consecutive frames to suppress pose jumps between frames.

[0054] During weighted fusion, the control unit calculates the average confidence level for each of the four ear canals. and the average confidence level of all valid corner edges .

[0055] ; ; in This is the number of valid corner edges that reach a preset threshold in terms of confidence. When the number of valid corner edges is 0, weighted fusion is not performed, and the pose of the previous frame is maintained.

[0056] Furthermore, during the low-position fine alignment stage (for example, when the frame descends to a distance of <200mm from the top surface of the battery pack), the hoisting lugs have moved out of the camera's field of view, and the pose calculation is completely switched to using corner edge features. The four corner cameras each acquire a vertical corner edge of the battery pack, and the extracted effective corner edge pixels are fitted with least-squares lines to obtain the corner edge line equations in the image coordinate system. Based on the normal distance between the current corner edge line and the standard line, the normal offset of each corner edge in the image is calculated; combined with the pixel equivalent of the camera at the current height, the normal offset of each corner edge is converted into a horizontal offset. Each of the four corner edges provides a set of observation data, forming an overdetermined system of equations, which is solved using the least-squares method to obtain the X-axis translational deviation and Y-axis translational deviation of the frame relative to the battery pack. The yaw angle deviation remains unchanged from the high-position stage calculation results.

[0057] Understandably, the pixel equivalent is derived from the pinhole imaging model, that is, the actual physical size of a unit pixel at the current depth value; specifically, for the i-th camera, at the current height Z, the pixel equivalent S i =Z / f i , where f i The equivalent focal length of the i-th camera is provided by the camera intrinsic parameter matrix.

[0058] It should be noted that confidence verification is also performed in the low-level phase, i.e., the fitting quality of the corner edges is judged. When there are four effective corner edges, each of the four corner edges provides a set of observation data, forming an overdetermined system of equations. The X-axis and Y-axis translational deviations of the frame relative to the battery pack are obtained by solving the least squares method. When there are three effective corner edges, the three corner edges can still form an overdetermined system of equations (three equations to solve for two unknowns). The optimal solution is obtained by the least squares method, with slightly lower accuracy than that of four corner edges, but still acceptable. When there are fewer than three effective corner edges, it is determined that the constraints obtained from the current vision are insufficient to simultaneously solve the X-axis and Y-axis translational deviations. The control unit maintains the effective pose output of the previous frame, pauses the descent, fine-tunes the horizontal field of view, and then re-acquires the image. Alternatively, if the number of effective corner edges is 0 for two consecutive frames, the global camera is automatically called to solve the coarse pose, guiding the frame to adjust the field of view and re-capture the corner edge features; if it still cannot be captured for five consecutive frames, an alarm is triggered. Understandably, the global camera can also continuously provide auxiliary identification of yaw angle deviation.

[0059] Furthermore, when the robot arm is far from the battery pack, the four corner cameras cannot fully capture the four ear holes. In this case, the global camera identifies the four corner points of the overall rectangular outline of the battery pack and uses the same four-point coordinate registration method as in the high-level stage to calculate the coarse pose deviation. Moreover, when the ear hole features gradually disappear and the corner and edge features are not yet stable, the coarse pose deviation is weighted and fused with the local pose of the corners and edges to suppress pose jumps. Initial coarse guidance: When the distance between the battery swapping robot 100 and the battery pack 400 is greater than one meter, the four corner cameras 31 cannot fully capture the lifting ear holes 401. At this time, the global camera 32 identifies the rectangular groove outline or overall outline of the top surface of the battery pack 400, calculates the coarse pose deviation of the frame 1 relative to the battery pack 400, and guides the walking mechanism 200 to move the robot to directly above the battery pack 400, so that the four lifting ear holes 401 enter the field of view of the four corner cameras 31 respectively. Transition interval assistance: When the frame descends to the 200~500mm transition interval, and the ear hole features and corner edge features are not completely reliable, the coarse pose deviation calculated based on the rectangular groove outline is introduced as an auxiliary constraint into the weighted fusion to suppress the fluctuation of the pose output.

[0060] For example, suppose that in the current frame, the average confidence of the ear hole is 0.5, the average confidence of the corner edge is 0.4, and the global coarse position confidence is 0.6. Since the average confidence of the corner edge is lower than a preset threshold, the control unit first fuses the local pose of the corner edge with the global coarse pose: the local pose of the corner edge accounts for 70%, and the global coarse pose accounts for 30%, resulting in a corrected corner edge pose. Subsequently, the control unit calculates the fusion weight between the ear hole confidence and the corrected corner edge confidence. In this example, the ear hole confidence is 0.5, and the corrected corner edge confidence is also 0.5, so the weight of both types of features is 0.5. The ear hole pose result and the corner edge pose result each account for half, jointly determining the final pose output. As the frame continues to descend, the corner edges gradually become clear and complete, and their confidence level rises to 0.7. At this point, the global coarse pose no longer participates in the correction of the corner edge pose. The corner edge pose is completely determined by its local pose solution, while the global coarse pose is only used as a monitoring reference until it is completely withdrawn.

[0061] Understandably, the capture phase employs a three-stage visual localization strategy: high-level earhole registration, transition zone weighted fusion, and low-level corner and edge fitting. When the frame is at a high position, the earhole with a complete outline and significant features is used as the localization target, and coarse pose can be used to quickly guide the position. As the frame descends and the earhole gradually moves out of the field of view, the system simultaneously performs earhole recognition and corner and edge recognition within a preset transition zone, and achieves seamless and smooth switching between the two features through confidence-driven dynamic weighting. Global coarse pose can also be added for auxiliary weighted fusion. When the earhole is completely out of the camera's field of view, the system automatically switches to corner and edge features with stronger robustness and noise resistance for localization. This strategy solves the problem of single visual features being easily lost during descent in a mining environment, ensuring that visual localization remains effective throughout the entire descent process.

[0062] It should be noted that feature results with confidence levels below a preset threshold are not included in the pose fusion calculation, effectively excluding low-quality recognition results caused by environmental factors such as mine dust and low illumination. Within the transition range, the fusion weights of the two types of features are dynamically allocated based on the relative magnitudes of the ear hole confidence and corner / edge confidence, with features having higher confidence levels receiving higher decision weights. The preset threshold for ear hole confidence is 0.4, and the preset threshold for corner / edge confidence is 0.5. These thresholds can be adaptively adjusted according to the mine's ambient lighting conditions, dust concentration, and recognition accuracy requirements; their specific values ​​can be obtained by those skilled in the art through a limited number of experiments or calibrations.

[0063] Secondly, the present invention also provides a heavy-duty battery swapping robot for mines; please refer to [link to relevant documentation]. Figure 1 It includes a walking mechanism 200, a lifting mechanism 300, a battery swapping robot 100 as described in any of the above embodiments, and a whole machine controller 500.

[0064] Specifically, the walking mechanism 200 is used to translate horizontally along a preset path. The lifting mechanism 300 is mounted on the walking mechanism 200 and is used to output vertical lifting drive and rotation drive around a vertical axis. The top connecting part 11 of the frame 1 of the battery swapping robot 100 is fixedly connected to the lifting end of the lifting mechanism 300. The whole machine controller is connected to the walking mechanism 200, the lifting mechanism 300, and the battery swapping robot 100 via signals. Please refer to [link / reference]. Figure 4The overall controller 500 acts as the host computer, and the control unit 6 acts as the slave computer, exchanging data in real time via a fieldbus. The control unit 6 and the overall controller adopt a master-slave control architecture, with the control unit 6 acting as the slave computer and the overall controller as the host computer. The control unit 6 reports posture deviations and safety signals in real time, and the overall controller makes motion decisions and issues execution commands based on this information. Based on the posture deviation reported by the control unit of the battery swapping robot, the controller adjusts the posture of the walking mechanism and the lifting mechanism to complete alignment, and controls the lifting mechanism to lower the battery swapping robot; the lifting mechanism only performs the lifting action when it receives a lifting permission signal output by the control unit of the battery swapping robot.

[0065] In one embodiment, a rotary drive device is provided between the lifting end of the lifting mechanism 300 and the connection 11 at the top of the frame 1. When the control unit 6 calculates the yaw angle deviation, the overall controller controls the rotary drive device to drive the frame 1 to rotate around the vertical axis until the yaw angle deviation is eliminated. The rotary drive device uses a worm gear reducer as the transmission element, and utilizes its reverse self-locking characteristic to automatically lock after the yaw angle adjustment is completed, preventing passive yaw caused by the shift of the center of gravity during the lifting process.

[0066] Understandably, the lifting mechanism 300 can adopt a winch-driven sling structure. A platform is installed at its lifting end, and a slewing drive is mounted on this platform. The frame 1 is mounted via the slewing drive, thus achieving lifting and slewing. In other words, the lifting mechanism 300 includes a winch-driven sling structure and a slewing drive, both connected to the overall machine controller. The slewing drive can be a heavy-duty worm gear slewing drive, a WH series coal mine slewing drive, a slewing reducer, or other existing mature equipment that meets the load requirements. When the lifting mechanism 300 adopts a winch-driven sling structure, the yaw angle is adjusted through the slewing drive. When the slewing drive operates, the rope system twists accordingly. To avoid excessive swaying caused by rope twisting during yaw adjustment, the control unit 6 employs an angle closed-loop iterative fine-tuning strategy, with each adjustment step not exceeding 0.5° to ensure smooth and controllable slewing. If the yaw angle deviation is large, slewing adjustment can be completed before descent to avoid severe twisting in the suspended state. The lifting mechanism 300 can also use existing equipment with both lifting and slewing functions, such as a fixed-boom rotary crane. There is no single limitation here, as long as the functional requirements are met.

[0067] It should be noted that the overall controller adopts a mine-grade explosion-proof controller, and the control unit 6 adopts a mine-grade explosion-proof PLC controller. The two communicate via CANopen or PROFINET fieldbus, or via WiFi. The overall controller is connected to the servo drivers of the traveling mechanism 200, the winch driver of the hoisting mechanism 300, and the servo drivers of the slewing drive device via explosion-proof cables. The control unit 6 is connected to the four corner cameras 31, the global camera 32, the explosion-proof servo motor 231, the telescopic lock stop pin 42, the position sensor 44, and the four force sensors 5 via explosion-proof cables, meeting the safety requirements of GB3836 series standards for mine electrical equipment. Understandably, the overall controller 500 is responsible for task scheduling and work process management; issuing X / Y direction travel commands to the traveling mechanism 200; controlling the lifting, starting, stopping, and speed of the hoisting mechanism 300; issuing yaw angle adjustment commands to the slewing drive device; receiving safety signals reported by the control unit and executing motion interlocks. The control unit is responsible for image acquisition and preprocessing of the four corner cameras 31 and the global camera 32; visual positioning and pose calculation (all image processing is completed inside the control unit 6); real-time generation and reporting of pose deviation signals; rotation control of the explosion-proof servo motor 231; locking / unlocking control of the telescopic lock stop pin 42; signal acquisition, threshold judgment and off-center load calculation of the four force sensors 5; and output of locking position signal and lifting permission signal.

[0068] In one embodiment, the camera 31 also participates in positioning guidance during the battery pack installation phase, but its identification target differs from that in the grasping phase. In the grasping phase, the lifting lugs 401 and corner edges 402 of the battery pack 400 are used as identification targets to achieve precise insertion of the lifting claw 22 into the lifting lugs 401. However, in the installation phase, the camera 31's identification target switches to the installation positioning features on the vehicle body to be swapped. These installation positioning features include one or more combinations of the outline edge of the vehicle body mounting frame, positioning holes or pins provided on the vehicle body mounting frame, and visual targets or QR code markings provided on the vehicle body mounting frame.

[0069] Specifically, after the battery pack 400 is lifted and transported to the vehicle to be swapped, the control unit 6 controls the four corner cameras 31 to acquire images of the vehicle mounting area, identify the four contour edges of the vehicle mounting frame, and extract the rectangular contour features of the mounting frame through edge detection and line fitting algorithms. If the vehicle body also has positioning holes or pins, the control unit 6 uses the same Hough circle detection and contour matching algorithm as the one used for high-position ear hole recognition to identify their center positions. Using the pre-stored standard physical coordinates of the vehicle mounting position as a reference, the control unit 6 calculates the pose deviation of the battery pack 400 relative to the vehicle mounting position (including X-axis translation deviation, Y-axis translation deviation, and yaw angle deviation), and reports the deviation signals to the overall controller in real time.

[0070] The controller controls the walking mechanism 200 to adjust its horizontal position based on the positional deviation signal of the battery pack 400 relative to the vehicle body mounting position, guiding the battery pack 400 to the area approximately above the vehicle body mounting position, so that the mating structure at the bottom of the battery pack 400 enters the effective range of the mechanical positioning structure on the vehicle body mounting frame.

[0071] It should be noted that the visual positioning during the installation phase only needs to guide the battery pack within the effective range of the mechanical positioning structure, and does not require millimeter-level precision positioning as in the grasping phase. The final accurate positioning and placement are accomplished by the mechanical positioning structure on the vehicle body in conjunction with the corresponding structure on the bottom of the battery pack. Edge detection and line fitting algorithms, as well as Hough circle detection and contour matching algorithms, are existing technologies in the field of image processing and have been widely applied in various visual inspection and positioning scenarios. Those skilled in the art can implement the corresponding functions by following the guidance in this manual and combining it with well-known algorithm principles.

[0072] Thirdly, the present invention also provides a control method for the heavy-duty battery swapping robot in the mine described in the above embodiments, and the steps are described below.

[0073] Step S1: Collect target positioning feature images through multiple cameras 31 at each corner of the frame 1, identify the positioning reference points of the lifting lugs 401 at each corner of the battery pack 400 and obtain pixel coordinates. Based on the pre-stored standard physical coordinates, and combined with the internal and external parameters of the cameras 31, calculate the X-direction translation deviation, Y-direction translation deviation and yaw angle deviation of the frame relative to the battery pack in the horizontal plane.

[0074] Specifically, when frame 1 is in a high position (≥500mm above the top of the battery pack), the four corner cameras 31 acquire images of the four lifting lugs respectively. The control unit 6 calculates the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation around the vertical axis of frame 1 relative to battery pack 400 in the horizontal plane using a four-point lug registration algorithm. The overall controller 500 controls the walking mechanism 200 to adjust its horizontal posture according to the deviation signals, so that the four lifting claws 22 are aligned with the four lifting lugs 401 respectively.

[0075] Step S2: Based on the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation, drive the traveling mechanism to adjust the position of the lifting mechanism and the frame in the horizontal plane, and drive the lifting mechanism to adjust the yaw angle of the frame around the vertical axis so that each lifting unit is vertically aligned with the corresponding lifting lug.

[0076] Step S3: Drive the lifting mechanism to lower the frame. During the descent, continuously collect target positioning feature images and calculate and obtain the pose deviation in real time. Adjust the frame pose in real time according to the pose deviation so that each lifting unit is vertically aligned with the corresponding lifting lug during the descent.

[0077] Step S31: During the descent, when the hoisting lug is within the camera's field of view, the hoisting lug of the battery pack is used as the target positioning feature, and the pose deviation is calculated by four-point coordinate registration and perspective projection conversion.

[0078] Step S32: After descending to the preset transition interval height, simultaneously perform hoisting ear hole recognition and corner edge recognition. Based on the confidence levels of the hoisting ear hole features and the corner edge features, output continuous pose deviations through weighted fusion.

[0079] Step S33: After the mounting ear hole moves out of the camera's field of view, the vertical angle edge of the battery pack is used as the target positioning feature. The real-time pose deviation is calculated by fitting the straight line of the angle edge, and the frame pose is continuously fine-tuned.

[0080] In this embodiment, a three-stage visual localization strategy was designed for the grasping stage, consisting of high-position ear hole registration, transition interval weighted fusion, and low-position angle edge fitting. This strategy addresses the problem of single visual features being easily lost during descent in a mining environment. Simultaneously, feature results with confidence levels below a preset threshold are not included in pose fusion calculations, effectively eliminating low-quality recognition results caused by environmental factors such as mine dust and low illumination.

[0081] Step S4: After each lifting unit is inserted into the corresponding lifting lug, each lifting unit is driven to perform hooking and grabbing within the lifting lug. After hooking and grabbing, each lifting unit is locked by the locking structure, and the locking position signal is detected and output.

[0082] Specifically, after frame 1 is lowered into position, all lifting claws 22 are inserted into the corresponding lifting lugs 401, in the insertion / removal position. Control unit 6 controls four explosion-proof servo motors 231 to start synchronously, driving the main shaft 21 to rotate 90° via worm gear reducer 232, causing the lifting claws 22 to rotate to the lifting position and engage with the inside of the lifting lugs 401; the reverse self-locking characteristic of worm gear reducer 232 constitutes the first locking. Subsequently, control unit 6 controls four telescopic lock pins 42 to extend synchronously and insert into the limiting grooves 43 of wheel 41, constituting the second locking; after position sensor 44 detects that the pins are in place, it outputs a locking signal to control unit 6.

[0083] Step S5: The lifting mechanism performs a pre-lifting action, supports the frame and attaches the grabbed battery, and checks the axial load of each lifting unit.

[0084] Specifically, the lifting mechanism 300 tightens slightly upwards, causing the lifting claw 22 to begin bearing part of the weight of the battery pack 400, and the four force sensors 5 detect the axial load value of each lifting unit 2 in real time.

[0085] Step S6: If all locking signals are received and the axial load values ​​of all lifting units meet the preset conditions, output a lifting permission signal and execute the lifting action through the lifting mechanism.

[0086] Specifically, the control unit 6 performs a safety check, verifying whether all four position sensors 44 have received valid locking signals; and simultaneously verifying whether the detection value of each force sensor 5 is within a preset threshold range. The lower limit confirms that the lifting claw has been effectively stressed, the upper limit prevents overload of a single claw, and the force difference between the four sensors does not exceed a preset off-center load threshold. Only when both checks pass, the control unit 6 outputs a lifting permission signal to the machine controller 500; upon receiving the signal, the machine controller 500 controls the lifting mechanism 300 to lift the battery pack 400 at a set speed, completing the grabbing and lifting action.

[0087] Furthermore, the process includes the following steps: after lifting and transporting the battery pack above the vehicle to be swapped, images of the vehicle mounting area are captured by cameras at the four corners, the outline edges and / or positioning marks of the vehicle mounting frame are identified, and the coarse pose deviation of the battery pack relative to the vehicle mounting position is calculated; the horizontal pose of the walking mechanism is adjusted according to the coarse pose deviation, so that the bottom mating structure of the battery pack enters the effective range of the mechanical positioning structure on the vehicle mounting frame, and the lifting mechanism is controlled to descend, and the battery pack is guided to complete its final placement through the mechanical positioning structure.

[0088] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A battery swapping robot for grasping a battery pack having a plurality of lifting lugs, characterized in that, include: The frame has a connecting part at its top for connecting the lifting mechanism; Multiple lifting units are respectively located at the corresponding corners of the frame and correspond one-to-one with the lifting lugs of the battery pack, for attaching, gripping or detaching from the corresponding lifting lugs. A visual positioning unit includes multiple cameras respectively located at corresponding corners of the frame and facing downwards, used to acquire target positioning feature images and output image data; A locking assembly, which is installed on the frame in a one-to-one correspondence with the lifting unit, is used to lock the lifting unit when it is attached to the corresponding lifting lug and output a locking signal. Multiple force sensors are installed one-to-one on the axial force path of the lifting unit to detect axial force and output detection data; The control unit is signal-connected to the lifting unit, visual positioning unit, locking components, and force sensors, and is configured to calculate the pose deviation between the frame and the target object based on image data and output a pose adjustment signal in real time; and outputs a lifting permission signal when it receives locking signals from all the locking components and when the detection data from all the force sensors meet the preset threshold conditions.

2. The battery swapping robot according to claim 1, characterized in that, The lifting unit includes a main shaft, lifting claws, and a drive mechanism. The main shaft is vertically rotatably connected to the frame. The lifting claws are fixed to the lower end of the main shaft. The drive mechanism is connected to the main shaft and drives the main shaft to rotate, so that the lifting claws have an insertion / removal position for inserting and removing from the corresponding lifting lugs, and a lifting position for being hooked onto the inside of the lifting lugs after rotation.

3. The battery swapping robot according to claim 2, characterized in that, The drive mechanism includes an explosion-proof servo motor and a worm gear reducer. The explosion-proof servo motor is fixedly mounted on the frame, and its output shaft is connected to the input end of the worm gear reducer. The output end of the worm gear reducer is connected to the main shaft.

4. The battery swapping robot according to claim 2, characterized in that, The locking assembly includes a wheel and a telescopic locking pin. The wheel is fixed to the main shaft and has a limit groove. The telescopic locking pin has a retractable pin end. When the main shaft rotates to the lifting position, the pin end extends and inserts into the limit groove, preventing the main shaft from rotating. When the main shaft rotates to the insertion / removal position, the pin end retracts and disengages from the limit groove. A position sensor is provided in the limit groove to detect whether the pin end is inserted into the limit groove and outputs a locking signal.

5. The battery swapping robot according to claim 1, characterized in that, The frame is rectangular, and there are four cameras, which are respectively mounted on the outer edges of the four corners of the frame. The control unit is configured to receive four channels of image data output by the four cameras when the frame is in a high position, and calculate the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation of the frame relative to the target object in the horizontal plane. Furthermore, the control unit is configured to stop recognizing the hoisting ear hole and start recognizing the edge features of the target object when the hoisting ear hole or positioning feature moves out of the field of view of any of the cameras during the descent of the frame, and continuously output the real-time position deviation of the frame relative to the target object based on the recognized features.

6. The battery swapping robot according to claim 5, characterized in that, The control unit calculates the pose deviation of the frame relative to the target object, including: When the frame is in a high position, the positioning reference points of the hoisting lugs in the images captured by the four cameras are identified respectively, and the pixel coordinates of the four hoisting lugs at the four corners of the battery pack are obtained. Based on the pre-stored standard physical coordinates of the hoisting lugs at the four corners of the battery pack, and combined with the internal and external parameters of the camera, the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation around the vertical axis of the frame relative to the battery pack in the horizontal plane are obtained through four-point coordinate registration and perspective projection conversion. After the frame is lowered until the hoisting lugs are out of the camera's field of view, the vertical corner edges of the battery pack in the images captured by the four cameras are identified respectively, and the straight line features of the four corner edges are fitted. Based on the standard relative positions of the four corner edges, combined with the camera's internal and external parameters, and according to the image normal offset of each edge, the X-axis translational deviation and Y-axis translational deviation of the frame relative to the battery pack in the horizontal plane are calculated.

7. The battery swapping robot according to claim 5, characterized in that, The visual positioning unit also includes a global camera, which is installed at the bottom center of the frame to acquire positioning feature images of the battery pack and output image data.

8. The battery swapping robot according to claim 7, characterized in that, The control unit is also configured to calculate the coarse pose deviation between the frame and the target based on the image data output by the global camera, in order to assist in guiding the four corner cameras to lock the hoisting ear hole area.

9. A heavy-duty battery swapping robot for mines, characterized in that, include: The traveling mechanism is used to translate along a preset path in the horizontal plane. A lifting mechanism, mounted on the traveling mechanism, is used to output vertical lifting drive and rotation drive around a vertical axis; and The battery swapping robot as described in any one of claims 1-8, wherein the top of the frame of the battery swapping robot is connected to the lifting end of the lifting mechanism; The overall controller is connected to the walking mechanism, the lifting mechanism, and the battery swapping robot via signals. Based on the positional deviation reported by the control unit of the battery swapping robot, it controls the walking mechanism and the lifting mechanism to adjust their positions to complete alignment, and controls the lifting mechanism to drive the battery swapping robot to descend. It only controls the lifting mechanism to perform the lifting action when it receives the lifting permission signal output by the control unit of the battery swapping robot.

10. A control method, characterized in that, The heavy-duty battery swapping robot for mining as described in claim 9 includes the following steps: S1. Collect target positioning feature images through multiple cameras at each corner of the frame, identify the positioning reference points of the lifting lugs at each corner of the battery pack and obtain pixel coordinates. Based on the pre-stored standard physical coordinates, and combined with the internal and external parameters of the cameras, calculate the X-direction translation deviation, Y-direction translation deviation and yaw angle deviation of the frame relative to the battery pack in the horizontal plane. S2. Based on the X-axis translational deviation, Y-axis translational deviation, and yaw angle deviation, drive the traveling mechanism to adjust the position of the lifting mechanism and frame in the horizontal plane, and drive the lifting mechanism to adjust the yaw angle of the frame around the vertical axis so that each lifting unit is vertically aligned with the corresponding lifting lug. S3. Drive the lifting mechanism to lower the frame. During the descent, continuously collect target positioning feature images and calculate and obtain the pose deviation in real time. Adjust the frame pose in real time according to the pose deviation so that each lifting unit is vertically aligned with the corresponding lifting lug during the descent. S31. During the descent, when the hoisting lug is within the camera's field of view, the hoisting lug of the battery pack is used as the target positioning feature, and the pose deviation is calculated by four-point coordinate registration and perspective projection conversion. S32. After descending to the preset transition zone height, simultaneously perform hoisting ear hole recognition and corner edge recognition. Based on the confidence levels of the hoisting ear hole features and the corner edge features, output continuous pose deviations through weighted fusion. S33. After the mounting lugs move out of the camera's field of view, the vertical angle edge of the battery pack is used as the target positioning feature. The real-time pose deviation is calculated by fitting the straight line of the edge and the frame pose is continuously fine-tuned. S4. After each lifting unit is inserted into the corresponding lifting lug, each lifting unit is driven to perform hooking and grabbing in the lifting lug. After hooking and grabbing, each lifting unit is locked by the locking structure, and the locking position signal is detected and output. S5. The lifting mechanism performs a pre-lifting action, supports the frame and attaches the grabbed battery, and detects the axial load of each lifting unit; S6. If all locking signals are received and the axial load values ​​of all lifting units meet the preset conditions, output a lifting permission signal and execute the lifting action through the lifting mechanism.

Images