Full-coating putty spraying path planning method for rail passenger car side wall plate

Abstract

The application discloses a full-coverage putty spraying path planning method for a rail passenger car side wall plate, and comprises the following steps: obtaining three-dimensional point cloud data of the rail passenger car side wall plate, extracting base surface information and constructing a grid map, and placing a neuron at each grid; setting a virtual point position and a starting advancing direction of an end effector of a spraying robot; optimizing a steering parameter term in a moving control equation of a biological excitation neural network algorithm, establishing a feedback adjustment coefficient by using topological neurons of a traversed area, and completing full-coverage path planning of a to-be-sprayed area based on an improved biological excitation neural network; in the movement of the end effector of the spraying robot, the connectivity between areas is detected in real time by using an area decomposition detection algorithm, and an optimal movement point is selected according to a risk avoidance mechanism, so that the probability of the end effector of the spraying robot falling into a dead zone is reduced; and the application meets the putty spraying process requirements of the existing rail vehicle, and the generated spraying path is coherent, orderly and has a low repetition rate.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent robot control, and specifically relates to a track passenger train side wall plate full-coverage putty spraying path planning method based on an improved biological excitation neural network. BACKGROUND

[0002] With the development of the track vehicle manufacturing industry, the traditional manual spraying production means cannot meet the requirements, robot automatic spraying as a new intelligent spraying technology has gradually matured in the coating field, and is limited by the high viscosity of putty. At present, there is no successful case in the putty spraying of the train body. Therefore, the traditional coating still uses the manual scraping method, and unsaturated polyester putty is used for full scraping of the train body. Even so, the manual spraying and polishing method still loses 30% of the material.

[0003] Due to the influence of the viscosity and curing reaction of the putty, when the robot spray gun is turned on each time, the putty is sprayed in a lump or columnar form, and if the spray gun is frequently turned on and off, the putty may be blocked. Therefore, it is necessary to reduce the number of times of turning on and off the spray gun as much as possible to reduce material waste and improve the quality of the putty coating. The realization of the full-coverage spraying path planning requires that a collision-free continuous path from the starting position of the robot end effector to the global coverage is planned in the specified area, and the global path total length and the path overlap rate of the putty spraying task need to be optimized. SUMMARY

[0004] The technical problem to be solved by the application is to provide a full-coverage putty spraying path planning method for a track passenger train side wall plate, which effectively reduces the path repetition rate and the total path length by using an improved biological excitation neural network algorithm, so as to reduce the number of times of turning on the spray gun, save putty, and improve the quality of the putty coating.

[0005] To achieve the above technical purposes, the technical scheme adopted by the application is as follows:

[0006] A full-coverage putty spraying path planning method for a track passenger train side wall plate, comprising:

[0007] Step 1: obtaining three-dimensional point cloud data of the track passenger train side wall plate, extracting base surface information and constructing a grid map, placing a neuron at each grid and initializing it;

[0008] Step 2, introducing a feedback adjustment coefficient and a steering function based on sequential bits, improving the path point selection function, and then constructing a decision strategy of path optimal points, and obtaining the grid unit that the robot end effector can move at the next time according to the decision strategy of the path optimal points;

[0009] Step 3, detecting the connectivity of the uncovered grid cells in the grid map by using the sequence repetition value coding technology;

[0010] Step 4, judging whether the uncovered grid region in the grid map is connected according to the result obtained in step 3; if not, obtaining the grid cell actually moved by the robot end effector at the next time according to the established regional decomposition risk avoidance mechanism (i.e. steps 4.2-4.3 hereinafter); if yes, directly taking the grid cell possibly moved by the robot end effector at the next time calculated in step 2 as the grid cell actually moved by the robot end effector at the next time;

[0011] Step 5, performing the spraying path planning based on the improved biologically inspired neural network full-coverage spraying path planning algorithm of steps 1-4;

[0012] Step 6, when the robot end effector falls into the dead zone, suspending the execution of the improved biologically inspired neural network full-coverage spraying path planning algorithm and performing the escape by using the biologically inspired neural network point-to-point path planning algorithm until the robot end effector reaches the unsprayed region;

[0013] Step 7, continuing to execute the improved biologically inspired neural network full-coverage spraying path planning algorithm of step 5 until there is no uncovered grid cell in the grid map, and generating the planning path.

[0014] As a further improved technical solution of the application, the step 1 specifically comprises:

[0015] 1.1, collecting the vehicle body to be sprayed by using a line laser three-dimensional scanning instrument, obtaining the point cloud data of the vehicle body surface, and then flattening the point cloud data according to the spatial geometric information of the vehicle body to obtain new point cloud data P;

[0016] 1.2, determining the spray gun fog diameter σ and the coating overlap width d0, setting the coating overlap width as d0=σ / 3, and setting the spray point interval width d of the two parallel spraying paths as d=σ-2d0=σ / 3;

[0017] 1.3, constructing a global spraying work map of the vehicle body, the map being composed of discretized grid cells, the size of each grid cell in the grid map being kept consistent, and the size of the grid map being:

[0018]

[0019] wherein, X max (P) represents the maximum value of the point cloud data P along the X axis, X min (P) represents the minimum value of the point cloud data P along the X axis, Y max (P) represents the maximum value of the point cloud data P along the Y axis, Y min (P) represents the minimum value of the point cloud data P along the Y axis, represents the number of grids in the X-axis direction, represents the number of grids in the Y-axis direction;

[0020] 1.4, Each grid cell in the grid map represents the surface of the vehicle body; the vehicle windows, doors and vehicle body defect areas are regarded as non-spraying areas, and the grid cells (m, n) of the non-spraying areas are set as -1, i.e. obstacles; the grid cells (m, n) of the spraying areas are set as 0, i.e. uncovered grids, and the grid cells (m, n) of the planned paths are set as 1, i.e. covered grids;

[0021] 1.5, The robot end effector is a spray gun, and the path movement of the spray gun on each grid cell needs to spend 1 unit time, and the entry position of the spray gun in the grid map is determined and the initial moving direction

[0022]

[0023] wherein, represents the grid cell where the spray gun is at time t, t≥0, represents the grid cell where the spray gun is at the initial time, is the grid cell where the spray gun is at the next time, at this time, the representation property of the grid cell (m, n) (0) and the representation property of the grid cell (m, n) (1) are set to 1, i.e. covered grids;

[0024] 1.6, The grid cell is regarded as an independent neuron N, and each neuron has a topological connection relationship with the adjacent neurons;

[0025] 1.7, The activity value change of each neuron is represented by a shunt formula:

[0026]

[0027] wherein, x i represents the activity value of the i-th neuron, x j represents the activity value of the j-th neuron in the neighborhood neurons around the i-th neuron, A, B and D are non-negative constants, A represents the decay rate, B represents the upper limit of the neuron activity value, and D represents the lower limit of the neuron activity value; wherein [I i ] - represents the inhibitory excitation, represents the excitatory excitation; I i represents the external input of the i-th neuron, which is determined by the environmental information, and is defined as:

[0028]

[0029] where F(m i ,n i ) represents the characteristic property of the grid cell (m i ,n i ) corresponding to the i-th neuron;

[0030] The functions [a] + and [a] - are linear threshold functions defined as [a] + = max{a, 0} and [a] - = max{-a, 0}, respectively; w ij = f(d ij ), d ij represents the Euclidean distance between the i-th neuron and the j-th neuron, E is a constant value, E » B, each neuron only calculates the neurons in the range of (0, r0), r0 is defined as the receptive field; f(d ij ) is:

[0031]

[0032] where u is the attenuation coefficient;

[0033] 1.8, set all neuron activity values in the grid map to 0, and update the neuron activity values of all grid cells in the grid map once using formula (3).

[0034] As a further improved technical solution of the application, the step 2 specifically comprises:

[0035] 2.1, limit the spray gun to move only horizontally and vertically, and reduce the size of the neuron receptive field r0, and modify the number of neighboring neurons to be judged to 4;

[0036] 2.2, introduce a feedback adjustment coefficient, the calculation formula is as follows:

[0037]

[0038] where w sj = |N s -N j | represents the Euclidean distance between the s-th neuron N s and the j-th neuron N j , wherein the neuron F c is a normal number, used to control the feedback intensity, is the characteristic property of the grid cell (m j ,n j) is the activity value of the jth neuron; x j s F(m s ,n s ) is the characteristic property of the grid cell (m s ,n s ) corresponding to the s th neuron in the neighborhood of the j th neuron;

[0039] 2.3, when the surrounding environment information of two neurons is the same, the same activity value is calculated, resulting in the inability to determine the optimal point position; according to the initial moving direction of the spray gun a steering function based on the sequential position is designed:

[0040]

[0041] wherein, O c is a non-negative constant; S j is the sequential number of the j th neuron in the current neuron neighborhood set, which is defined as:

[0042]

[0043] wherein, is determined by , and the two vectors are perpendicular, denotes the vector between the i th neuron and the j th neuron; the coefficients c j and Y j are substituted into the path point selection function, and finally the decision formula of the path optimal point is obtained:

[0044]

[0045] 2.4, the grid cell (m, n) (2) that the spray gun can possibly move at the next moment is calculated based on formula (12).

[0046] As a further improved technical solution of the present application, the step 3 specifically comprises:

[0047] 3.1, the characteristic property of the grid cell (m, n) (2) is set to -1, that is, an obstacle;

[0048] 3.2, the characteristic properties of each row of grid cells in the grid map are sequentially searched, when an un-covered grid cell with the same characteristic property is found, the search is continued until a grid cell with a different characteristic property is found;the continuously searched grid cells with the same characteristic property are encoded as a connected sequence r ik (s, e), abbreviated as r i k s is a grid cell position index on the i-th row of the grid map, e is a grid cell position index on the i-th row of the grid map, k is the index value of the connected sequence r on the i-th row of the grid map, the connected sequence set of the i-th row grid cell is expressed as:

[0049]

[0050] The connected sequence coding of the uncovered grid on the grid map is completed, and the connected sequence set of all rows is obtained:

[0051]

[0052] 3.3, starting from the connected sequence set of the first row of the grid map When the connected sequence set that is not empty set appears , L=i+1 is taken as the starting row of the subsequent region connectivity test; for the first connected sequence r i 1 the label γ is assigned, γ=0; then, the remaining connected sequences in are assigned the label γ respectively, and γ=γ+1 every time a connected sequence is assigned;

[0053] 3.4, determine whether L exceeds the number of rows of the grid map If it exceeds, skip steps 3.5 to 3.8, and execute step 3.9; otherwise, execute step 3.5;

[0054] 3.5, extract the connected sequences in set in turn wherein denotes the cardinality of set ;

[0055] 3.6, create an empty set T i+1 ; extract each connected sequence r i j in set and perform connectivity determination with , if indicates that r i j encoded grid cell and encoded grid cell are not connected, assign the label γ, and γ=γ+1; otherwise, r ij Add the label to set T i+1 Then, a connected sequence is constructed using formula (15). Connected parent node structure tree

[0056]

[0057] Among them, Ξ(r) i j ) represents a connected sequence r i j The connected parent node structure tree, i.e., the connected sequence r i j The set of position indices of all connected sequences; {i+1,k} represents a connected sequence. Location index; via structure tree It can be traced back to rows 1 through i and All connected sequences;

[0058] 3.7 When set T i+1 When the cardinality is greater than 1, it indicates that there are more than two connected sequence encoded raster units in the i-th row. There is connectivity between the encoded raster cells, in which case it is necessary to set T. i+1 The smallest label value in is assigned Right now Then through query Connected parent node structure tree Update the labels of all connected sequences included:

[0059]

[0060] Where S represents the connected parent node structure tree. The set of position indices of a connected sequence, for example, S = {H1, C1, H2, C2, ..., H} n C n}, where C n H represents the index of a connected sequence. n The index of the row containing the connected sequence allows us to locate the connected sequence. Represents a connected sequence The labels; S(2t) represents the number at the 2t-th position in set S, and S(2t-1) represents the number at the (2t-1)-th position in set S;

[0061] When set T i+1 When the radix is ​​1, it means that there is only one connected sequence encoded raster cell in the i-th row. The encoded raster cells have connectivity, so it is only necessary to set T. i+1 The elements in the middle are assigned to

[0062] 3.8. i = i + 1, and then L = i + 1 is used as the starting row for subsequent regional connectivity checks, and step 3.4 is executed;

[0063] 3.9 Obtain the set of labels for all connected sequences in the raster map. If the number of elements in the set is not 1 after removing duplicate elements, it means that the uncovered grid cells in the grid map are not connected. The connectivity of the grid map needs to be rechecked. This is done by backpropagating the labels based on the connected parent node structure tree of the connected sequences. Specifically, for each connected sequence r in the set R, a conditional judgment is made. When the label of r is not 0 and there is a connected sequence with a label of 0 in the connected parent node structure tree of r, 0 is assigned to r and all connected sequences in the connected parent node structure tree.

[0064] 3.10 Calculate the set without labels The number of elements in the set is obtained by removing duplicate elements.

[0065] 3.11 Grid cell (m,n) (2) The characterization property is set to 0, which means it is an uncovered raster.

[0066] As a further improvement to the present invention, step 4 specifically includes:

[0067] 4.1 Based on the calculation results of step 3.10 To determine whether uncovered raster areas in a raster map are connected, if If the value is not 1, it indicates that the connection is not established, and steps 4.2 to 4.3 are executed.

[0068] if If the value is 1, it indicates connectivity. In this case, skip steps 4.2 to 4.3 and directly use the grid cells (m,n) that the spray gun may move at the next moment, calculated in step 2.4. (2) As the grid unit that the spray gun actually moves in the next moment, it is added to the path set. Grid cell (m,n) (2) The characterization property is set to 1, which means that the raster is covered;

[0069] 4.2. Calculate the grid cells (m, n) that the spray gun may move at the next moment, as obtained in step 2.4. (2) That is, the grid cell (m,n) (2) The characterization property is set to -1, which means it is an obstacle;

[0070] 4.3, obtaining the position (m, n) of the spray gun at the previous time (1) 4-neighborhood neuron and the 4-neighborhood neuron consisting of the grid cells whose characteristic property is 0

[0071] If the set is not empty, record the position (m, n) (1) of the neuron whose activity value is x * Update all the activity values of the neurons in the grid map using formula (3), and calculate the grid cell that the spray gun can move to at the next time using formula (12) Assign the recorded activity value x * of the neuron to the neuron at position (m, n) (1) , and take the position as the grid cell that the spray gun actually moves to at the next time, and add it to the path set Set the characteristic property of the grid cell to 1, i.e. as a covered grid, and set the characteristic property of the grid cell (m, n) (2) to 0, i.e. as an uncovered grid.

[0072] If the set is empty, take the grid cell (m, n) (2) that the spray gun can move to at the next time calculated in step 2.4 as the grid cell that the spray gun actually moves to at the next time, and add it to the path set , and set the characteristic property of the grid cell (m, n) (2) to 1, i.e. as a covered grid.

[0073] As a further improved technical solution of the present application, the step 5 specifically comprises:

[0074] 5.1, updating all the activity values of the neurons in the grid map using formula (3);

[0075] 5.2, calculating the grid cell that the spray gun can move to at the next time using formula (12);

[0076] 5.3, calculating the connectivity of the uncovered grid area in the grid map using steps 3.1 to 3.11;

[0077] 5.4, calculating the grid cell that the spray gun actually moves to at the next time using steps 4.1 to 4.3.

[0078] The present application has the following beneficial effects:

[0079] (1) In order to solve the problem of poor path in the full coverage path planning task of the biological stimulation neural network, the feedback adjustment coefficient is established according to the activity value of the decay neuron in the perception domain of the central neuron, which can effectively solve the problem that the path cannot be completely covered at the narrow area, and a simple and coherent path is planned.

[0080] (2) The turning parameter term in the moving path judgment formula in the original algorithm is modified, which can solve the problem of the same activity value of the neighbor of the central neuron and select the optimal point of the path.

[0081] (3) The sequence repetition value coding technology is used for the grid of the uncovered area, a fast detection algorithm of regional connectivity is designed, and the weakness of the biological stimulation neural network algorithm in global planning is made up. The regional decomposition avoidance mechanism based on the algorithm greatly reduces the probability of the dead zone. BRIEF DESCRIPTION OF DRAWINGS

[0082] Figure 1 It is a planar expansion effect diagram of a side wall plate of a rail passenger car.

[0083] Figure 2 It is a construction of a grid map.

[0084] Figure 3 It is a schematic diagram of coating overlapping cumulative deposition model and grid division.

[0085] Figure 4 It is a neighbor neuron structure of the central neuron N i .

[0086] Figure 5 It is a coding process of the first four rows of the grid map.

[0087] Figure 6 It is a label transmission process of the connected sequence.

[0088] Figure 7 It is the path planning result of the improved biological stimulation neural network on the side wall plate of the rail passenger car.

[0089] Figure 7 In (a), it is the path planning result of the high-resolution grid map.

[0090] Figure 7 In (b), it is the path planning result of the low-resolution grid map.

[0091] Figure 8 It is the path planning result of the improved biological stimulation neural network on the complex side wall plate of the passenger car. DETAILED DESCRIPTION

[0092] The specific embodiments of the application will be further described below according to the drawings:

[0093] A full-coverage putty spraying path planning method for rail passenger car side wall plates, comprising:

[0094] Step 1: Obtain the three-dimensional point cloud data of the rail passenger car side wall plate, extract the base surface information and construct a grid map, place a neuron at each grid and initialize it;

[0095] Step 2, introduce feedback adjustment coefficient and steering function based on sequential position, improve the path point selection function, and then construct the decision strategy of path optimal point, according to the decision strategy of path optimal point, get the grid unit that the robot end effector can move at the next time;

[0096] Step 3, adopt sequence repetition value coding technology to detect the connectivity of the uncovered grid units in the grid map;

[0097] Step 4, establish regional decomposition risk avoidance mechanism; determine whether the uncovered grid region in the grid map is connected according to the result obtained in step 3; if not, obtain the grid unit that the robot end effector actually moves at the next time according to the established regional decomposition risk avoidance mechanism; if connected, directly take the grid unit that the robot end effector can move at the next time calculated in step 2 as the grid unit that the robot end effector actually moves at the next time;

[0098] Step 5, full-coverage spraying path planning algorithm based on improved biologically inspired neural network of steps 1-4 is carried out;

[0099] Step 6, when the robot end effector is trapped in the dead zone (when the robot end effector has no uncovered grid unit to choose from in the next time, i.e. it has entered the dead zone), pause the full-coverage spraying path planning algorithm based on improved biologically inspired neural network, and use the existing point-to-point path planning algorithm based on biologically inspired neural network to escape until the robot end effector reaches the unsprayed area;

[0100] Step 7, continue to execute the full-coverage spraying path planning algorithm based on improved biologically inspired neural network described in step 5 until there is no uncovered grid unit in the grid map, and generate the planning path as shown in Figure 7 or 8.

[0101] Wherein, the step 1 specifically comprises:

[0102] 1.1, adopt line laser three-dimensional scanning instrument to collect the to-be-sprayed car body, obtain the point cloud data of the car body surface, and then flatten the point cloud data according to the spatial geometric information of the car body to obtain new point cloud data P, as shown in Figure 1

[0103] ​1.2, determine the spray gun fog diameter σ and the coating overlap width d0, as shown in Figure 3 The coating overlap width is set to d0=σ / 3 according to the coating superposition principle and the putty deposition model, and the spray point interval width d of the two parallel spraying paths is d=σ-2d0=σ / 3;

[0104] 1.3, build a global spraying work map (lattice map, edge filling a circle of lattice units, indicating obstacles) of the vehicle body, which is fully covered, Figure 2 The lattice map of the spraying area is hidden for the convenience of subsequent path drawing and observation. The map is composed of discrete lattice units, and the size of each lattice in the lattice map remains the same. The size of the lattice map is:

[0105]

[0106] Where, X max (P) represents the maximum value of point cloud data P along the X axis, X min (P) represents the minimum value of point cloud data P along the X axis, Y max (P) represents the maximum value of point cloud data P along the Y axis, Y min (P) represents the minimum value of point cloud data P along the Y axis, represents the number of lattices in the X axis direction, represents the number of lattices in the Y axis direction;

[0107] 1.4, each lattice unit in the lattice map represents the surface of the vehicle body; the vehicle window, door and vehicle body defect area are all non-spraying areas. The characteristic property F(m,n) of the lattice unit (m,n) of the non-spraying area is set to-1, i.e. obstacle. The characteristic property F(m,n) of the lattice unit (m,n) of the to-be-sprayed area is set to 0, i.e. uncovered grid. The characteristic property F(m,n) of the lattice unit (m,n) that has been path planned is set to 1, i.e. covered grid;

[0108] 1.5, the robot end effector is a spray gun, and the path movement of the spray gun on each lattice unit needs to spend 1 unit time. The entry position of the spray gun in the lattice map is determined and the initial moving direction

[0109]

[0110] Where, represents the lattice unit where the spray gun is at time t, t≥0, represents the lattice unit where the spray gun is at the initial time, is the lattice unit where the spray gun is at the next time, at this time, the lattice unit (m,n)(0) Characterization properties and raster cells (m,n) (1) The characterization properties are all set to 1, which means that the grid is covered; where m in the grid cell (m,n) represents the horizontal coordinate position of the grid cell and n represents the vertical coordinate position of the grid cell.

[0111] 1.6 Treat each grid cell as an independent neuron N, with each neuron having a topological connection with its neighboring neurons;

[0112] 1.7 The activity value of a neuron is affected by other neurons within the receptive field, manifesting as excitation or inhibition. The change in activity value of each neuron is expressed using the shunt formula:

[0113]

[0114] Where, x i x represents the activity value of the i-th neuron. j Let A, B, and D represent the activity value of the j-th neuron in the neighborhood of the i-th neuron, where A, B, and D are non-negative constants, A represents the decay rate, B represents the upper limit of neuron activity value, and D represents the lower limit of neuron activity value; where [I i ] - This is represented as an inhibitory incentive. This is represented as excitatory stimulation; I i The external input of the i-th neuron, determined by environmental information, is defined as:

[0115]

[0116] Where F(m) i ,n i ) represents the grid cell (m) corresponding to the i-th neuron. i ,n i The characterization properties of );

[0117] Function [a] + and [a] - These are linear threshold functions, defined as [a], [b], [c], [d], [e], [b] + =max{a,0} and [a] - =max{-a,0};w ij =f(d ij ), d ij This represents the Euclidean distance between the i-th and j-th neurons, where E is a constant value, E >> B. For each neuron, only neurons within the interval (0, r0) are calculated, and r0 is defined as the receptive field; f(d ij )for:

[0118]

[0119] wherein u is an attenuation coefficient;

[0120] 1.8, set all the neuron activity values in the grid map to 0, and update the neuron activity values of all grid cells in the grid map once using formula (3).

[0121] wherein the step 2 specifically comprises:

[0122] 2.1, in the spraying process, if a diagonal spraying path appears, the local coating thickness will be increased, therefore, in order to meet the requirements of the spraying process, it is limited that the spray gun can only move horizontally and vertically, and at the same time, the size of the neuron receptive field r0 is reduced, and the number of adjacent neurons is modified to 4, as shown in Figure 4 ;

[0123] 2.2, the neuron activity value of the covered grid cell will gradually decrease with the increase of the time step, and there is a certain degree of stimulation effect on the neurons in the adjacent area; the neuron activity of the obstacle grid cell is a small positive value, which will be repelled and moved away during path planning. By introducing a feedback adjustment coefficient based on this characteristic, the calculation formula is as follows:

[0124]

[0125] wherein w sj = |N s -N j | represents the Euclidean distance between the s-th neuron N s and the j-th neuron N j , wherein the neuron F c is a normal number for controlling the feedback strength, is an eight-neighborhood neuron set centered on the grid cell (m j ,n j ) corresponding to the j-th neuron, x j represents the activity value of the j-th neuron; x s represents the activity value of the s-th neuron in the neighborhood of the j-th neuron; F(m s ,n s ) represents the characteristic property of the grid cell (m s ,n s ) corresponding to the s-th neuron;

[0126] 2.3, when the surrounding environment information of two neurons is the same, the same activity value will be calculated through formula (3), which leads to the inability to determine the optimal point position; according to the initial moving direction of the spray gun , a turning function based on the sequential position is designed:

[0127]

[0128] where O c is a non-negative constant; S j is the order number of the jth neuron in the current neuron neighborhood set, which is defined as:

[0129]

[0130] where, is determined by , and the two vectors are perpendicular, represents the vector between the ith neuron and the jth neuron; the coefficient c j and Y j are substituted into the path point selection function; wherein the path point selection function is:

[0131]

[0132] where, represents the neuron activity value at the selected moving path point, p n is the next position of the robot movement. c is a normal number, y j is a function related to the movement steering, and the function is defined as:

[0133]

[0134] where Δθ j represents the angle between the current path movement direction θ c and the next possible path movement direction θ j , and the value range is [0, π], and the formula is:

[0135]

[0136] represents the grid position of the spray gun at the current time, represents the grid position of the spray gun at the next time, (x pp , y pp ) represents the grid position of the spray gun at the previous time;

[0137] The decision formula of the path optimal point is finally obtained:

[0138]

[0139] 2.4, based on formula (12), when j is 1, 2, 3, and 4 respectively, the maximum neuron activity value x j +c jThe grid cell corresponding to Y is used as the possible movement path point of the spray gun in the next moment, thus obtaining the grid cell (m,n) that the spray gun may move in the next moment. (2) .

[0140] Specifically, step 3 includes:

[0141] 3.1 Grid cell (m,n) (2) The characterization property is set to -1, which means it is an obstacle;

[0142] 3.2. Sequentially search the representation properties of each row of raster cells in the raster map. When a raster cell with the representation property of "uncovered" is found... Continue searching until a raster cell with different characterization properties is found. The consecutive raster cells with the same representational properties retrieved by encoding are connected sequences r. i k (s,e), abbreviated as r i k ,like Figure 5 As shown, s is a grid cell. The position index on the i-th row of the raster map, where e is the raster cell. The position index in the i-th row of the raster map, k is the index value of the connected sequence r in the i-th row of the raster map, and the set of connected sequences of the raster cells in the i-th row. Represented as:

[0143]

[0144] Complete the encoding of the connected sequences of uncovered rasters on the raster map, thereby obtaining the set of connected sequences for all rows:

[0145]

[0146] 3.3. The set of connected sequences from the first row of the raster map Begin the check when a set of connected sequences that is not empty is found. When L = i + 1 is used as the starting row for subsequent regional connectivity tests; for The first connected sequence r in i 1 Assign the label γ, γ = 0; then, respectively... The remaining connected sequences are labeled with γ, and for each connected sequence assigned a value, γ = γ + 1;

[0147] 3.4 Determine if L exceeds the number of rows in the raster map. If the condition is exceeded, skip steps 3.5 to 3.8 and proceed to step 3.9; otherwise, proceed to step 3.5.

[0148] 3.5, Extracting connected sequences from set where denotes the cardinality of set

[0149] 3.6, Creating empty set T i+1 ; Extracting each connected sequence r i j from set and performing connectedness determination, if (i.e. if there is no subset a in the kth connected sequence r in the i+1th row that is equal to any subset b in the jth connected sequence r i j in the i-th row, which means that the two connected sequences in adjacent rows are not connected), it means that r i j If neither the encoded grid cell of r nor the encoded grid cell of r is connected, assign label γ to r and γ = γ + 1; otherwise, add the label of r i j to set T i+1 , and then construct the connected parent node structure tree of connected sequence r

[0150]

[0151] where Ξ(r i j ) denotes the connected parent node structure tree of connected sequence r i j , i.e. the position index set of all connected sequences that are connected to connected sequence r i j ; {i+1, k} denotes the position index of connected sequence r ; through structure tree , all connected sequences that are connected to r in rows 1 to i can be traced back;

[0152] 3.7, When the cardinality of set T i+1 is greater than 1, it means that there are more than two connected sequences in the i-th row whose encoded grid cells are connected to the encoded grid cell of r , at this time, the smallest label value in set T i+1 is assigned to r , i.e. Then, by querying the connected parent node structure tree of r ​​​​Update the labels of all connected sequences included, such as Figure 6 As shown:

[0153]

[0154] Where S represents the connected parent node structure tree. The set of position indices of a connected sequence, for example, S = {H1, C1, H2, C2, ..., H} n C n}, where C n H represents the index of a connected sequence. n The index of the row containing the connected sequence allows us to locate the connected sequence. Represents a connected sequence The labels; S(2t) represents the number at the 2t-th position in set S, and S(2t-1) represents the number at the (2t-1)-th position in set S;

[0155] When set T i+1 When the radix is ​​1, it means that there is only one connected sequence encoded raster cell in the i-th row. The encoded raster cells have connectivity, so it is only necessary to set T. i+1 The elements in the middle are assigned to

[0156] 3.8. i = i + 1, and then L = i + 1 is used as the starting row for subsequent regional connectivity checks, and step 3.4 is executed;

[0157] 3.9 Obtain the set of labels for all connected sequences in the raster map. If removing duplicate elements from the set results in a set with a number of elements other than 1 (i.e., assuming...) Removing duplicate elements at this point yields {0, 1, 2}, with a total of 3 elements. This indicates that uncovered raster cells in the raster map are not connected, requiring a re-check of the raster map's connectivity. This involves backpropagation of labels based on the connected parent node structure tree of the connected sequence, specifically by checking the set... For each connected sequence r in the algorithm, a condition is evaluated. If the label of r is not 0 and there is a connected sequence with the label 0 in the connected parent node structure tree of r, then 0 is assigned to r and all connected sequences in the connected parent node structure tree.

[0158] 3.10 Calculate the set without labels The number of elements in the set is obtained by removing duplicate elements.

[0159] 3.11 Grid cell (m,n) (2) The characterization property is set to 0, which means it is an uncovered raster.

[0160] wherein the step 4 specifically comprises:

[0161] 4.1, according to the calculation result of step 3.10 to determine whether the uncovered grid area in the grid map is connected, if is not 1, it means not connected, and steps 4.2 to 4.3 are executed;

[0162] if is 1, it means connected, then steps 4.2 to 4.3 are skipped, and the grid cell (m, n) calculated by step 2.4 as the next time possible movement of the spray gun (2) is taken as the next time actual movement of the spray gun, and is added to the path set (connected means that the next time spray gun position calculated by step 2.4 will not cause dead zone, and the absence of dead zone ensures that the planned spray gun movement path will not appear overlapping path, so it can be directly added to the path set), the characteristic property of grid cell (m, n) (2) is set to 1, that is, the covered grid;

[0163] 4.2, the grid cell (m, n) calculated by step 2.4 as the next time possible movement of the spray gun (2) , that is, the characteristic property of grid cell (m, n) (2) is set to -1, that is, the obstacle;

[0164] 4.3, the position (m, n) of the previous time spray gun (1) is taken as the center of the 4-neighborhood neuron and the 4-neighborhood neuron is taken as the set of neurons

[0165] if the set is not empty, the neuron activity value of position (m, n) (1) is recorded as x * , all neuron activity values in the grid map are updated by formula (3), and the next time possible movement of the spray gun is calculated by formula (12) The recorded neuron activity value x * is assigned to the neuron at position (m, n) (1) , and the position is taken as the next time actual movement of the spray gun, and is added to the path set The characteristic property of grid cell is set to 1, that is, the covered grid, and the characteristic property of grid cell (m, n) (2)The characteristic property of the grid cell (m, n) is set to 0, i.e. the uncovered grid;

[0166] If the set is empty, the grid cell (m, n) calculated in step 2.4 is taken as the grid cell actually moved by the spray gun at the next time and is added to the set of paths (2) The characteristic property of the grid cell (m, n) is set to 1, i.e. the covered grid. (2) The characteristic property of the grid cell (m, n) is set to 1, i.e. the covered grid.

[0167] The step 5 specifically comprises:

[0168] 5.1, updating all neuron activity values in the grid map by using formula (3);

[0169] 5.2, calculating the grid cell possibly moved by the spray gun at the next time by formula (12);

[0170] 5.3, calculating the connectivity of the uncovered grid region in the grid map by steps 3.1 to 3.11;

[0171] 5.4, calculating the grid cell actually moved by the spray gun at the next time by steps 4.1 to 4.3.

[0172] The protection scope of the present application includes but is not limited to the above embodiments, the protection scope of the present application is subject to the claims, any replacement, deformation, improvement of the present technology easily thought by the skilled in the art falls into the protection scope of the present application.​

Claims

1. A method for planning the path of full-coverage putty spraying for the side wall panels of rail passenger vehicles, characterized in that, include: Step 1: Obtain the 3D point cloud data of the side wall panel of the rail passenger vehicle, extract the base surface information and construct a grid map. Place a neuron at each grid and initialize it. Step 2: Introduce feedback adjustment coefficient and sequence position-based steering function to improve path point selection function, and then construct decision strategy for optimal path point. Based on the decision strategy for optimal path point, obtain the grid cell that the robot end effector may move in the next moment. Step 3: Use sequential repeat value encoding technology to detect the connectivity of uncovered raster cells in the raster map; Step 4: Determine whether the uncovered grid areas in the grid map are connected based on the results obtained in Step 3. If they are not connected, execute Steps 4.2 to 4.3 to obtain the grid unit that the robot end effector will actually move in the next moment according to the established region decomposition risk avoidance mechanism. If they are connected, directly use the grid unit that the robot end effector may move in the next moment calculated in Step 2 as the grid unit that the robot end effector will actually move in the next moment. 4.

2. The grid cells that the spray gun may move at the next moment, calculated in step 2.

4. , i.e., grid unit The characterization property is set to -1, which means it is an obstacle; 4.3 Obtain the spray gun position at the previous moment 4-neighbor neurons centered And 4 neighboring neurons A set of neurons is composed of grid units with a representation property of 0. ; If set Not an empty set, record position Neuron activity value The activity values ​​of all neurons in the grid map are updated using formula (3), and the grid cells that the spray gun may move at the next moment are calculated using formula (12). The recorded neuron activity values Assign a value to a position The neurons at that location will As the grid unit that the spray gun actually moves in the next moment, it is added to the path set. Grid unit The characterization property is set to 1, which means that the raster is covered, and the raster cell is... The characterization property is set to 0, which means it is an uncovered raster; If set If the set is empty, then the grid cells that the spray gun may move at the next moment, calculated in step 2.4, will be used. As the grid unit that the spray gun actually moves in the next moment, it is added to the path set. and grid cells The characterization property is set to 1, which means that the raster is covered; Step 5: Perform spraying path planning based on the improved biologically stimulated neural network full-coverage spraying path planning algorithm of Steps 1-4; Step 6: When the robot end effector gets stuck in the dead zone, pause the execution of the full-coverage spraying path planning algorithm based on the improved bio-stimulated neural network, and use the point-to-point path planning algorithm of the bio-stimulated neural network to escape until the robot end effector reaches the unsprayed area. Step 7: Continue executing the full-coverage spraying path planning algorithm based on the improved biologically stimulated neural network described in Step 5 until there are no uncovered grid cells in the grid map, and generate a planned path; Step 1 specifically includes: 1.

1. A line laser 3D scanner is used to acquire point cloud data of the vehicle body to be painted. After obtaining the point cloud data of the vehicle body surface, the point cloud data is flattened according to the spatial geometry information of the vehicle body to obtain new point cloud data. ; 1.2 Determine the spray nozzle diameter Coating overlap width Set the coating overlap width to The spray dot spacing width between two parallel spraying paths ; 1.3 Construct a global painting work map for the vehicle body. The map consists of discretized grid cells, with each cell maintaining a consistent size. The grid map size is: ； in, Representing point cloud data along The maximum value of the axis. Representing point cloud data along Minimum value of the axis, Representing point cloud data along The maximum value of the axis. Representing point cloud data along Minimum value of the axis, express The number of grid cells along the axial direction. express The number of grid cells along the axial direction; 1.

4. Each grid cell in the grid map represents the vehicle body surface; windows, doors, and defective areas are all treated as unpainted areas, and the grid cells for unpainted areas... The characterization property is set to -1, which means it is an obstacle; the grid unit of the area to be sprayed. The characterization property is set to 0, which means it is an uncovered raster cell that has been path-planned. The characterization property is set to 1, which means that the raster is covered; 1.

5. The robot's end effector is a spray gun. It is stipulated that the spray gun's path movement on each grid cell takes one unit of time. Determine the spray gun's entry position in the grid map. and initial movement direction : ； in, This represents the grid cell where the spray gun is located at time t. , , , This indicates the grid cell where the spray gun is located at the initial moment. For the grid cell where the spray gun will be located at the next moment, at this time, the grid cell... Characterization properties and raster units The characterization properties are all set to This indicates that the grid has been covered. 1.6 Treating grid units as independent neurons Each neuron has a topological connection with its neighboring neurons; 1.7 The change in activity value of each neuron is expressed by the shunt formula: ； in, This represents the activity value of the i-th neuron. This represents the activity value of the j-th neuron among the neighboring neurons of the i-th neuron. It is a non-negative constant. Indicates the attenuation rate. This indicates the upper limit of neuron activity. Indicates the lower limit of neuronal activity; where This is represented as an inhibitory incentive. This is represented as excitatory motivation; The external input of the i-th neuron, determined by environmental information, is defined as: ； in, This represents the grid cell corresponding to the i-th neuron. Characteristic properties; function and It is a linear threshold function, defined as follows: and ; , Indicates the first The first neuron and the second The Euclidean distance between neurons, where E is a constant value. Each neuron only calculates Neurons within the interval range, Defined as a receptive field; for: ； in, The attenuation coefficient; 1.8 Set the activity value of all neurons in the raster map to 0, and update the neuron activity value of all raster cells in the raster map once using formula (3); Step 2 specifically includes: 2.

1. Limit the spray gun to only horizontal and vertical movement, and simultaneously reduce the size of the neuronal receptive field. The number of judgments for neighboring neurons is changed to 4; 2.2 Introducing the feedback adjustment coefficient, the calculation formula is as follows: ； in, Indicates the first one neuron With the one neuron The Euclidean distance between neurons , It is a positive constant used to control the feedback strength. It is the grid unit corresponding to the j-th neuron The central eight-neighbor set of neurons This represents the activity value of the j-th neuron; This represents the activity value of the s-th neuron among the neighboring neurons of the j-th neuron; This represents the grid unit corresponding to the s-th neuron. Characteristic properties; 2.3 When two neurons have the same environmental information, they will calculate the same activity value, making it impossible to determine the optimal location; based on the initial movement direction of the spray gun. Design a turn function based on the sequence bit: ； in, It is a non-negative constant; For the first The sequential number of a neuron in the current neuron's neighborhood set is defined as follows: ； in, Depend on The decision is made, and the two vectors are perpendicular. Indicates the first The first neuron and the second The vector between neurons; the coefficients and Substituting these values ​​into the path selection function, we obtain the final decision formula for the optimal path point: ； 2.

4. Based on formula (12), the grid cells that the spray gun may move at the next moment are calculated. .

2. The method for full-coverage putty spraying path planning for the side wall panels of rail passenger vehicles according to claim 1, characterized in that, Step 3 specifically includes: 3.1 Grid Unit The characterization property is set to -1, which means it is an obstacle; 3.

2. Sequentially search the representation properties of each row of raster cells in the raster map. When a raster cell with the representation property of "uncovered" is found... Continue searching until a raster cell with different characterization properties is found. The consecutive raster cells with the same characterizing properties retrieved by encoding are connected sequences. abbreviation , For grid cells In the grid map The position index on the row, For grid cells In the grid map The position index on the row, For a connected sequence r in a raster map, the first... The index value on the row, the first The set of connected sequences of row raster cells Represented as: ； Complete the encoding of the connected sequences of uncovered rasters on the raster map, thereby obtaining the set of connected sequences for all rows: ； 3.

3. The set of connected sequences from the first row of the raster map Begin the check when a set of connected sequences that is not empty is found. At that time, As the starting line for subsequent regional connectivity checks; The first connected sequence in Assignment label , =0; then, respectively for Label the remaining connected sequences in The assignment, and for each connected sequence assigned, ; 3.

4. Judgment Does it exceed the number of rows in the raster map? If the condition is exceeded, skip steps 3.5 to 3.8 and proceed to step 3.9; otherwise, proceed to step 3.

5. 3.5 Extracting sets sequentially Connected sequences in , ,in Represents a set The cardinality; 3.6 Creating an empty collection Extracting the set Each connected sequence in and Perform connectivity determination, if ,illustrate Encoded raster units and If the encoded raster cells are not connected, then give Assignment label ,and Otherwise, Add the label to the collection Then, a connected sequence is constructed using formula (15). Connected parent node structure tree : ； in, Represents a connected sequence The connected parent node structure tree, i.e., the connected sequence The set of position indices of all connected sequences; Represents a connected sequence Location index; via structure tree It can be traced back to In the middle of the line and All connected sequences; 3.7 When the set When the base number is greater than 1, it indicates that the first... A raster cell in a row that contains two or more connected sequence encodings and There is connectivity between the encoded raster cells, in which case the set needs to be... The smallest label value in is assigned ,Right now Then through query Connected parent node structure tree Update the labels of all connected sequences included: ； in, Represents a connected parent node structure tree The set of position indices of a connected sequence, for example ,in Indices representing connected sequences The index of the row containing the connected sequence allows us to locate the connected sequence. ; Represents a connected sequence The label; This indicates taking the number at the 2t-th position in set S. This represents the number taken at the (2t-1)th position in set S; When set When the base is 1, it means that the first... There is only one connected sequence encoded raster cell in the row and The encoded raster cells have connectivity, so it is only necessary to set... The elements in the middle are assigned to ; 3.8, i = i + 1, then... As the starting line for subsequent regional connectivity checks, proceed to step 3.4; 3.9 Obtain the set of labels for all connected sequences in the raster map. If removing duplicate elements results in a set with a non-zero number of elements, it indicates that the uncovered raster cells in the raster map are not connected. The connectivity of the raster map needs to be rechecked, specifically by backpropagating the labels based on the connected parent node structure tree of the connected sequence. For each connected sequence r in the equation, a conditional judgment is made, when... The label is not 0 and If a connected sequence with label 0 exists in the connected parent node structure tree, assign 0 to it. And all connected sequences in the parent node's tree structure; 3.10 Calculate the set without labels The number of elements in the set is obtained by removing duplicate elements. ; 3.11 Grid Unit The characterization property is set to 0, which means it is an uncovered raster.

3. The method for full-coverage putty spraying path planning for the side wall panels of rail passenger vehicles according to claim 2, characterized in that, In step 4 mentioned above include: 4.1 Based on the calculation results of step 3.10 To determine whether uncovered raster areas in a raster map are connected, if If the value is not 1, it indicates that the connection is not made; if If the value is 1, it indicates connectivity, and the grid cells that the spray gun may move at the next moment, calculated in step 2.4, can be directly used. As the grid unit that the spray gun actually moves in the next moment, it is added to the path set. Grid unit The characterization property is set to 1, which means that the raster is covered.

4. The method for full-coverage putty spraying path planning for the side wall panels of rail passenger vehicles according to claim 3, characterized in that, Step 5 specifically includes: 5.

1. Update the activity values ​​of all neurons in the raster map using formula (3); 5.2 Calculate the grid cells that the spray gun may move at the next moment using formula (12); 5.3 Calculate the connectivity of uncovered raster areas in the raster map using steps 3.1 to 3.11; 5.

4. Calculate the actual grid cell movement of the spray gun at the next moment using steps 4.1 to 4.3.

Images