A long tunnel environment under the inspection robot positioning method

Abstract

The application relates to the field of inspection robots, in particular to a long-tunnel-environment inspection robot positioning method, which comprises the following steps: an incremental rotary encoder outputs incremental odometer information along with the movement of the robot; the content of the RFID tag is absolute position information measured by using a total station; the cumulative error of the odometer is corrected through the absolute position information in the RFID tag; the robot starts a laser start gmapping algorithm for positioning when passing through a fusion point; the robot closes the gmapping algorithm and enters a correction step when passing through a next adjacent fusion point; and the robot position is obtained after the correction of the cumulative error of the odometer. The application combines multiple technologies such as laser SLAM, odometer and RFID, solves the problems of laser SLAM degeneration caused by the single structural characteristics of the tunnel environment and the large cumulative error caused by long-term odometer positioning, and realizes the accurate positioning of the inspection robot in the tunnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of inspection robots, in particular to a positioning method of an inspection robot in a long tunnel environment. BACKGROUND

[0002] Currently, the main positioning methods in tunnel environment mainly include inertial positioning navigation method, wireless positioning method, SLAM technology, etc. The inertial positioning navigation method is a positioning navigation method through the combination of odometer and inertial device. However, the odometer and inertial device will produce cumulative error in long time and long distance, so it is impossible to complete accurate positioning in long tunnel environment. The wireless positioning method is a method of positioning the mobile robot in the tunnel by arranging wireless base stations in the tunnel. This method needs to modify the tunnel and has high cost.

[0003] The SLAM technology can obtain accurate positioning information by means of laser sensor or visual sensor. For example, a robot rapid repositioning method and system based on visual dictionary disclosed in Chinese patent CN110533722A includes the following steps: obtaining a current image frame by using an image acquisition module, comparing the current image frame with a key frame saved in a visual map to find the closest similar key frame; performing feature matching on the current image frame and the similar key frame to obtain the pose relationship of the current image frame relative to the pose of the similar key frame; obtaining the pose information of the current robot according to the pose of the similar key frame in the laser map and the pose relationship of the current image frame relative to the similar key frame, and completing repositioning. The coordinate system transformation matrix T between the laser radar sensor and the image acquisition module is used to make the starting positions the same, record the pose of each key frame in the visual SLAM corresponding to the pose in the laser SLAM, and complete the alignment of the two maps.

[0004] However, the above-mentioned SLAM technology in long tunnel environment cannot obtain enough features due to the single structure of the tunnel scene, which will cause degeneration phenomenon, resulting in failure to successfully position and map. Therefore, the existing technology has the above-mentioned shortcomings in long tunnel environment, and cannot complete accurate positioning in long tunnel environment. SUMMARY

[0005] The present application provides a positioning method of an inspection robot in a long tunnel environment. The present application combines laser SLAM, odometer and RFID and other technologies to solve the problems of laser SLAM degeneration caused by the single structure of the tunnel environment and the large cumulative error caused by long-term odometer positioning, and to realize accurate positioning of the inspection robot in the tunnel.

[0006] A positioning method of an inspection robot in a long tunnel environment, comprising the following steps:

[0007] S100: Set an incremental rotary encoder on the robot, the incremental rotary encoder outputs incremental odometry information (Δx, Δy) as the robot moves;

[0008] S200: Step 2: Fix an RFID tag on the robot track in the tunnel every n meters, the content of the RFID tag is the absolute position information (x rfid ,y rfid ) measured by the total station.

[0009] S300: Step 2: Set an RFID reader on the robot to read the information of the RFID tag during the movement of the robot, and correct the cumulative error of the odometry by the absolute position information in the RFID tag.

[0010] S400: Set a laser sensor on the robot, set a fusion point every M meters, the fusion point position is (x switch ,y switch ), and the robot starts the gmapping algorithm to locate when passing through the fusion point;

[0011] S500: The robot passes through the next adjacent fusion point, closes the gmapping algorithm, enters step S300, and corrects the cumulative error of the odometry to obtain the position of the robot.

[0012] The present application combines laser SLAM, odometry and RFID and other technologies, solves the problems of laser SLAM degradation due to single tunnel environment structure characteristics and large cumulative error caused by long-term odometry positioning, and realizes accurate positioning of the inspection robot in the tunnel.

[0013] Specifically, the step of correcting the cumulative error of the odometry by the absolute position information in the RFID tag in step S200 includes:

[0014] S210: Read the RFID tag at time t, calculate the uncorrected position of the robot at time t, and the specific expression is as follows:

[0015]

[0016] Where (x t-1 ,y t-1 ) is the position of the robot at the last time, and (Δx, Δy) is the incremental odometry information output by the rotary encoder.

[0017] S220: The absolute position information of the RFID tag is (x rfid ,y rfid ). Assuming that the expression of the covariance matrix of the robot position at time t-1 is as follows:

[0018]

[0019] The expression of the measured noise variance is:

[0020]

[0021] The weight coefficient K at time t is calculated, and the specific expression is as follows:

[0022] K=(K x ,K y )=(∑xx / (∑xx+R xx ),∑yy / (∑yy+R yy )) (2-2),

[0023] S230: Calculate the corrected robot position, and the specific expression is as follows:

[0024] (x t ,y t )=(x t-1 +K x *(x rfid -x odom ),y t-1 +K y *(y rfid -y odom )) (2-3),

[0025] Specifically, in step S400 of the present application, when the Euclidean distance between the robot position (x robot ,y robot ) and the fusion point position (x switch ,y switch ) is less than d, it is judged that the robot has passed the fusion point, and the specific expression of d is as follows:

[0026] (y robot -y switch ) 2 +(x robot -x switch ) 2 ≤d 2 .

[0027] Specifically, in step S400 of the present application, the gmapping algorithm obtains the incremental odometer information (Δx, Δy) to update the particle set of the gmapping algorithm and scans the surrounding environment through the laser sensor, and finally outputs accurate robot position information.

[0028] Specifically, in the present application, the value range of the setting interval n of the RFID tag is 4-6 meters.

[0029] Specifically, the interval m of the laser fusion point in the application is 18-25 meters.

[0030] Specifically, the Euclidean distance d between the robot position and the fusion point position in the application is 0.4-0.6 meters.

[0031] The application has the beneficial effect that the combination positioning method is used in the application, different positioning methods are used in different areas, and the problems of SLAM algorithm degradation and large accumulated error of the odometer caused by single tunnel environment in a long tunnel environment are solved. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a flowchart of a long tunnel environment under the robot positioning method of the application.

[0033] Figure 2 is a schematic diagram of the principle of the long tunnel environment under the robot positioning method of the application.

[0034] Figure 3 is the opening and closing rule of the gMapping algorithm when the robot passes through the fusion point. DETAILED DESCRIPTION

[0035] The application will be further described below in combination with the drawings and specific embodiments, and the schematic embodiments of the application and the description thereof are used to explain the application and do not constitute an improper limitation on the application.

[0036] Example 1:

[0037] As shown in Figure 1 and Figure 2 , a long tunnel environment under the robot positioning method, comprising the following steps:

[0038] S100: An incremental rotary encoder is arranged on the robot, and the incremental rotary encoder outputs incremental odometer information (Δx, Δy) as the robot moves;

[0039] S200: Step 2: An RFID tag is fixed every n meters on the robot track in the tunnel, and the content of the RFID tag is the absolute position information (x rfid ,y rfid ) measured by using a total station,

[0040] S300: Step 2: An RFID card reader is arranged on the robot for reading the information of the RFID tag during the movement of the robot, and the accumulated error of the odometer is corrected by the absolute position information in the RFID tag,

[0041] As shown in Figure 3As shown, S400: The robot is equipped with laser sensors, and fusion points are set at intervals of M meters. The fusion point positions are (x... switch ,y switch When the robot passes through the fusion point, it activates the laser-based gmapping algorithm for localization.

[0042] like Figure 3 As shown, in step S500, when the robot passes the next adjacent fusion point, the gmapping algorithm is turned off, and the robot position is obtained after correcting the cumulative error of the odometry.

[0043] Example 2:

[0044] like Figure 1 and Figure 2 As shown, a method for locating an inspection robot in a long tunnel environment includes the following steps:

[0045] S100: An incremental rotary encoder is installed on the robot. The incremental rotary encoder outputs incremental odometry information (Δx, Δy) as the robot moves.

[0046] S200: Step 2: Fix an RFID tag at n-meter intervals along the robot track inside the tunnel. The content of the RFID tag is the absolute position information (x, y) obtained using a total station. rfid ,y rfid ),

[0047] S300: Step 2: An RFID reader is installed on the robot to read the information from the RFID tags during robot movement. The cumulative error of the odometer is corrected by using the absolute position information in the RFID tags.

[0048] like Figure 3 As shown, S400: The robot is equipped with laser sensors, and fusion points are set at intervals of M meters. The fusion point positions are (x... switch ,y switch When the robot passes through the fusion point, it activates a laser-based gmapping algorithm for localization.

[0049] like Figure 3 As shown, in step S500: when the robot passes the next adjacent fusion point, the gmapping algorithm is turned off, and the process proceeds to step S300. After correcting the accumulated error of the odometry, the robot position is obtained.

[0050] In this embodiment, step S200, which involves correcting the cumulative error of the odometer using the absolute position information in the RFID tag, includes:

[0051] S210: record the time t when the RFID tag is read, and calculate the uncorrected position of the robot at time t, which is expressed as follows:

[0052]

[0053] where (x t-1 ,y t-1 ) is the position of the robot at the last time, and (Δx, Δy) is the incremental odometry information output by the rotary encoder,

[0054] S220: read the absolute position information of the RFID tag as (x rfid ,y rfid ), and the covariance matrix P is a matrix composed of the covariances between states. Since the covariance between the same variables is the variance, the elements on the diagonal are the variances of x and y, respectively, and the other two elements are the covariances between the two elements, which can be estimated according to the state variance. Assuming that the expression of the covariance matrix of the robot position at time t-1 is as follows:

[0055]

[0056] The measurement noise covariance matrix R is the error of the sensor measurement value. In use, the sensor will generally give an accuracy index, which can be directly converted to the matrix R. The measurement variances of each measurement can be read from the sensor specification. In addition, the covariances between states cannot be determined and can be set to 0. In this embodiment, the expression of the measurement noise covariance matrix is as follows:

[0057]

[0058] Calculate the weight coefficient K at time t, which is expressed as follows:

[0059] K = (K x ,K y ) = (∑xx / (∑xx+R xx ),∑yy / (∑yy+R yy )) (2-2),

[0060] S230: calculate the corrected robot position, which is expressed as follows:

[0061] (x t ,y t ) = (x t-1 + K x *(x rfid -x odom ),y t-1 + K y *(y rfid -y odom)) (2-3),

[0062] In this embodiment, in step S400, when the Euclidean distance between the robot position (x robot ,y robot ) and the fusion point position (x switch ,y switch ) is less than d, it is determined that the robot has passed the fusion point, and the specific expression of d is as follows:

[0063] (y robot -y switch ) 2 +(x robot -x switch ) 2 ≤d 2 .

[0064] In this embodiment, in step S400, the gmapping algorithm obtains the incremental odometer information (Δx, Δy) to update the particle set of the gmapping algorithm and scans the surrounding environment through the laser sensor, and finally outputs accurate robot position information.

[0065] In this embodiment, the value of n is 5 meters, the value of m is 20 meters, and the value of d is 0.5 meters.

[0066] As shown in Figure 2 , the specific working method of this embodiment is as follows: the tunnel inspection robot is equipped with a rotary encoder, a laser sensor and an RFID card reader;

[0067] As shown in Figure 1As shown, the robot internal industrial computer installs ROS operating system, and continuously reads information of robot rotary encoder to obtain incremental mileage information of robot movement. RFID tags are fixed to the tunnel every 5 meters, and total station is used to measure position of each RFID tag relative to tunnel entrance, and position information is written into RFID tag through card reader for solidification. Then, the tunnel is divided according to 20 meters, and a fusion point is set every 20 meters. Finally, the robot is started, and the industrial computer continuously obtains odometer information during movement of the robot, and when RFID tag is detected during movement, the content of the RFID tag is corrected to the obtained odometer information to eliminate cumulative error in the rotary encoder; when the robot moves to the vicinity of the first fusion point, the laser SLAM algorithm gmapping is started, and the odometer information output by the rotary encoder is provided to the gmapping algorithm as initial position information and updates the particle set in the gmapping algorithm, and during the gmapping algorithm is started, the robot position information is provided by the gammping, and since the key section area is small in size, the gmamping can provide accurate positioning information in the key section area. When the robot moves to the adjacent fusion point, the gmapping algorithm is closed, and after the gmapping algorithm is closed, the position information output by the rotary encoder and corrected by the RFID is used, and through the above method, positioning failure caused by degradation of the gmapping in the long tunnel environment can be solved, and cumulative error of the rotary encoder can be solved.

[0068] To verify the technical effect of the positioning method of the embodiment, three tests are carried out in the tunnel by using three methods, the first positioning method only uses the odometer output, the second positioning method uses the odometer and RFID tag, and the third positioning method uses the long tunnel environment patrol robot positioning method described in the embodiment. In the three tests, we select a section between the start point position (0, 0) and the end point position (400, 2) as a test section, the start point coordinate A of the test section is (300, 2), the end point D coordinate is (340, 2), the RFID tag point B is set at (310, 2), and the fusion point C is set at (320, 2).

[0069] First test: using the first positioning method, only collect the positioning data from the rotary encoder, the robot starts at the origin (0, 0), at this time the robot position is (0, 0), during the robot running, the rotary encoder constantly outputs incremental information (Δx, Δy), the robot position is constantly accumulated. When the robot reaches the starting point A of the test section, the robot positioning position information is (308.34, 1.8), when the robot reaches B point, the robot shows the positioning (315.90, 1.9), the cumulative error has reached (15.90, 0.1).

[0070] Second test: using the second positioning method, that is, using the fusion method of odometer and RFID tag. We set the RFID tag at B point, and write the position information (310, 2). We observe that before the robot reaches B point, the robot positioning information is (314.0, 1.88). Assuming that at time t-1, the robot position is (314.5, 1.88), due to measurement error, the content written in the RFID tag is (310, 2). At time t, the robot detects the RFID tag and starts to correct the robot position information. At this time, the covariance matrix of the position information The measurement noise matrix refers to the error of calibrating the content of the RFID tag, that is, the error of calibrating the positioning data (320, 2). We use the measuring wheel to measure, according to the instruction manual, the accuracy of the measuring wheel is 0.5%, so we set the measurement matrix variance Substituting the above data into the formula K = (K x ,K y ) = (∑xx / (∑xx+R xx ),∑yy / (∑yy+R yy )) (2-2)、(x t ,y t ) = (x t-1 +K x *(x rfid -x odom ),y t-1 +K y *(y rfid -y odom )) (2-3) can be obtained, after detecting the RFID tag, the robot position information is (310.02, 1.999). Compared with the first test result, the error is greatly reduced.

[0071] Third test: using the third positioning method, that is, the long tunnel environment under the inspection robot positioning method described in this embodiment, on the basis of the second test, the performance is advanced to point C, when approaching the fusion point, that is, the Euclidean distance between the robot positioning information and the position information of point C is less than 0.5 meters, we start the gammping positioning algorithm.

[0072] The gmapping algorithm is a RBPF-based SLAM solution. This solution is to first estimate the robot motion trajectory, and then estimate the environment map according to the state of the robot; then this environment map can be used to update the robot motion trajectory in turn. Each particle in the algorithm independently records a possible robot trajectory and its corresponding environment map. Gmapping is based on a filtering framework, and each calculation will have a lot of particles, and each particle will carry its own map, which is equivalent to sacrificing space (storage) to win time (computing efficiency). Therefore, when constructing a small scene map, the gmapping calculation is small and the precision is not too much. However, the problem of gmapping is that it is not suitable for large scenes, and the storage is easy to be blown up. The algorithm steps are as follows:

[0073] First, update the particle pose by the odometer;

[0074] Second, perform laser matching, optimize the pose of each particle updated by the odometer according to the matching degree of the current frame laser and the map, and calculate its score and likelihood value to update the parameters related to the particle;

[0075] Third, update the weight of each particle in the node. According to the map and a frame of laser, the score of each particle is calculated. The score means: if the frame of laser is obtained at the robot pose represented by the particle, how well does it match the map, the higher the score, the better the matching degree.

[0076] Fourth, judge whether to resample.

[0077] If yes: resample according to the particle weight, and the particles remaining after resampling update the map according to the laser frame, and a node corresponding to the current state is added to the node member.

[0078] If no: update the map of each particle according to the laser frame, and add a node corresponding to the current state to the node member.

[0079] Fifth, a series of initialization operations:

[0080] Update the map of each particle according to the laser data, add the first node to the particle, etc.

[0081] The sixth step is to perform the same operation as the fifth step. Because: if it is initialized, the weight is updated; if it is resampled, the weight is also updated because the weight is changed during the resampling process

[0082] In a long tunnel environment, the accumulated error of the odometer is large and the laser features in the tunnel environment are single, which can easily cause the gmapping algorithm to fail to locate. In this method, we take segmented start of the gmapping algorithm according to the tunnel environment, thereby solving the problem of data explosion of the gmapping algorithm in a large scene, and continuously obtaining good positioning effect during the running of the gmapping algorithm.

[0083] After multiple tests, the third test, a long tunnel environment under the positioning method of the inspection robot in this embodiment, uses the odometer and RFID tag correction fusion data to provide a better motion trajectory estimation for the gmapping algorithm. After starting the gmapping algorithm, the robot positioning information at point D is (340±0.5, 2±0.5), and the error is greatly reduced. At the same time, the positioning information between points C and D is also very accurate.

[0084] Example 3:

[0085] The difference between this embodiment and the above-mentioned embodiment 2 is that the value of n is 6 meters, the value of m is 25 meters, and the value of d is 0.6 meters.

Claims

1. A method for positioning a patrol robot in a long tunnel environment, characterized in that The method comprises the following steps: S100: An incremental rotary encoder is provided on the robot, which outputs incremental odometry information as the robot moves ; S200: Fix an RFID tag on the robot track in the tunnel every n meters, the content of the RFID tag is the absolute position information measured by the total station ; S300: An RFID reader is arranged on the robot to read the information of the RFID tag during the movement of the robot, and the accumulated error of the odometer is corrected by the absolute position information in the RFID tag; S400: The laser sensor is arranged on the robot, a fusion point is arranged every M meters, the fusion point position is , and the robot starts the laser to start the gmapping algorithm for positioning when passing the fusion point. S500: When the robot passes the next adjacent fusion point, the gmapping algorithm is turned off, and the robot position is obtained after the accumulated error of the odometer is corrected in step S300; The step of correcting the accumulated error of the odometer by the absolute position information in the RFID tag in step S300 comprises: S310: The RFID tag is read at time t, and the uncorrected position of the robot at time t is calculated, and the expression is as follows: ， wherein, is the position of the robot at the last time instant, is incremental odometry information output by the rotary encoder; S320: read the absolute position information of the RFID tag as Assume that the expression of the covariance matrix of the robot position at time t-1 is as follows: ， The expression of the measured noise variance is: ， The weight coefficient K at time t is calculated, and the expression is as follows: ， S330: The corrected robot position is calculated, and the expression is as follows: 。 2. The method of claim 1, wherein: In step S400, when the Euclidean distance between the robot position and the fusion point position is less than d, it is determined that the robot has passed the fusion point, and the expression of d is as follows: 。 3. The method of claim 1, wherein: In step S400, the gmapping algorithm obtains incremental odometry information to update the particle set of the gmapping algorithm and scan the surrounding environment through the laser sensor, and finally output accurate robot position information.

4. The long tunnel environment-based inspection robot positioning method according to claim 1, characterized in that: The value range of n is 4-6 meters.

5. The method of claim 1, wherein: The value range of m is 18-25 meters.

6. The long tunnel environment robot positioning method according to claim 1, wherein: The value range of d is 0.4-0.6 meters.

Images