Three-dimensional path planning method for unmanned aerial vehicle based on 26-direction moving operator compression encoding

Abstract

The application relates to the technical field of unmanned aerial vehicle path planning, and specifically provides a three-dimensional path planning method for an unmanned aerial vehicle based on 26-direction mobile operator compression coding, which comprises the following steps: based on the starting point and the target endpoint of the unmanned aerial vehicle, an initial path set of the path of the unmanned aerial vehicle is formed; the displacement between adjacent grid coordinate points in the initial path is mapped into corresponding mobile operators through a mapping relationship table, so that a compression coding path is generated; the compression coding path is subjected to an iterative optimization algorithm, so that an optimized compression coding path is obtained; in the case that there are conflict points in the optimized compression coding, the conflict points are subjected to trend guidance through the mobile operators, so that feasible points are obtained; the optimized compression coding path is updated according to the feasible points, and the final optimized path is determined after the iterative optimization algorithm is terminated. The problems that the path planning of the unmanned aerial vehicle is low in efficiency in a large-scale map and the collision repair and re-planning are long in time in the related art are solved.

Description

Technical Field

[0001] This invention relates to the field of UAV path planning technology, and in particular to a UAV three-dimensional path planning method based on 26-axis motion operator compression coding. Background Technology

[0002] Path planning for unmanned aerial vehicles (UAVs) in complex three-dimensional environments is a key technology for ensuring their autonomous operation, and it is widely used in urban inspection, disaster relief, and large-area surveying. The way a path is encoded during path optimization directly determines the scale of the decision space and the efficiency of the algorithm search, representing a core bottleneck in three-dimensional path planning technology.

[0003] In related technologies, the computational load for UAV path planning is enormous when expanding nodes in a 3D grid, easily leading to the curse of dimensionality. Although path planning methods possess global search capabilities, directly optimizing continuous coordinate points results in an excessively high dimensional solution space, slow convergence speed, and a tendency to get trapped in local optima. Furthermore, if a collision point is detected after path optimization, existing solutions typically require global replanning or random jitter correction. The collision detection and replanning processes are complex and time-consuming, making it difficult to meet the real-time response requirements of large-scale maps.

[0004] There is currently no effective solution to the problems of low efficiency in UAV path planning on large-scale maps and time-consuming collision repair and replanning in related technologies. Summary of the Invention

[0005] The present invention provides a UAV 3D path planning method based on 26-axis motion operator compression coding, which at least solves the problems of low efficiency of UAV path planning in large-scale maps and long time consumption of collision repair and replanning in related technologies.

[0006] According to one aspect of the present invention, a method for UAV 3D path planning based on 26-direction movement operator compressed coding is provided, comprising: constructing an initial path set for the UAV path based on the UAV's starting point and target endpoint; wherein the UAV path is composed of continuous grid coordinate points; the initial path set includes multiple initial paths; mapping the displacements between adjacent grid coordinate points in the initial path to corresponding movement operators through a mapping table to generate a compressed coded path; wherein the movement operators are integers, each integer uniquely corresponding to a 3D displacement direction; obtaining an optimized compressed coded path through an iterative optimization algorithm; in the case of conflict points in the optimized compressed coding, trend guidance is provided to the conflict points through movement operators to obtain feasible points; wherein the conflict points are grid coordinate points in the infeasible region after the optimized compressed coding is decompressed, and the feasible points are grid coordinate points in the feasible region; updating the optimized compressed coded path according to the feasible points, and determining the final optimized path after the iterative optimization algorithm terminates.

[0007] As an optional approach, when conflict points exist in the optimized compression coding, a trend guidance method is used to guide the conflict points to obtain feasible points. This includes: decompressing the optimized compression coding path to obtain a grid coordinate path in three-dimensional space; determining the nearest neighboring feasible points in the path before and after the conflict point based on the conflict point; wherein, the neighboring feasible points include the grid coordinate points from the conflict point to the previous feasible region and to the grid coordinate points to the next feasible region; calculating the trend guidance vector from the conflict point to the two neighboring feasible points using a distance formula, and obtaining candidate points through calculation; if the candidate point is in a feasible region, the candidate point is taken as a feasible point; if the candidate point is in an infeasible region, a new conflict point is obtained through the movement operator, and iterative optimization is performed until a feasible point is obtained.

[0008] As an optional approach, obtaining new conflict points through the movement operator includes: moving the candidate points one by one using the movement operator to obtain a set of updated points; calculating the absolute difference between each updated point and the grid coordinate point in the next feasible domain along each coordinate axis, and summing the absolute differences along each coordinate axis to obtain the displacement distance corresponding to each updated point; and selecting the updated point with the smallest displacement distance as the new conflict point.

[0009] As an optional approach, the displacements between adjacent grid coordinate points in the initial path are mapped to corresponding movement operators through a mapping table to generate a compressed coded path. This further includes: dividing the three-dimensional space of the UAV flight into cubic grids of equal side length to obtain a three-dimensional grid map; wherein the three-dimensional grid map includes the grid coordinates of the center point of the cubic grid and the feasible and infeasible regions of the three-dimensional space; determining the grid displacement between adjacent grid points based on adjacent grid points in the initial path set; wherein the adjacent grid points include the vertices, midpoints of edges, and midpoints of faces of a cube centered on the target grid point; mapping the grid displacement to movement operators through a mapping table; and determining the compressed coded path based on the movement operators.

[0010] As an optional approach, the compressed coding path is optimized through an iterative optimization algorithm, including: obtaining current iteration information and calculating dynamic inertia weights; wherein the current iteration information includes the current iteration count, the maximum iteration count, and the minimum and maximum values ​​of the inertia weights; based on the dynamic inertia weights and the algorithm parameters of the optimization algorithm, combined with the velocity vector, position vector, and historical best solutions of the current iteration cycle, the next-generation velocity vector of each path in the compressed coding path is calculated using a velocity update formula; wherein the algorithm parameters include two random numbers and two acceleration coefficients; based on the next-generation velocity vector, the next-generation position vector is calculated using a position vector calculation formula; wherein the calculated result of the position vector is rounded up to the nearest integer and used to represent multiple movement operators in the path; based on the next-generation position vector, an optimized compressed coding path is determined, wherein the optimized compressed coding path consists of multiple movement operators, each movement operator corresponding to a decision variable dimension; the position vector is used to represent the value of each decision variable dimension.

[0011] As an optional approach, the expression for the position vector calculation formula is:

[0012] ;

[0013] in, Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The velocity vector of the decision variables;

[0014] ;

[0015] in, c1 and c2 represent the inertia weight; c1 and c2 represent the acceleration coefficients. r1 and r2 represent random numbers. ; Represented as a group in the th In the nth iteration, the 1st The historical optimal solution for the decision variable;

[0016] ;

[0017] in, , ; For the current algebra, This represents the maximum number of iterations.

[0018] As an optional approach, the compressed coding path is optimized through an iterative optimization algorithm. The method further includes: mapping the optimized compressed coding path to a grid coordinate path using a mapping table; wherein the grid coordinate path is the grid coordinate representation of the UAV flight path in three-dimensional space; calculating the fitness of the grid coordinate path for each optimization objective using a fitness formula for each objective; wherein the fitness formula is used to evaluate the quality of the optimized compressed coding path; sorting the fitness values ​​using a non-dominated sorting algorithm, selecting the optimal solution, and archiving it to obtain archived particles; generating a next-generation optimized compressed coding path using the iterative optimization algorithm based on the archived particles, and accumulating the iteration count; and determining the final optimized compressed coding path based on the next-generation optimized compressed coding path when the number of iterations reaches a preset number.

[0019] As an alternative approach, the optimization objectives include: path length, number of turns, and safety.

[0020] As an optional approach, an initial path set for the UAV's path is constructed based on the UAV's starting point and target endpoint, including: obtaining a feasible path based on the flight parameters and terrain environment data acquired by the UAV using a pathfinding algorithm; randomly generating intermediate nodes between the starting point and the target endpoint based on the UAV's starting point and target endpoint; generating multiple random paths based on the starting point, the target endpoint, and the intermediate nodes; and using the feasible path and the multiple random paths as the initial path set.

[0021] According to another aspect of the present invention, an electronic device is also provided, comprising: a processor, and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the method described in any of the preceding claims.

[0022] This invention provides a UAV 3D path planning method based on 26-axis mover compression encoding. By encoding the 3D path into a compact integer sequence using 26-axis movers, the search space and computational load of subsequent iterative optimization algorithms are greatly reduced, thereby effectively improving the path planning efficiency on large-scale maps. Furthermore, when collision points appear during the decompression of the optimized encoded path, the algorithm does not need to perform time-consuming global replanning. Instead, it utilizes the geometric semantics of the movers to provide local trend guidance and rapid repair at the collision point, generating feasible points and updating the path. This significantly shortens the collision repair time, thus solving the problems of low efficiency and time-consuming collision repair replanning in related technologies for UAV path planning on large-scale maps. Attached Figure Description

[0023] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other embodiments based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a UAV three-dimensional path planning method based on 26-axis motion operator compressed coding, according to an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the displacement direction of the displacement operator in the grid according to an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the trend-guided iterative optimization process of an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of the structure of the electronic device created by this invention. Detailed Implementation

[0028] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0029] Related technologies, such as UAV 3D grid path planning technology, all use a 3D coordinate array of path points as the encoding carrier in large-scale maps. If the path contains N path points, the overall search space in an L×W×H 3D grid map is (L×W×H). N As the map size increases, it is prone to the curse of dimensionality, and the intelligent optimization algorithm has a large number of iterations and convergences, resulting in extremely low search efficiency.

[0030] Collision repair in infeasible regions using UAV 3D grid path planning is mostly aimed at global trajectories or sudden collisions, with low efficiency in repairing locally infeasible path points. If only a few path points fall into the infeasible region, the relevant technology will directly determine the entire path as infeasible and discard it, ignoring the effective information that most path points are already in the feasible region. This results in extremely low utilization of feasible solutions, and repair or replanning is time-consuming, making it unsuitable for the local obstacle avoidance needs of UAVs during dynamic flight.

[0031] To address the issues of low efficiency and time-consuming collision recovery and replanning in related technologies for UAV path planning on large-scale maps, this invention provides a UAV 3D path planning method based on 26-axis motion operator compressed coding, comprising:

[0032] Step S101: Based on the UAV's starting point and target endpoint, an initial path set for the UAV's path is constructed; wherein, the UAV path is composed of continuous grid coordinate points; the initial path set includes multiple initial paths.

[0033] Step S102: The displacements between adjacent grid coordinate points in the initial path are mapped to corresponding translation operators through a mapping table to generate a compressed coded path; wherein, the translation operators are integers, and each integer uniquely corresponds to a three-dimensional displacement direction;

[0034] Step S103: The compressed encoding path is optimized by using an iterative optimization algorithm.

[0035] Step S104: In the case of conflict points in the optimized compression coding, the conflict points are guided by the movement operator to obtain feasible points; wherein, the conflict point is the grid coordinate point in the infeasible region after the optimized compression coding is decompressed, and the feasible point is the grid coordinate point in the feasible region.

[0036] Step S105: Update and optimize the compression coding path based on the feasible points, and determine the final optimized path after the iterative optimization algorithm terminates.

[0037] The above steps can be executed by the flight control computer or airborne computing unit on the UAV itself, which is used to plan the three-dimensional path in real time before flight or during dynamic flight; or by the control system of the ground station, which generates the optimized path offline or online and then uploads it to the UAV for execution; the method can also be extended to path planning scenarios of intelligent carriers such as unmanned vehicles and mobile robots, and executed by the autonomous navigation control system of the corresponding carrier.

[0038] The 3D map of the drone flight scene is rasterized into L×W×H grids. p(l,w,h) represents the position of a grid coordinate point in space. A path P from the starting point to the target endpoint can be composed of M grid path points, such as: P={(l1,w1,h1),(l2,w2,h2),……,(l M ,w M ,h M )}.

[0039] Multiple initial paths can be generated based on the drone's starting point and target destination, forming an initial path set. Multiple initial paths provide diverse candidate solutions, increasing the coverage of the search space, preventing subsequent optimization from getting trapped in local optima, and thus improving global optimization capabilities.

[0040] The initial path set can be flexibly constructed. For example, an A* algorithm or a fast-exploration random tree can be used to generate a feasible reference path, which can then be combined with multiple randomly generated paths to form the initial set. Alternatively, all paths can be generated randomly, or based on heuristic rules. The number of initial paths can be adjusted according to the map size and computing resources to ensure population diversity.

[0041] The displacements between adjacent grid coordinate points in the initial path are mapped to corresponding translation operators using a mapping table, generating a compressed coded path. This step aims to transform path encoding from a coordinate point description to a translation direction description, thereby achieving essential compression of the decision space.

[0042] A set of standardized movement operators is predefined, such as 26 directions. Each movement operator corresponds to a unique movement direction of the UAV from the current point to an adjacent point in a 3D grid. By establishing a mapping table between movement operators and grid displacements, any path consisting of continuous grid coordinate points can be converted into a compressed coded sequence consisting of movement operator identifiers.

[0043] In the three-dimensional coordinate encoding method, each waypoint needs to be selected from L×W×H grids of the entire map, resulting in a path search space of up to (L×W×H) containing N waypoints. N When the map size reaches 100×100×100, this value is on the order of 10^120.

[0044] This invention employs a movement operator compression encoding, representing the path as the movement direction between adjacent points. Each movement step has only 26 possible values, and the search space is fixed at 26. (N-1) At the map size mentioned above, it is approximately on the order of 10^26, which is 10^94 times larger than the previous value.

[0045] This shift in coding approach, from describing where the waypoints are to describing how to take the path, completely decouples the search space from the map size, thereby achieving a leapfrog compression of the decision space from astronomical figures to a controllable scale.

[0046] This encoding method, identified by the movement operator, compresses the encoding length of the UAV path from 3M to M-1, significantly reducing the dimensionality of the decision variables; the search space is decoupled from the map grid size and fixed as a power of the number of operators, avoiding the curse of dimensionality; at the same time, this encoding method is naturally compatible with iterative optimization algorithms and can be directly used as the position vector input of particles.

[0047] Multi-objective optimization search is performed on the path within the compressed coding space. Since the compressed coding path is already composed of a sequence of movers, the space for multi-objective optimization search is much smaller than that of traditional coordinate coding methods, thus significantly improving the convergence speed and search efficiency of the iterative optimization algorithm.

[0048] When implementing iterative optimization, swarm intelligence optimization algorithms such as multi-objective particle swarm optimization, genetic algorithms, and ant colony optimization can be employed. These algorithms treat compressed coded paths as individuals, generating new generations of individuals through operations such as velocity updates, crossover, and mutation during iteration. Fitness is calculated based on multiple optimization objectives, guiding the population towards the Pareto front. The Pareto front, in multi-objective optimization, describes the set of optimal solutions where further improvement is impossible among multiple conflicting objectives.

[0049] The problem of local repair when a few path points fall into the infeasible region can be solved by making full use of the trend information of existing feasible path points to achieve efficient targeted repair and avoid global replanning.

[0050] After the optimized compressed encoding path is decompressed into a raster coordinate path, if a path point is detected to be located in an infeasible region (i.e., a conflict point), a local repair process is triggered. The basic idea of ​​repair is to construct a trend guidance vector using the nearest feasible points before and after the conflict point to generate candidate points; if the candidate point is in the feasible region, the repair is complete; otherwise, further search is performed based on the movement operator until a feasible point is obtained.

[0051] This repair mechanism only adjusts the conflict point locally, without changing other feasible path points; it utilizes the trend information of the path itself, and the repair direction has physical meaning; in conjunction with the compression coding method, the coded sequence can be directly updated after repair. The repaired feasible points are re-encoded as movement operators, the compressed coded path is updated, and the final optimization result is output after the iteration termination condition is met. In related technologies, the entire path is discarded and a global replanning is triggered when infeasible points are encountered. This embodiment only repairs local points, significantly reducing the repair time.

[0052] The repaired feasible points are re-encoded, the compressed coding path is updated, and the final path is output after the iteration terminates. The local repair results are solidified into the compressed coding path to form a complete closed loop of encoding, optimization, repair, and update, thereby ensuring that the repaired path can be used for subsequent iterations. The final output path is the result of the synergistic effect of global optimization and local repair, and has both global optimality and feasibility.

[0053] The present invention provides a UAV 3D path planning method based on 26-axis translation operator compression coding. By encoding the 3D path into a compact integer sequence through 26-axis translation operator, the search space and computational load of subsequent iterative optimization algorithms are greatly reduced, thereby effectively improving the path planning efficiency under large-scale maps.

[0054] Meanwhile, when a collision occurs during the decompression of the optimized encoded path, the algorithm does not need to perform time-consuming global replanning. Instead, it uses the geometric semantics of the mover operator to guide trends and quickly repair the collision at the local point of conflict, generating feasible points and updating the path. This significantly shortens the collision repair time, thus solving the problems of low efficiency in UAV path planning on large-scale maps and long collision repair replanning time in related technologies.

[0055] As an alternative solution, such as Figure 3 As shown, in the case of conflict points in the optimized compression coding, a trend guidance method is used to guide the conflict points to obtain feasible points. This includes: decompressing the optimized compression coding path to obtain the grid coordinate path in three-dimensional space; determining the nearest neighboring feasible points before and after the conflict point in the path; wherein, the neighboring feasible points include the grid coordinate points from the conflict point to the previous feasible region and to the grid coordinate points to the next feasible region; calculating the trend guidance vector from the conflict point to the two neighboring feasible points using the distance formula, and obtaining candidate points through calculation; if the candidate point is in a feasible region, the candidate point is taken as a feasible point; if the candidate point is in an infeasible region, a new conflict point is obtained through the movement operator, and iterative optimization is performed until a feasible point is obtained.

[0056] Optimized compressed coding path refers to the path representation composed of a sequence of mover operators after optimization by an iterative optimization algorithm. Decompression refers to the process of converting the sequence of mover operators into a continuous sequence of raster coordinate points through a mapping table; it is the inverse operation of compressed coding.

[0057] The optimized encoded path is restored to a coordinate form that allows collision detection in physical space. Therefore, collision point determination must be performed on actual raster coordinates and cannot be directly completed in the compressed encoding domain. By decompressing, a conversion between the compressed encoding space and the physical space is established, providing a foundation for subsequent feasible region determination and repair operations.

[0058] The complete raster coordinate path is obtained for point-by-point collision detection. The decompression and encoding processes are inverse operations to ensure lossless information conversion. Decompression can be implemented using a reverse lookup corresponding to the mapping table or by coordinate recursion calculation based on displacement operators, as long as the raster coordinate sequence can be accurately reconstructed.

[0059] A conflict point refers to a grid coordinate point located in the infeasible region after decompression, i.e., a path point that falls into an obstacle or no-fly zone. The nearest neighbor feasible point refers to the first grid coordinate point located in the feasible region encountered when searching forward (towards the starting point) and backward (towards the ending point) from the conflict point position in the grid coordinate path.

[0060] Local trend information of the path near the conflict point is extracted because the conflict point is sandwiched between two feasible points, which represent the basic direction of the path in that area, providing a reference benchmark for subsequent targeted repair. By utilizing only local information rather than global information, the repair process is kept lightweight.

[0061] The nearest feasible point can be determined by a bidirectional linear search along the path starting from the conflict point, or by using a preset index marker for quick location. In special cases, such as when the conflict point is located at the start or end point, a single-sided reference point can be defined or adjusted flexibly according to the actual situation.

[0062] Assume that the path point in the infeasible region is p. F The distance formula here specifically refers to the mathematical operation used to calculate the coordinate difference vector between two points, that is, subtracting the coordinates of the previous point from the coordinates of the later point to obtain the displacement vector. Calculate p F Displacement vector to the path point in the previous feasible region ; Calculate p F Displacement vector to the path point in the next feasible region .

[0063] A trend-guiding vector is a vector synthesized from two displacement vectors from the conflict point to the preceding and following feasible points, used to indicate the direction of repair. It integrates the path trends in both directions. .

[0064] Candidate points refer to new grid coordinate points obtained by adding the trend-guiding vector to the coordinates of the conflict point. If the calculation result falls on a non-integer coordinate, it needs to be rounded to meet the integer requirement of the raster coordinate.

[0065] Simply moving in one direction may deviate from the path trend. Therefore, this embodiment combines information from both directions to ensure that candidate points simultaneously meet the requirements of the preceding and following path segments. This achieves a shift from blind search to targeted repair, significantly reducing the number of iterations required for repair.

[0066] This embodiment obtains the grid coordinate path through decompression, locates feasible points before and after the conflict point to extract the local trend of the path, calculates the trend guidance vector to generate candidate points, and performs direct replacement or iterative search according to the feasibility of the candidate points, forming a complete local directional repair mechanism. This mechanism transforms conflict repair from traditional global replanning to local fine-tuning only for conflict points. It achieves efficient directional guidance by utilizing the path's own trend information. When guidance fails, it performs a finite neighborhood search and iterates until success through a movement operator. Thus, it completes the repair with extremely low computational complexity while maintaining path continuity, significantly improving the repair efficiency of locally infeasible points and effectively solving the problem of long replanning time in collision repair in related technologies.

[0067] As an optional approach, new conflict points are obtained through a movement operator, including: moving candidate points one by one using the movement operator to obtain a set of updated points; calculating the absolute difference between each updated point and the grid coordinate point in the next feasible region along each coordinate axis, and summing the absolute differences along each coordinate axis to obtain the displacement distance corresponding to each updated point; and selecting the updated point with the smallest displacement distance as the new conflict point.

[0068] Candidate points are grid coordinates that are still located in the infeasible region after the initial attempt to guide the trend failed. Each candidate point is moved using a movement operator; that is, the current coordinates of the candidate point are sequentially added to the displacements corresponding to 26 movement operators, generating 26 new grid coordinates that form an update point set. Each update point represents the neighborhood location reachable by taking one step from the candidate point along a certain movement direction.

[0069] For example, suppose candidate points If the grid is moved by the mover operator R1, and the grid displacement corresponding to mover operator R1 is (1,0,0), then the update point is... .

[0070] When the trend-guiding vector fails to bring a candidate point into the feasible region in one attempt, it indicates that the local obstacle distribution at the current point is relatively complex, requiring a more detailed neighborhood search. By generating 26 neighboring points, a complete set of all possible movement directions around the candidate point is constructed, providing a complete candidate space for subsequent optimal selection. This ensures both the comprehensiveness of the search and avoids unlimited blind diffusion.

[0071] Assume the grid coordinates of the next feasible region are p. B The absolute difference along each coordinate axis refers to the difference between the x-coordinate and p-coordinate calculated for each update point. B The absolute value of the difference between the x-coordinates, the y-coordinate and p B The absolute value of the difference between the y-coordinates, the z-coordinate and p B The displacement distance is the absolute value of the difference between the z-coordinates of the updated point and the target point p. The displacement distance is calculated by summing the absolute values ​​of the differences along each coordinate axis, resulting in a scalar value. B The spatial distance between them.

[0072] The above displacement distance calculation process is simple and efficient, requiring only 26 additions and absolute value operations, consistent with the lightweight goal of the overall repair method. Calculating the absolute differences of each coordinate axis separately before summing avoids the problem of positive and negative values ​​canceling each other out, ensuring the accuracy of the distance measurement. Simultaneously, the displacement distance assigns a quantifiable evaluation index to each neighborhood candidate point, accurately reflecting the actual number of steps moved between grids, making the performance in different directions comparable. By selecting the direction closest to the path's orientation, the continuity and smoothness of the repaired path are guaranteed.

[0073] Among the 26 calculated displacement distance values, the update point with the smallest displacement distance is taken as the new conflict point. This update point is used as the starting point for the next iteration to guide the trend and continue iterative repair.

[0074] This embodiment strictly limits the local search scope to 26 standardized directions and replaces blind random search with directional optimization based on path trends. This ensures the completeness of the search while maintaining simple computational complexity. It ensures that the repair process can converge to a feasible point through a finite number of iterations and maintains the continuity between the repaired path and the original path, forming a complete closed loop of directional guidance and neighborhood optimization repair.

[0075] As an optional approach, the displacements between adjacent grid coordinates in the initial path are mapped to corresponding movement operators through a mapping table to generate a compressed coded path. This also includes: dividing the three-dimensional space of the UAV flight into cubic grids of equal side length to obtain a three-dimensional grid map; wherein the three-dimensional grid map includes the grid coordinates of the center point of the cubic grid and the feasible and infeasible regions of the three-dimensional space; determining the grid displacement between adjacent grid points based on adjacent grid points in the initial path set; wherein adjacent grid points include the vertices, midpoints of edges, and midpoints of faces of a cube centered on the target grid point; mapping the grid displacements to movement operators through a mapping table; and determining the compressed coded path based on the movement operators.

[0076] The three-dimensional space in which the drone flies is divided into cubic grids of equal side length, forming a three-dimensional grid map. The grid can be preset according to engineering requirements, with the side length of the grid set as a fixed step size d, such as 1m, 5m, or 10m. The grid step size is the minimum moving distance between adjacent path points of the drone.

[0077] The position of any grid is the center point of the grid, represented by a three-digit integer, denoted as p(x,y,z), where x, y, and z are all integers, representing the horizontal, vertical, and axial coordinates of the grid in three-dimensional space, respectively. The grid point coordinates are the unique coordinate form of the UAV path point.

[0078] A grid with obstacles, such as buildings, mountains, or no-fly zones, is an infeasible region S(x,y,z)=0, denoted as F; a grid without obstacles is a feasible region S(x,y,z)=1, denoted as A; the path of the UAV must fall entirely within the feasible region.

[0079] The grid displacement vector between adjacent grid coordinate points in the initial path refers to the three-dimensional coordinate increment from the current grid coordinate point to the next grid coordinate point, i.e. (Δx, Δy, Δz). In the three-dimensional grid, each component can only take the values ​​-1, 0, or 1, resulting in 26 possible directions.

[0080] By using a pre-established mapping table, each grid displacement vector is mapped to a unique integer from 1 to 26, i.e., a movement operator. By sequentially converting all adjacent displacements in the entire path into an integer sequence, the compressed encoded path is obtained, which represents the path as a compact integer array.

[0081] The mapping relationship between displacement operators and grid displacement is shown in Table 1:

[0082]

[0083] The move operator is a 26-bit integer because in a 3D grid, there are 26 possible directions from the current grid coordinates to adjacent grid coordinates, and each direction can be uniquely identified by an integer. This encoding method significantly compresses the path data, making it easier to store and transmit, while preserving the geometric and topological information of the path. This allows subsequent optimization algorithms to perform efficient searches in the integer sequence space, reducing computational complexity.

[0084] like Figure 2 As shown, the possible scenarios for the drone moving from point p to the surrounding areas include: moving to the position of the 8 vertices of the current plane, as indicated by the purple arrows; moving to the position of the 8 vertices of the plane after adding 1 in the vertical direction, as indicated by the green arrows; moving to the position of the 8 vertices of the plane after subtracting 1 in the vertical direction, as indicated by the cyan arrows; and moving to the positions of the vertices directly above and below, as indicated by the orange arrows.

[0085] After the UAV moves from the current path point p(x,y,z) along the movement operator R, the next path point... The coordinates are uniquely determined, and the calculation formula is: Conversely, given the current coordinates p and the next path point... The coordinates can be determined The value of the displacement operator R is uniquely determined using Table 1.

[0086] In summary, based on the mapping relationship between the displacement operator and the grid displacement, the M*3 bit path vector P = {(l1,w1,h1),(l2,w2,h2),……,(l M ,w M ,h M Each path can be mapped one-to-one to an M-bit compressed coding path vector. =(R1,R2,……,R M-1 Similarly, when calculating distances in three-dimensional space, It can also be reverse mapped back to P.

[0087] This embodiment establishes a standardized environment model by dividing the three-dimensional space into equilateral cube grids, determines the grid displacement in 26 neighborhood directions centered on the current point, establishes a bidirectional mapping relationship between the grid displacement and the movement operator, and finally generates a compressed coded path composed of a sequence of movement operators.

[0088] This method transforms the path representation from a sequence of three-dimensional coordinate points to a sequence of movement directions, compressing the dimensionality of decision variables from 3M to M-1, and expanding the search space from (L×W×H) which is related to the map size. M Compressed to a fixed 26 MThis completely decouples the encoding method from the map grid size, thus solving the problem of the curse of dimensionality faced by encoding methods in large-scale maps, and providing standardized encoding input for subsequent iterative optimization and local repair.

[0089] As an optional approach, the compressed coding path is optimized through an iterative optimization algorithm, including: obtaining the current iteration information and calculating the dynamic inertia weight; wherein the current iteration information includes the current iteration number, the maximum iteration number, and the minimum and maximum values ​​of the inertia weight; based on the dynamic inertia weight and the algorithm parameters of the optimization algorithm, combined with the velocity vector, position vector, and historical best solution of the current iteration cycle, the next-generation velocity vector of each path in the compressed coding path is calculated using a velocity update formula; wherein the algorithm parameters include two random numbers and two acceleration coefficients; based on the next-generation velocity vector, the next-generation position vector is calculated using a position vector calculation formula; wherein the calculated position vector is rounded up to the nearest integer and used to represent multiple movement operators in the path; based on the next-generation position vector, the optimized compressed coding path is determined, wherein the optimized compressed coding path consists of multiple movement operators, each movement operator corresponding to a decision variable dimension; the position vector is used to represent the value of each decision variable dimension.

[0090] This embodiment introduces the global exploration and local development capabilities of the dynamic inertial weighted particle swarm algorithm, treats the compressed coding path as a swarm particle, adaptively adjusts the search step size based on the current iteration information, and uses the velocity update formula and position vector calculation formula to iteratively optimize the movement operator sequence in the compressed coding space.

[0091] Simultaneously, by rounding up the position vector calculation result, it is ensured that each generated translation operator is a valid integer. If the calculation result exceeds 1 to 26, it is constrained to the boundary value. Thus, from 26... M This embodiment achieves efficient multi-objective path search within a finite decision space. The seamless collaboration of the compression coding method significantly improves the algorithm's convergence speed and optimization quality while ensuring path feasibility, further addressing the problem of low path planning efficiency in large-scale 3D raster maps.

[0092] As an alternative approach, the expression for the position vector calculation formula is:

[0093] ;

[0094] in, Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The velocity vector of the decision variables;

[0095] ;

[0096] in, c1 and c2 represent the inertia weight; c1 and c2 represent the acceleration coefficients. r1 and r2 represent random numbers. ; Represented as a group in the th In the nth iteration, the 1st The historical optimal solution for the decision variable.

[0097] This velocity update formula balances global exploration and local exploitation capabilities by introducing a dynamic inertia weight w, and utilizes the group's historical optimal solution. The system guides particles to converge towards the current optimal region and enhances the randomness and diversity of the search by combining acceleration coefficients c1 and c2 with random numbers r1 and r2, thereby achieving a higher degree of randomness and diversity in the search process. M Efficient path iterative optimization is achieved within a finite decision space.

[0098] Formula 2 The design of using the population's historical optimal solution simplifies the algorithm structure, enabling particles to quickly utilize the current population's optimal information for collaborative search. This drives the population to approach the Pareto front in the compressed coded path space, effectively improving the convergence speed and optimization quality of multi-objective path planning.

[0099] As an alternative approach, the speed update formula can also be adopted. Combining individual optimal and group optimal Generate the next generation of particles, This represents the individual optimality. The individual optimality can be achieved by comparing the current position with its own historical optimal position after each iteration. If the current position dominates the historical optimal position, then the individual optimality is updated.

[0100] pass Individual Optimal This guides particles toward their historical best positions, reflects the accumulation of individual experience, and prompts particles to conduct a fine search in the high-quality areas they discover. middle This indicates that the optimal guidance particle moves closer to the global optimal position of the population, realizes information sharing, and enables the entire population to converge towards the current optimal region in a coordinated manner.

[0101] The three parts of the velocity update formula work together to balance the particle's exploration of its own experience with the wisdom of the group, thus enabling it to fly efficiently in the solution space and gradually approach the global optimum. Random numbers introduce randomness and enhance the ability to escape local optima; acceleration coefficients c1 and c2 adjust the weights of the cognitive and social components, usually set to 2 to balance their influence.

[0102] ;

[0103] in, , ; For the current algebra, This represents the maximum number of iterations.

[0104] The dynamic inertia weight is calculated using a linearly decreasing strategy to balance the algorithm's global exploration capability and local exploitation capability. In the early stages of iteration, the dynamic inertia weight is relatively large, allowing particles to maintain high motion inertia and enabling them to search new regions over a large area, enhancing global exploration and preventing premature convergence. In the later stages of iteration, the dynamic inertia weight gradually decreases, reducing the influence of particle motion on its own velocity and allowing it to move more towards individual and swarm optima, strengthening local fine-grained search and accelerating convergence. This adaptive adjustment mechanism allows the algorithm to automatically match appropriate search step sizes at different stages, improving convergence accuracy and stability.

[0105] As an optional approach, the compressed coding path is optimized through an iterative optimization algorithm. This further includes: mapping the optimized compressed coding path to a grid coordinate path using a mapping table; where the grid coordinate path is the grid coordinate representation of the UAV flight path in three-dimensional space; calculating the fitness of the grid coordinate path for each optimization objective using a fitness formula; where the fitness formula is used to evaluate the quality of the optimized compressed coding path; sorting the fitness using a non-dominated sorting algorithm, selecting the optimal solution, and archiving it to obtain archived particles; generating the next generation of optimized compressed coding paths using an iterative optimization algorithm based on the archived particles, and accumulating the number of iterations; and determining the final optimized compressed coding path based on the next generation of optimized compressed coding paths when the number of iterations reaches a preset number.

[0106] Compressed paths exist as sequences of movers, which are convenient for optimization algorithms, but cannot be directly used for feasibility assessments, fitness calculations, or actual flight in physical space. They need to be restored to grid coordinate paths in order to detect whether a point is within the feasible region, calculate path length, number of turns, and other optimization objectives on a 3D map, and provide usable track coordinates for the final UAV execution.

[0107] Based on the starting coordinates and the sequence of movement operators, the calculation proceeds point by point. Starting from the starting point (x1, y1, z1), the corresponding displacement (Δx, Δy, Δz) is looked up from the mapping table according to the first movement operator to obtain the second point (x1+Δx, y1+Δy, z1+Δz), and so on until the endpoint to obtain the raster coordinate path. This lossless reconstruction from the compressed coding space to the physical space yields the complete raster coordinate path, providing a data foundation for subsequent fitness calculations and feasibility verification.

[0108] Each optimization objective measures the quality of a path from different dimensions. A predefined mathematical formula for each objective is called a fitness formula, used to quantify the geometric or topological characteristics of the path into comparable numerical values. In multi-objective optimization, each path corresponds to a set of fitness values. For example, the fitness formula for path length is the sum of the Euclidean distances between all adjacent path points; the fitness formula for the number of turns is the number of times the path direction changes; and the fitness formula for safety is the minimum distance from a path point to the nearest obstacle.

[0109] The fitness formula can be flexibly defined according to specific application scenarios, and the calculated fitness is used to quantify the performance of the path in a certain dimension. For example, the path length can be measured by Euclidean distance, Manhattan distance, or other distance metrics; the safety can be defined as the distance to the nearest obstacle, the minimum safety margin with respect to obstacles, or a penalty function can be introduced; other optimization objectives such as path smoothness, energy consumption, and time can also be added.

[0110] In the iterative optimization process, it is necessary to quantitatively evaluate the merits of each path to guide the population towards a better direction. By calculating the fitness of each optimization objective, the geometric characteristics of the paths are transformed into numerical indicators, providing a quantitative basis for subsequent ranking, selection, and archiving. A characteristic of multi-objective optimization is that multiple objectives may conflict with each other; therefore, it is necessary to simultaneously retain compromise solutions that perform well on all objectives.

[0111] Non-dominated ranking is a classic ranking method in multi-objective optimization, used to stratify individuals in a population based on Pareto dominance. Individual A is said to dominate B if it is not inferior to individual B on all objectives and is superior to B on at least one objective. Non-dominated ranking divides the population into the first non-dominated stratum, assigning individuals not dominated by any other individual to this first non-dominated stratum, then assigning the remaining undominated individuals to the first non-dominated stratum, and so on.

[0112] In the context of non-dominated sorting, "archived particles" typically refer to individuals located in the first non-dominated layer, representing the set of compromise-optimal solutions found in the population. Archived particles refer to the selection of optimal solutions stored separately in an external set. These archived particles serve as candidates for the global optimum in subsequent iterations and guide population evolution. The archiving mechanism prevents the degeneration of the Pareto front and preserves historical optimal solutions.

[0113] The iterative optimization algorithm utilizes the optimal information in the current population and archives to generate a new generation of individuals, i.e., the next generation of optimized compressed coding paths, through operations such as speed updates and crossover mutations. These paths will continue to be optimized in the next iteration. The iteration count is accumulated by incrementing the current iteration count by 1, which is used to determine whether the termination condition has been met.

[0114] Iteration stops when the cumulative number of iterations equals the preset maximum number of iterations. The new population generated in the last iteration contains multiple optimized compressed coding paths. The final result selected from the population of the last iteration can usually be one or more solutions as the final output. The specific selection method can be determined according to actual needs, such as selecting the point closest to the ideal point, or selecting according to the decision-maker's preference.

[0115] This embodiment establishes a complete evaluation feedback loop between the compressed coding space and the physical space, effectively handles the conflict relationships between multiple optimization objectives through non-dominated sorting and archiving strategies, and drives the population to approach the true Pareto front while maintaining the diversity of solution sets. This enables the efficient generation of a set of feasible paths that meet the requirements of multiple objectives such as path length, number of turns, and safety in large-scale 3D raster maps, and systematically solves the problem that related technologies are difficult to balance the objectives and are prone to losing excellent solutions in multi-objective optimization.

[0116] As an alternative approach, the optimization objectives include: path length, number of turns, and safety.

[0117] Path length, as an optimization objective, minimizes the UAV's flight distance. By summing the Euclidean distances between all adjacent path points and using this sum as the fitness formula, the population can be guided to converge towards paths with shorter total distances. Shorter flight paths directly translate to reduced energy consumption and shorter flight times, thereby improving the UAV's operational efficiency and endurance. In conjunction with other optimization objectives, path length forms a fundamental dimension that is interdependent with safety and the number of turns, enabling the algorithm to seek an optimal balance between taking shortcuts and taking safe, smooth paths, avoiding obstacle crossing or sharp turning problems caused by simply pursuing shorter distances.

[0118] Suppose (l, w, h) represents the coordinates of a grid point in space. Define a feasible region S: if the grid point coordinates (l, w, h) are feasible, then S(l, w, h) = 1; otherwise, S(l, w, h) = 0. A path from the starting point to the ending point can consist of M grid points, such as: P = {(l1, w1, h1), (l2, w2, h2), ..., (l... M ,w M ,h M )}.

[0119] The path length is the sum of the Euclidean distances between all adjacent path control points. The fitness formula for the path length is as follows:

[0120] ;in, The fitness value represents the path length; M represents the number of grid points. , , These represent the values ​​of the i-th grid point on the x, y, and z axes, respectively. , , +1 represents the value of the (i+1)th grid point on the x, y, and z axes, respectively.

[0121] Optimizing the number of turns can reduce the frequency of attitude adjustments during UAV flight. By statistically analyzing the number of path points where direction changes occur and minimizing the number of turns, path planning tends to generate straight, smooth flight trajectories. Reducing the number of turns reduces the burden on the flight control system and decreases additional energy consumption; simultaneously, a smoother trajectory helps the UAV maintain stable flight, reduces the risk of loss of control, and better meets the minimum turning radius constraints of fixed-wing UAVs. In a multi-objective framework, the number of turns and path length have a positive but non-linear correlation; optimizing this objective can effectively avoid jagged, invalid paths created to shorten distances.

[0122] In path P, if path point (l) i ,w i ,h i ) and (l i-1 ,w i-1 ,h i-1 ) and (l i+1 ,w i+1 ,h i+1 If they are not on a straight line, then the number of turns is B. Pi =1. Conversely, B Pi =0. The sum of the number of turns at all path points in path P is the total number of turns for that path. The fitness formula for the number of turns is: Where i ranges from 2 to M-1.

[0123] Safety, as an optimization objective, maximizes the obstacle avoidance margin between the path and obstacles. By calculating the minimum distance from all points on the path to the nearest obstacle and using it as the fitness formula, the path can be guided to actively avoid dangerous areas. A higher safety level means the path has stronger fault tolerance; even with positioning errors or external disturbances, the drone is less likely to collide with obstacles.

[0124] Safety is often in competition with path length and number of turns; that is, the safest path is often not the shortest or smoothest path. By incorporating safety into multi-objective optimization, we can seek the optimal efficiency while ensuring flight safety, so that the final output path not only meets feasibility requirements but also has good engineering practicality.

[0125] The fitness of safety is represented by the minimum distance from all path points in path P to the nearest obstacle. The fitness formula for safety is as follows:

[0126] ;

[0127] in, Let K be the coordinates of the k-th obstacle, i.e., the infeasible grid point, and K be the total number of obstacles.

[0128] As an optional approach, an initial path set for the UAV is constructed based on the UAV's starting point and target endpoint. This includes: obtaining a feasible path using a pathfinding algorithm based on the flight parameters and terrain environment data acquired by the UAV; randomly generating intermediate nodes between the starting point and target endpoint based on the UAV's starting point and target endpoint; generating multiple random paths based on the starting point, target endpoint, and intermediate nodes; and using the feasible path and multiple random paths as the initial path set.

[0129] Flight parameters refer to the performance constraints and flight requirements of the UAV itself, such as maximum turning angle, minimum turning radius, flight altitude limit, and maximum range. These parameters affect the feasibility assessment of the path. Terrain environment data refers to the geographic information data of the UAV's flight area, including digital elevation models, building distribution, no-fly zone coordinates, and obstacle locations. This data is used to construct feasible and infeasible regions in a 3D raster map.

[0130] Pathfinding algorithms are classic algorithms that can search for feasible paths from a starting point to a destination in a grid map, such as the A* algorithm and the Fast Expanding Random Tree algorithm. These algorithms can find a collision-free connected path based on environmental data. A path found by a pathfinding algorithm, where all path points are within the feasible region and satisfy the UAV flight parameter constraints, consists of continuous grid coordinate points and can serve as a reference benchmark for subsequent optimization of the population.

[0131] The quality of the initial population directly affects the convergence speed and the quality of the final solution of the iterative optimization algorithm. By using a pathfinding algorithm to pre-guarantee that at least one feasible path is included in the initial population, effective genes and evolutionary directions are provided to guide other random paths toward the feasible region. Leveraging the maturity and determinism of traditional pathfinding algorithms, a high-quality benchmark solution can be obtained with lower computational cost.

[0132] A path originating from a starting point, passing through several intermediate nodes, and finally reaching the destination. These paths consist of continuous raster coordinate points, potentially including points located in infeasible regions. Sub-paths can be generated between adjacent anchor points using random walks or random interpolation, and then stitched together to form a complete path. Alternatively, paths can be randomly generated based on movement operators, starting from the starting point, randomly selecting a movement operator at each step, and adjusting the direction near intermediate nodes to ensure passage through those nodes. A smooth path can also be generated using B-spline curves before rasterization. The number of random paths, N-1, can be preset according to population size requirements.

[0133] Iterative optimization algorithms require a diverse initial population to effectively explore the solution space. By randomly generating multiple paths, different orientations, curvatures, and spatial distributions can be covered, increasing the diversity of the population. Even if infeasible points exist in these random paths, they still carry valuable information that can be transformed into feasible solutions through path repair.

[0134] This embodiment maximizes the coverage of the solution space while ensuring that the initial population contains at least one valid solution. This allows the population to have both the quality advantage of utilizing the benchmark solution and the diversity advantage of exploring random solutions, providing a high-quality evolutionary starting point for subsequent iterative optimization algorithms. It effectively avoids the problem of slow convergence or getting trapped in local optima due to low quality or insufficient diversity of the initial population, thereby improving the efficiency of the overall path planning and the quality of the solution.

[0135] According to another aspect of the invention, an electronic device is also provided, comprising: a processor, and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform any of the methods described above.

[0136] refer to Figure 4 The present invention will now describe a structural block diagram of an electronic device that can serve as an embodiment of the present invention, serving as an example of a hardware device applicable to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0137] like Figure 4As shown, the electronic device includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. The RAM 403 may also store various programs and data required for the operation of the electronic device. The computing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0138] Multiple components in the electronic device are connected to I / O interface 405, including: input unit 406, output unit 407, storage unit 408, and communication unit 409. Input unit 406 can be any type of device capable of inputting information into the electronic device. Input unit 406 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 407 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 408 may include, but is not limited to, disks and optical discs. Communication unit 409 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, and / or wireless communication transceivers, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0139] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, CPUs, graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention can be implemented as computer programs tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on an electronic device via ROM 402 and / or communication unit 409. In some embodiments, the computing unit 401 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0140] Computer programs for implementing the methods of embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0141] In the context of embodiments of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, or infrared systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0142] It should be noted that the term "comprising" and its variations used in the embodiments of this invention are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "a plurality" mentioned in the embodiments of this invention are illustrative and not restrictive, and those skilled in the art should understand that unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0143] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this invention are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0144] The steps described in the method embodiments provided by the present invention can be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0145] The term "embodiment" in this specification refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily imply the same embodiment, nor does it imply independence or alternativeity from other embodiments. The various embodiments in this specification are described in a related manner, with reference to each other for similar or identical parts. In particular, for apparatus, device, and system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and relevant details are referred to in the description of the method embodiments.

[0146] The above embodiments merely illustrate several implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A UAV 3D path planning method based on 26-axis motion operator compressed coding, characterized in that, include: An initial path set is formed based on the drone's starting point and target endpoint; wherein, the drone path is composed of continuous grid coordinate points; the initial path set includes multiple initial paths; The displacements between adjacent grid coordinate points in the initial path are mapped to corresponding translation operators through a mapping table to generate a compressed coded path; wherein, the translation operators are integers, and each integer uniquely corresponds to a three-dimensional displacement direction; The compressed encoding path is then optimized using an iterative optimization algorithm. In the case of conflict points in the optimized compression coding, the moving operator guides the conflict points to obtain feasible points; wherein, the conflict point is the grid coordinate point in the infeasible region after the optimized compression coding is decompressed, and the feasible point is the grid coordinate point in the feasible region. The compressed encoding path is updated and optimized based on the feasible points, and the final optimized path is determined after the iterative optimization algorithm terminates.

2. The UAV 3D path planning method based on 26-axis motion operator compressed coding according to claim 1, characterized in that, In the case of collision points in the optimized compression coding, feasible points are obtained by using a moving operator to guide the trend of the collision points, including: Decompress the optimized compressed encoding path to obtain the raster coordinate path in three-dimensional space; Based on the conflict point, determine the nearest neighboring feasible points before and after the conflict point in the path; wherein, the neighboring feasible points include the grid coordinate points from the conflict point to the previous feasible region and to the grid coordinate points to the next feasible region; The trend guidance vector from the conflict point to two feasible neighboring points is calculated using the distance formula, and candidate points are obtained through calculation. If the candidate point is within the feasible region, the candidate point is considered a feasible point. When the candidate point is in the infeasible region, a new conflict point is obtained through the movement operator, and the optimization is iteratively performed until a feasible point is obtained.

3. The UAV 3D path planning method based on 26-axis motion operator compressed coding according to claim 2, characterized in that, The new conflict point is obtained through the movement operator, including: The candidate points are moved using the moving operator to obtain the updated point set; By calculating the absolute difference between each update point and the grid coordinate point in the next feasible domain in each coordinate axis direction, and adding the absolute differences in each coordinate axis, the displacement distance corresponding to each update point is obtained. The update point with the smallest displacement distance is taken as the new conflict point.

4. The UAV 3D path planning method based on 26-axis motion operator compressed coding according to claim 1, characterized in that, The displacements between adjacent grid coordinate points in the initial path are mapped to corresponding translation operators through a mapping table to generate a compressed encoded path, which also includes: The three-dimensional space in which the UAV flies is divided into cubic grids of equal side length to obtain a three-dimensional grid map; wherein, the three-dimensional grid map includes the grid coordinates of the center point of the cubic grid and the feasible and infeasible regions of the three-dimensional space; Based on the adjacent grid points in the initial path set, determine the grid displacement between adjacent grid points; wherein, the adjacent grid points include the vertices, the midpoints of the edges, and the midpoints of the faces of a cube centered on the target grid point; The grid displacement is mapped to a movement operator through a mapping table; The compression encoding path is determined based on the moving operator.

5. The UAV 3D path planning method based on 26-axis motion operator compressed coding according to claim 1, characterized in that, The compressed encoding path is then optimized using an iterative optimization algorithm, including: Obtain the current iteration information and calculate the dynamic inertia weight; wherein, the current iteration information includes the current iteration number, the maximum iteration number, and the minimum and maximum values ​​of the inertia weight; Based on the dynamic inertia weight and the algorithm parameters of the optimization algorithm, combined with the velocity vector, position vector and historical best solution of the current iteration cycle, the next-generation velocity vector of each path of the compressed coding path is calculated through the velocity update formula; wherein, the algorithm parameters include two random numbers and two speedup coefficients; Based on the next-generation velocity vector, the next-generation position vector is calculated using the position vector calculation formula; wherein, the calculation result of the position vector is rounded up to the nearest integer and used to represent multiple movement operators in the path; Based on the next-generation position vector, an optimized compression coding path is determined, wherein the optimized compression coding path consists of multiple move operators, each move operator corresponding to a decision variable dimension; the position vector is used to characterize the value of each decision variable dimension.

6. The UAV three-dimensional path planning method based on 26-axis motion operator compressed coding according to claim 5, characterized in that, The expression for the position vector calculation formula is: ； in, Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The position vector of the decision variable; Indicates the compressed encoding path In the In the nth iteration, the 1st The velocity vector of the decision variables; ； in, c1 and c2 represent the inertia weight; c1 and c2 represent the acceleration coefficients. r1 and r2 represent random numbers. ; Represented as a group in the th In the nth iteration, the 1st The historical optimal solution for the decision variable; ； in, , ; For the current algebra, This represents the maximum number of iterations.

7. The UAV three-dimensional path planning method based on 26-axis motion operator compressed coding according to claim 5, characterized in that, The optimized compressed encoding path is obtained by iteratively optimizing the compressed encoding path using an optimization algorithm, and the method further includes: The optimized compression encoding path is mapped to a raster coordinate path through a mapping table; wherein, the raster coordinate path is the raster coordinate representation of the UAV flight path in three-dimensional space; Based on the grid coordinate path, the fitness of the grid coordinate path for each optimization objective is calculated using the fitness formula for each optimization objective; wherein, the fitness formula is used to evaluate the merits of the optimized compression coding path; The fitness values ​​are sorted using a non-dominated sorting algorithm, and the optimal solution is selected and archived to obtain archived particles. Based on the archived particles, the next generation of optimized compression coding paths is generated through the iterative optimization algorithm, and the number of iterations is accumulated. When the number of iterations reaches a preset number, the final optimized compression coding path is determined based on the next-generation optimized compression coding path.

8. The UAV three-dimensional path planning method based on 26-axis motion operator compressed coding according to claim 7, characterized in that, The optimization objectives include: path length, number of turns, and safety.

9. The UAV three-dimensional path planning method based on 26-axis motion operator compressed coding according to claim 1, characterized in that, Based on the drone's starting point and target destination, an initial path set is formed to constitute the drone's path, including: Based on the flight parameters and terrain environment data acquired by the UAV, a feasible path is obtained through a pathfinding algorithm; Based on the drone's starting point and target endpoint, an intermediate node between the starting point and the target endpoint is randomly generated; Based on the starting point, the target endpoint, and the intermediate nodes, multiple random paths are generated; The feasible path and the multiple random paths are used as the initial path set.

10. An electronic device, comprising: A processor and a memory storing a program, characterized in that the program includes instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 9.

Images