Method for automatically planning spraying track of tunnel wet spraying trolley

Abstract

The invention discloses a tunnel wet spraying trolley spraying track automatic planning method, which comprises the following steps: constructing a rasterized spraying target matrix G by using a tunnel to-be-sprayed surface point cloud data set obtained by scanning and a design section model, constructing an accumulation grid matrix S under parallel track overlapping by using a concrete spraying accumulation model as a basis, and calculating the spraying track of a tunnel to be sprayed according to the accumulation grid matrix S; and then the optimal injection time matrix T on the basic injection path is calculated, cumulative model parameters are determined only through an initial injection test, a large number of training samples or existing experience do not need to be collected, a large database does not need to be constructed, the calculation efficiency is higher, and the cost is lower.

Description

Technical Field

[0001] The present invention belongs to the technical field of tunnel construction, and in particular relates to a method for automatically planning the spraying trajectory of a tunnel wet spraying trolley. Background Art

[0002] The wet spraying trolley is a key equipment for wet concrete spraying in tunnel, mine, slope support and other projects. It integrates the functions of concrete transportation, spraying and robotic arm operation.

[0003] Mechanized and intelligent tunnel construction technology is an important development trend in the future. As an important tool for tunnel primary lining construction, the degree of intelligence of the wet shotcrete trolley greatly affects the quality of lining construction.

[0004] Compared to manually controlled spraying methods, the automated spraying technology of wet spray trolleys and robotic arms offers greater accuracy and stability in controlling spray layer flatness and thickness. This effectively addresses current construction quality issues caused by uneven operator skills. The automated spray trajectory planning method is the core technology for automated spraying using wet spray trolleys and robotic arms.

[0005] Currently, CN118933860A discloses an automatic control method for a wet-spraying trolley in a mine tunnel. This method uses a three-dimensional laser radar to obtain geometric information and features of the area to be sprayed. Based on the results of path planning and inverse calculations, the method controls the movement of the wet-spraying manipulator, causing it to perform spraying operations along a predetermined trajectory and posture. However, a detailed description of the path planning algorithm is lacking. This method adds an industrial computer and sensors to the existing wet-spraying trolley to monitor the trolley's status in real time. Control instructions are then sent to the trolley's main controller via an extended CAN communication interface to control the trolley's movements.

[0006] For example, CN118933860A discloses an automatic control method for a wet-spraying trolley in a mining tunnel. This method uses a three-dimensional laser radar to acquire geometric information and features of the area to be sprayed. Based on the results of path planning and inverse calculations, the method controls the movement of the wet-spraying manipulator, causing it to perform spraying operations along a predetermined trajectory and posture. However, a detailed description of the path planning algorithm is lacking. This method controls the movement of the wet-spraying trolley by adding an industrial computer and sensors to the existing trolley to monitor the trolley's status in real time. Furthermore, the method sends control commands to the trolley's main controller via an extended CAN communication interface.

[0007] For example, CN117404108A discloses an intelligent control method for a wet spraying trolley based on a large database. This method collects construction condition information and operational data during wet spraying operations to establish a cloud-based large database. During operation, the optimal wet spraying formula, including the spraying path and wet spraying parameters for each sub-area, is searched in the database based on the positioning information and construction condition information of the wet spraying trolley. The wet spraying trolley's arm and pumping linkage unit are then controlled to spray according to the current spraying path. The data of the surface to be sprayed is collected in real time and compared with the working database to adjust the wet spraying formula.

[0008] That is, the applicant found that most of the existing technologies revolve around construction parameters and scene perception technology and wet spraying trolley automatic control technology. Decision-making issues such as spraying path planning and medium parameter optimization still require human participation or determination through a large number of experiments, and the degree of intelligence is not high. Summary of the Invention

[0009] The purpose of the present invention is to overcome the defects of the prior art and provide a method for automatically planning the spraying trajectory of a tunnel wet spraying trolley. The method uses the scanned point cloud dataset of the tunnel surface to be sprayed and the designed cross-section model to construct a rasterized spraying target matrix G. Based on the concrete spraying accumulation model, the cumulative grid matrix S under the overlapping parallel trajectories is constructed, and then the optimal spraying time matrix T on the basic spraying path is calculated. Only the initial spraying test is required to determine the cumulative model parameters. There is no need to collect a large number of training samples or existing experience, and there is no need to build a large database. The calculation efficiency is higher and the cost is lower.

[0010] The object of the present invention is achieved through the following technical solutions:

[0011] A method for automatically planning the spraying trajectory of a tunnel wet spraying trolley comprises the following steps:

[0012] S1. Construct a design section model based on the tunnel's planned central axis, design section profile, and reserved deformation;

[0013] S2, collect the original point cloud data set of the tunnel surface to be sprayed;

[0014] S3, pre-processing the original point cloud data set, and then reconstructing the surface to be sprayed to build a cross-section model of the surface to be sprayed;

[0015] S4. In the polar coordinate system, the cross-section model to be sprayed is gridded. The grid size is determined according to the spray radius and accuracy requirements.

[0016] S5. Based on the designed cross-sectional model, a reference fitting surface of the point cloud dataset of the spraying surface is constructed using the least squares method. The fitting surface is compared with the cross-sectional model to be sprayed, and the over-excavation and under-excavation information of each grid of the spraying surface is calculated. According to the maximum layer thickness, the gridded spraying target matrix G is determined.

[0017] S6. Assuming that β distribution is the cumulative model of concrete spraying, the distribution parameters are determined through trial spraying experiments, and the spray gun cumulative grid matrix P is constructed;

[0018] S7, moving the spray gun horizontally and linearly from bottom to top layer by layer to construct a basic spray path, and using the gradient descent algorithm to calculate the spray time of each grid on the spray gun movement path, so that the mean square error between the spray accumulation matrix S of the spray surface to be sprayed and the spray target matrix G is minimized to obtain the optimal spray time matrix T;

[0019] S8. After a single spraying operation is completed, steps S2 to S7 are repeated until the contour of the surface to be sprayed matches the designed cross-sectional model.

[0020] In one embodiment, in step S1, the designed cross-section model is represented by polar coordinates as follows:

[0021]

[0022] Among them, r0 is the tunnel design section profile, (x0, y0, z0) is the centerline coordinate, r Δ Reserved deformation value.

[0023] In one embodiment, in step S2, a three-dimensional laser scanning device is used to collect an original point cloud data set of the tunnel surface to be sprayed.

[0024] In one embodiment, step S3 includes:

[0025] Step S301: Registration: Using the tunnel centerline control point corresponding to the current mileage as a reference, set up a scanning station and register the point cloud dataset through coordinate transformation;

[0026] Step S302: De-noising, using statistical filtering method to remove noise points;

[0027] Step S303: Thinning, using a voxel downsampling algorithm to uniformly sample and reduce the size of the point cloud dataset;

[0028] Step S304: Segmentation: using a random sampling consensus method to identify and segment the point cloud data set of the surface to be sprayed;

[0029] Step S305 , surface reconstruction, using the moving least squares method to smooth and reconstruct the segmented surface point cloud data set to construct a section model to be sprayed.

[0030] In one embodiment, in step S4, the section model to be sprayed is subjected to rasterization processing, including:

[0031] Step S401: Assume that a vertical section intersects the tunnel centerline at point O, and establish the polar coordinates of the surface to be sprayed with O as the origin:

[0032]

[0033] Step S402: Divide the tunnel into equal grids in terms of length and angle. Assuming the grid size is s×s, then:

[0034] The dimensions in the long and short directions are:

[0035]

[0036] The ceil function means rounding up;

[0037] Step S403: All data points of the point cloud dataset of the surface to be sprayed are classified into corresponding grids according to the polar coordinate angle θ and radius r. For example, the row and column number of the grid to which a certain point belongs is:

[0038]

[0039] In one embodiment, in step S5, determining the gridded injection target matrix G includes:

[0040] Step S501: r0 = f(θ, y) represents the design surface, and the least square method is used to fit the fitting surface of the spraying surface. The fitting surface is parallel to the design surface, and r s =f(θ,y) means, r s is the optimal fitting radius;

[0041] Step S502: traverse all grids, calculate the distance between the data point in each grid and the fitting surface, and set the maximum thickness of a single injection as d max , then the target thickness of the spray under a single spray is d = r s +d max -r;

[0042] Step S503: Set is the average spray thickness of all data points in the grid, then A grid matrix of dimensions rows×cols is constructed as the grid value, which is the target injection matrix G.

[0043] In one embodiment, step S6 includes:

[0044] Step S601: Determine the value of β:

[0045]

[0046] Among them, β is the distribution parameter, F is the cohesion force, α is the horizontal angle of the spray axis, R is the spray radius of the spray gun, Q is the spray velocity, and t is the spray time;

[0047] Step S602: Consider the thickness overlap accumulation of two parallel spraying trajectories.

[0048] Calculate the current β value command Take the minimum theoretical optimal track spacing L;

[0049] Step S603: Grid the cumulative model. The size of the grid is the same as that in step S4. Calculate the cumulative thickness per unit time P of all grid center points within the radiation radius. kl , the stroke spray gun accumulation grid matrix P, with dimension m×m;

[0050] in,

[0051] In one embodiment, in step S7, obtaining the optimal injection time matrix T includes:

[0052] Step S701: Constructing the objective function:

[0053]

[0054] G is the injection target matrix obtained in step S4;

[0055] Step S702: Initialize the time matrix T(t ij ), the dimension of T is (rows+m)×floor(cols / L), where L is the distance between two parallel tracks;

[0056] Step S703: The calculation formula of the thickness accumulation matrix S of the surface to be sprayed is as follows:

[0057]

[0058] Step S704: Use the gradient descent algorithm to iteratively find the optimal time matrix T;

[0059] Step S705: Obtain the actual motion trajectory of the nozzle through interpolation calculation.

[0060] The beneficial effects of the present invention are:

[0061] The automatic spraying trajectory planning method provided by the present invention mainly seeks optimization on the basic spraying path, constructs a rasterized spraying target matrix G based on the scanned point cloud dataset of the tunnel surface to be sprayed and the designed cross-section model, and constructs a cumulative grid matrix S under the overlapping parallel trajectories based on the concrete spraying accumulation model. Then, the optimal spraying time matrix T on the basic spraying path is calculated. Only the initial spraying test is required to determine the cumulative model parameters, and the layered spraying adjustment is carried out by combining theoretical calculations and detection feedback. The optimal spraying path can be obtained without collecting a large number of training samples or existing experience, which has higher computational efficiency and lower cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The present invention will be described in more detail below based on embodiments and with reference to the accompanying drawings, wherein:

[0063] Figure 1 Shows a schematic flow diagram of the present invention;

[0064] Figure 2 shows a schematic diagram of the injection path planning of the present invention;

[0065] In the drawings, like reference numerals are used for like parts, but the drawings are not necessarily true to scale. DETAILED DESCRIPTION

[0066] The present invention will be further described below with reference to the accompanying drawings.

[0067] The present invention provides a method for automatically planning the spraying trajectory of a tunnel wet spraying trolley. Figure 1 As shown, the following steps are included:

[0068] Step S1: constructing a design section model, including:

[0069] Construct a design section model based on the tunnel's planned central axis, design section profile, and reserved deformation;

[0070] The design section model is expressed in polar coordinates as follows:

[0071]

[0072] Among them, is the tunnel design section profile, is the centerline coordinate, and is the reserved deformation value;

[0073] Step S2: collecting the original point cloud data set of the tunnel surface to be sprayed by a three-dimensional laser scanning device;

[0074] Step S3: pre-process the original point cloud data set, and then reconstruct the surface to be sprayed to construct a cross-sectional model of the surface to be sprayed, including:

[0075] Step S301: Registration: Using the tunnel centerline control point corresponding to the current mileage as a reference, set up a scanning station and register the point cloud dataset through coordinate transformation;

[0076] Step S302: De-noising, using statistical filtering method to remove noise points;

[0077] Step S303: Thinning, using a voxel downsampling algorithm to uniformly sample and reduce the size of the point cloud dataset;

[0078] Step S304: Segmentation: using a random sampling consensus method to identify and segment the point cloud data set of the surface to be sprayed;

[0079] Step S305: surface reconstruction, using the moving least squares method to smooth and reconstruct the segmented surface point cloud data set to construct a section model to be sprayed;

[0080] Step S4, rasterization processing of the surface to be sprayed, including:

[0081] In the polar coordinate system, the section model to be sprayed is gridded, and the grid size is determined according to the spray radius and accuracy requirements;

[0082] Specifically, in step S401, assuming that a vertical section intersects the tunnel centerline at point O, the polar coordinates of the surface to be sprayed are established with O as the origin:

[0083]

[0084] Step S402: Divide the tunnel into equal grids in terms of length and angle. Assuming the grid size is s×s, then:

[0085] The dimensions in the long and short directions are:

[0086]

[0087] The ceil function means rounding up;

[0088] Step S403: All data points of the point cloud dataset of the surface to be sprayed are classified into corresponding grids according to the polar coordinate angle θ and radius r. For example, the row and column number of the grid to which a certain point belongs is:

[0089]

[0090] Step S5: construct a reference fitting surface for the surface to be sprayed, compare the cross-section model to be sprayed and calculate the over-excavation and under-excavation information to obtain the spray target matrix, including:

[0091] Based on the designed cross-section model, the least squares method is used to construct a reference fitting surface of the point cloud dataset of the spraying surface. The fitting surface is compared with the cross-section model to be sprayed, and the over-excavation and under-excavation information of each grid of the spraying surface is calculated. According to the maximum layer thickness, the gridded spraying target matrix G is determined.

[0092] Specifically:

[0093] Step S501: r0 = f(θ, y) represents the design surface, and the least square method is used to fit the fitting surface of the spraying surface. The fitting surface is parallel to the design surface, and r s =f(θ,y) means, r s is the optimal fitting radius;

[0094] Step S502: traverse all grids, calculate the distance between the data point in each grid and the fitting surface, and set the maximum thickness of a single injection as dmax , then the target thickness of the spray under a single spray is d = r s +d max -r;

[0095] Step S503: Set is the average spray thickness of all data points in the grid, then As the grid value, a grid matrix of dimensions rows×cols is constructed, which is the target injection matrix G;

[0096] Step S6: constructing a spray gun accumulation grid matrix, including:

[0097] Assuming that β distribution is the cumulative model of concrete spraying, the distribution parameters are determined through trial spraying experiments, and the spray gun cumulative grid matrix P is constructed;

[0098] Specifically, step S6 includes:

[0099] Step S601: Determine the value of β:

[0100]

[0101] Among them, β is the distribution parameter, F is the cohesion force, α is the horizontal angle of the spray axis, R is the spray radius of the spray gun, Q is the spray velocity, and t is the spray time;

[0102] Step S602: Consider the thickness overlap accumulation of two parallel spraying trajectories.

[0103] Calculate the current β value command Take the minimum theoretical optimal track spacing L;

[0104] Step S603: Grid the cumulative model. The size of the grid is the same as that in step S4. Calculate the cumulative thickness per unit time P of all grid center points within the radiation radius. kl , the stroke spray gun accumulation grid matrix P, with dimension m×m;

[0105] in,

[0106] Step S7: According to the basic route of horizontal straight-line movement from bottom to top, the optimal injection time matrix T on the injection path is calculated using the gradient descent algorithm with the mean square error between the injection accumulation matrix S and the injection target matrix G as the loss function. The injection parameters are obtained by smoothing to complete a single injection cycle, as shown in the following example: Figure 2 Shown, including:

[0107] Step S701: Constructing the objective function:

[0108]

[0109] G is the injection target matrix obtained in step S4;

[0110] Step S702: Initialize the time matrix T(t ij ), the dimension of T is (rows+m)×floor(cols / L), where L is the distance between two parallel tracks;

[0111] The floor() function means rounding down;

[0112] Step S703: The calculation formula of the thickness accumulation matrix S of the surface to be sprayed is as follows:

[0113]

[0114] Step S704: Use the Adam gradient descent algorithm (Adaptive Moment Estimation) to iteratively find the optimal time matrix T;

[0115] Specifically, the pseudo code is as follows:

[0116] #==================Model structure pseudo code==================

[0117] Class PaintingTimeModel:

[0118] initialization:

[0119] Inherited from the neural network module;

[0120] Define the trainable parameter T (track time matrix) ← all zero matrix (size is tracks_rows × (tracks_cols));

[0121] Forward propagation (m, input, repeated_cols_tensor):

[0122] t_pos←apply softplus transformation (beta=10) to t;

[0123] Transpose t_pos to get transposed_t;

[0124] Perform window sliding expansion operation on transposed_w:

[0125] The first dimension is expanded by the window dimension 2 → the second dimension is expanded by the window dimension m (the gun matrix dimension);

[0126] Flatten and transpose → unfolded_t;

[0127] Repeated expansion along dim=1 dimension by repeated_cols_tensor →

[0128] repeated_unfolded_t;

[0129] Calculate the weighted input: multiplied_t←repeated_unfolded_t⊙input;

[0130] Output prediction value: out←sum along the last dimension (multiplied_t);

[0131] return out;

[0132]

[0133]

[0134] Step S705: Obtain the actual motion trajectory of the nozzle through a smoothing method, such as interpolation calculation;

[0135] Step S8: After the single spraying operation is completed, repeat steps S2 to S7 until the contour of the surface to be sprayed matches the designed cross-sectional model. That is, after the single spraying operation is completed, check whether the designed thickness is reached. If the designed thickness is not reached, repeat steps S2 to S7 until the contour of the surface to be sprayed matches the designed cross-sectional model.

[0136] It should be noted that, in this embodiment, optimization is performed on the basic spraying path, and a rasterized spraying target matrix G is constructed using the scanned point cloud dataset of the tunnel surface to be sprayed and the designed cross-section model. Based on the concrete spraying accumulation model, a cumulative grid matrix S under overlapping parallel trajectories is constructed, and then the optimal spraying time matrix T on the basic spraying path is calculated. Only the initial spraying test is required to determine the cumulative model parameters, and the layered spraying is adjusted by combining theoretical calculations and detection feedback. The optimal spraying path can be obtained without collecting a large number of training samples or existing experience, which has higher computational efficiency and lower cost.

[0137] In the description of the present invention, it should be understood that the terms "upper", "lower", "bottom", "top", "front", "back", "inside", "outside", "left", "right", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention.

[0138] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be employed in conjunction with other described embodiments.

Claims

1. A method for automatically planning the spraying trajectory of a tunnel wet spraying trolley, characterized in that: The steps include: S1. Construct a design section model based on the tunnel's planned central axis, design section profile, and reserved deformation; S2, collect the original point cloud data set of the tunnel surface to be sprayed; S3, pre-processing the original point cloud data set, and then reconstructing the surface to be sprayed to build a cross-section model of the surface to be sprayed; S4. In the polar coordinate system, the cross-section model to be sprayed is gridded. The grid size is determined according to the spray radius and accuracy requirements. S5. Based on the designed cross-sectional model, a reference fitting surface of the point cloud dataset of the spraying surface is constructed using the least squares method. The fitting surface is compared with the cross-sectional model to be sprayed, and the over-excavation and under-excavation information of each grid of the spraying surface is calculated. According to the maximum layer thickness, the gridded spraying target matrix G is determined. S6. Assuming that β distribution is the cumulative model of concrete spraying, the distribution parameters are determined through trial spraying experiments, and the spray gun cumulative grid matrix P is constructed; S7, moving the spray gun horizontally and linearly from bottom to top layer by layer to construct a basic spray path, and using the gradient descent algorithm to calculate the spray time of each grid on the spray gun movement path, so that the mean square error between the spray accumulation matrix S of the spray surface to be sprayed and the spray target matrix G is minimized to obtain the optimal spray time matrix T; S8. After a single spraying operation is completed, steps S2 to S7 are repeated until the contour of the surface to be sprayed matches the designed cross-sectional model.

2. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 1 is characterized in that: In step S1, the designed cross-section model is expressed in polar coordinates as follows: Among them, r0 is the tunnel design section profile, (x0, y0, z0) is the centerline coordinate, r Δ Reserved deformation value.

3. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 1 is characterized in that: In step S2, a three-dimensional laser scanning device is used to collect the original point cloud data set of the tunnel surface to be sprayed.

4. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 1 is characterized in that: In step S3, it includes: Step S301: Registration: Using the tunnel centerline control point corresponding to the current mileage as a reference, set up a scanning station and register the point cloud dataset through coordinate transformation; Step S302: De-noising, using statistical filtering method to remove noise points; Step S303: Thinning, using a voxel downsampling algorithm to uniformly sample and reduce the size of the point cloud dataset; Step S304: Segmentation: using a random sampling consensus method to identify and segment the point cloud data set of the surface to be sprayed; Step S305 , surface reconstruction, using the moving least squares method to smooth and reconstruct the segmented surface point cloud data set to construct a section model to be sprayed.

5. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 4 is characterized in that: In step S4, the section model to be sprayed is subjected to rasterization processing, including: Step S401: Assume that a vertical section intersects the tunnel centerline at point O, and establish the polar coordinates of the surface to be sprayed with O as the origin: Step S402: Divide the tunnel into equal grids in terms of length and angle. Assuming the grid size is s×s, then: The dimensions in the long and short directions are: The ceil function means rounding up; Step S403: All data points of the point cloud dataset of the surface to be sprayed are classified into corresponding grids according to the polar coordinate angle θ and radius r. For example, the row and column number of the grid to which a certain point belongs is:

6. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 5 is characterized in that: In step S5, a gridded injection target matrix G is determined, including: Step S501: r0 = f(θ, y) represents the design surface, and the least square method is used to fit the fitting surface of the spraying surface. The fitting surface is parallel to the design surface, and r s =f(θ,y) means, r s is the optimal fitting radius; Step S502: traverse all grids, calculate the distance between the data point in each grid and the fitting surface, and set the maximum thickness of a single injection as d max , then the target thickness of the spray under a single spray is d = r s +d max -r; Step S503: Set is the average spray thickness of all data points in the grid, then A grid matrix of dimensions rows×cols is constructed as the grid value, which is the target injection matrix G.

7. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 6 is characterized in that: Step S6 includes: Step S601: Determine the value of β: Among them, β is the distribution parameter, F is the cohesion force, α is the horizontal angle of the spray axis, R is the spray radius of the spray gun, Q is the spray velocity, and t is the spray time; Step S602: Consider the thickness overlap accumulation of two parallel spraying trajectories. Calculate the current β value command Take the minimum theoretical optimal track spacing L; Step S603: Grid the cumulative model. The size of the grid is the same as that in step S4. Calculate the cumulative thickness per unit time P of all grid center points within the radiation radius. kl , the stroke spray gun accumulation grid matrix P, with dimension m×m; in, 8. The method for automatically planning the spraying trajectory of a tunnel wet spraying trolley according to claim 7 is characterized in that: In step S7, the optimal injection time matrix T is obtained, including: Step S701: Constructing the objective function: G is the injection target matrix obtained in step S4; Step S702: Initialize the time matrix T(t ij ), the dimension of T is (rows+m)×floor(cols / L), where L is the distance between two parallel tracks; Step S703: The calculation formula of the thickness accumulation matrix S of the surface to be sprayed is as follows: Step S704: Use the gradient descent algorithm to iteratively find the optimal time matrix T; Step S705: Obtain the actual motion trajectory of the nozzle through interpolation calculation.

Images