Method and device for earthwork balance and construction path planning of low hilly broken field block

Abstract

This invention discloses a method and apparatus for earthwork balancing and construction path planning in fragmented fields in low hills and mountains. The method includes: S1: collecting the original topographic data of the fields to be merged and automatically dividing them into merging units using a region growing algorithm; S2: establishing a topsoil thickness distribution map, determining the topsoil stripping scheme, and determining the topsoil storage location with the shortest transportation distance after stripping as the objective; S3: establishing an optimization model with the dual objectives of minimizing earthwork transportation distance and achieving the most uniform topsoil backfill thickness, and using the constraint that the deviation of excavation and filling volume balance does not exceed a preset threshold to solve for the optimal merging design elevation; S4: after determining the optimal design elevation, establishing an earthwork allocation network model and solving for the earthwork allocation scheme of the transportation direction and transportation volume from excavation to filling between each grid; S5: converting the earthwork allocation scheme into grader operation path planning and generating path node instructions executable by the vehicle controller; S6: the grader is under closed-loop control during construction.

Description

Technical Field

[0001] This invention relates to the field of agricultural land consolidation and construction technology, specifically to a method and apparatus for earthwork balance and construction path planning in fragmented fields in low hills. Background Technology

[0002] In hilly areas, the terrain is fragmented with significant elevation differences between fields, resulting in massive earthwork excavation and backfilling volumes during field consolidation projects. Currently, earthwork balance calculations mainly rely on manual estimation using CAD software, and the design elevation after consolidation is primarily determined based on the experience and judgment of construction personnel. In practice, this presents the following main technical shortcomings:

[0003] 1. Severe damage to the topsoil. The topsoil is the core carrier of soil fertility, typically 15-30 cm thick and rich in organic matter and nutrients. Current construction methods do not incorporate topsoil stripping and backfilling as constraints in earthwork balance calculations, often resulting in a mixture of topsoil and subsoil during backfilling, leading to a significant decrease in topsoil fertility after consolidation. While some construction units recognize the importance of topsoil protection, the determination of stripping thickness still relies on experience, lacking scientific calculation methods based on sampling data. The stripping range and storage location are also not optimized, resulting in excessively long topsoil transportation distances and severe secondary handling.

[0004] 2. Design elevation relies on experience. The design elevation of the consolidated fields is a core parameter for earthwork balance, determining the distribution of excavation and filling areas and the total earthwork volume. In current practice, the design elevation is mainly determined by construction or design personnel based on experience, without incorporating key indicators such as earthwork haulage distance and topsoil backfill uniformity into the quantitative optimization model. This experience-driven approach leads to the following consequences: First, it is difficult to accurately balance excavation and filling volumes, often resulting in a large amount of surplus soil being transported off-site or borrowed for backfilling during actual construction, increasing project costs; second, excessively long earthwork haulage distances increase machinery operating time and construction period; and third, uneven topsoil thickness in different backfill areas affects the consistency of subsequent crop growth.

[0005] 3. Construction path relies on manual judgment. After the earthwork allocation plan is determined, how to convert the abstract allocation data into the specific working path of the grader is a key link connecting planning and execution. In current construction, earthwork allocation plans are usually presented in the form of two-dimensional drawings or tables. The working path of the grader (such as where to start leveling, in which direction to travel, when to raise and lower the shovel, etc.) is still judged by the driver based on the site conditions and experience. This manual judgment mode has the following shortcomings: First, the working path lacks global optimization, resulting in large empty driving mileage, repeated compaction, and low work efficiency; second, the leveling accuracy depends on the driver's skill level, and the flatness of the field surface fluctuates greatly; third, it is difficult to meet the format requirements of intelligent equipment such as Beidou unmanned graders for path instructions, which restricts the improvement of construction automation level.

[0006] 4. Lack of real-time quality feedback and dynamic correction during construction. Field consolidation is a dynamic process, influenced by various factors such as terrain undulations, soil differences, and mechanical vibrations. Unavoidable deviations exist between the actual leveled elevation and the design elevation. Under current construction methods, leveling quality is typically inspected once all construction is completed using RTK-GNSS, a "post-construction inspection" model. The problem with this model is that once deviations occur, they are difficult to remedy, and some areas may require rework, delaying the project and increasing costs. The lack of real-time elevation monitoring and dynamic correction mechanisms during construction makes it impossible to achieve closed-loop quality control of "construction, inspection, and correction simultaneously."

[0007] In summary, the existing technology lacks a complete technical solution that organically integrates five functions: quantitative constraint of topsoil protection, dual-objective optimization of design elevation, grid-level optimization of earthwork allocation, intelligent planning of grader paths, and dynamic correction during construction. In particular, there is a significant technical gap in the connection between "algorithm optimization and mechanical execution". Summary of the Invention

[0008] The purpose of this invention is to provide a method and apparatus for earthwork balancing and construction path planning in fragmented fields in low hills, so as to at least solve the problems of existing earthwork calculation methods that do not protect the topsoil, rely on experience for design elevation, and rely on manual judgment for construction path.

[0009] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for earthwork balance and construction path planning in fragmented fields in low hills, comprising the following steps:

[0010] S1: Collect the original topographic data of the fields to be merged and generate a high-precision digital elevation model; identify natural field ridges, roads and water system boundaries according to the field merging target, and automatically divide the merging units using a region growing algorithm;

[0011] S2: Sample the topsoil thickness in each merged unit, establish a topsoil thickness distribution map, determine the topsoil stripping scheme, and determine the topsoil storage location with the shortest transportation distance after stripping as the target.

[0012] S3: Using the merged unit as the basic calculation unit, an optimization model is established with the dual objectives of minimizing earthwork haulage distance and achieving the most uniform backfill thickness of the topsoil layer. With the constraint that the balance deviation of excavation and filling volume does not exceed a preset threshold, the improved particle swarm optimization algorithm is used to solve for the optimal merged design elevation.

[0013] S4: After determining the optimal design elevation, an earthwork allocation network model is established, and an improved ant colony algorithm is used to solve it, so as to obtain the earthwork allocation scheme of the transportation direction and transportation volume from excavation to filling between each grid.

[0014] S5: Convert the earthwork allocation plan into a grader operation path plan and generate path node instructions that can be executed by the vehicle controller;

[0015] S6: The grader operates according to the path node instructions and collects the elevation data of the leveled area in real time through the Beidou RTK positioning system during the construction process. It compares the measured elevation with the design elevation grid by grid and triggers the correction process for areas where the deviation exceeds the preset threshold, forming a closed-loop control.

[0016] Furthermore, in step S1, the method for generating a high-precision digital elevation model includes:

[0017] S11: Collect the original topographic data of the fields to be merged, use UAV oblique photography or LiDAR to acquire point cloud data, and after filtering, noise reduction and ground point classification, generate a high-precision digital elevation model and set the grid size of the high-precision digital elevation model.

[0018] Furthermore, in step S1, the specific method for the region growing algorithm to automatically divide and merge units is as follows:

[0019] S12: Seed point selection: In the high-precision digital elevation model, the slope value and relative elevation difference of each grid cell are used as the initial classification criteria. Grid cells with a slope not higher than the first slope threshold and an elevation difference not higher than the first elevation difference threshold are selected as candidate seed points. The boundaries of adjacent fields are used as the dividing constraints. In each independent closed area enclosed by natural boundaries, the grid cell with the lowest elevation near the geometric center is selected as the initial seed point of that field.

[0020] S13: Region Growth Criteria: Design growth criteria, starting from the initial seed point, incorporate adjacent grids that meet the growth criteria into the current region and use them as new growth starting points, expanding outward layer by layer in the 8-neighborhood direction; grids that do not meet the growth criteria are marked as boundary points and growth in that direction is stopped; the growth criteria are: the slope value of adjacent grid cells is not higher than the second slope threshold, the elevation difference with the seed point is not higher than the second elevation difference threshold, and they do not belong to natural boundary grids;

[0021] S14: Region merging: After the region growth of all seed points is completed, the adjacent small regions with an area less than S are merged. If the elevation difference between the boundaries of two adjacent regions is not higher than the first elevation difference threshold and the total area after merging is not less than S mu, then the two regions are merged into one merged unit.

[0022] S15: Merged Cell Output: Outputs the boundary coordinate sequence and area statistics for each merged cell.

[0023] Further, in step S2, the topsoil stripping scheme calculates the stripping depth according to the following formula:

[0024] ;

[0025] In the formula, D strip d represents the peeling depth. i The topsoil thickness is denoted by ; n is the total number of sampling points; the thickness of each stripping operation is no more than 30 cm; for areas with deep soil layers, the stripping and storage are performed in layers.

[0026] Furthermore, in step S3, the expression for the objective function is:

[0027] F = min(w1·F1 + w2·F2);

[0028] Earthwork transport distance function: ;

[0029] Topsoil backfill uniformity function: ;

[0030] Among them, h k Let be the backfill thickness of the k-th backfill grid. is the average backfill thickness; p is the total number of backfill grids; w1 and w2 are weighting coefficients, and w1+w2=1.

[0031] Furthermore, in step S3, the specific steps of the improved particle swarm optimization algorithm to solve for the optimal merged design elevation are as follows:

[0032] S31: Parameter initialization: Set the number of particles N and the maximum number of iterations T. max The learning factor c1=c2, the inertia weight w decreases linearly, and the design elevation H is encoded as the particle position vector x. i The initial position is in [H min H max Randomly generated within the interval;

[0033] S32: Fitness assessment: For each particle x i Substitute the corresponding candidate design elevations into the objective function, calculate the fitness value, and check the balance deviation of cut and fill volumes. The constraints are applied, and for particles that do not meet the constraints, a penalty term is added to their fitness value. ;;

[0034] S33: Individual Optimal and Global Optimal Update: This involves updating each particle's current fitness value with its historical best position p. Best The fitness values ​​are compared, and if the current value is better, then p is updated. Best The position corresponding to the optimal fitness value among all particles is denoted as the global optimal position g. Best ;

[0035] S34: Particle velocity and position update: Update the velocity and position of each particle according to the following formula:

[0036] v i (t+1) = w(t)·v i (t) + c1·rand1·(p Besti - x i (t))+ c2·rand2·(g Best - x i (t));

[0037] x i (t+1) = x i (t) + v i (t+1);

[0038] Where w(t) is the inertia weight; rand1 and rand2 are random numbers in the interval [0,1].

[0039] Apply boundary constraints to the updated velocity and position;

[0040] S35: Iteration Termination Judgment: Repeat steps S32 to S34 until the maximum number of iterations is reached, or the global optimal solution g is obtained for N consecutive generations. Best The iteration terminates when the change is less than a preset threshold ε, and the global optimal position g is output. Best The optimal merging design elevation H.

[0041] Furthermore, in step S4, the specific steps of the improved ant colony algorithm for solving the earthwork allocation scheme are as follows:

[0042] S41: Network construction: Divide the merged unit into a set size of network grid, calculate the cut or fill volume of each grid according to the optimal design elevation H, assign all cut grids to the source set S, assign all fill grids to the sink set T, and construct a complete bipartite network graph G = (S, T, E).

[0043] S42: Ant colony initialization: Set the number of ants K and the maximum number of iterations I. termax Pheromones evaporation coefficient ρ, pheromone intensity Q, pheromone heuristic factor α, expected heuristic factor β, and initial pheromone concentration τ along each path. ij (0) = τ0;

[0044] S43: Path Selection and Earthwork Allocation: Each ant starts from the first cut grid and selects the corresponding fill grid for each cut grid in turn. The ant is at node s i Select the next fill grid. j The transition probability is:

[0045] ;

[0046] in, This is heuristic information; allowed is the set of fill grids that have not yet been fully allocated.

[0047] S44: Pheromone Update: After all ants have completed one iteration, the pheromone concentration on each path is updated according to the following formula:

[0048] ;

[0049] Wherein, if ant k chooses path ij in this iteration, then ;otherwise, =0; The total transportation distance of the allocation plan constructed for ant k;

[0050] S45: Iteration Termination and Optimal Solution Output: Repeat steps S43 to S44 until the maximum number of iterations I is reached. termax The iteration terminates when the total transport distance change of the optimal solution for N consecutive generations is less than the preset threshold δ; the global optimal allocation scheme is output.

[0051] Furthermore, in step S5, the specific method for converting the earthwork allocation plan into a grader operation path plan is as follows:

[0052] S51: Classification and boundary extraction of work area: According to the allocation plan output in step S4, the net earthwork volume of each grid is calculated. Grids with a net earthwork volume greater than zero are classified as excavation areas, grids with a net earthwork volume less than zero are classified as fill areas, and grids with a net earthwork volume equal to zero are classified as leveling areas. The outer boundary polygons of each type of area are extracted respectively.

[0053] S52: Path type matching: Match spiral paths for cut areas, reciprocating paths for fill areas, and zigzag paths for leveling areas;

[0054] S53: Path Node Generation: Using the outer boundary polygon of the excavation area as the starting outline, and with the width of the leveling shovel as the step size, generate an inward shrinking polygon sequence layer by layer and connect them to form a spiral path node sequence; using the long side direction of the circumscribed rectangle of the fill area as the direction of travel, and with the width of the leveling shovel as the row spacing, generate a reciprocating polyline path node sequence; using the circumscribed rectangle of the leveling area as the range, and with the short side direction as the step size, generate a "Z"-shaped polyline path node sequence.

[0055] S54: Path Node Assignment: Assign each path node with coordinates, target elevation, travel speed, and shovel lifting / lowering command attributes;

[0056] S55: Command Output: Encodes the path node sequence in sequence into an instruction format executable by the vehicle controller; the path node instruction includes the path node coordinate sequence, the target elevation of each node, the travel speed instruction, and the shovel lifting instruction.

[0057] Furthermore, in step S6, the dynamic correction process is specifically as follows:

[0058] The BeiDou RTK positioning system on the grader collects elevation data of the leveled area in real time at a set sampling frequency and uploads it to a cloud server via wireless network; the server then compares the measured elevation with the design elevation grid by grid.

[0059] If the deviation is less than or equal to the first set value, it is marked as qualified;

[0060] If the deviation is between the first set value and the second set value, it is marked as a warning and the deviation value is recorded; wherein, the first set value is less than the second set value;

[0061] If the deviation is greater than the second set value, it is marked as unqualified and a correction process is triggered. A correction area of ​​a set size is delineated with the deviation grid as the center, the excavation and filling volume of the area is recalculated, a local correction instruction is generated and sent to the vehicle controller for execution via the cloud.

[0062] Secondly, the present invention provides an earthwork balancing and construction path planning device for fragmented fields in low hills and mountains, comprising: a processor and a memory; wherein, the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory to enable the device to perform the above-mentioned method.

[0063] The beneficial effects of this invention are as follows: by optimizing the stripping of the topsoil beforehand, the loss of topsoil is reduced, effectively protecting soil fertility; by optimizing the design elevation with the dual objectives of minimizing earthmoving distance and achieving the most uniform backfill thickness of the topsoil, the total earthmoving volume is reduced compared to experience-based construction, and the accuracy of excavation-filling balance is significantly improved, which can significantly reduce project costs; by linking the earthmoving allocation scheme with the Beidou unmanned grader's operating path, integrated intelligent construction of "optimization calculation → construction execution" is achieved, reducing manual judgment and shortening the construction period; and by using a dynamic closed-loop correction mechanism, the flatness is continuously controlled during construction to ensure the pass rate of field flatness and meet the stringent requirements of high-standard farmland construction for field flatness (error ≤ ±3cm). Attached Figure Description

[0064] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, use the same reference numerals to denote the same or similar parts. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0065] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0066] Figure 2 This is a typical plan view of the merged unit;

[0067] Figure 3 for Figure 2 A schematic diagram of the A-A' cross-section;

[0068] Figure 4 This is a flowchart of a bi-objective optimization algorithm according to an embodiment of the present invention;

[0069] Figure 5 This is a modified flowchart of an embodiment of the present invention. Detailed Implementation

[0070] Firstly, such as Figure 1 As shown, this invention discloses a method for earthwork balance and construction path planning in fragmented fields in low hills, including the following steps:

[0071] S1: Collect the original topographic data of the fields to be merged and generate a high-precision digital elevation model (DEM); identify natural field ridges, roads and water system boundaries according to the field merging target, and automatically divide the merging units using a region growing algorithm;

[0072] S2: Sample the topsoil thickness in each merged unit, establish a topsoil thickness distribution map, determine the topsoil stripping scheme, and determine the topsoil storage location with the shortest transportation distance after stripping as the target.

[0073] S3: Using the merged unit as the basic calculation unit, an optimization model is established with the dual objectives of minimizing earthwork haulage distance and achieving the most uniform backfill thickness of the topsoil layer. With the constraint that the balance deviation of excavation and filling volume does not exceed a preset threshold, the improved particle swarm optimization algorithm is used to solve for the optimal merged design elevation.

[0074] S4: After determining the optimal design elevation, an earthwork allocation network model is established, and an improved ant colony algorithm is used to solve it, so as to obtain the earthwork allocation scheme of the transportation direction and transportation volume from excavation to filling between each grid.

[0075] S5: Convert the earthwork allocation plan into a grader operation path plan and generate path node instructions that can be executed by the vehicle controller;

[0076] S6: The grader operates according to the path node instructions and collects the elevation data of the leveled area in real time through the Beidou RTK positioning system during the construction process. It compares the measured elevation with the design elevation grid by grid and triggers the correction process for areas where the deviation exceeds the preset threshold, forming a closed-loop control.

[0077] This method is applicable to high-standard farmland construction, land consolidation, and sloping farmland transformation projects in low mountain and hilly areas nationwide, with broad application prospects and significant economic and social benefits. By optimizing the topsoil stripping process beforehand, the loss of topsoil is reduced, effectively protecting soil fertility. Optimizing the design elevation with the dual objectives of minimizing earthmoving distance and achieving the most uniform topsoil backfill thickness results in less total earthwork and significantly improved cut-fill balance accuracy compared to experience-based construction, thus significantly reducing project costs. Linking the real-time earthwork allocation scheme with the BeiDou-based unmanned grader's operating path enables integrated intelligent construction from "optimization calculation to construction execution," reducing manual judgment and shortening the construction period. A dynamic closed-loop correction mechanism ensures continuous control of flatness during construction, guaranteeing a high pass rate for field flatness and meeting the stringent requirements of high-standard farmland construction for field flatness (error ≤ ±3cm).

[0078] According to one embodiment of this application, in step S1, the high-precision digital elevation model generation method includes:

[0079] S11: Collect the original topographic data of the fields to be merged, acquire point cloud data using UAV oblique photography or LiDAR, generate a high-precision digital elevation model (DEM) after filtering, denoising and ground point classification, and set the grid size of the high-precision digital elevation model.

[0080] A high-precision digital elevation model (DEM) is essentially an ordered array of numerical values, using the mathematical function Z=f(x,y) to describe the Earth's surface morphology. It contains only surface elevation information, such as the elevation of bare ground, but not the height of buildings, vegetation, or other coverings. It is typically stored as a regular grid, with each grid cell containing an elevation value at its center. In this example, the grid size can be set to 0.3–1.0 m.

[0081] According to one embodiment of this application, the specific method for automatically dividing and merging units using the region growing algorithm in step S1 is as follows:

[0082] S12: Seed point selection: In the high-precision digital elevation model, the slope value and relative elevation difference of each grid cell are used as the initial classification criteria. Grid cells with a slope not higher than the first slope threshold (e.g., ≤3°) and an elevation difference not higher than the first elevation difference threshold (e.g., ≤0.5m) are selected as candidate seed points. Based on this, the boundaries of adjacent fields (natural field ridges, field roads, ditches) are used as the dividing constraints. In each independent closed area enclosed by natural boundaries, the grid cell with the lowest elevation near the geometric center is selected as the initial seed point of that field.

[0083] S13: Region Growth Criteria: Design growth criteria, starting from the initial seed point, incorporate adjacent grids that satisfy the growth criteria into the current region and use them as new growth starting points, and expand outward layer by layer in the 8-neighborhood directions (east, south, west, north, southeast, southwest, northwest, northeast); grids that do not satisfy the growth criteria (if any one of the conditions is not met, then the growth criteria are not satisfied) are marked as boundary points, and growth in that direction is stopped;

[0084] The growth criteria are: the slope value of adjacent grid cells is not higher than the second slope threshold (e.g., ≤5°), the elevation difference with the seed point is not higher than the second elevation difference threshold (e.g., ≤0.8m), and they do not belong to natural boundary grids (field ridges, roads, water systems).

[0085] S14: Region Merging: After all seed points have completed their regional growth, adjacent small regions with an area smaller than S (e.g., 5 mu) are merged. If the elevation difference between the boundaries of two adjacent regions is ≤0.3m and the total area after merging is not less than 5 mu, then the two regions are merged into one merged unit. If the merged area is still less than 5 mu, then the merged unit is merged with the next adjacent region until all merged units have an area ≥5 mu.

[0086] S15: Merged Unit Output: Output the boundary coordinate sequence (longitude, latitude) and area statistics of each merged unit as the basic data unit for subsequent step 2 topsoil sampling and step 3 elevation optimization.

[0087] According to one embodiment of this application, in step S2, topsoil thickness is sampled within each merging unit (no less than 3 sampling points per 5-10 mu), and the topsoil thickness at each point is measured using the profile method (from the surface to the upper interface of the plow layer), thus establishing a topsoil thickness distribution map; the topsoil stripping amount is set as the average thickness of each sampling point, and the stripping depth is calculated according to the following formula:

[0088] ;

[0089] In the formula, D strip d represents the peeling depth (cm). i Let be the topsoil thickness (cm) at the i-th sampling point; n be the total number of sampling points; the thickness of a single stripping operation shall not exceed 30cm; for areas with deep soil layers, stripping and storage can be carried out in layers; determine the stripping boundary and storage location, and determine the storage point with the shortest transportation distance after stripping as the objective.

[0090] According to one embodiment of this application, in step S3, the expression of the objective function is:

[0091] F = min(w1·F1 + w2·F2);

[0092] Earthwork transport distance function: ;

[0093] Topsoil backfill uniformity function: ;

[0094] Among them, h k Let be the backfill thickness of the k-th backfill grid. is the average backfill thickness; p is the total number of backfill grids; w1 and w2 are weighting coefficients, and w1+w2=1, which are determined according to the actual situation of the project (generally w1=0.6, w2=0.4).

[0095] According to one embodiment of this application, in step S3, the specific steps of the improved particle swarm optimization algorithm to solve for the optimal merged design elevation are as follows:

[0096] S31: Parameter initialization: Set the number of particles N=80 (range 50~100), and the maximum number of iterations T. max =500 (value range 300~1000), learning factor c1=c2=1.5, inertia weight w linearly decreases from 0.9 to 0.4, and the design elevation H is encoded as the particle position vector x. i (i=1,2,…,N), initial position is in [H min H max Randomly generated within the interval; initial velocity v i In [-0.1×(H max -H min ), 0.1×(H max -H min Randomly generated within the interval;

[0097] S32: Fitness assessment: For each particle x i Substitute the corresponding candidate design elevations into the objective function F = min(w1·F1 + w2·F2) to calculate the fitness value; simultaneously check the balance deviation of cut and fill volumes. The constraints are applied, and for particles that do not meet the constraints, a penalty term is added to their fitness value. Where M is the penalty coefficient, which is set to 10. 6 This significantly degrades the fitness of infeasible solutions, guiding the algorithm to search for feasible solutions.

[0098] S33: Individual Optimal and Global Optimal Update: This involves updating each particle's current fitness value with its historical best position p. Best The fitness values ​​are compared, and if the current value is better, then p is updated. Best The position corresponding to the optimal fitness value among all particles is denoted as the global optimal position g. Best ;

[0099] S34: Particle velocity and position update: Update the velocity and position of each particle according to the following formula:

[0100] v i (t+1) = w(t)·v i (t) + c1·rand1·(p Besti - x i (t))+ c2·rand2·(g Best - x i (t));

[0101] x i (t+1) = x i (t) + v i (t+1);

[0102] Where, w(t) = 0.9 - (0.9 - 0.4) × t / T max w(t) is the inertia weight (which decreases linearly from 0.9 to 0.4 with the iteration number t); rand1 and rand2 are random numbers in the interval [0,1].

[0103] Apply boundary constraints to the updated velocity and position; specifically, apply boundary constraints after the velocity update: if vi exceeds [-0.2×(H max -H min ), 0.2×(H max -H min If xi exceeds the range of [H], then truncate it to the boundary value; if xi exceeds the range of [H], then truncate it to the boundary value. min H max If the range is not specified, then pull it back to the nearest boundary value.

[0104] This invention makes the following improvements based on existing particle swarm optimization algorithms:

[0105] (1) The topsoil stripping thickness is embedded as an independent constraint in the objective function, and the weight coefficients w1 and w2 can be adjusted according to the actual situation of the project;

[0106] (2) In view of the characteristics of large topographic relief and irregular field boundaries in low hilly and fragmented fields, the inertial weight of the particle swarm algorithm is dynamically adjusted from 0.9 to 0.4 linearly, and the number of particles is set to 50 to 100 and the maximum number of iterations is set to 300 to 1000 to adapt to the calculation scale of DEM grid (0.3 to 1.0 m) of hilly fields;

[0107] (3) The algorithm output results are directly coupled with the operation path planning module of the Beidou unmanned leveling machine, filling the technical gap between algorithm optimization and mechanical execution.

[0108] S35: Iteration Termination Judgment: Repeat steps S32 to S34 until the maximum number of iterations Tmax = 500 is reached, or the global optimal solution g is obtained by N consecutive iterations (which can take the value 50). Best The iteration terminates when the change is less than the preset threshold ε = 0.001m, and the global optimal position g is output. Best The optimal merging design elevation H.

[0109] According to one embodiment of this application, in step S4, the specific steps of the improved ant colony algorithm to solve the earthwork allocation scheme are as follows:

[0110] S41: Allocation Network Construction: Divide the merging unit into allocation grids of a set size (e.g., 10m × 10m). Calculate the cut volume (positive value) or fill volume (negative value) for each grid based on the optimal design elevation H. Assign all cut grids to the source point set S = {s1, s2, ..., s}. m All fill grids are assigned to the sink set T = {t1, t2, ..., t}. n Construct a complete bipartite network graph G=(S,T,E); where E is the set of all possible dispatch paths, and the path weight is the Euclidean distance L between the two points. ij ;

[0111] S42: Ant colony initialization: Set the number of ants K (range 30-80), and the maximum number of iterations I. termax =300 (range 200-500), pheromone volatility coefficient ρ=0.1 (range 0.05-0.2), pheromone intensity Q=100, pheromone heuristic factor α=1, expected heuristic factor β=2, initialize pheromone concentration τ on each path. ij (0) = τ0 (τ0 is a constant 1.0), indicating that the pheromone concentration of each path is equal at the initial time;

[0112] S43: Path Selection and Earthwork Allocation: Each ant starts from the first cut grid s1 and selects the corresponding fill grid for each cut grid in turn. The ant is at node s1. i Select the next fill grid. j The transition probability is:

[0113] ;

[0114] in, The heuristic information is (the shorter the distance, the higher the probability of being selected); allowed is the set of filling grids that have not yet been fully allocated; after each ant completes a full allocation plan construction, the total transportation distance D = ΣΣ L is calculated. ij ·V ij .

[0115] S44: Pheromone Update: After all ants have completed one iteration, the pheromone concentration on each path is updated according to the following formula:

[0116] ;

[0117] Wherein, if ant k chooses path ij in this iteration, then ;otherwise, =0; The total transportation distance of the allocation plan constructed for ant k;

[0118] S45: Iteration Termination and Optimal Solution Output: Repeat steps S43 to S44 until the maximum number of iterations I is reached. termax =300, or the iteration terminates when the total change in the optimal solution for N consecutive generations (which can be 50) is less than the preset threshold δ=1.0m·m³; output the global optimal allocation scheme, which includes the transportation direction and transportation volume from each cut grid to each fill grid.

[0119] According to one embodiment of this application, the specific method for converting the earthwork allocation plan into a grader operation path plan in step S5 is as follows:

[0120] S51: Classification and Boundary Extraction of Work Areas: Based on the allocation plan output in step S4, the net earthwork volume of each grid is calculated. Grids with a net earthwork volume greater than zero are classified as "excavation areas", grids with a net earthwork volume less than zero are classified as "fill areas", and grids with a net earthwork volume equal to zero are classified as "leveling areas". The outer boundary polygons of each type of area are extracted (using the convex hull algorithm or the Alpha-shape algorithm to extract the boundary point sequence).

[0121] S52: Path type matching: Match the path type according to the geometric characteristics of various areas. For excavation areas, which require cutting soil layer by layer from the edge to the center, a spiral path is matched; for fill areas, which require spreading row by row along the long side, a reciprocating path is matched; for leveling areas, which require full coverage and leveling, a "Z" shaped path is matched.

[0122] S53: Path node generation: For the spiral path in the excavation area, take the outer boundary polygon as the starting outline and the width of the flat shovel (2.5m) as the step size to generate a sequence of inward-shrinking polygons layer by layer; connect the corresponding vertices of adjacent polygons in sequence to form a spiral path node sequence.

[0123] For the reciprocating path in the fill area, the direction of travel is the long side of the rectangle circumscribed in the fill area, and the row spacing is the width of the leveling shovel (2.5m). A sequence of reciprocating broken line path nodes parallel to the long side is generated.

[0124] For the "Z"-shaped path in the flat area, the node sequence of the "Z"-shaped polyline path is generated with the bounding rectangle of the flat area as the range and the short side direction as the stepping direction.

[0125] S54: Path Node Assignment: Assign the following attributes to each path node: coordinates (longitude and latitude, obtained by interpolation from the center coordinates of the DEM grid), target elevation (in the excavation area, the elevation gradually decreases from the initial elevation to the design elevation H, with a change rate ≤ 3cm / m; in the fill area, the elevation gradually increases from the design elevation H to the target elevation; in the leveling area, H is taken directly), travel speed (1.5km / h in the excavation area, 2.0km / h in the fill area, and 2.5km / h in the leveling area), and shovel lifting command (calculated based on the elevation difference between the current node and the previous node).

[0126] S55: Command output: The path node sequence is encoded sequentially into an instruction format executable by the vehicle controller, and executed in real time via Beidou RTK differential positioning;

[0127] The field grid is divided into "cutting area", "filling area" and "leveling area":

[0128] Excavation area: A spiral path is adopted, advancing spirally from the edge of the excavation area towards the center to reduce empty driving mileage;

[0129] Fill area: Use a reciprocating path, with the direction of travel parallel to the long side of the field;

[0130] Leveling area: Use a "Z" shaped path.

[0131] The generated vehicle controller commands include: path node coordinate sequence (longitude, latitude), target elevation of each node, travel speed commands (1.5km / h in excavation area, 2.0km / h in filling area, and 2.5km / h in leveling area), and shovel lifting commands, which can be executed in real time via Beidou RTK differential positioning.

[0132] According to one embodiment of this application, the dynamic correction process in step S6 is specifically as follows:

[0133] The BeiDou RTK positioning system on the grader collects elevation data of the leveled area in real time at a set sampling frequency and uploads it to a cloud server via wireless network; the server then compares the measured elevation with the design elevation grid by grid.

[0134] If the deviation is less than or equal to the first set value, it is marked as "qualified" and no correction is triggered.

[0135] If the deviation is between the first set value and the second set value, it is marked as a warning and the deviation value is recorded; wherein, the first set value is less than the second set value;

[0136] If the deviation is greater than the second set value, it is marked as "unqualified" and a correction process is triggered. A correction area of ​​a set size (such as 3m×3m) is delineated with the deviation grid as the center, the excavation and filling volume of the area is recalculated, a local correction command is generated, and it is sent to the vehicle controller through the cloud for execution, forming a closed-loop control of "collection-comparison-correction-execution-recollection".

[0137] The following example, a high-standard farmland construction project in Gaoping District, Nanchong City, Sichuan Province, further illustrates this method:

[0138] In the high-standard farmland construction project in Gaoping District, Nanchong City, Sichuan Province, a typical consolidation unit was selected for construction. This consolidation unit consists of 12 scattered small plots, with a total area of ​​approximately 18 mu (about 1.2 hectares). The terrain is gentle hills with a natural slope of 5° to 12° and a surface elevation variation of about 1.2 meters. The consolidation goal is to form a rectangular field (120m long side and 60m short side) with an area of ​​no less than 10 mu (about 0.67 hectares) for large-scale planting of silage corn.

[0139] Step 1: Data Acquisition and DEM Modeling

[0140] Aerial surveying of the merged units was conducted using a UAV equipped with a lidar at an altitude of 80m, with a forward overlap of 80% and a lateral overlap of 60%. After acquiring point cloud data, the data was filtered, denoised, and classified to generate a digital elevation model (DEM) with a grid resolution of 0.5m × 0.5m. Simultaneously, the elevations of at least 30 feature points at the four corners and within the fields were measured using RTK-GNSS to verify the accuracy of the DEM, ensuring that the elevation error was ≤ ±3cm.

[0141] Step 2: Topsoil thickness survey and stripping scheme determination

[0142] Sampling points were set up within the merged unit at a density of no less than 3 sampling points per 5 mu (approximately 0.33 hectares), for a total of 12 sampling points. The topsoil thickness (from the surface to the upper interface of the plow layer) at each point was measured using the profile method. The measured data are as follows:

[0143]

[0144] The average topsoil thickness D was calculated. strip = 19.3cm, rounded to 20cm. The stripping boundary is determined to be a range of 1m outside the merged unit. The topsoil storage area (approximately 200m²) is located near the geometric center of the merged unit with higher terrain and better drainage conditions. The ground of the storage area is covered with impermeable geotextile, and simple drainage ditches are excavated around it.

[0145] Step 3: Bi-objective optimization of merging design elevations

[0146] Using an 18-mu (approximately 1.2 hectares) unit as the calculation range, the DEM was gridded into approximately 48,000 grids (0.5m × 0.5m). The target field was set as a rectangle (120m × 60m), and the design elevation H was the variable to be optimized.

[0147] Construct the objective function F = min(0.6·F1 + 0.4·F2), with the constraint that the balance deviation of cut and fill volumes is ≤5%, H min ≤ H ≤ H max .

[0148] An improved particle swarm optimization algorithm is used to solve the problem: 80 particles, 500 maximum iterations, inertia weight decreasing linearly from 0.9 to 0.4, and learning factors c1=c2=1.5.

[0149] After optimization calculation, the optimal design elevation H = 356.2m (Yellow Sea elevation), and the total excavation volume V 挖= 2860m³, total fill volume V 填 = 2780m³, balance deviation 2.8% (meets ≤5% constraint), weighted average earthwork haul distance L avg = 42m, Standard deviation σ of topsoil backfill thickness h = 1.8cm.

[0150] Step 4: Optimization of Earthwork Allocation Plan

[0151] The merging unit was divided into a 10m × 10m grid (approximately 720 grids in total). An improved ant colony algorithm was used to solve the problem: 50 ants, 300 iterations, pheromone evaporation coefficient ρ = 0.1, and pheromone intensity Q = 100.

[0152] Optimized calculations show that the excavation area is mainly concentrated on the southwest side of the merged unit, with a total area of ​​approximately 3.2 acres and a total excavation volume of 2860 m³; the fill area is mainly concentrated on the northeast side of the merged unit, with a total area of ​​approximately 3.8 acres and a total fill volume of 2780 m³. A total of 86 allocation routes are formed, with the farthest allocation distance being 68 m and the average allocation distance being 42 m, reducing the total earthwork transportation distance by approximately 22% compared to the empirical scheme.

[0153] Step 5: Planning the working path of the Beidou automatic grader

[0154] According to the earthwork allocation plan, the field is divided into excavation, filling, and leveling areas. The excavation area uses a spiral path (progressing spirally from the edge to the center), the filling area uses a reciprocating path (parallel to the long side of the field), and the leveling area uses a zigzag path. The generated vehicle controller commands include the path node coordinate sequence, the target elevation of each node, the travel speed command (1.5 km / h for the excavation area, 2.0 km / h for the filling area, and 2.5 km / h for the leveling area), and the shovel lifting command.

[0155] Step 6: Dynamic Correction During Construction

[0156] During construction, the BeiDou RTK positioning system on the grader collects real-time elevation data of the leveled area (sampling frequency 10Hz) and uploads it to a cloud server. The server compares the measured elevation with the design elevation grid by grid. When the deviation exceeds ±3cm, a warning is issued; when it exceeds ±5cm, a correction process is triggered. A total of 23 correction commands were triggered, covering a total area of ​​approximately 120m², accounting for 1.1% of the total area.

[0157] Implementation results: The total construction period was 13 days (2 days ahead of schedule), and the total earthwork volume was 5640 m³, which is 17.1% less than the estimated 6800 m³ of the experienced construction method. After the construction was completed, the RTK-GNSS full coverage elevation verification (with a measuring point spacing of 5m×5m, totaling approximately 1440 measuring points) showed a pass rate (deviation ≤±3cm) of 96.3%, meeting the technical requirements for high-standard farmland construction.

[0158] Secondly, the present invention discloses an earthwork balancing and construction path planning device for fragmented fields in low hills, comprising: a processor and a memory; wherein, the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory to enable the device to perform the above-mentioned method.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for earthwork balance and construction path planning in fragmented fields in low hills, characterized in that, Includes the following steps: S1: Collect the original topographic data of the fields to be merged and generate a high-precision digital elevation model; identify natural field ridges, roads and water system boundaries according to the field merging target, and automatically divide the merging units using a region growing algorithm; S2: Sample the topsoil thickness in each merged unit, establish a topsoil thickness distribution map, determine the topsoil stripping scheme, and determine the topsoil storage location with the shortest transportation distance after stripping as the target. S3: Using the merged unit as the basic calculation unit, an optimization model is established with the dual objectives of minimizing earthwork haulage distance and achieving the most uniform backfill thickness of the topsoil layer. With the constraint that the balance deviation of excavation and filling volume does not exceed a preset threshold, the improved particle swarm optimization algorithm is used to solve for the optimal merged design elevation. S4: After determining the optimal design elevation, an earthwork allocation network model is established, and an improved ant colony algorithm is used to solve it, so as to obtain the earthwork allocation scheme of the transportation direction and transportation volume from excavation to filling between each grid. S5: Convert the earthwork allocation plan into a grader operation path plan and generate path node instructions that can be executed by the vehicle controller; S6: The grader operates according to the path node instructions and collects the elevation data of the leveled area in real time through the Beidou RTK positioning system during the construction process. It compares the measured elevation with the design elevation grid by grid and triggers a correction process for areas where the deviation exceeds the preset threshold, thus forming a closed-loop control.

2. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S1, the high-precision digital elevation model generation method includes: S11: Collect the original topographic data of the fields to be merged, use UAV oblique photography or LiDAR to acquire point cloud data, and after filtering, noise reduction and ground point classification, generate a high-precision digital elevation model and set the grid size of the high-precision digital elevation model.

3. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1 or 2, characterized in that, In step S1, the specific method for the region growing algorithm to automatically divide and merge units is as follows: S12: Seed point selection: In the high-precision digital elevation model, the slope value and relative elevation difference of each grid cell are used as the initial classification criteria. Grid cells with a slope not higher than the first slope threshold and an elevation difference not higher than the first elevation difference threshold are selected as candidate seed points. Adjacent field boundaries are used as dividing constraints. In each independent closed area enclosed by natural boundaries, the grid cell with the lowest elevation near the geometric center is selected as the initial seed point of that field. S13: Region Growth Criteria: Design growth criteria, starting from the initial seed point, incorporate adjacent grids that meet the growth criteria into the current region and use them as new growth starting points, expanding outward layer by layer in the 8-neighborhood direction; mark grids that do not meet the growth criteria as boundary points and stop growth in that direction; the growth criteria are: the slope value of adjacent grid cells is not higher than the second slope threshold, the elevation difference with the seed point is not higher than the second elevation difference threshold, and they do not belong to natural boundary grids; S14: Region merging: After the region growth of all seed points is completed, the adjacent small regions with an area less than S are merged. If the elevation difference between the boundaries of two adjacent regions is not higher than the first elevation difference threshold and the total area after merging is not less than S mu, then the two regions are merged into one merged unit. S15: Merged Cell Output: Outputs the boundary coordinate sequence and area statistics for each merged cell.

4. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S2, the topsoil stripping scheme calculates the stripping depth according to the following formula: ； In the formula, D strip d represents the peeling depth. i The topsoil thickness is denoted by ; n is the total number of sampling points; the thickness of each stripping operation is no more than 30 cm; for areas with deep soil layers, the stripping and storage are performed in layers.

5. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S3, the expression for the objective function is: ； Earthwork transport distance function: ; Topsoil backfill uniformity function: ; Among them, h k Let be the backfill thickness of the k-th backfill grid. is the average backfill thickness; p is the total number of backfill grids; w1 and w2 are weighting coefficients, and w1+w2=1.

6. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1 or 5, characterized in that, In step S3, the specific steps of the improved particle swarm optimization algorithm to solve for the optimal merged design elevation are as follows: S31: Parameter initialization: Set the number of particles N and the maximum number of iterations T. max The learning factor c1=c2, the inertia weight w decreases linearly, and the design elevation H is encoded as the particle position vector x. i The initial position is in [H min H max Randomly generated within the interval; S32: Fitness assessment: For each of the stated particles x i Substitute the corresponding candidate design elevation into the objective function, calculate the fitness value, and check the balance deviation of cut and fill volumes. For particles that do not meet the constraints, a penalty term is added to their fitness value. ; S33: Individual Optimal and Global Optimal Update: This involves updating each particle's current fitness value with its historical best position p. Best The fitness values ​​are compared, and if the current value is better, then p is updated. Best The position corresponding to the optimal fitness value among all particles is denoted as the global optimal position g. Best ; S34: Particle velocity and position update: Update the velocity and position of each particle according to the following formula: v i (t+1) = w(t)·v i (t) + c1·rand1·(p Besti - x i (t))+ c2·rand2·(g Best - x i (t)); x i (t+1) = x i (t) + v i (t+1); Where w(t) is the inertia weight; rand1 and rand2 are random numbers in the interval [0,1]. Apply boundary constraints to the updated velocity and position; S35: Iteration Termination Judgment: Repeat steps S32 to S34 until the maximum number of iterations is reached, or the global optimal solution g is obtained for N consecutive generations. Best The iteration terminates when the change is less than a preset threshold ε, and the global optimal position g is output. Best The optimal merging design elevation H.

7. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S4, the specific steps of the improved ant colony algorithm for solving the earthwork allocation scheme are as follows: S41: Network construction: Divide the merged unit into a set size of network grid, calculate the cut or fill volume of each grid according to the optimal design elevation H, assign all cut grids to the source set S, assign all fill grids to the sink set T, and construct a complete bipartite network graph G = (S, T, E). S42: Ant colony initialization: Set the number of ants K and the maximum number of iterations I. termax Pheromones evaporation coefficient ρ, pheromone intensity Q, pheromone heuristic factor α, expected heuristic factor β, and initial pheromone concentration τ along each path. ij (0) = τ0; S43: Path Selection and Earthwork Allocation: Each ant starts from the first cut grid and selects the corresponding fill grid for each cut grid in turn. The ant is at node s i Select the next fill grid. j The transition probability is: ； in, This is heuristic information; allowed is the set of fill grids that have not yet been fully allocated. S44: Pheromone Update: After all ants have completed one iteration, the pheromone concentration on each path is updated according to the following formula: ; Wherein, if ant k chooses path ij in this iteration, then ;otherwise, =0; The total transportation distance of the allocation plan constructed for ant k; S45: Iteration Termination and Optimal Solution Output: Repeat steps S43 to S44 until the maximum number of iterations I is reached. termax The iteration terminates when the total transport distance change of the optimal solution for N consecutive generations is less than the preset threshold δ; the global optimal allocation scheme is output.

8. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S5, the specific method for converting the earthwork allocation plan into a grader operation path plan is as follows: S51: Classification and boundary extraction of work area: According to the allocation plan output in step S4, the net earthwork volume of each grid is calculated. Grids with a net earthwork volume greater than zero are classified as excavation areas, grids with a net earthwork volume less than zero are classified as fill areas, and grids with a net earthwork volume equal to zero are classified as leveling areas. The outer boundary polygons of each type of area are extracted respectively. S52: Path type matching: Match spiral paths for cut areas, reciprocating paths for fill areas, and zigzag paths for leveling areas; S53: Path Node Generation: Using the outer boundary polygon of the excavation area as the starting outline, and with the width of the leveling shovel as the step size, generate an inward shrinking polygon sequence layer by layer and connect them to form a spiral path node sequence; using the long side direction of the circumscribed rectangle of the fill area as the direction of travel, and with the width of the leveling shovel as the row spacing, generate a reciprocating polyline path node sequence; using the circumscribed rectangle of the leveling area as the range, and with the short side direction as the step size, generate a "Z"-shaped polyline path node sequence. S54: Path Node Assignment: Assign each path node with coordinates, target elevation, travel speed, and shovel lifting / lowering command attributes; S55: Command output: Encode the path node sequence in sequence into an instruction format executable by the vehicle controller; the path node instruction includes the path node coordinate sequence, the target elevation of each node, the travel speed instruction, and the shovel lifting instruction.

9. The method for earthwork balance and construction path planning in fragmented fields in low hills and mountains according to claim 1, characterized in that, In step S6, the dynamic correction process specifically includes: The BeiDou RTK positioning system on the grader collects elevation data of the leveled area in real time at a set sampling frequency and uploads it to a cloud server via wireless network; the server then compares the measured elevation with the design elevation grid by grid. If the deviation is less than or equal to the first set value, it is marked as qualified; If the deviation is between the first set value and the second set value, it is marked as a warning and the deviation value is recorded; wherein, the first set value is less than the second set value; If the deviation is greater than the second set value, it is marked as unqualified and a correction process is triggered. A correction area of ​​a set size is delineated with the deviation grid as the center, the excavation and filling volume of the area is recalculated, a local correction instruction is generated and sent to the vehicle controller for execution via the cloud.

10. A device for earthwork balancing and construction path planning in fragmented fields in low hills and mountains, characterized in that, include: A processor and a memory; wherein the memory is used to store computer execution instructions, and when the device is running, the processor executes the computer execution instructions stored in the memory to cause the device to perform the method of any one of claims 1-9.

Images