Multi-strategy unmanned aerial vehicle rapid obstacle avoidance path planning method and device based on dynamic ellipsoid sampling

Abstract

The application discloses a multi-strategy unmanned aerial vehicle rapid obstacle avoidance path planning method based on dynamic ellipsoid sampling, which comprises the following steps: constructing an ellipsoid according to the starting point and the ending point of the unmanned aerial vehicle path planning, mapping a sampling point to the ellipsoid space to obtain a sampling point in the ellipsoid, and dynamically adjusting ellipsoid parameters through cyclic sampling to obtain an initial planning path and path nodes, which can focus on an effective area and improve sampling efficiency; a multi-strategy expansion mechanism is used to optimize the initial planning path nodes, and the convergence speed and obstacle avoidance ability are improved; potential parent nodes meeting the feasibility condition are searched in the neighborhood of the optimized path nodes, and the node with the minimum path cost is selected as a new parent node, and the new parent node is used to update the optimized planning path; a three-stage path smoothing strategy is used to smooth the updated optimized planning path, and an unmanned aerial vehicle planning path is obtained, so that the path feasibility and safety are ensured, and the path smoothness is improved.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) path planning technology, and more specifically, to a method and apparatus for rapid obstacle avoidance path planning for multi-strategy UAVs based on dynamic ellipsoid sampling. Background Technology

[0002] Currently, drones are widely used in scenarios such as logistics delivery, power line inspection, and emergency rescue, and related drone path planning methods are constantly evolving.

[0003] Early algorithms such as A* and Dijkstra, which were based on the assumption of a static environment, could effectively find the shortest or better path in simple static environments, but they faced limitations such as excessive computation and poor adaptability to dynamic changes in the environment when faced with complex and ever-changing environments.

[0004] Sampling-based path planning algorithms (such as RRT, RRT*, etc.) have attracted widespread attention due to their good search efficiency in high-dimensional spaces. The core process of the Rapidly-Exploring Random Trees (RRT) algorithm is as follows: First, initialize a random tree with the starting point as the root node. Then, continuously generate sampling points randomly in the environment. Find the node closest to the sampling point in the existing tree, and expand the tree towards the sampling point with a fixed step size to generate a new node. If the new node does not collide with an obstacle, it is added to the tree. Repeat this process until the tree expands to near the destination. Finally, backtrack from the destination to the starting point to form a feasible path. This method relies on random sampling to expand the search range and can find a feasible path relatively quickly in high-dimensional spaces, especially suitable for complex static environments. However, it still has shortcomings in dynamic obstacle avoidance and multi-objective optimization (such as simultaneously considering distance, energy consumption, and safety).

[0005] The RRT* algorithm improves upon RRT, retaining its basic framework of random sampling and tree expansion. Its core improvement lies in the addition of a path optimization mechanism. The core process is as follows: when a new node is generated, RRT* not only adds it to the tree but also searches existing tree nodes in the new node's neighborhood, calculating the path cost (e.g., distance) from these nodes to the new node. It selects the node with the lowest cost and no collisions as the new parent node. If a lower cost can be achieved through the new node, the parent nodes of other nodes in the neighborhood are updated. Through this "parent node reconnection" operation, RRT* can progressively optimize the path, making the generated path more favorable in terms of cost metrics such as length, thus getting closer to the optimal solution than RRT. However, this also increases computational complexity and slightly reduces real-time performance.

[0006] Traditional RRT and its improved RRT* algorithms can find a better path in the environment by continuously optimizing the path cost during the sampling process. However, in dynamic environments, due to the untimely update of sampling points, the path is prone to collisions with dynamic obstacles. In addition, most existing methods consider only a single optimization objective (such as the shortest path), making it difficult to balance multiple constraints (such as energy consumption, obstacle avoidance safety, and task timeliness).

[0007] However, this traditional RRT and its improved RRT* algorithms still have the following drawbacks:

[0008] First, poor adaptability to dynamic environments: It cannot quickly perceive the movement status (such as speed and direction) of dynamic obstacles in the environment, which makes the planned path prone to conflict with dynamic obstacles.

[0009] Second, multi-constraint optimization imbalance: In path planning, only a single objective (such as the shortest distance) is focused on, while other key constraints such as the energy consumption limit of the UAV and the obstacle avoidance safety distance are ignored, which leads to a decrease in the practicality of the path.

[0010] Third, insufficient path smoothness: The planned path may contain many broken lines or sharp turns, which does not conform to the dynamic characteristics of the UAV, increasing the difficulty of flight control and energy consumption. Summary of the Invention

[0011] To address at least one deficiency or improvement need in existing technologies, this invention provides a multi-strategy UAV rapid obstacle avoidance path planning method and apparatus based on dynamic ellipsoidal sampling. An ellipsoid is constructed using two mathematical definitions to achieve uniform sampling and dynamically adjust parameters, focusing on the effective area, improving sampling and expansion efficiency, and enabling adaptive linkage with the optimal path. Three strategies—direct target connection, conical obstacle avoidance, and random expansion towards sampling points—are integrated and activated sequentially according to priority, balancing exploration and utilization, and improving convergence speed and obstacle avoidance capability. Through greedy direct connection simplification, large-segment interpolation encryption, and B-spline smoothing, the smoothness and dynamic friendliness of the path are significantly improved while ensuring path feasibility and safety.

[0012] To achieve the above objectives, according to a first aspect of the present invention, a fast obstacle avoidance path planning method for multi-strategy unmanned aerial vehicles based on dynamic ellipsoid sampling is provided, the method comprising:

[0013] S1. Based on the starting point and ending point of the UAV path planning, construct an ellipsoid in the form of focal point and linear transformation. Map the sampling points in the unit sphere to the ellipsoid space to obtain the sampling points in the ellipsoid. Then, dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes.

[0014] S2, in order of priority, the multi-strategy expansion mechanism of direct expansion toward the target point, conical obstacle-around expansion and random expansion are used to optimize the initial planned path nodes, and the optimized path nodes and optimized planned path are obtained.

[0015] S3, search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, and select the node with the lowest path cost as the new parent node, and use the new parent node to update the optimized planned path;

[0016] S4. A three-stage path smoothing strategy, consisting of greedy direct connection simplification, large segment interpolation encryption, and B-spline smoothing, is used to smooth the updated optimized planning path to obtain the UAV planning path.

[0017] Furthermore, step S1 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0018] S11, constructing a focal ellipsoid by taking the start and end points of the UAV path planning as the foci of the ellipsoid, and constructing a linearly transformed ellipsoid based on the unit sphere, orthogonal basis and scale matrix;

[0019] S12, determine the center and orientation of the sphere, uniformly sample within the unit sphere, and map the sampling points within the unit sphere to the ellipsoidal space to obtain the sampling points within the ellipsoid;

[0020] S13, through iterative updating of the major and minor axes of the ellipsoid by cyclic sampling, to obtain the initial planned path and path nodes.

[0021] Furthermore, step S12 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0022] Determine the center and orientation of the sphere, and sample uniformly within a unit sphere:

[0023] The sampling points are mapped onto an ellipsoid, and the ellipsoid's focus parameters are used to scale the sampling points. Based on an orthogonal basis constructed from the direction vectors from the start point to the end point, the scaled sampling points are rotated onto the ellipsoid. The rotated sampling points are then translated to their spatial positions on the ellipsoid using addition, thus obtaining the sampling points within the ellipsoid.

[0024] Furthermore, step S2 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0025] S21, use the expansion strategy directly towards the target point to expand the path nodes. If the expansion strategy directly towards the target point is not feasible, proceed to step S22. If the expansion strategy directly towards the target point is feasible, obtain the optimized path nodes and the optimized planned path based on the expanded path nodes.

[0026] S22, use the cone obstacle bypass expansion strategy to expand the path nodes. If the cone obstacle bypass expansion strategy is not feasible, proceed to step S23. If the cone obstacle bypass expansion strategy is feasible, obtain the optimized path nodes and the optimized planned path based on the expanded path nodes.

[0027] S23, use a random direct connection expansion strategy towards the sampling point to expand the path nodes, and obtain the optimized path nodes and optimized planned path based on the expanded path nodes.

[0028] Furthermore, step S21 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0029] The target point direction is determined based on the current parent node and the global target, and new nodes are expanded based on a preset step size;

[0030] Determine the feasibility of the new node. If it is feasible, accept the new node and expand the path. Based on the expanded path nodes, obtain the optimized path nodes and the optimized planned path.

[0031] Furthermore, step S22 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0032] The cone axis is determined based on the current parent node and the global target, and a cone is constructed by combining the unit basis orthogonal to the cone axis.

[0033] Divide the azimuth angle equally on the conical surface of the cone and search for nodes that meet the feasibility conditions in ascending order of deflection angle as new nodes for path expansion. Based on the expanded path nodes, obtain the optimized path nodes and the optimized planned path.

[0034] Furthermore, the aforementioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0035] The new node is within the workspace, the path from the current parent node to the new node does not collide with any obstacles, and the new node maintains a safe distance from obstacles.

[0036] Furthermore, step S3 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0037] S31, with the new node as the center, set a neighborhood with a preset radius, and filter out the set of existing tree nodes within the neighborhood as the neighborhood range;

[0038] S32, for each node in the tree node set, calculate the path cost from the starting point to the current node and the Euclidean distance from the current node to the new node, and check whether the path from the current node to the new node satisfies the collision-free and safe distance constraints from obstacles, and obtain the potential parent node.

[0039] S33, Select the node with the minimum total path cost from the potential parent nodes as the optimal parent node, and update the path cost of the optimal parent node to the minimum total path cost;

[0040] S34. For other nodes in the neighborhood, if the path cost from the optimal parent node to other nodes is less than the minimum total path cost and the feasibility is satisfied, then update the parent node of other nodes to the optimal parent node and recursively update the path cost of their subsequent nodes.

[0041] Furthermore, step S4 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0042] The greedy direct connection simplification includes: for the original discrete path, starting from the current node, taking the farthest node that can be directly connected to the current node in one go and satisfies the security constraints as the new current node and updating the original discrete path to obtain a new path, iterating until the current node is the endpoint;

[0043] The large-segment interpolation encryption includes: if the distance between adjacent nodes is greater than the maximum allowed no-interpolation distance, then configure linear interpolation points to meet the interpolation distance requirement and perform interpolation encryption;

[0044] The B-spline smoothing includes performing cubic spline interpolation in the three coordinate axes of the node to obtain a smooth path output.

[0045] According to a second aspect of the present invention, a multi-strategy UAV rapid obstacle avoidance path planning device based on dynamic ellipsoid sampling is also provided, comprising:

[0046] The sampling point generation module is configured to construct an ellipsoid in focal form and linear transformation form based on the start and end points of the UAV path planning, map the sampling points in the unit sphere to the ellipsoid space to obtain the sampling points in the ellipsoid, and dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes.

[0047] The new node generation module is configured to use a multi-strategy expansion mechanism, namely direct expansion toward the target point, conical obstacle-around expansion, and random expansion, in order of priority to optimize the initial planned path nodes and obtain the optimized path nodes and the optimized planned path.

[0048] The parent node optimization module is configured to search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, select the node with the lowest path cost as the new parent node, and update the optimized planned path using the new parent node.

[0049] The path smoothing module is configured to use a three-stage path smoothing strategy of greedy direct connection simplification, large segment interpolation encryption, and B-spline smoothing to smooth the updated optimized planning path and obtain the UAV planning path.

[0050] According to a third aspect of the present invention, a multi-strategy UAV rapid obstacle avoidance path planning device based on dynamic ellipsoid sampling is also provided, which includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit performs the steps of any of the methods described above.

[0051] According to a fourth aspect of the invention, a storage medium is also provided that stores a computer program executable by an access authentication device, which, when run on the access authentication device, causes the access authentication device to perform the steps of any of the methods described above.

[0052] According to a fifth aspect of the present invention, a computer program product is also provided, comprising a computer program / instructions, characterized in that, when executed by a processor, the computer program / instructions implement the steps of any of the methods described above.

[0053] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0054] (1) The multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling provided by the present invention constructs an ellipsoid through two definitions: focal type and linear transformation type, to achieve uniform sampling from the unit sphere to the ellipsoid and dynamically adjust the ellipsoid parameters. This strategy focuses on more promising areas, has uniformity and unbiasedness, can adaptively link with the best path, and is simple and low-cost to implement.

[0055] (2) The multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling provided by the present invention first adopts the direct target strategy. If it fails, it switches to the cone obstacle bypass expansion. If both of the first two fail, it adopts the random expansion towards the sampling point as a backup strategy. The direct target strategy is greedy, efficient, globally oriented and controllable bias. The cone obstacle bypass strategy can achieve local minimum escape, provide directional diversity and early stopping efficiency. The two strategies work together to balance exploration and utilization and improve accessibility.

[0056] (3) The multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling provided by the present invention is greedy direct connection to simplify and greatly reduce redundancy and is safe and feasible. Large-segment interpolation encryption improves numerical stability. B-spline smoothing makes the path continuous. It is dynamically friendly and has high geometric quality. The three-stage strategy takes into account feasibility, safety and dynamic executability. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 A flowchart illustrating a fast obstacle avoidance path planning method for multi-strategy UAVs based on dynamic ellipsoid sampling, provided for an embodiment of this application;

[0059] Figure 2 This is a schematic diagram of an unmanned aerial vehicle (UAV) urban inspection provided in an embodiment of this application;

[0060] Figure 3 This application provides a comparison chart of the average output path length of four models under different environments in the embodiments of this application;

[0061] Figure 4 A comparison chart of the average number of nodes, average number of iterations, and computation time of different algorithms provided in the embodiments of this application;

[0062] Figure 5 A comparison chart of the average safe distance of different algorithms provided in the embodiments of this application in multiple sets of experiments;

[0063] Figure 6 Distribution of average steering angles for different algorithms provided in embodiments of this application across multiple sets of experiments;

[0064] Figure 7 A 3D path visualization comparison diagram between the algorithm provided in the embodiments of this application and existing algorithms;

[0065] Figure 8 This is a schematic diagram of a multi-strategy UAV fast obstacle avoidance path planning device based on dynamic ellipsoid sampling, provided in an embodiment of this application. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0067] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0068] According to the first aspect of the invention, such as Figure 1 As shown, a fast obstacle avoidance path planning method for multi-strategy UAVs based on dynamic ellipsoid sampling is provided. This method includes:

[0069] S1: Based on the start and end points of the UAV path planning, construct ellipsoids in focal form and linear transformation form. Map the sampling points within the unit sphere to the ellipsoidal space to obtain the sampling points within the ellipsoid. Then, dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes. This step constructs a dynamic ellipsoidal sampling strategy: constructing the ellipsoid using two mathematical definitions to achieve uniform sampling and dynamically adjust parameters, focusing on the effective area, improving sampling and expansion efficiency, and enabling adaptive linkage with the optimal path. That is, constructing the ellipsoid using both focal and linear transformation definitions achieves uniform sampling from the unit sphere to the ellipsoid and dynamically adjusts the ellipsoid parameters a and b. This strategy focuses on more promising areas, utilizes geometric priors early on, possesses uniformity and unbiasedness, can adaptively link with the best path, and is simple and low-cost to implement.

[0070] S2 optimizes the initial planned path nodes by sequentially employing a multi-strategy expansion mechanism: direct target-point expansion, cone-around-obstacle expansion, and random expansion, resulting in optimized path nodes and an optimized planned path. This step constructs a multi-strategy expansion mechanism that integrates three strategies: direct target-point expansion, cone-around-obstacle expansion, and random expansion towards sampling points. These strategies are activated sequentially according to priority, balancing exploration and utilization, and improving convergence speed and obstacle avoidance capabilities. Specifically, the direct target-point expansion strategy is used first; if it fails, it switches to cone-around-obstacle expansion. If both of the first two fail, random expansion towards sampling points is used as a backup strategy. The direct target-point expansion strategy is greedy, efficient, globally oriented, and has controllable bias; the cone-around-obstacle strategy can achieve local minimum escape, provides directional diversity, and is efficient with early stopping. The two strategies work together to balance exploration and utilization, improving reachability.

[0071] S3. Search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, and select the node with the lowest path cost as the new parent node. Use the new parent node to update the optimized planned path. This step builds a reconnection parent node optimization function. After the new node is generated, the connection relationship between nodes is optimized to reduce the path cost.

[0072] S4, a three-stage path smoothing strategy—greedy direct connection simplification, large-segment interpolation encryption, and B-spline smoothing—is employed to smooth the updated optimized planned path, yielding the UAV planned path. This step constructs a three-stage path smoothing strategy, which, through greedy direct connection simplification, large-segment interpolation encryption, and B-spline smoothing, significantly improves the smoothness and dynamic friendliness of the path while ensuring feasibility and safety. Specifically, greedy direct connection simplification greatly reduces redundancy and is safe and feasible; large-segment interpolation encryption improves numerical stability; and B-spline smoothing gives the path C... 2 It features continuity, dynamic friendliness, and high geometric quality; the three-stage strategy balances feasibility, safety, and dynamic executability.

[0073] As can be seen, the four steps of this invention are closely linked, forming a complete path planning process. The dynamic ellipsoid sampling module, as the starting point of the entire process, is responsible for generating targeted sampling points, providing high-quality candidate points for subsequent expansion strategies. The multi-strategy expansion module flexibly selects appropriate expansion strategies to generate new nodes based on different environmental conditions, and is the core link in path growth. The reconnecting parent node optimization module optimizes the connection relationships between nodes after new nodes are generated to reduce path costs. The path smoothing module processes the generated original path, improving its smoothness and usability, making it more suitable for UAV flight.

[0074] Furthermore, step S1 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0075] S11, constructing a focal-form ellipsoid using the start and end points of the UAV path planning as foci, and constructing a linearly transformed ellipsoid based on the unit sphere, orthogonal basis, and scale matrix; as one possible implementation, constructing the focal-form ellipsoid includes: given the start point... and the finish line Distance between two focal points ,center ,by , As the focal point, the major half-axis The slender triaxial ellipsoid is Among them, the short half-shaft (The two minor axes are equal:) ), eccentricity Constructing an ellipsoid of linear transformation involves: taking a unit sphere. Then the ellipsoid can be written as .

[0076] S12, determine the center and orientation of the sphere, uniformly sample within the unit sphere, and map the sampled points within the unit sphere to the ellipsoidal space to obtain the sampled points within the ellipsoid; as a possible implementation, when the loop begins, random nodes are generated through a dynamic ellipsoidal sampling strategy. First, determine the center and orientation of the sphere, then uniformly sample within the unit sphere: until convergence. The sampling points are obtained and mapped onto an ellipsoid. Then, scaling, rotation, and translation are performed using set parameters. Finally, the rotated points are translated to their spatial positions on the ellipsoid using addition, resulting in the sampling points within the ellipsoid.

[0077] S13, the semi-major and semi-minor axes of the ellipsoid are updated iteratively through cyclic sampling to obtain the initial planned path and path nodes. As one possible implementation, the main loop updates the path every fixed number of iterations. and ,in For parameters As the major axis increases, the increase in the minor axis is smaller than that of the major axis, thus prioritizing the exploration of the major axis and resulting in a more slender ellipse. This sampling point setting focuses on a more probable range, reduces exploration of regions far from the target point, and transforms the mapping into a single operation, resulting in relatively low computational complexity and improved algorithm speed while meeting requirements.

[0078]

[0079]

[0080] Furthermore, step S12 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0081] The center and orientation of the sphere are determined, and uniform sampling is performed within the unit sphere: the sampling points are mapped onto an ellipsoid, and the sampling points are scaled using the ellipsoid's focus parameters. Based on an orthogonal basis constructed from the direction vectors from the start to the end point, the scaled sampling points are rotated onto the ellipsoid. Addition is then used to translate the rotated sampling points to their spatial positions within the ellipsoid, resulting in sampling points within the ellipsoid. As one possible implementation, random nodes are generated using a dynamic ellipsoid sampling strategy at the start of the loop.

[0082] First, determine the center and orientation of the ball; the corresponding code is (here). , Let c be the direction vector. ):

[0083]

[0084]

[0085] Then, uniform sampling is performed within the unit sphere: repeatedly at... Repeat sampling until The sampled points are obtained and mapped onto an ellipsoid. The first line of code below uses the parameters set in S11 to perform a scaling transformation, and the second line of code implements the rotation and translation transformation. It is an orthogonal basis constructed from the direction vectors from the start point to the end point, used to rotate the scaled points to the pose of the ellipsoid (the major axis is along the direction from the start point to the end point); ' and ' respectively represent the matrix transpose operation, ensuring that the vector dimensions match; It is the center of the ellipsoid , Starting from, (with the endpoint as the reference point), the rotated point is finally translated to its spatial position on the ellipsoid through addition, thus obtaining the sampling point within the ellipsoid. :

[0086]

[0087]

[0088] Furthermore, due to boundary constraints, if x is not in the workspace If the sample is inside, then resampling is performed.

[0089] Furthermore, step S2 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0090] S21, expand the path nodes using a direct-to-target-point expansion strategy. If the direct-to-target-point expansion strategy is not feasible, proceed to step S22. If the direct-to-target-point expansion strategy is feasible, obtain the optimized path nodes and optimized planned path based on the expanded path nodes. As one possible implementation, the current parent node is known. With global goal ,direction Single-step expansion of new nodes ,in,

[0091]

[0092]

[0093] Step size set , utilizing factors Control and balance direction and speed. When the direction is feasible, quickly advance to the next node. When it is not feasible, control the speed to reduce the risk of collision.

[0094] The feasibility assessment is as follows: In the workspace Inside; line segment No collision ( Maintain a safe distance between new nodes and obstacles. If the above conditions are met, then accept and proceed. Reconnect; otherwise, enter the cone obstacle avoidance strategy.

[0095] S22, the path nodes are expanded using a cone-shaped obstacle-around expansion strategy. If the cone-shaped obstacle-around expansion strategy is not feasible, step S23 is executed. If the cone-shaped obstacle-around expansion strategy is feasible, the optimized path nodes and optimized planned path are obtained based on the expanded path nodes. As a possible implementation, when the direct-charge strategy is not feasible, from... around The axis, i.e. the deflection angle on the conical surface, is tested in ascending order of the deflection angle. When the first alternative direction is encountered, the value is returned, thus reducing the calculation time.

[0096] Cone parameter and basis settings: Axial unit vector Construct a unit basis orthogonal to u. }, , ; .

[0097] Direction generation: given deflection angle With azimuth Unit direction Single-step alternative nodes .

[0098] Discrete search order: Take the set of increasing deflection angles Divide the set of azimuth angles into equal parts Double-layer loop outer layer Inner layer scan, the first point that meets the feasibility conditions is accepted.

[0099] S23, the path nodes are expanded using a random direct connection expansion strategy toward the sampling points, and the optimized path nodes and optimized planned paths are obtained based on the expanded path nodes. As a possible implementation, random direct connection expansion toward the sampling points (Rand-line) is adopted: this is enabled when the previous two strategies fail.

[0100] First, select the parent node from all possible tree nodes. Find the sampling point The most recent one :

[0101]

[0102]

[0103] Obtaining direction and amplitude limiting steps:

[0104]

[0105]

[0106] By using random direct connections, the scenario where both strategies fail is filled in. In this case, the new node is located on a sphere with radius s and the original node as its center.

[0107]

[0108]

[0109] The feasibility assessment is as follows: It is still... Within the workspace; line segment No collision ( ); If the conditions are met, the connection is inserted and reconnected according to the RRT* rules; otherwise, this strategy fails.

[0110] Furthermore, step S21 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0111] The target point direction is determined based on the current parent node and the global target, and new nodes are expanded based on a preset step size. The feasibility of the new nodes is assessed; if feasibility is satisfied, the new nodes are accepted and the path is expanded. Optimized path nodes and an optimized planned path are obtained based on the expanded path nodes. As one possible implementation, the current parent node is known. With global goal ,direction Single-step expansion of new nodes ,in,

[0112]

[0113]

[0114] Step size set , utilizing factors Control and balance direction and speed. When the direction is feasible, quickly advance to the next node. When it is not feasible, control the speed to reduce the risk of collision.

[0115] The feasibility assessment is as follows: In the workspace Inside; line segment No collision ( Maintain a safe distance between new nodes and obstacles. If the above conditions are met, then accept and proceed. Reconnect; otherwise, enter the cone obstacle avoidance strategy.

[0116] Furthermore, step S22 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0117] The cone axis is determined based on the current parent node and the global target. A cone is constructed using unit bases orthogonal to the cone axis. The azimuth angle is equally divided on the cone surface, and nodes satisfying the feasibility conditions are searched in ascending order of deflection angle as new nodes for path expansion. Optimized path nodes and optimized planned paths are obtained based on the expanded path nodes. As one possible implementation, when a direct rush strategy is not feasible, [the following steps are taken]. around The axis, i.e. the deflection angle on the conical surface, is tested in ascending order of the deflection angle. When the first alternative direction is encountered, the value is returned, thus reducing the calculation time.

[0118] Cone parameter and basis settings: Axial unit vector Construct a unit basis orthogonal to u. }, , ; .

[0119] Direction generation: given deflection angle With azimuth Unit direction Single-step alternative nodes .

[0120] Discrete search order: Take the set of increasing deflection angles Divide the set of azimuth angles into equal parts Double-layer loop outer layer Inner layer scan, the first point that meets the feasibility conditions is accepted.

[0121] As can be seen, the multi-strategy expansion function of this invention first uses the most aggressive direct expansion strategy toward the target point, which can greatly improve the convergence speed. If the direct expansion toward the target point fails, it switches to conical obstacle-around expansion. Both of these methods have relatively high initial speeds, which means that the path has fewer nodes and a faster convergence speed.

[0122] Furthermore, regardless of the expansion strategy used to generate a new node, an attempt is made to find a better parent node within the neighborhood of the new node to optimize the path cost, and reconnection is performed according to the RRT* rule. Step S3 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0123] S31, using the new node as the center, define a neighborhood with a preset radius, and select the set of existing tree nodes within the neighborhood as the neighborhood range; that is, determine the neighborhood range: using the new node... Centered on a node, define a neighborhood with a certain radius (usually related to the step size), and then select the set of existing tree nodes within that neighborhood. .

[0124] S32, for each node in the tree node set, calculate the path cost from the starting point to the current node and the Euclidean distance from the current node to the new node, and check whether the path from the current node to the new node satisfies the collision-free and safe distance constraints from obstacles to obtain potential parent nodes; that is, evaluate potential parent nodes: for the set Each node in Calculate from the starting point to Path cost plus arrive Euclidean distance (i.e., total path cost) ), and check the line segments Does it meet the no-collision and safe distance constraints (distance from obstacles)? ).

[0125] S33, select the node with the minimum total path cost from the potential parent nodes as the optimal parent node, and update the path cost of the optimal parent node to the minimum total path cost; that is, determine the optimal parent node: select the node with the minimum total path cost as... The new parent node, updated The path cost is the minimum total cost.

[0126] S34, for other nodes in the neighborhood, if the path cost from the optimal parent node to other nodes is less than the minimum total path cost and feasibility is satisfied, then update the parent node of the other nodes to the optimal parent node, and recursively update the path cost of their subsequent nodes. That is, reconnection optimization: for other nodes in the neighborhood... If from arrive Path cost ( (less than) Current path cost, and line segment If feasible, then The parent node is updated to It then recursively updates the path cost of its subsequent nodes.

[0127] Furthermore, step S4 of the above-mentioned multi-strategy UAV fast obstacle avoidance path planning method based on dynamic ellipsoid sampling also includes:

[0128] This process has three stages: polyline simplification under feasibility constraints → numerical encryption → The constraints for continuous curve fitting are:

[0129] No collision:

[0130] : Represents two consecutive points on the smoothed path (or any two points on the path segment). and The path parameter is the position index on the path.

[0131] O: The set of obstacles, which includes all obstacles in the environment (such as spheres, cubes, cylinders, etc.).

[0132] Safe distance:

[0133] Its implicit optimization objective is:

[0134]

[0135] This is a length weight; a larger value indicates that the path should be shortened. For bending weight, a larger value tends to reduce bending. Represents the Euclidean distance between two adjacent nodes, used to measure the length of a path segment. This involves considering the combined cost of "length + second-order difference (curvature)". By optimizing the objective function value, the total length and curvature of the path are comprehensively reflected. The ultimate goal is to minimize J, achieving a balance between "short path + high smoothness".

[0136] The greedy direct-connect simplification includes: for the original discrete path, starting from the current node, taking the farthest node that can be directly connected to the current node in one step and satisfies the safety constraints as the new current node, and updating the original discrete path to obtain a new path, iterating until the current node becomes the destination. Specifically, the original discrete path... From the current index Let's set off to find a safe, direct connection (no collisions) The farthest node (with certain distance constraints) ,Will Add to the new path and make Iterate to .

[0137] A certain distance is ,

[0138] Safety distance constraints:

[0139] The above formula corresponds to the collision-free determination code. `res` represents the sampling resolution, which determines the density of sampling points on the line segment during the collision detection phase. Higher resolution results in more accurate collision detection. This method addresses the greedy subproblem of finding the furthest feasible direct connection, skipping as many removable nodes as possible in each step, or in other words, directly reducing the computation of implicit costs.

[0140] Path length:

[0141]

[0142] Broken line curvature (second difference):

[0143]

[0144] By reducing computation and skipping nodes, redundancy is significantly eliminated, often resulting in a substantial reduction in path length, and the system remains safe and feasible, always operating under safety constraints.

[0145] The large-segment interpolation encryption includes: if the distance between adjacent nodes is greater than the maximum allowed no-interpolation distance, then configuring linear interpolation points to meet the interpolation distance requirement and performing interpolation encryption; specifically, for adjacent points... ,like (dmax is the maximum allowed distance between adjacent nodes without interpolation, set to 100 according to the experimental environment; exceeding this requires encrypted interpolation), then in the parameter Linear interpolation points at the location T represents the number of interpolation points. This step provides more uniform support points for the next stage of smoothing, effectively avoiding path oscillations or over-smoothing caused by excessively large point spans.

[0146] The B-spline smoothing involves performing cubic spline interpolation along the three coordinate axes of the nodes to obtain a smoothed path output. Specifically, the encrypted discrete points are: (Equal spacing parameters) , (For three-dimensional coordinates), cubic spline interpolation is performed along the three coordinate axes, with the following basis functions: order ( )

[0147] Three-axis interpolation formula: ( The interpolation coefficients are obtained by satisfying... The conditions can be met to find out. (where B is the interpolation parameter and B is the spline basis function)

[0148] Then add comments to the finer mesh: , To summarize the points;

[0149] Obtain a smooth path output: ;

[0150] Cubic spline interpolation has Continuity is maintained; position, velocity, and acceleration are continuous, curvature and acceleration are smoother, and parameter images are reduced, making it more suitable for UAV turning acceleration and facilitating trajectory tracking. Interpolation splines are used here, strictly passing through the original points (including densified points) to ensure reliability. The final overall smoothness comes from inter-segment continuity and high-density fine sampling. Natural boundary / node vectors at endpoints are processed to refine the sampling density. It only affects the output resolution, does not change the geometric curve, has reliability, and more importantly, removes the sharp corners of the broken lines, improving smoothness. The smooth curve is actually closer to low power consumption.

[0151] To verify the effectiveness of the improved RRT algorithm, we conducted comparative experiments. Combined with... Figure 2-7As shown, 3D map models were first created for two classic environments. Obstacle types supported included spheres (parameters: center coordinates + radius), cubes (parameters: center coordinates + side length), and cylinders (parameters: base center coordinates + radius + height). Complex scenes were simulated by adjusting the number, position, and size of the obstacles. Each environment contained twenty different dynamic obstacle maps. Environment one had a larger proportion of cubes, while environment two had a greater proportion of spheres and cylinders. Repeated comparative experiments were conducted in these environments.

[0152] The evaluation metrics used are commonly used predictive performance metrics. Path length is calculated by summing the lengths of all line segments on the path, reflecting the optimality of the path generated by the algorithm; the shorter the value, the better the path. Computation time is the time taken from algorithm startup to finding the path, reflecting the algorithm's time efficiency; less time indicates better real-time performance. Total number of nodes is the number of all nodes generated during the search process; for a single-tree algorithm, it is the number of nodes in a single tree, and for a bidirectional algorithm, it is the total number of nodes in both trees. This reflects the algorithm's space complexity; fewer nodes indicate lower resource consumption. Number of iterations is the number of iterations required to find the path, reflecting the algorithm's convergence speed; fewer iterations indicate faster convergence. Average obstacle distance is the average distance between all points on the path and obstacles, reflecting the path's safety; a larger distance indicates higher obstacle avoidance redundancy and a safer path. Average turning angle is the average angle between adjacent line segments on the path, reflecting the path's smoothness; a smaller angle indicates a smoother path, which is more conducive to actual execution.

[0153] The benchmark models selected for comparison include RRT*, BI-RRT, and BI-APF-RRT. RRT* is an algorithm that optimizes the path through single-tree expansion, relying on random sampling and reconnection mechanisms. BI-RRT is a bidirectional version that grows two trees simultaneously from the start and end points to improve search efficiency. BI-APF-RRT* is an improved algorithm that integrates an artificial potential field on top of bidirectional expansion, introducing attraction and repulsion to guide the expansion direction. Table 1 shows the average performance comparison of each model in the two environments.

[0154]

[0155] Table 1

[0156] By comparing experimental data and visual analysis, the multi-strategy UAV fast obstacle avoidance method based on dynamic ellipsoid sampling proposed in this invention has achieved multi-dimensional breakthroughs in path planning performance. The specific technical effects and reasons are as follows:

[0157] 1. Improved Path Length, Significantly Enhancing Planning Efficiency: Table 1 shows that the improved RRT achieves an average path length of 1679.4765 grid cells in Environment 1, more than 20% shorter than RRT* (2100.1155), and 13% and 12% shorter than BI-RRT (1951.8400) and BI-APF-RRT (1920.4590), respectively. In Environment 2, the improved RRT achieves an average path length of 1384.9450 grid cells, 30% shorter than RRT* (1991.6320), and approximately 27% shorter than BI-RRT (1904.3680) and BI-APF-RRT (1888.5080). This advantage stems from the precise focusing of the search area by the dynamic ellipsoid sampling strategy—through dynamically adjusting the semi-major axis of the ellipsoid. and short half shaft This prioritizes the distribution of sampling points near the potentially optimal path from the start to the end, reducing the exploration of invalid regions. Simultaneously, the direct-target strategy in the multi-strategy expansion further accelerates convergence to the end point, thereby shortening the final path length. (Appendix) Figure 3 It also demonstrates that the improved algorithm performs better on obstacles containing spherical or cylindrical shapes.

[0158] 2. Significantly reduced computation time and enhanced real-time performance: The average computation time of the improved RRT in Environment 1 is 0.0830 seconds, only 16.5% of RRT* (0.5035 seconds), slightly lower than BI-RRT (0.1065 seconds) and BI-APF-RRT (0.1865 seconds); in Environment 2, its computation time is further reduced to 0.0610 seconds, only 8.5% of RRT* (0.7175 seconds), and 51.7% of BI-RRT (0.1180 seconds). This is due to two improvements: First, dynamic ellipsoid sampling achieves the mapping from a unit sphere to an ellipsoid through a single linear transformation, resulting in low computational complexity and avoiding the blindness of random sampling in traditional RRT; second, multi-strategy expansion is enabled sequentially according to priority (direct target → cone obstacle bypass → random expansion), reducing invalid exploration and enabling the algorithm to find feasible paths in fewer iterations.

[0159] 3. Reduced number of nodes and iterations, resulting in lower resource consumption: The improved RRT in Environment 1 has an average of 24 nodes and 30 iterations, far lower than RRT*'s 334 nodes and 419 iterations, and also significantly less than BI-RRT's 67 nodes and 101 iterations. In Environment 2, its number of nodes (19) and iterations (18) are also significantly better than the comparison algorithms. This is because the candidate points generated by dynamic ellipsoid sampling are more targeted, and with the reconnection of parent nodes optimization module, high-quality paths can be built with fewer nodes. The multi-strategy expansion's "direct rush" strategy (prioritizing direct connection to the target) further reduces the number of iterations required for path growth, reducing memory usage and computational resource consumption.

[0160] 4. Improved safety distance and superior path security: The average obstacle distance of the improved RRT reaches 52.7505 in Environment 1 and 40.1560 in Environment 2, both higher than all compared algorithms. This result stems from the safety constraint design of the path smoothing module: during the greedy direct connection simplification stage, the distance between the path and the obstacle is strictly required to be greater than 0.3×d_min to ensure the security of the original path; in the subsequent encryption and spline smoothing processes, the interpolation algorithm maintains a safety redundancy for obstacles, avoiding the risk of traditional algorithms getting too close to obstacles in pursuit of short paths. In addition, the cone obstacle avoidance strategy uses deflection angle control to make the obstacle avoidance path naturally move away from the obstacle, further improving the overall security.

[0161] 5. Significantly improved path smoothness and enhanced dynamic adaptability: The average steering angle of the improved RRT is 16.7020 degrees in Environment 1 and 32.3025 degrees in Environment 2, representing only 41.6% and 69.9% of the RRT* (40.0845 degrees and 46.1595 degrees), respectively. This is attributed to the synergistic effect of the three-stage smoothing strategy: greedy direct connection simplification removes redundant polylines, large-segment interpolation densification provides uniform support points for smoothing, and cubic spline interpolation ensures the path's C-axis stability. 2 Continuity (position, velocity, and acceleration are continuous). As can be seen from the angular distribution visualization in the attached figure, the improved RRT's steering angle is concentrated in a small angle range, avoiding the sharp turns that frequently occur in traditional algorithms. This is more in line with the dynamic characteristics of UAVs and can reduce the difficulty of flight control and energy consumption.

[0162] In summary, this invention, through a collaborative design of dynamic ellipsoidal sampling, multi-strategy expansion, reconnection optimization, and three-stage smoothing, comprehensively surpasses traditional RRT and its improved algorithms in terms of path length, computational efficiency, resource consumption, security, and dynamic adaptability, as shown in the appendix. Figure 7 As shown, it is particularly suitable for drone operation scenarios with high requirements for real-time performance and safety in dynamic and complex environments.

[0163] According to a second aspect of the invention, such as Figure 8 As shown, a multi-strategy UAV rapid obstacle avoidance path planning device based on dynamic ellipsoid sampling is also provided, which includes:

[0164] The sampling point generation module is configured to construct an ellipsoid in focal form and linear transformation form based on the start and end points of the UAV path planning, map the sampling points in the unit sphere to the ellipsoid space to obtain the sampling points in the ellipsoid, and dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes.

[0165] The new node generation module is configured to use a multi-strategy expansion mechanism, namely direct expansion toward the target point, conical obstacle-around expansion, and random expansion, in order of priority to optimize the initial planned path nodes and obtain the optimized path nodes and the optimized planned path.

[0166] The parent node optimization module is configured to search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, select the node with the lowest path cost as the new parent node, and update the optimized planned path using the new parent node.

[0167] The path smoothing module is configured to use a three-stage path smoothing strategy of greedy direct connection simplification, large segment interpolation encryption, and B-spline smoothing to smooth the updated optimized planning path and obtain the UAV planning path.

[0168] The steps performed by each module in the above-mentioned multi-strategy UAV rapid obstacle avoidance path planning device based on dynamic ellipsoid sampling are the same as those of the aforementioned multi-strategy UAV rapid obstacle avoidance path planning device method based on dynamic ellipsoid sampling, and will not be elaborated here.

[0169] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0170] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0171] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0172] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0173] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0174] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0175] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0176] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0177] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0178] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0179] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fast obstacle avoidance path planning method for multi-strategy UAVs based on dynamic ellipsoid sampling, characterized in that, include: S1. Based on the starting point and ending point of the UAV path planning, construct an ellipsoid in the form of focal point and linear transformation. Map the sampling points in the unit sphere to the ellipsoid space to obtain the sampling points in the ellipsoid. Then, dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes. S2, in order of priority, the multi-strategy expansion mechanism of direct expansion toward the target point, conical obstacle-around expansion and random expansion are used to optimize the initial planned path nodes, and the optimized path nodes and optimized planned path are obtained. S3, search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, and select the node with the lowest path cost as the new parent node, and use the new parent node to update the optimized planned path; S4. A three-stage path smoothing strategy of greedy direct connection simplification, large segment interpolation encryption, and B-spline smoothing is adopted to smooth the updated optimized planning path to obtain the UAV planning path. Step S1 further includes: S11, constructing a focal ellipsoid by taking the start and end points of the UAV path planning as the foci of the ellipsoid, and constructing a linearly transformed ellipsoid based on the unit sphere, orthogonal basis and scale matrix; S12, determine the center and orientation of the sphere, uniformly sample within the unit sphere, and map the sampling points within the unit sphere to the ellipsoidal space to obtain the sampling points within the ellipsoid; S13, the major and minor axes of the ellipsoid are updated iteratively through cyclic sampling to obtain the initial planned path and path nodes; Step S2 also includes: S21, use the expansion strategy directly towards the target point to expand the path nodes. If the expansion strategy directly towards the target point is not feasible, proceed to step S22. If the expansion strategy directly towards the target point is feasible, obtain the optimized path nodes and the optimized planned path based on the expanded path nodes. S22, use the cone obstacle bypass expansion strategy to expand the path nodes. If the cone obstacle bypass expansion strategy is not feasible, proceed to step S23. If the cone obstacle bypass expansion strategy is feasible, obtain the optimized path nodes and the optimized planned path based on the expanded path nodes. S23, use a random direct connection expansion strategy towards the sampling point to expand the path nodes, and obtain the optimized path nodes and optimized planned path based on the expanded path nodes.

2. The multi-strategy UAV rapid obstacle avoidance path planning method as described in claim 1, characterized in that, Step S12 also includes: Determine the center and orientation of the sphere, and sample uniformly within a unit sphere: The sampling points are mapped onto an ellipsoid, and the ellipsoid's focus parameters are used to scale the sampling points. Based on an orthogonal basis constructed from the direction vectors from the start point to the end point, the scaled sampling points are rotated onto the ellipsoid. The rotated sampling points are then translated to their spatial positions on the ellipsoid using addition, thus obtaining the sampling points within the ellipsoid.

3. The multi-strategy UAV rapid obstacle avoidance path planning method as described in claim 1, characterized in that, Step S21 also includes: The target point direction is determined based on the current parent node and the global target, and new nodes are expanded based on a preset step size; Determine the feasibility of the new node. If it is feasible, accept the new node and expand the path. Based on the expanded path nodes, obtain the optimized path nodes and the optimized planned path.

4. The multi-strategy UAV rapid obstacle avoidance path planning method as described in claim 1, characterized in that, Step S22 also includes: The cone axis is determined based on the current parent node and the global target, and a cone is constructed by combining the unit basis orthogonal to the cone axis. Divide the azimuth angle equally on the conical surface of the cone and search for nodes that meet the feasibility conditions in ascending order of deflection angle as new nodes for path expansion. Based on the expanded path nodes, obtain the optimized path nodes and the optimized planned path.

5. The multi-strategy UAV rapid obstacle avoidance path planning method as described in any one of claims 3-4, characterized in that, The feasibility meets the following conditions: The new node is within the workspace, the path from the current parent node to the new node does not collide with any obstacles, and the new node maintains a safe distance from obstacles.

6. The multi-strategy UAV rapid obstacle avoidance path planning method as described in claim 1, characterized in that, Step S3 also includes: S31, with the new node as the center, set a neighborhood with a preset radius, and filter out the set of existing tree nodes within the neighborhood as the neighborhood range; S32, for each node in the tree node set, calculate the path cost from the starting point to the current node and the Euclidean distance from the current node to the new node, and check whether the path from the current node to the new node satisfies the collision-free and safe distance constraints from obstacles, and obtain the potential parent node. S33, Select the node with the minimum total path cost from the potential parent nodes as the optimal parent node, and update the path cost of the optimal parent node to the minimum total path cost; S34. For other nodes in the neighborhood, if the path cost from the optimal parent node to other nodes is less than the minimum total path cost and the feasibility is satisfied, then update the parent node of other nodes to the optimal parent node and recursively update the path cost of their subsequent nodes.

7. The multi-strategy UAV rapid obstacle avoidance path planning method as described in claim 1, characterized in that, Step S4 also includes: The greedy direct connection simplification includes: for the original discrete path, starting from the current node, taking the farthest node that can be directly connected to the current node in one go and satisfies the security constraints as the new current node and updating the original discrete path to obtain a new path, iterating until the current node is the endpoint; The large-segment interpolation encryption includes: if the distance between adjacent nodes is greater than the maximum allowed no-interpolation distance, then configure linear interpolation points to meet the interpolation distance requirement and perform interpolation encryption; The B-spline smoothing includes performing cubic spline interpolation in the three coordinate axes of the node to obtain a smooth path output.

8. A multi-strategy UAV rapid obstacle avoidance path planning device based on dynamic ellipsoid sampling, characterized in that, include: The sampling point generation module is configured to construct an ellipsoid in focal form and linear transformation form based on the start and end points of the UAV path planning, map the sampling points in the unit sphere to the ellipsoid space to obtain the sampling points in the ellipsoid, and dynamically adjust the ellipsoid parameters through cyclic sampling to obtain the initial planned path and path nodes. The new node generation module is configured to use a multi-strategy expansion mechanism, namely direct expansion toward the target point, conical obstacle-around expansion, and random expansion, in order of priority to optimize the initial planned path nodes and obtain the optimized path nodes and the optimized planned path. The parent node optimization module is configured to search for potential parent nodes that meet the feasibility conditions in the neighborhood of the optimized path node, select the node with the lowest path cost as the new parent node, and update the optimized planned path using the new parent node. The path smoothing module is configured to use a three-stage path smoothing strategy of greedy direct connection simplification, large segment interpolation encryption, and B-spline smoothing to smooth the updated optimized planning path and obtain the UAV planning path. The sampling point generation module specifically includes: The starting and ending points of the UAV path planning are used as the foci of the ellipsoid to construct an ellipsoid in the form of a focal sphere. An ellipsoid in the form of a linear transformation is constructed based on the unit sphere, orthogonal basis, and scale matrix. Determine the center and orientation of the sphere, sample uniformly within the unit sphere, and map the sampling points within the unit sphere to the ellipsoidal space to obtain the sampling points within the ellipsoid; The initial planned path and path nodes are obtained by iteratively updating the major and minor semi-axes of the ellipsoid through cyclic sampling. The new node generation module specifically includes: The path nodes are expanded using a direct-to-target-point expansion strategy. If the direct-to-target-point expansion strategy is not feasible, step S22 is executed. If the direct-to-target-point expansion strategy is feasible, the optimized path nodes and optimized planned path are obtained based on the expanded path nodes. The path nodes are expanded using a cone obstacle bypass expansion strategy. If the cone obstacle bypass expansion strategy is not feasible, step S23 is executed. If the cone obstacle bypass expansion strategy is feasible, the optimized path nodes and optimized planned path are obtained based on the expanded path nodes. The path nodes are expanded using a random direct connection expansion strategy oriented towards the sampling points, and the optimized path nodes and optimized planned paths are obtained based on the expanded path nodes.

Images