Visual positioning method and system for multiple types of battery packs of battery-swapping robots

Abstract

The present invention discloses a multi-type battery pack visual positioning method and system device for a battery swap robot, comprising the following steps: 1. visually measuring the battery pack to be replaced on the chassis of a vehicle; 2. constructing a three-dimensional point cloud view of the battery pack, performing voxel filtering and Euclidean clustering on the point cloud, calculating the initial position of the vehicle chassis, and using the RANSAC algorithm to fit the corresponding morphology of the battery pack on the vehicle chassis; 3. using HSV color threshold segmentation and Hough circle transform to identify and roughly locate the point cloud of the locking and unlocking holes on the battery pack; 4. performing least squares fitting on the segmented locking and unlocking hole point cloud to obtain the precise center position and normal direction of the locking and unlocking holes; 5. transforming the topographic plane of the vehicle chassis battery pack and the precise position of the locking and unlocking holes into the coordinate system of the battery swap station to guide the movement of the battery swap robot. 6. constructing a system device composed of a visual sensor, a visual information positioning processor, and a communication module. This positioning method and system device are suitable for the operation of replacing battery packs of vehicles in battery swap stations, and can accurately locate the battery packs and guide the battery swap robot in the battery swap operation.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent battery replacement for electric vehicles, and in particular relates to a visual positioning method and system device for multiple types of battery packs used in battery replacement robots. Background Art

[0002] Electric vehicles, with their energy-saving, environmentally friendly, and high energy conversion efficiency, are being actively promoted by the government. Currently, the most common way to supplement electric vehicle energy is to charge the battery, either through fixed charging stations or mobile charging vehicles. However, this method typically takes a long time and cannot meet users' needs for rapid recharging. Battery swapping platforms, as an alternative method of recharging, allow users to quickly recharge by directly replacing the vehicle's battery pack, thus compensating for the slow charging time and providing a better user experience.

[0003] During the battery swapping process in electric vehicles, precise positioning of the battery pack is crucial for efficient and rapid battery swapping. Existing battery swapping platforms require guiding the electric vehicle to park within a designated area and utilizing mechanical fine-tuning devices such as auxiliary supports, leveling mechanisms, and guide rails to fine-tune the vehicle body, bringing it into a parallel and horizontal position for the battery swapping operation. After this rough alignment process, the battery swapping robot rises and connects the locking and unlocking holes or positioning marks of the vehicle body and the battery pack via a floating platform, achieving the final, precise alignment.

[0004] Currently, battery-swapping robots have insufficient compatibility with multiple types and models of vehicle and battery pack identification and positioning. Since battery-swapping vehicles vary in size and model, the positions of the battery packs and the battery pack lock holes on different vehicle chassis may also be different. This requires the battery-swapping system to dynamically identify the topography of the vehicle chassis and obtain accurate position and posture information. At the same time, multi-brand and multi-model battery packs have different locking methods such as snap-on, bolt-on, and spin-on, which also poses a challenge to the compatibility of the battery-swapping robot positioning system. On the other hand, in order for the battery-swapping robot to accurately dock with the vehicle battery pack, the existing battery-swapping platform requires the vehicle body to be parked relatively precisely in a fixed position. The parking accuracy requirements for the vehicle body are too high, and additional devices are required for positioning and adjustment, resulting in increased battery-swapping costs and time.

[0005] In the field of battery replacement for multiple types of battery packs in electric vehicles, there is an urgent need for an efficient and accurate battery replacement platform positioning method. Summary of the Invention

[0006] In order to solve the above problems, the present invention proposes a multi-type battery pack visual positioning method and system device for battery swapping robots, which solves the problem that existing battery swapping stations cannot adapt to the positioning requirements of multi-model battery packs.

[0007] To achieve the above object, the technical solution adopted by the present invention is:

[0008] The visual positioning method for multiple types of battery packs used in battery swapping robots has the following specific steps:

[0009] (1) Obtain visual measurement information of the battery swapping scene by visual sensors;

[0010] (2) Back-projection is performed to obtain the three-dimensional point cloud of the battery pack, and then the point cloud is filtered using a voxel filter to obtain a point cloud set P1 containing the point cloud of the vehicle chassis battery pack and part of the environment point cloud;

[0011] For P1, Euclidean clustering is used, and the clustered point cloud is divided into point cloud subsets {Q1, Q2..Q n}; Select the largest cluster Q k For the chassis battery pack point cloud P2, use the PCA principal component analysis method to calculate the point cloud, obtain the normal direction n1 and its position t1 of the chassis plane l1, remove all point clouds in P2 that are greater than d from plane l1, and obtain the chassis point cloud P3; use the RANSAC method to perform a subset plane fitting of the P3 point cloud to obtain the fitting plane l2 of the chassis battery pack, whose corresponding normal vector is n2;

[0012] (3) Based on the color characteristics of the feature marker, HSV color space transformation is used to segment the red point cloud, and its center of mass is used as the location of the unlocking hole ξ1;

[0013] Use the Hough circle transform to detect the unlocking hole on the vehicle chassis battery pack, and set the center of the detected circle as the center of the hole ξ2. When the distance d = ||ξ1-ξ2|| between the two hole centers is less than the threshold ε, the unlocking hole is detected, and ξ2 is taken as the initial position q of the unlocking hole.

[0014] (4) Real-time segmentation of multiple point cloud subsets Q1, Q2, ...Q containing unlocked holes during robot operation n ; For each hole point cloud set Q k , using the least squares method to fit the hole normal direction η k and the center coordinate ρ k , complete the accurate pose estimation of the hole;

[0015] (5) According to the relative position R0, t0 of the visual sensor coordinate system relative to the absolute coordinate system of the battery swap station workspace, l2, n2, η k , ρ k Transform to the working space coordinate system of the battery swap station to provide guidance for the battery swap robot operation.

[0016] As a further improvement of the present invention, the method for setting the visual sensor in step (1) is as follows:

[0017] The visual sensor is set at a fixed position 0.5 meters from the bottom of the battery swap platform. It is set up in an eye-to-hand manner. The visual sensor needs to be calibrated and tested before use. The relative position of the visual sensor coordinate system relative to the battery swap station coordinate system is R0, t0. The visual sensor shoots the battery pack on the vehicle chassis above to obtain a point cloud image including the vehicle chassis battery pack and part of the battery swap station structure. The holes on the surface of the battery pack are feature marks used for positioning.

[0018] As a further improvement of the present invention, step (2) vehicle chassis battery pack plane fitting includes the following steps:

[0019] (2-1) Back-projecting the image pixels containing depth information into the three-dimensional coordinate system of the visual sensor to obtain a three-dimensional point cloud containing the battery pack, and filtering the point cloud using a voxel filter to obtain a point cloud set P1 containing the point cloud of the vehicle chassis battery pack and part of the environment point cloud;

[0020] (2-2) Perform Euclidean clustering of point clouds to obtain point cloud set P2. In order to avoid adhesion between the chassis point cloud and the battery swap station structure point cloud after clustering, the Euclidean clustering radius R should be the distance between the chassis and the battery swap station structure.

[0021] (2-3) Use the PCA principal component analysis method to obtain the initial estimated plane l1 of the vehicle chassis battery pack and its normal direction n1. Based on the position t1 of the plane l1 and the normal vector n1, remove all point clouds in P2 that are greater than d from the plane to obtain the point cloud set P3;

[0022] (2-4) Use the RANSAC method to perform plane fitting on P3 to obtain the fitting plane l2 of the vehicle chassis battery pack, whose corresponding normal vector is n2.

[0023] As a further improvement of the present invention, the identification of the locking and unlocking holes on the battery pack in step (3) includes the following steps:

[0024] (3-1) Using HSV color space transformation, the color of the chassis battery pack point cloud is converted from RGB space to HSV space, and the point cloud with H channel values within ±30 is selected, corresponding to the red feature mark, and its centroid is used as the position ξ1 of the unlocking hole;

[0025] (3-2) Use Hough circle transform to detect the unlocking holes of the chassis battery pack. Project the point cloud of the chassis battery pack onto the fitting plane l2 and perform Hough circle detection on the two-dimensional image. The center of the circle detected is ξ2;

[0026] (3-3) When the distance d = ||ξ1-ξ2|| between the centers of the two holes is less than the threshold ε, it is determined that the locking and unlocking hole is detected, and ξ2 is taken as the initial position q of the locking and unlocking hole.

[0027] As a further improvement of the present invention, step (4) of accurately estimating the pose of the unlocked hole includes the following steps:

[0028] (4-1) Point cloud set Q for each hole k , the area where the unlocking hole is located is regarded as a small plane, and the plane normal η is fitted using the least squares method k , the hole plane corresponding to the normal is I k ;

[0029] (4-2) Perform least squares circle fitting on the circular point cloud of the hole to calculate the precise center coordinates ρ of the unlocked hole k .

[0030] The present invention provides a system device for a visual positioning method of multiple types of battery packs using a battery swapping robot, which is characterized in that it includes a visual sensor, a visual information positioning processor, and a communication module. The visual sensor is connected to the communication module of the visual information positioning processor, and the main control computer of the battery swapping station and the battery swapping robot controller are connected to the communication module of the visual information positioning processor.

[0031] Beneficial effects:

[0032] The present invention discloses a visual positioning method and system device for multiple types of battery packs for battery-swapping robots. The method uses a visual sensor to photograph the battery pack to be replaced on the chassis of a car, obtains a three-dimensional point cloud view containing the positioning features of the battery pack, performs voxel filtering and Euclidean clustering on the point cloud, removes noise points and removes point clouds other than the chassis part in the point cloud, uses the PCA principal component analysis method to calculate the initial pose of the pre-processed chassis battery pack point cloud, further segments the point cloud of the non-chassis part, and fits the topographic plane corresponding to the chassis battery pack using the RANSAC plane fitting algorithm. Use HSV color threshold segmentation and Hough circle transform method to identify and roughly locate the unlocking hole point cloud on the chassis battery pack, perform least squares fitting on the segmented unlocking hole point cloud, and obtain the precise center position and normal direction of the unlocking hole. Construct a system device consisting of a visual sensor, a visual positioning information processor, and a communication module. The battery-swapping robot can use the topographical plane of the battery pack on the vehicle chassis, the precise position of the unlocking hole, and the normal direction to complete the positioning of the battery pack to be replaced, thereby performing the battery swap. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of the method disclosed in the present invention;

[0034] Figure 2 This is a schematic diagram of the camera setting of the battery swap station disclosed in the present invention. DETAILED DESCRIPTION

[0035] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0036] The present invention discloses a multi-type battery pack visual positioning method and system device for a battery-swapping robot, the flow chart of which is as follows: Figure 1 As shown, the following steps are included:

[0037] Step 1: How to obtain visual measurement information of the battery swapping scenario.

[0038] like Figure 2 The working space of the battery swap platform is directly below the vehicle parking area, and the battery swap robot is set on the working platform at the bottom of the space to perform battery swap operations on the battery pack on the chassis of the vehicle to be swapped. In the positioning scheme of the present invention, the visual sensor is set at a fixed position 0.5 meters from the bottom of the battery swap platform, and its setting method is eye-to-hand type. The visual sensor needs to be calibrated and tested before use. The relative pose of the visual sensor coordinate system relative to the battery swap coordinate system is R0, t0. The visual sensor shoots the battery pack of the car chassis above, and obtains a point cloud image including the vehicle chassis battery pack and part of the battery swap station structure, wherein the holes on the surface of the battery pack are feature marks that can be used for positioning. After subsequent computer processing, the positioning information of the captured image is calculated to obtain the absolute pose of the battery pack in the battery swap station space, which is used to guide the docking of the battery swap robot with the battery pack.

[0039] Step 2: Calculate the vehicle chassis battery pack fitting plane. This includes the following steps:

[0040] (2-1) Given the vision sensor position and model, back-project the image pixels containing depth information into the vision sensor's 3D coordinate system to obtain a 3D point cloud containing the battery pack. Use a voxel filter to filter the point cloud, removing noise and some outliers, and reducing the number of points to obtain point cloud set P1, which includes the vehicle chassis point cloud and some surrounding point clouds.

[0041] (2-2) Perform Euclidean clustering of the point cloud. Randomly select a certain sub-point s1 from the chassis point cloud P1, build a KD-tree to search the radius R of the seed point s1, generate cluster Q1, select a new seed point in Q1, and continue to perform the domain search until the number of clusters in Q1 no longer increases. Randomly select seed point s2 for the remaining point cloud in P1, and continue to perform the above steps to obtain Q2. Repeat this process and after multiple iterations, all point clouds are classified into {Q1, Q2..Q n In order to avoid the adhesion between the chassis point cloud and the battery swap station structure point cloud after clustering, R should be the distance between the chassis and the battery swap station structure.

[0042] (2-3) Select the largest cluster Q kThe chassis point cloud P2 is used. The PCA principal component analysis method is used to obtain the preliminary pose information of the chassis. First, the centroid and covariance matrix of the 3D point cloud set are calculated:

[0043]

[0044]

[0045] Perform SVD decomposition on the covariance matrix H to obtain its eigenvalues λ1, λ2, λ3 and eigenvectors u1, u2, u3;

[0046] The largest eigenvalue λ1 and the second largest eigenvalue λ2 and their eigenvectors u1 and u2 correspond to the direction with the most points in the point cloud, i.e., the XY direction of the chassis, and the plane of the chassis, which is labeled as l1. The smallest eigenvalue λ3 and its eigenvector u3 correspond to the direction with the least points in the point cloud, i.e., the Z-axis direction, which can be regarded as the normal direction n1 of the chassis plane l1. Construct the rotation matrix R1 and the translation vector t1 as the initial estimated pose of the chassis point cloud in the visual sensor coordinate system:

[0047]

[0048]

[0049] In order to remove some non-chassis battery pack point clouds again, according to the position t1 and normal vector n1 of plane l1, all point clouds in P2 with a distance greater than d from the plane are removed to obtain a more accurate chassis point cloud P3.

[0050] (2-4) Use the RANSAC method to perform plane fitting on P3. The RANSAC method includes multiple iterations. In the kth iteration, a point cloud subset M is randomly selected from P3. k , calculate M using the minimum variance estimate k The fitting plane parameters of the subset are then calculated in P3 except M k The deviation of all inliers from the model is then compared with the set threshold and the deviation is recorded. The error rate, number of inliers, total number of samples, and current number of iterations are used to set an iteration end judgment condition. After the iteration is completed, the point cloud estimated plane with the largest number of inliers is taken as the fitting plane l2 of the chassis battery pack, and its corresponding normal vector is n2

[0051] Step 3: Add unlocking hole identification. This includes the following steps:

[0052] (3-1) The locking and unlocking holes of the vehicle chassis battery pack are surrounded by pre-sprayed red feature marks. The locking and unlocking holes are located according to the color characteristics of the feature marks. The HSV color space transformation is used to convert the color of the vehicle chassis battery pack point cloud from RGB space to HSV space. In the HSV space, the H parameter represents the color information, that is, the position of the spectral color, ranging from -180 to 180, the purity S is a proportional value ranging from 0 to 1, indicating the contrast, and V represents the brightness of the color, ranging from 0 to 1. The point cloud with an H channel value within ±30 is selected, corresponding to the red feature mark, and its center of mass is used as the position ξ1 of the locking and unlocking hole.

[0053] (3-2) Since the locking and unlocking holes are circular, the Hough circle transform is used to detect the locking and unlocking holes on the vehicle chassis battery pack. The point cloud of the vehicle chassis is projected onto the fitting plane l2, and Hough circle detection is performed on the two-dimensional image. The Hough algorithm, based on the duality of points and lines, converts a circle in the XY two-dimensional coordinate system into a point in the abr three-dimensional coordinate system. The pixels of the entire image are converted to coordinates in the abr three-dimensional coordinate system. The point with a cumulative value greater than the set threshold is the center of the detected circle ξ2.

[0054] (3-3) Compare the results of the circular hole detection with the results of the color feature detection. When the distance d = ||ξ1-ξ2|| between the centers of the two holes is less than the threshold ε, it is determined that the locking and unlocking hole is detected, and ξ2 is taken as the initial position q of the locking and unlocking hole.

[0055] Step 4: Accurately fit the unlocked hole pose. This includes the following steps:

[0056] (4-1) Using the method in step 3, multiple point cloud subsets Q1, Q2, ...Q containing unlocked holes are segmented in real time during the robot operation. n .

[0057] Point cloud set Q for each hole k , the least square method is used to fit the normal direction of the hole and the coordinates of the circle center to complete the accurate pose estimation of the hole. The area where the unlocked hole is located is regarded as a small plane, and the equation of the plane is,

[0058] Ax+By+Cz+D=0

[0059] Among them, A, B, C, and D are the required plane parameters. Transform the equation into

[0060] z=a0x+a1y+a2

[0061] in

[0062] make

[0063] where p = [x n ,y n , z n ]∈ρ k

[0064] [a0, a1, a2] = (A T A) -1 A T b

[0065] Then η k =[a0, a1, 1] can be used as the precise direction of the hole normal, and the hole plane corresponding to the normal is I k .

[0066] (4-2) Perform least squares circle fitting on the circular point cloud of the hole. k All points in the projected hole are projected onto the calculated hole plane I k , and get its two-dimensional coordinates. The standard equation of a circle is

[0067] (XA) 2 +(YB) 2 =R 2

[0068] It can be expressed as

[0069] X 2 +Y 2 +aX+bY+c=0

[0070] in

[0071]

[0072] make

[0073] where p = [x n ,y n , z n ]∈ρ k . Solved as

[0074] [a, b, c] = (A T A) -1 A T b

[0075] The coordinates of the fitted circle center are calculated and back-projected into the visual sensor coordinate system to finally obtain the precise coordinates of the circle center of the unlocking hole ρ k .

[0076] Step 5: Coordinate information conversion.

[0077] Steps 3 and 4 obtain the precise position information η of the topographic plane l2, n2 of the chassis battery pack and the locking and unlocking holes on it in the visual sensor coordinate system. k , ρ k According to the relative position R0, t0 of the visual sensor coordinate system relative to the absolute coordinate system of the battery swap station workspace, the above information is transformed into the coordinate system of the battery swap station workspace to provide guidance for the battery swap robot operation.

[0078] Step 6: Battery swap positioning system installation.

[0079] A system device for positioning a battery-swapping robot includes a visual sensor, a visual positioning information sensor, and a communication module. The visual sensor is connected to the communication module of a visual positioning information processor, and the main control computer of the battery-swapping station and the battery-swapping robot controller are connected to the communication module of the visual positioning information processor. When a vehicle enters the station and docks, the battery-swapping station sends information to the visual positioning information processor, activating its visual processing function. The visual positioning information processor receives the visual information sent back by the visual sensor and processes it to obtain corresponding positioning information. It also receives status information from the battery-swapping robot. Combined with this status information, the visual positioning information processor sends the final positioning information to the battery-swapping robot controller.

[0080] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent variation based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A multi-type battery pack visual positioning method for battery swapping robots, characterized in that: The specific steps are as follows: (1) Obtain visual measurement information of the battery swapping scene by the visual sensor; (2) Back-projection is performed to obtain the three-dimensional point cloud of the battery pack, and then the point cloud is filtered using a voxel filter to obtain a point cloud set P1 containing the point cloud of the vehicle chassis battery pack and part of the environment point cloud; For P1, Euclidean clustering is used, and the clustered point cloud is divided into point cloud subsets {Q1, Q2..Q n }; Select the largest cluster Q k For the chassis battery pack point cloud P2, use the PCA principal component analysis method to calculate the point cloud, obtain the normal direction n1 and its position t1 of the chassis plane l1, remove all point clouds in P2 that are greater than d from plane l1, and obtain the chassis point cloud P3; use the RANSAC method to perform a subset plane fitting of the P3 point cloud to obtain the fitting plane l2 of the chassis battery pack, whose corresponding normal vector is n2; (3) Based on the color characteristics of the feature marker, HSV color space transformation is used to segment the red point cloud, and its centroid is used as the location of the unlocking hole ξ1; Use the Hough circle transform to detect the unlocking hole on the vehicle chassis battery pack, and set the center of the detected circle as the center of the hole ξ2. When the distance d = ||ξ1-ξ2|| between the two hole centers is less than the threshold ε, the unlocking hole is detected, and ξ2 is taken as the initial position q of the unlocking hole. (4) Real-time segmentation of multiple point cloud subsets Q1, Q2, ...Q containing unlocked holes during robot operation n ; For each hole point cloud set Q k , using the least squares method to fit the hole normal direction η k and the center coordinate ρ k , complete the accurate pose estimation of the hole; (5) According to the relative position R0, t0 of the visual sensor coordinate system relative to the absolute coordinate system of the battery swap station workspace, l2, n2, η k , ρ k Transform to the working space coordinate system of the battery swap station to provide guidance for the battery swap robot operation.

2. The multi-type battery pack visual positioning method for a battery swapping robot according to claim 1 is characterized in that: The method for setting the visual sensor in step (1) is as follows: The visual sensor is set at a fixed position 0.5 meters from the bottom of the battery swap platform; the visual sensor needs to undergo calibration testing, and the relative position of the visual sensor coordinate system relative to the battery swap station coordinate system is R0, t0. The visual sensor photographs the vehicle chassis battery pack above and obtains a point cloud image including the vehicle chassis battery pack and part of the battery swap station structure, where the holes on the battery pack surface are feature marks used for positioning.

3. The multi-type battery pack visual positioning method for a battery-swapping robot according to claim 1 is characterized in that: Step (2) The vehicle chassis battery pack plane fitting includes the following steps: (2-1) Back-projecting the image pixels containing depth information into the three-dimensional coordinate system of the visual sensor to obtain a three-dimensional point cloud containing the battery pack, and filtering the point cloud using a voxel filter to obtain a point cloud set P1 containing the point cloud of the vehicle chassis battery pack and part of the environment point cloud; (2-2) Perform Euclidean clustering of point clouds to obtain point cloud set P2. In order to avoid adhesion between the chassis point cloud and the battery swap station structure point cloud after clustering, the Euclidean clustering radius R should be the distance between the chassis and the battery swap station structure. (2-3) Use the PCA principal component analysis method to obtain the initial estimated plane l1 of the vehicle chassis battery pack and its normal direction n1. Based on the position t1 of the plane l1 and the normal vector n1, remove all point clouds in P2 that are greater than d from the plane to obtain the point cloud set P3; (2-4) Use the RANSAC method to perform plane fitting on P3 to obtain the fitting plane l2 of the vehicle chassis battery pack, whose corresponding normal vector is n2.

4. The multi-type battery pack visual positioning method for a battery swapping robot according to claim 1 is characterized in that: Identifying the unlocking hole on the battery pack in step (3) includes the following steps: (3-1) Using HSV color space transformation, the color of the chassis battery pack point cloud is converted from RGB space to HSV space, and the point cloud with H channel values within ±30 is selected, corresponding to the red feature mark, and its centroid is used as the position ξ1 of the unlocking hole; (3-2) Use Hough circle transform to detect the unlocking holes of the chassis battery pack. Project the point cloud of the chassis battery pack to the fitting plane l2 and perform Hough circle detection on the two-dimensional image. The detected center of the circle is ξ2. (3-3) When the distance d = ||ξ1-ξ2|| between the centers of the two holes is less than the threshold ε, it is determined that the locking and unlocking hole is detected, and ξ2 is taken as the initial position q of the locking and unlocking hole.

5. The multi-type battery pack visual positioning method for a battery-swapping robot according to claim 1 is characterized in that: Step (4) Accurate pose estimation of the unlocked hole includes the following steps: (4-1) Point cloud set Q for each hole k , the area where the unlocking hole is located is regarded as a small plane, and the plane normal η is fitted using the least squares method k , the hole plane corresponding to the normal is I k ; (4-2) Perform least squares circle fitting on the circular point cloud of the hole to calculate the precise center coordinates ρ of the unlocked hole k .

6. A system device using the multi-type battery pack visual positioning method of a battery-swapping robot according to any one of claims 1 to 5, characterized in that: It includes a visual sensor, a visual information positioning processor, and a communication module. The visual sensor is connected to the communication module of the visual information positioning processor, and the main control computer of the battery swap station and the battery swap robot controller are connected to the communication module of the visual information positioning processor.

Images