Distributing machine intelligent path planning method based on evolution constraint optimization

Abstract

The invention discloses a distribution machine intelligent path planning method based on evolutionary constraint optimization, which is used for planning a global path of a distribution machine working in a work target area map, and comprises the following steps: obtaining a plurality of sub-grid areas and obstacles divided by the work target area map; and obtaining an optimal access sequence of the sub-grid regions based on a genetic algorithm, obtaining sub-paths in the sub-grid regions based on a genetic annealing algorithm, and connecting all the sub-paths to obtain a global path. According to the invention, by means of the idea of the evolutionary algorithm, the path planning efficiency and the execution flexibility of the material distributing machine in a complex operation environment are remarkably improved, so that the material distributing machine can achieve optimal performance in both global and local path optimization.

Description

Technical Field

[0001] The present invention relates to the field of path planning for a concrete placing machine, and in particular to a path planning method for planning a global path for a concrete placing machine to operate in a work target area map. Background Art

[0002] In highly complex operating scenarios such as construction and automatic control of concrete placing booms, path planning is the key to ensuring efficient operation of equipment. However, in these scenarios, placing booms often need to perform delicate operations in an irregular working environment with dense obstacles, which poses a severe challenge to path planning. Traditional path planning algorithms are difficult to adapt to such complex environments and cannot guarantee accurate coverage of the entire working area and safe obstacle avoidance.

[0003] There are many deficiencies in the practical application of existing full coverage path planning algorithms in fields such as concrete placing machines and building construction. Although the ox-ploughing method is simple and efficient, it has poor adaptability to irregular areas and has difficulty handling complex obstacles; the spiral method has a compact path, but has low coverage and flexibility in environments with dense boundaries and obstacles; although the random walk method has certain exploratory characteristics, the path planning efficiency is low and the coverage is uneven; the graph-based path planning method can adapt to complex scenarios and achieve global optimization, but the computational complexity is high, especially in large-scale or dynamic environments. It is time-consuming. These deficiencies make it difficult for existing technologies to fully meet the needs in terms of precise coverage, path optimization, and adaptability to complex environments. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides an intelligent path planning method for a concrete placing machine based on evolutionary constraint optimization, which is used to plan the global path of the concrete placing machine operating in a work target area map, comprising the following steps:

[0005] Obtain a number of sub-grid areas and obstacles divided by the work target area map;

[0006] Acquire multiple random sorting sequences composed of sorting of all sub-grid regions, transform the sorting of sub-grid regions in the random sorting sequence by using a transformation operation to generate a variant sorting sequence, repeatedly select a preferred variant sorting sequence from the variant sorting sequence to replace the random sorting sequence, and repeat the transformation operation to obtain an access sequence, wherein, compared with the replaced random sorting sequence, the preferred variant sorting sequence has: fewer crossover times in the path composed of sequence nodes in sorting order, smaller span between two sub-grid regions with continuous sorting order in the sequence, and the sequence nodes are sub-grid regions;

[0007] Get the sub-path in each sub-grid area that covers the entire sub-grid area;

[0008] Reconstructing the sub-path to obtain an obstacle avoidance path, where the obstacle avoidance path can avoid obstacles and be separated from the obstacles by a safe distance;

[0009] Traversing the sub-grid areas that are adjacent to each other in sequence according to the access sequence, obtaining a connection path between an end point of a sub-path or obstacle avoidance path in a previous sub-grid area and a starting point of a sub-path or obstacle avoidance path in a subsequent sub-grid area, wherein the connection path has no intersection with the obstacle, and using the connection path to sequentially connect the sub-paths or obstacle avoidance paths that are adjacent to each other to obtain a candidate global path;

[0010] A global path is selected from the candidate global paths, wherein the number of intersections between the connection path of the global path and other connection paths, obstacle avoidance paths and sub-paths is minimal.

[0011] Furthermore, the step of transforming the order of the sub-grid regions in the random ordering sequence to generate a variant ordering sequence by using a transformation operation includes:

[0012] Region exchange: randomly select two sub-grid regions in the random sorting sequence and exchange their positions in the random sorting sequence;

[0013] Interval reversal: select an interval consisting of multiple continuous sub-grid areas in the random sorting sequence, and reverse the order of the sub-grid areas in the interval;

[0014] Neighborhood fine-tuning: swap the order of adjacent sub-grid areas in the random order sequence;

[0015] The transformation operation randomly selects one of the three transformation operations to generate a mutated sorted sequence, and the number of mutated sorted sequences before and after the transformation is equal to the number of random sorted sequences.

[0016] Furthermore, the step of selecting a preferred variant sequence from the variant sequence to replace the random sequence comprises:

[0017] Calculate the objective function of the random sorting sequence and the mutated sorting sequence; wherein, the smaller the objective function is, the fewer the number of crossovers in the path composed of the sorting order of the sequence nodes and / or the smaller the span of the two sub-grid areas with continuous sorting order in the sequence;

[0018] If the objective function of the mutated sorting sequence is smaller than that of the random sorting sequence, the mutated sorting sequence is selected as the preferred mutated sorting sequence to replace the random sorting sequence;

[0019] When the number of repetitions reaches a set number or the number of selected preferred variant sorted sequences converges, the repetition is terminated.

[0020] Furthermore, the objective function includes a path intersection penalty term and a region selection constraint penalty term.

[0021] F(τ)=u·Σ 1≤p<q≤k (intersect(τ0))+λ·Σ 1≤p<q≤k constraint(τ p ,τ q );

[0022] Where: u is the weight of the path crossing penalty, λ is the weight of the region selection constraint, Σ 1≤p<q≤k (intersect(τ0)) represents the number of intersections in the path composed of the sequence node sorting order, constraint(τ p ,τ q ) represents the region selection constraint penalty coefficient, τ p ,τ q is the pth and qth consecutive subgrid regions in the random sorting sequence or the mutated sorting sequence, q = p + 1;

[0023]

[0024]

[0025] τ p ,τ p Adjacent means that the boundary coordinates of two sub-grid areas are adjacent; Δ max is the maximum allowable span, d(τ p ,τ p ) are two sub-grid areas τ p and τ p The distance from the center point of , δ is greater than 1.

[0026] Further, obtaining the sub-path in each sub-grid area that covers the entire sub-grid area includes:

[0027] The sub-grid area is divided into a grid map composed of grid cells using perpendicular longitude and latitude lines; each grid cell without obstacles is marked as passable, and each grid cell with obstacles is marked as impassable;

[0028] Detect each sub-grid area and its next adjacent sub-grid area in sequence according to the access sequence, find the starting edge and the ending edge of each sub-grid area in turn, select the edge of each sub-grid area away from the obstacle as the starting edge, and select the edge of each sub-grid area close to the next sub-grid area as the ending edge;

[0029] A grid unit is selected from a column or a row of grid units near the starting edge of the sub-grid area as the starting point of the sub-path, and the sub-path is extended from a direction parallel to the starting edge in the sub-grid area to form a winding sub-path formed by connecting the path segments end to end, and the adjacent path segments of the sub-path at the bend are perpendicular to each other, and the end point of the sub-path is located at the terminating edge of the sub-grid area. The sub-path includes several rows of path segments parallel to the starting edge, and the sub-path covers the entire sub-grid area and does not generate an intersection in the sub-grid area.

[0030] Further, reconstructing the sub-path to obtain an obstacle avoidance path includes:

[0031] If an intersection portion of the subpath that overlaps or touches a grid unit where an obstacle exists is detected, the intersection portion is offset to obtain an offset path, and the offset path maintains a safe distance from the obstacle;

[0032] Replaces the path segment with an offset path.

[0033] Furthermore, the step of sequentially traversing the sub-grid areas that are adjacent to each other in sequence according to the access sequence to obtain a connection path between an end point of a sub-path or obstacle avoidance path in a previous sub-grid area and a starting point of a sub-path or obstacle avoidance path in a subsequent sub-grid area further includes:

[0034] Expand the grid map of the sub-grid area to the entire work target area map, mark the grid cells without obstacles as passable, and mark the grid cells with obstacles as impassable;

[0035] If it is detected that the connection path passes through an inaccessible grid unit, the connection path is shifted to bypass the inaccessible grid unit, and a safe distance is maintained between the connection path and the inaccessible grid unit.

[0036] Furthermore, the obtaining of the sub-path covering the entire sub-grid area in each sub-grid area further includes: preferably setting the step length between the plurality of rows of path segments parallel to the starting edge so as to maximize the fitness of the sub-path set according to the step length;

[0037] Fitness is: F(l) = αF coverage (l)+γv avg (R i )

[0038] Among them, α and γ are the weight coefficients of coverage and speed respectively;

[0039] F coverage (l) is the coverage rate,

[0040] A covered (R i )=L(Ri )·W,L(R i ) is the subpath length, W is the coverage width;

[0041] The total area of ​​the sub-grid region is A(R i ), the covered area is A covered (R i );

[0042] v avg (R i ) is the regional path speed,

[0043] The amount of cement that needs to be poured in the sub-grid area is M(R i ), the total length of the subpath is L(R i ).

[0044] Furthermore, the optimization includes the step of optimizing the step length based on the genetic annealing algorithm:

[0045] Population initialization: From the step size set {l1,l2,…,l n} randomly selects a set of step lengths as the initial population, and each step length l i As an individual, create subpaths according to the subpath acquisition step and calculate the fitness F(l i );

[0046] Selection operation: In each generation of genetic algorithm operation, the individuals in the current population are sorted from large to small according to the fitness value, and the individuals in the top 10% of the fitness value ranking are taken as excellent individuals and put into the elite population Q;

[0047] A good individual is randomly selected from the elite population Q as the parent to perform crossover and mutation operations with the current individual of the current population to generate a new individual; the crossover operation includes: exchanging the step length between the good individual and the path segment of the current individual that is partially parallel to the starting edge; the mutation operation: applying a random bias to the step length between the good individual and the path segment of the current individual that is partially parallel to the starting edge;

[0048] According to the simulated annealing acceptance criteria, replace the current individual;

[0049] Merge excellent individuals and new individuals to generate the next generation population;

[0050] If the evolutionary generation reaches the set maximum value, the iteration ends, otherwise the genetic algorithm operation is repeated for iteration;

[0051] The simulated annealing acceptance criteria are:

[0052]

[0053] Where T represents the current temperature. As the number of iterations increases, the temperature gradually decreases. T = T0·α n , T0 is the initial temperature, α∈(0,1) is the temperature attenuation coefficient, n is the current iteration number; F(l new ) is the fitness of the new individual, F(l current ) is the fitness of the current individual, l new Is a new individual, current is the current individual, and p is a number greater than 0 and less than 1 randomly generated during each iteration.

[0054] The beneficial effects of the present invention are as follows: through a path planning method for a fabric crane based on a hierarchical genetic annealing algorithm, an innovative solution for efficient path planning of a fabric crane in a complex industrial environment is proposed. This method can not only effectively sort sub-areas at the global level to reduce path intersections, but also achieve efficient coverage and obstacle avoidance functions in local path planning through genetic and simulated annealing algorithms. Finally, this method achieves high precision and high adaptability in path planning for the fabric crane through global and local collaborative optimization, and can give full play to its advantages in operating scenarios with dense and irregular obstacles, providing reliable technical support for the efficient operation of the fabric crane. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is the sub-grid area path pattern diagram;

[0056] Figure 2 It is the path obstacle avoidance map of the sub-grid area. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution of the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings of the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the described embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0058] Example 1

[0059] An intelligent path planning method for a concrete placing machine based on evolutionary constraint optimization is used to plan the global path of a concrete placing machine operating in a work target area map, comprising the following steps:

[0060] Divide sub-areas: Obtain the boundary information of the target area through map data, and divide the entire target area into multiple sub-grid areas based on the location of obstacles, ensuring that each sub-grid area has no intersection with obstacles;

[0061] Sorting of sub-grid area access sequence: obtaining multiple random sorting sequences composed of all sub-grid area sortings, using transformation operations to transform the sorting of sub-grid areas in the random sorting sequences to generate variant sorting sequences, repeatedly selecting preferred variant sorting sequences from the variant sorting sequences to replace the random sorting sequences, and repeating the transformation operations to obtain access sequences, wherein the preferred variant sorting sequences have smaller objective function values ​​compared to the replaced random sorting sequences, and smaller objective function values ​​represent fewer intersections in the path composed of sequence nodes in sorting order, and smaller spans between two sub-grid areas with continuous sorting order in the sequence, where sequence nodes are sub-grid areas; span refers to the center distance of non-adjacent sub-grid areas in the map data.

[0062] Optimizing sub-paths within sub-grid areas: obtaining sub-paths within each sub-grid area that cover the entire sub-grid area, and reconstructing the sub-paths to obtain obstacle avoidance paths, wherein the obstacle avoidance paths have no intersection with the obstacles and are separated from the obstacles by a safe distance;

[0063] Optimizing the connection path between sub-grid areas: traversing the sub-grid areas that are adjacent to each other in sequence according to the access sequence, obtaining a connection path between the end point of a sub-path or obstacle avoidance path in a previous sub-grid area and the starting point of a sub-path or obstacle avoidance path in a subsequent sub-grid area, wherein the connection path has no intersection with the obstacle, and using the connection path to sequentially connect the sub-paths or obstacle avoidance paths that are adjacent to each other to obtain a candidate global path;

[0064] A global path is selected from the candidate global paths, wherein the number of intersections between the connection path of the global path and other connection paths, obstacle avoidance paths and sub-paths is minimal.

[0065] Specifically, the sub-grid region access sequence sorting aims to optimize the sorting of sub-grid regions so that they are connected in series in a certain order so that the paths composed of sequence nodes in the sorted order form the least intersection. To achieve this goal, the present invention proposes a region sorting optimization method based on the domain large-scale search (LNS) algorithm. By introducing region selection constraints, the sub-grid region order is gradually adjusted to ensure the rationality of the sorting path and minimize intersections.

[0066] The specific optimization sorting process includes the following key steps:

[0067] 1. Initialize the sorting scheme: First, abstract each sub-grid area as a point and generate a random sorting scheme τ0 = (τ1, τ2, ..., τ k ), as the parent individual. This initial solution is used as the starting point of the optimization process, and the objective function value F(τ0) of the random sorting solution is calculated. This value will be used to evaluate the pros and cons of subsequent optimization solutions.

[0068] In the design of the objective function, path intersection penalty and region selection constraint penalty are included. The objective function is set as follows:

[0069]

[0070] Where: τ p ,τ q are the consecutive p-th and q-th sub-grid areas in the sorting scheme, q = p + 1; u is the weight of the path crossing penalty, which is used to control the impact of the crossing on the path; λ is the weight of the area selection constraint, which ensures that the sub-grid area combination meets the actual application scenario requirements.

[0071] The components of the objective function are explained as follows: ∑ 1≤p<q≤k (intersect(τ0)): represents the number of intersections in the path composed of the sequence nodes of the random sorting scheme in the sorting order, which is used to measure the intersection of the paths in the sorting scheme. After each sub-grid area is abstracted as a point, it is connected into a complete path through sorting, and the number of intersections in the path is counted as a measure of path complexity. The more intersections, the higher the complexity of the path, so the penalty for the scheme will be increased during the optimization process to guide the path to be optimized in a smoother and less cross-cutting direction.

[0072] constraint(τ p ,τ q ): represents the region selection constraint penalty coefficient, which is used to evaluate whether there are any violations of constraints such as span in the sorting of the random sorting scheme. If the sorting scheme exceeds the maximum allowed span limit, this item will impose a larger penalty to guide the optimization process to converge in the direction of satisfying the constraints.

[0073] Region selection constraint penalty setting:

[0074] In order to improve the rationality of sorting and meet the requirements of actual applications, the following region selection constraint penalty items are defined:

[0075] Adjacent region priority constraints:

[0076] To ensure that sub-grid areas with close physical distances are arranged adjacently in the sorting, the adjacency function is defined as follows:

[0077]

[0078] Calculate two consecutive sub-regions τ in the sequence in sequence p and τ q The adjacency of the two sub-grid areas τ p and τ q The boundary coordinates of the two sub-grid areas τ p and τ qAre they completely adjacent? If so, then adjacency(τ p ,τ q )=0; if not adjacent, then adjacency(τ p ,τ q )=1;

[0079] Set the maximum allowed span constraint for the case where the sequence is continuous but not adjacent to the sub-grid area:

[0080] To avoid excessive span in the path, set the maximum span constraint Δ max , define d(τ p ,τ q ) are two consecutive but non-adjacent sub-grid regions in the sequence τ p and τ q The distance from the center point;

[0081]

[0082] For two consecutive sub-regions τ p and τ p If they are adjacent, the penalty coefficient is 0; if they are not adjacent, further calculate d(τ p ,τ p )

[0083] Among them, δ is the penalty coefficient corresponding to exceeding the allowed span. δ is greater than 1 to ensure that the regions in the sorting are not too dispersed. The span refers to the center distance of the sub-grid areas that are continuous but not adjacent in the sequence sorting in the map data.

[0084] 2. Local search and adjustment: Based on the current parent individuals, the random sorting scheme is finely transformed to obtain a variant sorting sequence, including the following operations:

[0085] Region exchange: Randomly select two sub-grid regions within the parent individual and And exchange their positions in the parent individuals. Through this sub-grid area exchange, the intersection can be reduced while ensuring the stability of the overall structure, balancing the coherence and rationality of the path.

[0086] Interval reversal: Select an interval τ consisting of multiple continuous sub-grid areas in the parent individual i:j , and reverse the order of the sub-grid areas within the interval. This operation is intended to break the inherent path intersection of the interval, reduce the path complexity through reverse order adjustment, and improve the global rationality of the sorting.

[0087] Neighborhood fine-tuning: Fine-tune the order of adjacent sub-grid areas in the path based on neighborhood information. This operation is especially suitable for situations where there are many local path intersections. By making small adjustments to the adjacent sub-grid areas, the number of path intersections can be further reduced and the region selection constraints can be met.

[0088] During local search and adjustment, one of the three transformation operations is randomly selected for mutation, and the number of mutated sorted sequences before and after the transformation is equal to the number of random sorted sequences.

[0089] 3. Objective function recalculation and constraint checking: After each local adjustment, the objective function value F(τ) of the new variant sorting sequence is recalculated and the constraint condition is checked. Specifically, ensure that the order adjustment of all sub-grid regions meets the requirements of the adjacent region priority constraint and the maximum span constraint. If the variant sorting sequence does not meet the constraints as well as the parent individual, a penalty weight is applied so that the variant sorting sequence is reasonably restricted by the constraint mechanism in the evaluation.

[0090] 4. Adaptive optimization strategy: Based on the comparison of the objective function value between the mutated sorting sequence and the parent individual, decide whether to update the sorting scheme. If the mutated sorting sequence has a lower objective function value than the parent individual, replace the parent individual with the mutated sorting sequence as the new parent individual. Otherwise, keep the original parent individual unchanged to ensure the convergence of the optimization process.

[0091] 5. Termination and output of the optimal solution: When the set number of iterations is reached or the optimization effect of the solution tends to be stable, the algorithm stops and outputs the optimal sorting solution τ that meets all constraints * As the access sequence, this scheme maximizes the rationality of regional sorting on the basis of minimizing path intersections, which meets the needs of application scenarios.

[0092] This method establishes a good balance between local search and global optimization, so that the region sorting is not only optimized in terms of path intersections, but also has higher practicality under the constraints of actual applications.

[0093] Specifically, the sub-path optimization within the sub-grid area includes:

[0094] The sub-grid area is divided into a grid map composed of grid cells by perpendicular longitude and latitude lines; each grid cell is marked as passable or impassable, that is, F(x,y)=0 indicates that the grid cell is passable, and F(x,y)=1 indicates that the grid cell is impassable.

[0095] According to the access sequence order, the starting edge and the ending edge of each sub-grid area are found in turn, the edge of each sub-grid area away from the obstacle is selected as the starting edge, and the edge of each sub-grid area close to the next sub-grid area is selected as the ending edge. The next sub-grid area of ​​the sub-grid area is the sub-grid area adjacent to the sub-grid area in sequence according to the access sequence order.

[0096] A grid unit is selected from a column or row of grid units near the starting edge of the sub-grid area as the starting point of the sub-path. The sub-path is usually planned in a "bow-shaped" path. When planning the bow-shaped path, it starts from the starting edge of the sub-grid area and moves forward in parallel to the row of the starting edge. When reaching the boundary of the row, the fabricator turns and enters the adjacent row, and continues to move in the opposite direction, moving forward row by row until the entire sub-grid area is covered and the end point at the end edge is reached, forming a winding sub-path formed by connecting each path segment end to end, and the spacing between adjacent rows is the step length. This path mode is suitable for the working characteristics of the fabricator, that is, cement irrigation is carried out simultaneously during the process of moving. Since cement will naturally flow to both sides after spraying, choosing a bow-shaped path can ensure that each movement is parallel and adjacent to the previous row, reducing omissions and overlaps, and helping to cover the entire sub-area.

[0097] like Figure 1 As shown, by adjusting different step sizes, different direction selections have different sub-path combinations. The purpose of this step is to select a suitable sub-grid area planning mode.

[0098] Adjacent path segments of the sub-path at the bend are perpendicular to each other, and the sub-path covers the entire sub-grid area and does not generate intersections in the sub-grid area.

[0099] Acquiring a sub-path in each sub-grid area that covers the entire sub-grid area includes: optimizing the step length between a plurality of rows of path segments parallel to the starting edge based on a genetic annealing algorithm.

[0100] In the current sub-grid area R i The sub-path planning of the cloth placing machine involves the selection of the step length l, which directly affects the coverage efficiency and connection smoothness of the cloth placing machine in the sub-grid area. To this end, the genetic annealing algorithm is introduced to find the optimal step length l through the population search ability of the genetic algorithm and the acceptance criterion optimization of the simulated annealing. * .

[0101] In the step size optimization process, firstly in the step size range l∈[l min ,l max ] takes multiple discrete step values ​​l1, l2, ..., l n , each step generates a subpath. The placing machine follows these subpaths, from the subgrid area R i The starting point S(R i) and proceeds in a reciprocating motion until the entire sub-grid area is covered.

[0102] To find the optimal step size, the steps of using the genetic annealing algorithm are as follows:

[0103] 1. Population initialization: From the step size set {l1,l2,…,l n} randomly selects a set of step lengths as the initial population, and each step length individual represents a candidate subpath solution. i , create sub-paths according to the sub-path generation rules of the fabric placing machine, and calculate its fitness function F(l i ).

[0104] 2. Fitness function evaluation: To ensure that the subpaths have good coverage and smooth connection between regions, the fitness function F(l) comprehensively considers the following factors: coverage rate and regional path speed;

[0105] The overall formula of the fitness function is:

[0106] F(l)=αF coverage (l)+γv avg (R i ) (5)

[0107] Where α and γ are the weight coefficients of coverage and speed respectively.

[0108] F coverage (l) is the coverage rate: the spreading machine is at R i The higher the coverage rate within the grid, the greater the fitness function value of the subpath. Let the total area of ​​the subgrid region be A(R i ), the covered area is A covered (R i ), the coverage can be expressed as:

[0109]

[0110] Assume that the pouring width of the concrete placing machine is W,A covered (R i )=L(R i )·W;L(R i ) is the subpath length, and W is the coverage width.

[0111] v avg (R i ) is the regional path speed: Assume that each sub-region R i The amount of cement required is M(R i ), the speed of the placing machine should be adjusted according to the demand for cement. i ) divided by the total length of the subpath L(R i) to calculate the average speed v of the fabric placing machine in this area avg (R i ), the formula is as follows:

[0112]

[0113] in:

[0114] M(R i ) is the fabric placing machine in sub-area R i The amount of cement that needs to be laid.

[0115] L(R i ) is the total path length of the concrete placing machine in this sub-area.

[0116] 3. Selection operation: In each generation of genetic algorithm operation, the individuals in the current population are sorted from large to small according to the fitness value, and the individuals in the top 10% of the fitness value ranking are taken as excellent individuals and put into the elite population Q; while retaining the individuals with better step length, the individuals with low fitness are moderately retained to maintain the diversity of the population.

[0117] 4. Crossover and mutation operations: Crossover: Randomly select a good individual from the elite population Q as the parent and perform crossover and mutation operations with the current individual of the current population to generate a new individual, exchanging the step length between the good individual and the path segment that is partially parallel to the starting edge in the current individual. The crossover operation uses the existing gene combination to generate a new step length, which may find a better solution. Mutation: Make a slight variation on the step length between the good individual and the path segment that is partially parallel to the starting edge in the current individual, that is, apply a random bias to the step length value. This introduces randomness in the search space and prevents the algorithm from falling into a local optimum.

[0118] 5. Simulated annealing acceptance criteria: In each generation of the genetic algorithm, the acceptance criteria of simulated annealing are used to decide whether to replace the current individual with a new individual. The acceptance criteria are as follows:

[0119]

[0120] Where T represents the current temperature, and the temperature gradually decreases as the number of iterations increases. p is a randomly generated number greater than 0 and less than 1 during each iteration. This acceptance criterion allows the algorithm to accept moderately degraded solutions in the early stages to escape the local optimal solution area, and gradually tends to only accept optimal solutions in the later stages.

[0121] 6. Temperature reduction strategy: The temperature T decreases algebraically to control the randomness of the acceptance criteria. The temperature reduction formula is:

[0122] T=T0·α n (6)

[0123] Where T0 is the initial temperature, α∈(0,1) is the temperature attenuation coefficient, and n is the current iteration number. The temperature drop makes the system gradually accept the optimal solution, reducing the probability of poor solutions, and finally converges to the global optimal solution.

[0124] 7. Termination condition and step length determination: After multiple generations of genetic evolution and simulated annealing optimization, when the population fitness value tends to converge or reaches the maximum number of iterations, the algorithm terminates. At this time, the step length l with the highest fitness is selected. * As the optimal step length, this step length performs well in terms of coverage and regional connectivity, meeting the requirements of sub-path planning for the placing machine.

[0125] Specifically, reconstructing the subpath to obtain the obstacle avoidance path includes:

[0126] If it is detected that a path segment in the sub-path intersects with an obstacle in the sub-grid area, the path segment is offset to obtain an offset path, and the offset path maintains a safe distance from the obstacle;

[0127] Specifically, if a path segment overlaps or contacts a grid unit where an obstacle exists, the path segment is deemed to intersect with the obstacle in the sub-grid area.

[0128] The intersection of the path segment and the grid cells where the obstacle is located is located. The intersection is the path portion where the concrete placing machine needs to adjust in this area to ensure that the obstacle is avoided.

[0129] The intersection portion is replaced by an offset path.

[0130] The goal of reconstructing the path segment is to minimize the deviation from the atomic path shape and ensure that the new path segment seamlessly connects with the obstacle-free part of the atomic path. Generate offset path segments: When bypassing the obstacle position, each point (x′ p ,y′ p ) should keep a certain safe distance from the grid cells where obstacles exist to avoid the path being close to obstacles. The shape and direction of the offset path segment are adjusted according to the travel direction of the fabric placing machine and the relative position of the grid cells where obstacles exist to ensure that the reconstructed offset path is naturally connected with the atomic path.

[0131] Path connection: After the offset path segment that bypasses the obstacle is generated, the offset path segment is reconnected with the obstacle-free area of ​​the atomic path to form a continuous obstacle avoidance path to ensure full coverage of the placing machine within the sub-grid area.

[0132] Through the above path obstacle avoidance processing, the placing machine can effectively bypass obstacles beside the sub-grid area, ensure that the obstacle avoidance path is not interfered by obstacles, and maintain the smoothness and overall coverage of the obstacle avoidance path during the obstacle avoidance process. Figure 2 shown.

[0133] Specifically, the connection path optimization between sub-grid regions aims to: after determining the access order of sub-grid regions σ=(σ1,σ2,…,σ k ), generate different complete sub-path combinations based on this order, and select the best candidate global path path total .

[0134] Each subgrid area Its internal subpaths have been generated Therefore, it is necessary to connect the sub-paths of each sub-grid area in series to ensure that the concrete placing machine can move smoothly and fully cover the entire working area.

[0135] For two adjacent sub-grid areas in order and The cloth placing machine needs to start from the end point of the i-th sub-grid area Move to the start point of the next subgrid area The coordinates of the end point and the starting point are and By generating connecting paths between sub-grid areas A complete candidate global path can be gradually formed.

[0136] Since different sub-grid area orders correspond to different sub-grid area connection path combinations, each sorting result will generate a unique candidate global path. In order to select the best solution, the fitness function F(path total ), the rationality of the path is measured by the number of intersections between the connection path of the candidate global path and other connection paths, obstacle avoidance paths and sub-paths. The fitness function is expressed as follows:

[0137] F(path total )=cross(path total )

[0138] Among them, cross(path total ) represents the candidate global path path total The fewer the number of intersections, the smoother and more efficient the candidate global path. By minimizing the number of intersections of the candidate global path, unnecessary path interference can be avoided, thereby improving the working efficiency of the fabric placing machine.

[0139] By evaluating and screening the number of intersections of sub-paths and connection path combinations in different sub-grid areas, the candidate global path with the least number of intersections is finally selected to ensure smooth connection and barrier-free coverage of the fabric placing machine in the entire working area.

[0140] The connection path optimization between sub-grid areas also includes:

[0141] The grid map of the sub-grid area is expanded to the map of the entire working target area, and the grid cells without obstacles are marked as passable, and the grid cells with obstacles are marked as impassable; that is, F(x, y) = 0 indicates that the cell is passable, and F(x, y) = 1 indicates an impassable obstacle.

[0142] Intersection detection: First, each path segment in the connection path is detected to determine whether it intersects with an obstacle position in the obstacle set. If a path segment coincides with or touches an obstacle position, the connection path is considered to intersect with the obstacle.

[0143] Intersection positioning: Once the intersection between the connection path and an obstacle is detected, the path segment l is located intersect The intersection path segment is the part of the path where the concrete placing machine needs to adjust in this area to ensure that it avoids obstacles.

[0144] Path reconstruction: The offset intersection path forms an offset intersection path segment to bypass the impassable grid unit, and keep the offset intersection path segment at a safe distance from the impassable grid unit. The shape and direction of the offset intersection path segment are adjusted according to the travel direction of the concrete placing machine and the relative position of the obstacle to ensure that the reconstructed offset intersection path segment is naturally connected with the original connection path.

[0145] Path connection: After the offset connection path segment that bypasses the obstacle is generated, the offset connection path segment is reconnected with the obstacle-free area of ​​the original connection path to form a continuous connection path.

[0146] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0147] In summary, the method of the present invention can not only effectively sort sub-areas at the global level to reduce path intersections, but also achieve efficient coverage and obstacle avoidance functions in local path planning through genetic and simulated annealing algorithms. Finally, this method achieves high precision and high adaptability in the path planning of the concrete placing machine through global and local collaborative optimization, and can give full play to its advantages in operating scenes with dense and irregular obstacles, providing reliable technical support for the efficient operation of the concrete placing machine.

Claims

1. An intelligent path planning method for a concrete placing machine based on evolutionary constraint optimization, used to plan the global path of a concrete placing machine operating in a work target area map, characterized in that: The steps include: Obtain a number of sub-grid areas and obstacles divided by the work target area map; Acquire multiple random sorting sequences composed of sorting of all sub-grid regions, transform the sorting of sub-grid regions in the random sorting sequence by using a transformation operation to generate a variant sorting sequence, repeatedly select a preferred variant sorting sequence from the variant sorting sequence to replace the random sorting sequence, and repeat the transformation operation to obtain an access sequence, wherein, compared with the replaced random sorting sequence, the preferred variant sorting sequence has: fewer crossover times in the path composed of sequence nodes in sorting order, smaller span between two sub-grid regions with continuous sorting order in the sequence, and the sequence nodes are sub-grid regions; Get the sub-path in each sub-grid area that covers the entire sub-grid area; Reconstructing the sub-path to obtain an obstacle avoidance path, where the obstacle avoidance path can avoid obstacles and be separated from the obstacles by a safe distance; Traversing the sub-grid areas that are adjacent to each other in sequence according to the access sequence, obtaining a connection path between an end point of a sub-path or obstacle avoidance path in a previous sub-grid area and a starting point of a sub-path or obstacle avoidance path in a subsequent sub-grid area, wherein the connection path has no intersection with the obstacle, and using the connection path to sequentially connect the sub-paths or obstacle avoidance paths that are adjacent to each other to obtain a candidate global path; A global path is selected from the candidate global paths, wherein the number of intersections between the connection path of the global path and other connection paths, obstacle avoidance paths and sub-paths is minimal.

2. The method according to claim 1, characterized in that The method of transforming the order of the sub-grid regions in the random ordering sequence by using the transformation operation to generate a variant ordering sequence comprises: Region exchange: randomly select two sub-grid regions in the random sorting sequence and exchange their positions in the random sorting sequence; Interval reversal: select an interval consisting of multiple continuous sub-grid areas in the random sorting sequence, and reverse the order of the sub-grid areas in the interval; Neighborhood fine-tuning: swap the order of adjacent sub-grid areas in the random order sequence; The transformation operation randomly selects one of the three transformation operations to generate a mutated sorted sequence, and the number of mutated sorted sequences before and after the transformation is equal to the number of random sorted sequences.

3. The method according to claim 1, characterized in that The step of selecting a preferred variant sequence from the variant sequence to replace the random sequence comprises: Calculate the objective function of the random sorting sequence and the mutated sorting sequence; wherein, the smaller the objective function is, the fewer the number of crossovers in the path composed of the sorting order of the sequence nodes and / or the smaller the span of the two sub-grid areas with continuous sorting order in the sequence; If the objective function of the mutated sorting sequence is smaller than that of the random sorting sequence, the mutated sorting sequence is selected as the preferred mutated sorting sequence to replace the random sorting sequence; When the number of repetitions reaches a set number or the number of selected preferred variant sorted sequences converges, the repetition is terminated.

4. The method according to claim 3, characterized in that The objective function includes a path intersection penalty term and a region selection constraint penalty term. F(τ)=u·∑ 1≤p<q≤k (intersect(τ0))+λ·∑ 1≤p<q≤k constraint(t p ,t q ); Where: u is the weight of the path crossing penalty, λ is the weight of the region selection constraint, Σ 1≤p<q≤k (intersect(τ0)) represents the number of intersections in the path composed of the sequence node sorting order, constraint(τ p ,τ q ) represents the region selection constraint penalty coefficient, τ p ,τ q is the pth and qth consecutive subgrid regions in the random sorting sequence or the mutated sorting sequence, q = p + 1; τ p ,τ q Adjacent means that the boundary coordinates of two sub-grid areas are adjacent; Δ max is the maximum allowable span, d(τ p ,τ q ) are two sub-grid areas τ p and τ q The distance from the center point of , δ is greater than 1.

5. The method according to claim 1, characterized in that The obtaining of the sub-paths in each sub-grid area that cover the entire sub-grid area includes: The sub-grid area is divided into a grid map composed of grid cells using perpendicular longitude and latitude lines; each grid cell without obstacles is marked as passable, and each grid cell with obstacles is marked as impassable; Detect each sub-grid area and its next adjacent sub-grid area in sequence according to the access sequence, find the starting edge and the ending edge of each sub-grid area in turn, select the edge of each sub-grid area away from the obstacle as the starting edge, and select the edge of each sub-grid area close to the next sub-grid area as the ending edge; A grid unit is selected from a column or a row of grid units near the starting edge of the sub-grid area as the starting point of the sub-path, and the sub-path is extended from a direction parallel to the starting edge in the sub-grid area to form a winding sub-path formed by connecting the path segments end to end, and the adjacent path segments of the sub-path at the bend are perpendicular to each other, and the end point of the sub-path is located at the terminating edge of the sub-grid area. The sub-path includes several rows of path segments parallel to the starting edge, and the sub-path covers the entire sub-grid area and does not generate an intersection in the sub-grid area.

6. The method according to claim 1, characterized in that The reconstructing the sub-path to obtain the obstacle avoidance path comprises: If an intersection portion of the subpath that overlaps or touches a grid unit where an obstacle exists is detected, the intersection portion is offset to obtain an offset path, and the offset path maintains a safe distance from the obstacle; Replaces the path segment with an offset path.

7. The method according to claim 1, characterized in that The step of sequentially traversing the mutually adjacent sub-grid areas according to the access sequence to obtain a connection path between an end point of a sub-path or obstacle avoidance path in a previous sub-grid area and a start point of a sub-path or obstacle avoidance path in a subsequent sub-grid area further includes: Expand the grid map of the sub-grid area to the entire work target area map, mark the grid cells without obstacles as passable, and mark the grid cells with obstacles as impassable; If it is detected that the connection path passes through an inaccessible grid unit, the connection path is shifted to bypass the inaccessible grid unit, and a safe distance is maintained between the connection path and the inaccessible grid unit.

8. The method according to claim 5, characterized in that The obtaining of the sub-path covering the entire sub-grid area in each sub-grid area further comprises: preferably determining the step length between the plurality of rows of path segments parallel to the starting edge so as to maximize the fitness of the sub-path set according to the step length; Fitness is: F(l) = αF coverage (l)+γv avg (R i ) Among them, α and γ are the weight coefficients of coverage and speed respectively; F coverage (l) is the coverage rate, A covered (R i )=L(R i )·W,L(R i ) is the subpath length, W is the coverage width; The total area of ​​the sub-grid region is A(R i ), the covered area is A covered (R i ); v avg (R i ) is the regional path speed, The amount of cement that needs to be poured in the sub-grid area is M(R i ), the total length of the subpath is L(R i ).

9. The method according to claim 8, characterized in that The optimization includes the step of optimizing the step length based on the genetic annealing algorithm: Population initialization: From the step size set {l1,l2,…,l n } randomly selects a set of step lengths as the initial population, and each step length l i As an individual, create subpaths according to the subpath acquisition step and calculate the fitness F(l i ); Selection operation: In each generation of genetic algorithm operation, the individuals in the current population are sorted from large to small according to the fitness value, and the individuals in the top 10% of the fitness value ranking are taken as excellent individuals and put into the elite population Q; A good individual is randomly selected from the elite population Q as the parent to perform crossover and mutation operations with the current individual of the current population to generate a new individual; the crossover operation includes: exchanging the step length between the good individual and the path segment of the current individual that is partially parallel to the starting edge; the mutation operation: applying a random bias to the step length between the good individual and the path segment of the current individual that is partially parallel to the starting edge; According to the simulated annealing acceptance criteria, replace the current individual; Merge excellent individuals and new individuals to generate the next generation population; If the evolutionary generation reaches the set maximum value, the iteration ends, otherwise the genetic algorithm operation is repeated for iteration; The simulated annealing acceptance criteria are: Where T represents the current temperature. As the number of iterations increases, the temperature gradually decreases. T = T0·α n , T0 is the initial temperature, α∈(0,1) is the temperature attenuation coefficient, n is the current iteration number; F(l new ) is the fitness of the new individual, F(l current ) is the fitness of the current individual, l new Is a new individual, current is the current individual, and p is a number greater than 0 and less than 1 randomly generated during each iteration.

Images