Perceptive positioning method for supporting bracket of underground bracket transfer robot

Abstract

The invention discloses a sensing and positioning method for a supporting bracket of an underground bracket transfer robot, and belongs to the technical field of sensing and positioning, and the method comprises the following steps: improving the accuracy and robustness through the cooperative use of a laser radar and a camera, carrying out the positioning through a QR code, and carrying out the detection; according to the method, the multi-modal information from the radar and the visual sensor is fused, so that the detection precision and robustness of the supporting bracket are effectively improved. According to the method, an intelligent recognition and classification algorithm is adopted, and the detected supports can be accurately classified and recognized. According to the method, the recognition accuracy and stability of the system on the support are effectively improved, the false detection rate and the omission ratio are reduced, machine learning and an intelligent optimization algorithm are adopted, the positioning precision and efficiency are further improved, intelligent optimization and adjustment can be carried out according to the actual operation condition, and faster and more accurate support insertion and taking positioning is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of perception and positioning technology, and in particular relates to a perception and positioning method for a support support of an underground support handling robot. Background Art

[0002] The underground support handling robot is an intelligent handling equipment specially designed for underground mining environment. It has a small body size and is flexible and maneuverable. It is suitable for fast and efficient handling operations in narrow tunnels. The main engine adopts a high-power electronic injection engine, and the working mechanism has a variety of optional accessories such as shovels, buckets, shovel plates, breaker hammers, belt conveyors, etc., which can be quickly replaced. When the underground support handling robot is running, it is necessary to sense and locate its support bracket.

[0003] Although current support perception and positioning methods can also achieve positioning functions, interference and occlusion in heterogeneous application environments will affect the performance and data quality of the sensor; there are also problems with time synchronization and spatial calibration between different sensors; and there are problems with real-time processing and low computing efficiency in large-scale scenarios. Therefore, there is an urgent need for a support perception and positioning method for an underground support handling robot. Summary of the invention

[0004] The purpose of the present invention is to propose a support bracket perception and positioning method for an underground support handling robot in order to solve the problems that although the current support perception and positioning method can also achieve the positioning function, the interference and occlusion in the heterogeneous application environment will affect the performance and data quality of the sensor; there are also time synchronization and spatial calibration problems between different sensors; and there are problems of low real-time processing and computing efficiency in large-scale scenarios.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: a support support sensing and positioning method of an underground support handling robot, comprising the following steps:

[0006] S1, LiDAR is used in conjunction with cameras to improve accuracy and robustness and to use QR code positioning;

[0007] S2, using adaptive filters to compensate and eliminate the path;

[0008] S3, collecting sensor data;

[0009] S4, perform data processing and fusion;

[0010] S5. Positioning algorithm implementation.

[0011] As a further description of the above technical solution:

[0012] In S1, the laser radar is used in conjunction with the camera to improve accuracy and robustness, and QR code positioning is used, and Kalman filtering is used for data fusion. The camera selects one or more of a thermal imaging camera and a low-light camera.

[0013] As a further description of the above technical solution:

[0014] In S1, Kalman filtering is used to perform data fusion, and the specific steps are as follows:

[0015] A1. First determine the state variables to be fused and estimated, such as the position and speed of the target. The state vector can be defined as Where (x, y) is the location of the target, is the velocity of the target in the x and y directions;

[0016] A2. Establish a state transfer model to describe the evolution of the state between different time steps. For a linear system, the state transfer equation can be expressed as: k =F k x k-1 +w k-1 , where F k is the state transfer matrix, w k-1 is the process noise;

[0017] A3. Establish observation equations for the laser radar and camera respectively. Assume that the observation vector of the laser radar is z L，k =[x L，k ,y L，k ], the camera’s observation vector is z C，k =[x C，k ,y C，k ], then the observation equation can be expressed as: For laser radar: z L，k =H L，k x k +v L，k , for the camera: z C，k =H C，k x k +υ C，k , where H L，k and H C，k is the observation matrix, V L，k and C，k is the observation noise.

[0018] As a further description of the above technical solution:

[0019] In S1, the specific steps of the Kalman filter loop process are as follows:

[0020] B1. Given an initial state estimate and the initial covariance matrix P0;

[0021] B2. Prior state estimation:

[0022] B3. Prior covariance estimation: Where Q k-1 is the process noise covariance matrix;

[0023] B4. Perform Kalman gain calculation for LiDAR: For webcam: Among them, R L，k and R C，k are the observation noise covariance matrices of the lidar and camera, respectively;

[0024] B5. Update the status of LiDAR: For webcam: Covariance update: The update of camera data is similar;

[0025] B6. Use programming language (Python) to write Kalman filter and process and update the data according to the above process.

[0026] As a further description of the above technical solution:

[0027] In S2, the path is compensated and eliminated by using an adaptive filter, and the adaptive filter is used to process the sensor data to reduce noise, interference or error in the path, thereby improving the accuracy and stability of the path measurement.

[0028] As a further description of the above technical solution:

[0029] In S3, the sensor data is collected and a program is written to obtain the point cloud data of the laser radar and the camera data.

[0030] As a further description of the above technical solution:

[0031] In S4, data processing and fusion are performed, algorithms are developed to process the data, and lidar data is used to construct maps and positioning camera data is used to perform data detection and tracking.

[0032] As a further description of the above technical solution:

[0033] In S5, the positioning algorithm is implemented, a map is constructed using laser radar data, SLAM is used to realize robot positioning, and a camera is used to locate the bracket. The specific steps are as follows:

[0034] C1. Collect data about the surrounding environment using sensors equipped by the robot, including lidar, camera, and IMU;

[0035] C2. Perform pre-processing operations such as filtering, denoising, and time synchronization on the data collected by the sensor to improve data quality;

[0036] C3, extract representative features from the preprocessed data (geometric features in the lidar point cloud, corners and edges in the image);

[0037] C4. Perform feature matching on data at adjacent moments or in different frames to establish associations between the data;

[0038] C5. Estimate the robot's motion using motion models and optimization algorithms based on feature matching results. Common motion models include uniform motion models and acceleration motion models.

[0039] C6. Assume that the movement of the robot between two times t and t+1 can be represented by a transformation matrix T t，t+1 To describe, it can be estimated by the following formula, for the method based on feature points, the least squares method can be used to solve it: Among them, p t，i and p t+1，i are the coordinates of the matched feature points at time t and t+1.

[0040] As a further description of the above technical solution:

[0041] In S5, during map construction, as the robot moves, the collected data is integrated to construct a map, which is a feature-based map, which is one or more of a point cloud map and a grid map. The map construction method is one or more of an iterative closest point (ICP) algorithm and an occupancy grid mapping (occupancy grid map construction).

[0042] As a further description of the above technical solution:

[0043] In S5, after the map is constructed, closed-loop detection and optimization are performed to detect whether the robot returns to an area that has been visited before. Once a closed loop is detected, the entire posture and map are optimized using one of the graph optimization (g2o, Ceres and other libraries) or bundle adjustment methods to eliminate cumulative errors. The constructed map and estimated robot motion are used to determine the robot's posture in the map in real time.

[0044] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0045] In the present invention, the method integrates multimodal information from radar and visual sensors to effectively improve the accuracy and robustness of support bracket detection. To ensure that the system can accurately detect brackets in complex environments, the method uses intelligent recognition and classification algorithms to accurately classify and identify detected brackets. Effectively improve the system's recognition accuracy and stability of brackets, reduce false detection rates and missed detection rates, the method uses machine learning and intelligent optimization algorithms to further improve positioning accuracy and efficiency, and can perform intelligent optimization and adjustment according to actual operating conditions to achieve faster and more accurate bracket insertion and positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flow chart of a support support sensing and positioning method for an underground support handling robot. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example 1

[0049] See also Figure 1 The present invention provides a technical solution: a support support sensing and positioning method of an underground support handling robot, comprising the following steps:

[0050] S1. LiDAR is used in conjunction with a camera to improve accuracy and robustness, and QR code positioning is used. Kalman filtering is used for data fusion. A thermal imaging camera is selected as the camera, and Kalman filtering is used for data fusion. The specific steps are as follows:

[0051] A1. First determine the state variables to be fused and estimated, such as the position and speed of the target. The state vector can be defined as Where (x, y) is the location of the target, is the velocity of the target in the x and y directions;

[0052] A2. Establish a state transfer model to describe the evolution of the state between different time steps. For a linear system, the state transfer equation can be expressed as: k =F k x k-1 +w k-1 , where F k is the state transfer matrix, w k-1 is the process noise;

[0053] A3. Establish observation equations for the laser radar and camera respectively. Assume that the observation vector of the laser radar is z L，k =[x L，k ,y L，k ], the camera’s observation vector is z C，k =[x C，k ,y C，k ], then the observation equation can be expressed as: For laser radar: z L，k =H L，k x k +v L，k , for the camera: z C，k =H C，k x k +v C，k , where H L，k and H C，k is the observation matrix, v L，k and v C，k is the observation noise;

[0054] Among them, the specific steps of the Kalman filter cycle process are as follows:

[0055] B1. Given an initial state estimate and the initial covariance matrix P0;

[0056] B2. Prior state estimation:

[0057] B3. Prior covariance estimation: Where Q k-1 is the process noise covariance matrix;

[0058] B4. Perform Kalman gain calculation for LiDAR: For webcam: Among them, R L，k and R C，k are the observation noise covariance matrices of the lidar and camera, respectively;

[0059] B5. Update the status of LiDAR: For webcam: Covariance update: The update of camera data is similar;

[0060] B6. Use programming language (Python) to write Kalman filter and process and update the data according to the above process.

[0061] S2. Compensating and eliminating the path using an adaptive filter, where the adaptive filter is used to process sensor data to reduce noise, interference or errors in the path, thereby improving the accuracy and stability of the path measurement;

[0062] S3, collect sensor data and write programs to obtain laser radar point cloud data and camera data;

[0063] S4, perform data processing and fusion, develop algorithms to process data, use lidar data for map construction and positioning camera data for data detection and tracking;

[0064] S5, positioning algorithm implementation, using laser radar data to build a map, using SLAM to achieve robot positioning, and using the camera to position the bracket. The specific steps are:

[0065] C1. Collect data about the surrounding environment using sensors equipped by the robot, including lidar, camera, and IMU;

[0066] C2. Perform pre-processing operations such as filtering, denoising, and time synchronization on the data collected by the sensor to improve data quality;

[0067] C3, extract representative features from the preprocessed data (geometric features in the lidar point cloud, corners and edges in the image);

[0068] C4. Perform feature matching on data at adjacent moments or in different frames to establish associations between the data;

[0069] C5. Estimate the robot's motion using motion models and optimization algorithms based on feature matching results. Common motion models include uniform motion models and acceleration motion models.

[0070] C6. Assume that the movement of the robot between two times t and t+1 can be represented by a transformation matrix T t，t+1 To describe, it can be estimated by the following formula, for the method based on feature points, the least squares method can be used to solve it: Among them, p t，i and p t+1，i are the coordinates of the matched feature points at time t and t+1;

[0071] When building the map, as the robot moves, the collected data is integrated to build a map. It is a feature-based map, a point cloud map. The map construction method is the iterative closest point (ICP) algorithm. After the map is built, closed-loop detection and optimization are performed to detect whether the robot has returned to an area it has visited before. Once a closed loop is detected, graph optimization (libraries such as g2o and Ceres) is used to optimize the entire posture and map to eliminate cumulative errors. The constructed map and estimated robot motion are used to determine the robot's posture in the map in real time.

[0072] In this embodiment, the method integrates multimodal information from radar and visual sensors to effectively improve the accuracy and robustness of support bracket detection, thereby ensuring that the system can accurately detect brackets in complex environments.

[0073] Example 2

[0074] See also Figure 1 The present invention provides a technical solution: a support support sensing and positioning method of an underground support handling robot, comprising the following steps:

[0075] S1. LiDAR is used in conjunction with a camera to improve accuracy and robustness, and QR code positioning is used. Kalman filtering is used for data fusion. The camera selects a low-light camera, and Kalman filtering is used for data fusion. The specific steps are as follows:

[0076] A1. First determine the state variables to be fused and estimated, such as the position and speed of the target. The state vector can be defined as Where (x, y) is the location of the target, is the velocity of the target in the x and y directions;

[0077] A2. Establish a state transfer model to describe the evolution of the state between different time steps. For a linear system, the state transfer equation can be expressed as: k =F k x k-1 +w k-1 , where F k is the state transfer matrix, w k-1 is the process noise;

[0078] A3. Establish observation equations for the laser radar and camera respectively. Assume that the observation vector of the laser radar is z L，k =[x L，k ,y L，k ], the camera’s observation vector is z C，k =[x C，k ,y C，k ], then the observation equation can be expressed as: For laser radar: z L，k =H L，k x k +v L，k , for the camera: z C，k =H C，k x k +v C，k , where H L，k and H C，k is the observation matrix, v L，k and v C，k is the observation noise;

[0079] Among them, the specific steps of the Kalman filter cycle process are as follows:

[0080] B1. Given an initial state estimate and the initial covariance matrix P0;

[0081] B2. Prior state estimation:

[0082] B3. Prior covariance estimation: Where Q k-1 is the process noise covariance matrix;

[0083] B4. Perform Kalman gain calculation for LiDAR: For webcam: Among them, R L，k and R C，k are the observation noise covariance matrices of the lidar and camera, respectively;

[0084] B5. Update the status of LiDAR: For webcam: Covariance update: The update of camera data is similar;

[0085] B6. Use programming language (Python) to write Kalman filter and process and update the data according to the above process.

[0086] S2. Compensating and eliminating the path using an adaptive filter, where the adaptive filter is used to process sensor data to reduce noise, interference or errors in the path, thereby improving the accuracy and stability of the path measurement;

[0087] S3, collect sensor data and write programs to obtain laser radar point cloud data and camera data;

[0088] S4, perform data processing and fusion, develop algorithms to process data, use lidar data for map construction and positioning camera data for data detection and tracking;

[0089] S5, positioning algorithm implementation, using laser radar data to build a map, using SLAM to achieve robot positioning, and using the camera to position the bracket. The specific steps are:

[0090] C1. Collect data about the surrounding environment using sensors equipped by the robot, including lidar, camera, and IMU;

[0091] C2. Perform pre-processing operations such as filtering, denoising, and time synchronization on the data collected by the sensor to improve data quality;

[0092] C3, extract representative features from the preprocessed data (geometric features in the lidar point cloud, corners and edges in the image);

[0093] C4. Perform feature matching on data at adjacent moments or in different frames to establish associations between the data;

[0094] C5. Estimate the robot's motion using motion models and optimization algorithms based on feature matching results. Common motion models include uniform motion models and acceleration motion models.

[0095] C6. Assume that the movement of the robot between two times t and t+1 can be represented by a transformation matrix T t，t+1 To describe, it can be estimated by the following formula, for the method based on feature points, the least squares method can be used to solve it: Among them, p t，i and p t+1，i are the coordinates of the matched feature points at time t and t+1;

[0096] When constructing the map, as the robot moves, the collected data is integrated to construct a map. It is a feature-based map, a grid map, and the map construction method is occupancy grid mapping. After the map is constructed, closed-loop detection and optimization are performed to detect whether the robot has returned to an area it has visited before. Once a closed loop is detected, the bundle adjustment method is used to optimize the entire posture and map to eliminate cumulative errors. The constructed map and estimated robot motion are used to determine the robot's posture in the map in real time.

[0097] In this embodiment, the method adopts an intelligent recognition and classification algorithm, which can accurately classify and identify the detected stents, effectively improve the system's recognition accuracy and stability of stents, and reduce the false detection rate and missed detection rate.

[0098] Example 3

[0099] See also Figure 1 The present invention provides a technical solution: a support support sensing and positioning method of an underground support handling robot, comprising the following steps:

[0100] S1. LiDAR is used in conjunction with a camera to improve accuracy and robustness, and QR code positioning is used. Kalman filtering is used for data fusion. The camera selects a thermal imaging camera and a low-light camera, and Kalman filtering is used for data fusion. The specific steps are as follows:

[0101] A1. First determine the state variables to be fused and estimated, such as the position and speed of the target. The state vector can be defined as Where (x, y) is the location of the target, is the velocity of the target in the x and y directions;

[0102] A2. Establish a state transfer model to describe the evolution of the state between different time steps. For a linear system, the state transfer equation can be expressed as: k =F k x k-1 +w k-1 , where F k is the state transfer matrix, w k-1 is the process noise;

[0103] A3. Establish observation equations for the laser radar and camera respectively. Assume that the observation vector of the laser radar is z L，k =[x L，k ,y L，k ], the camera’s observation vector is z C，k =[x C，k ,y C，k ], then the observation equation can be expressed as: For laser radar: z L，k =H L，k x k +v L，k , for the camera: z C，k =H C，k x k +v C，k , where H L，k and H C，k is the observation matrix, v L，k and v C，k is the observation noise;

[0104] Among them, the specific steps of the Kalman filter cycle process are as follows:

[0105] B1. Given an initial state estimate and the initial covariance matrix P0;

[0106] B2. Prior state estimation:

[0107] B3. Prior covariance estimation: Where Q k-1 is the process noise covariance matrix;

[0108] B4. Perform Kalman gain calculation for LiDAR: For webcam: Among them, R L，k and R C，k are the observation noise covariance matrices of the lidar and camera, respectively;

[0109] B5. Update the status of LiDAR: For webcam: Covariance update: The update of camera data is similar;

[0110] B6. Use programming language (Python) to write Kalman filter and process and update the data according to the above process.

[0111] S2. Compensating and eliminating the path using an adaptive filter, where the adaptive filter is used to process sensor data to reduce noise, interference or errors in the path, thereby improving the accuracy and stability of the path measurement;

[0112] S3, collect sensor data and write programs to obtain laser radar point cloud data and camera data;

[0113] S4, perform data processing and fusion, develop algorithms to process data, use lidar data for map construction and positioning camera data for data detection and tracking;

[0114] S5, positioning algorithm implementation, using laser radar data to build a map, using SLAM to achieve robot positioning, and using the camera to position the bracket. The specific steps are:

[0115] C1. Collect data about the surrounding environment using sensors equipped by the robot, including lidar, camera, and IMU;

[0116] C2. Perform pre-processing operations such as filtering, denoising, and time synchronization on the data collected by the sensor to improve data quality;

[0117] C3, extract representative features from the preprocessed data (geometric features in the lidar point cloud, corners and edges in the image);

[0118] C4. Perform feature matching on data at adjacent moments or in different frames to establish associations between the data;

[0119] C5. Estimate the robot's motion using motion models and optimization algorithms based on feature matching results. Common motion models include uniform motion models and acceleration motion models.

[0120] C6. Assume that the movement of the robot between two times t and t+1 can be represented by a transformation matrix T t，t+1 To describe, it can be estimated by the following formula, for the method based on feature points, the least squares method can be used to solve it: Among them, p t，i and p t+1，iare the coordinates of the matched feature points at time t and t+1;

[0121] When building the map, as the robot moves, the collected data is integrated to build a map. It is a feature-based map, which is a point cloud map and a grid map. The map construction method is the iterative closest point (ICP) algorithm and occupancy grid mapping. After the map is built, closed-loop detection and optimization are performed to detect whether the robot has returned to an area it has visited before. Once a closed loop is detected, the entire posture and map are optimized using graph optimization (libraries such as g2o and Ceres) to eliminate cumulative errors. The constructed map and estimated robot motion are used to determine the robot's posture in the map in real time.

[0122] In this embodiment, the method uses machine learning and intelligent optimization algorithms to further improve positioning accuracy and efficiency, and can perform intelligent optimization and adjustment according to actual operating conditions to achieve faster and more accurate bracket insertion and positioning.

[0123] In Examples 1-3, it should be noted that:

[0124] B6. Use programming language (Python) to write Kalman filter, process and update the data according to the above process. The following is a simple Python sample code that demonstrates the basic Kalman filter implementation;

[0125]

[0126]

[0127]

[0128]

[0129] In C6, the constructed map and estimated robot motion are used to determine the robot's position in the map in real time. The following is a code example (a simple visual odometer example using Python and OpenCV):

[0130]

[0131]

[0132]

[0133] The laser radar (LiDAR) and QR code cited in Examples 1-3 are two different technologies, which have different characteristics and application scenarios.

[0134] Environmental perception and navigation: LiDAR can be used to obtain real-time three-dimensional information of the environment, including terrain, obstacles, etc. By combining QR codes with LiDAR, QR codes can be posted at specific locations or objects. When LiDAR scans the QR code, it can obtain relevant information stored in the QR code, such as location, direction, instructions, etc., thereby achieving more accurate navigation and positioning.

[0135] QR code assisted LiDAR positioning: QR codes are placed in the environment, and LiDAR can scan and identify these QR codes. The location information of the QR code can be used to calibrate and correct the positioning results of the LiDAR to improve the accuracy of positioning.

[0136] How to integrate QR code with LiDAR data? The following is the specific process, algorithm formula, and the method used in programming engineering:

[0137] First, in the data collection stage, it is necessary to obtain the QR code information and the lidar scanning data at the same time.

[0138] The processing of QR codes usually includes decoding to obtain key information such as location and identification contained therein.

[0139] During the fusion process, a coordinate transformation algorithm may be used to unify the coordinate system of the QR code with the coordinate system of the lidar.

[0140] In terms of algorithmic formulas, for example, in coordinate transformation, the following matrix operations may be involved:

[0141]

[0142] Among them, (x, y, z) is the original coordinate, (x', y', z') is the transformed coordinate, and r ij Construct the rotation matrix, t i is the translation vector.

[0143] In programming engineering, methods such as feature matching and data interpolation may be applied.

[0144] For the development of algorithms for data processing, common ones include Kalman filtering for predicting and correcting data, and least squares method for fitting curves or finding the optimal solution.

[0145] For example, the update step of the Kalman filter is formulated as follows:

[0146] predict:

[0147]

[0148] P k|k-1 =AP k-1 | k-1 AT +Q

[0149] renew:

[0150] K k =P k|k-1 H T (HP k|k-1 H T +R) -1

[0151]

[0152] P k|k =(IK k H)P k|k-1

[0153] in, is the state estimate, P is the estimation error covariance, A is the state transfer matrix, B is the control input matrix, u is the control input, Q is the process noise covariance, H is the observation matrix, R is the observation noise covariance, z is the observation, and K is the Kalman gain.

[0154] In the above scheme, the observation equations for the lidar and the camera are:

[0155]

[0156] The updated equation means that the data can be divided into 100 groups and calculated separately, which is faster. At the same time, if a group of data is abnormal, too high or lower than other groups of data, the group can be deleted to ensure robustness.

[0157] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for sensing and positioning a support support of an underground support handling robot, characterized in that: The steps include: S1, LiDAR is used in conjunction with cameras to improve accuracy and robustness and to use QR code positioning; S2, using adaptive filters to compensate and eliminate the path; S3, collecting sensor data; S4, perform data processing and fusion; S5. Positioning algorithm implementation.

2. A method for sensing and positioning support brackets of a downhole support handling robot according to claim 1, characterized in that: In S1, the laser radar is used in conjunction with the camera to improve accuracy and robustness, and QR code positioning is used, and Kalman filtering is used for data fusion. The camera selects one or more of a thermal imaging camera and a low-light camera.

3. A method for sensing and positioning a support support of an underground support handling robot according to claim 2, characterized in that: In S1, Kalman filtering is used to perform data fusion, and the specific steps are as follows: A1. First determine the state variables to be fused and estimated, such as the position and speed of the target. The state vector can be defined as Where (x, y) is the location of the target, is the velocity of the target in the x and y directions; A2. Establish a state transfer model to describe the evolution of the state between different time steps. For a linear system, the state transfer equation can be expressed as: k =F k x k-1 +w k-1 , where F k is the state transfer matrix, w k-1 is the process noise; A3. Establish observation equations for the laser radar and camera respectively. Assume that the observation vector of the laser radar is z L，k =[x L，k ,y L，k ], the camera’s observation vector is z C，k =[x C，k ,y C，k ], then the observation equation can be expressed as: For laser radar: z L，k =H L，k x k +v L，k , for the camera: z C，k =H C，k x k +v C，k , where H L，k and H C，k is the observation matrix, v L，k and v C，k is the observation noise.

4. A method for sensing and positioning a support support of an underground support handling robot according to claim 2, characterized in that: In S1, the specific steps of the Kalman filter loop process are as follows: B1. Given an initial state estimate and the initial covariance matrix P0; B2. Prior state estimation: B3. Prior covariance estimation: Where Q k-1 is the process noise covariance matrix; B4. Perform Kalman gain calculation for LiDAR: For webcam: Among them, R L，k and R C，k are the observation noise covariance matrices of the lidar and camera, respectively; B5. Update the status of LiDAR: For webcam: Covariance update: The update of camera data is similar; B6. Use programming language (Python) to write Kalman filter and process and update the data according to the above process.

5. The method for sensing and positioning a support support of a downhole support handling robot according to claim 1, characterized in that: In S2, the path is compensated and eliminated by using an adaptive filter, and the adaptive filter is used to process the sensor data to reduce noise, interference or error in the path, thereby improving the accuracy and stability of the path measurement.

6. The method for sensing and positioning a support support of a downhole support handling robot according to claim 1, characterized in that: In S3, the sensor data is collected and a program is written to obtain the point cloud data of the laser radar and the camera data.

7. The method for sensing and positioning a support support of a downhole support handling robot according to claim 1, characterized in that: In S4, data processing and fusion are performed, algorithms are developed to process the data, and lidar data is used to construct maps and positioning camera data is used to perform data detection and tracking.

8. The method for sensing and positioning a support support of a downhole support handling robot according to claim 1, characterized in that: In S5, the positioning algorithm is implemented, a map is constructed using laser radar data, SLAM is used to realize robot positioning, and a camera is used to locate the bracket. The specific steps are as follows: C1. Collect data about the surrounding environment using sensors equipped by the robot, including lidar, camera, and IMU; C2. Perform pre-processing operations such as filtering, denoising, and time synchronization on the data collected by the sensor to improve data quality; C3, extract representative features from the preprocessed data (geometric features in the lidar point cloud, corners and edges in the image); C4. Perform feature matching on data at adjacent moments or in different frames to establish associations between the data; C5. Estimate the robot's motion using motion models and optimization algorithms based on feature matching results. Common motion models include uniform motion models and acceleration motion models. C6. Assume that the movement of the robot between two times t and t+1 can be represented by a transformation matrix T t，t+1 To describe, it can be estimated by the following formula, for the method based on feature points, the least squares method can be used to solve it: Among them, p t，i and p t+1，i are the coordinates of the matched feature points at time t and t+1.

9. A method for sensing and positioning a support support by a downhole support handling robot according to claim 8, characterized in that: In S5, during map construction, as the robot moves, the collected data is integrated to construct a map, which is a feature-based map, which is one or more of a point cloud map and a grid map. The map construction method is one or more of an iterative closest point (ICP) algorithm and an occupancy grid mapping (occupancy grid map construction).

10. A support support sensing and positioning method for a downhole support handling robot according to claim 9, characterized in that: In S5, after the map is constructed, closed-loop detection and optimization are performed to detect whether the robot returns to an area that has been visited before. Once a closed loop is detected, the entire posture and map are optimized using one of the graph optimization (g2o, Ceres and other libraries) or bundle adjustment methods to eliminate cumulative errors. The constructed map and estimated robot motion are used to determine the robot's posture in the map in real time.

Images