Unmanned aerial vehicle path planning method and system based on improved artificial potential field algorithm

Abstract

The application discloses an unmanned aerial vehicle path planning method and system based on an improved artificial potential field algorithm, and relates to the technical field of unmanned aerial vehicle path planning.The method precisely constructs an airport fusion operation environment model through uniform grid division, comprehensively quantifies distance risks, air route risks and no-fly risks in combination with a Reich model, realizes fine evaluation of collision risks, and solves the problems of one-sided risk evaluation and rough environment description in traditional path planning;with the aid of a self-adaptive particle swarm optimization algorithm, a global virtual navigation path is generated to provide global guidance for the improved artificial potential field algorithm, effectively avoiding the technical bottleneck that the traditional artificial potential field algorithm is prone to local minimum values, and the global virtual navigation path is locally optimized with collision risks as the core, further relieving the problem of path correction lag in a dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) path planning technology, and in particular to a UAV path planning method and system based on an improved artificial potential field algorithm. Background Technology

[0002] With the rapid development of drone technology, its application in various fields is becoming increasingly widespread. Within airport controlled airspace, the integrated operation of drones and transport aviation is becoming an important future trend. However, the integrated operation of drones and manned transport aviation within airport controlled airspace faces core challenges such as high environmental complexity, stringent safety constraints, and strong real-time requirements. Traditional path planning algorithms struggle to simultaneously meet the triple requirements of "obstacle avoidance reliability, risk controllability, and real-time responsiveness." Traditional artificial potential field algorithms rely solely on local information about the attraction and repulsion forces at the target point, easily getting trapped in local minima of attraction and repulsion equilibrium in complex environments, leading to path stagnation or target unreachability. Repulsion force calculations are based only on distance thresholds, without considering quantitative assessments of collision risk, failing to ensure the path meets preset safety levels. Existing improved APF algorithms either only optimize local obstacle avoidance logic or rely on complex global search algorithms, resulting in excessive computational load and failing to meet the real-time requirements of airport scenarios. Summary of the Invention

[0003] The purpose of this invention is to provide a UAV path planning method and system based on an improved artificial potential field algorithm, so as to improve the technical problems of traditional APF, such as local minima, uncontrollable risks, poor adaptability to manned and unmanned aircraft, and the inability of the generated path plan to meet the constraints of the target safety level.

[0004] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0005] A method for UAV path planning based on an improved artificial potential field algorithm, comprising:

[0006] Collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and construct an airport integrated operation environment model through a uniform grid partitioning algorithm;

[0007] Based on the airport integrated operation environment model, the collision risk of drones at each grid point is calculated using the Reich model; the drone collision risk includes distance risk, route risk, and no-fly risk.

[0008] Set the start and end points of the drones, and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm;

[0009] Based on the collision risks of each UAV, an improved artificial potential field algorithm is used to optimize the global virtual navigation path, generating an optimized global virtual navigation path.

[0010] In the above process, this method accurately constructs an airport integrated operation environment model through uniform grid division, and comprehensively quantifies distance risk, route risk, and no-fly risk by combining the Reich model, thereby achieving a refined assessment of collision risk. This solves the problems of one-sided risk assessment and coarse environmental description in traditional path planning. By using an adaptive particle swarm optimization algorithm to generate a global virtual navigation path, it provides global guidance for the improved artificial potential field algorithm, effectively avoiding the technical bottleneck of traditional artificial potential field algorithms being prone to getting trapped in local minima. At the same time, it performs local optimization of the global virtual navigation path with collision risk as the core, ensuring that the path meets the stringent safety constraints of the airport while taking into account the real-time responsiveness in dynamic environments. This solves the core problem of the difficulty in balancing "obstacle avoidance reliability, risk controllability, and real-time adaptability" in the integrated operation scenario of airport airspace UAVs and transport aviation. It also improves the smoothness and executability of the path, adapts to the actual application needs of UAV flight control systems, and further alleviates the problem of path correction lag in dynamic environments (such as sudden changes in the frequency of manned aircraft flight routes and the addition of obstacles).

[0011] Furthermore, the calculation of drone collision risk at each grid point includes:

[0012] Based on the coordinate information of fixed obstacles and the position information of UAVs at each grid point in the airport integrated operation environment model, the distance risk between the UAV and the fixed obstacles is calculated by the Reich model.

[0013] Based on the route information in the airport integrated operation environment model, the route risk of each grid point is calculated using the Reich model.

[0014] Based on the no-fly zone range and drone location information in the airport integrated operation environment model, the no-fly risk of each grid point is set;

[0015] Based on the distance risk, flight path risk, and no-fly zone risk of each grid point, the drone collision risk of each grid point is calculated using a weighted summation minimum value function.

[0016] In the above process, this method accurately calculates the collision risk characteristics of each grid point in multiple dimensions, quantifies the distance risk of fixed obstacles and the route risk of manned aircraft routes based on the airport integrated operation environment model and the Reich model, and sets no-fly risks in a targeted manner. Then, it achieves the scientific fusion of multi-dimensional risks through a weighted sum minimum value function. This not only solves the technical problems of relying solely on a single distance indicator for collision risk assessment, and the one-sided consideration of multiple types of risks in the complex airport environment without quantitative basis, but also makes the risk values ​​of each grid point more consistent with the actual airport operation scenario. This provides accurate and comprehensive risk data support for the adaptive particle swarm optimization algorithm to generate a global virtual navigation path. At the same time, it avoids the problem of the global path falling into a high-risk area due to inaccurate risk assessment or the inappropriate obstacle avoidance decision when the subsequent artificial potential field algorithm is locally optimized.

[0017] Further, the generation of the global virtual navigation path includes:

[0018] Based on the drone's origin and destination, multiple candidate paths are generated;

[0019] Each candidate path is treated as a particle, and the initial position and initial velocity vector of each particle are initialized, and adaptive parameters are set; the adaptive parameters include inertia weight and group learning factor;

[0020] Based on the airport integrated operation environment model and the UAV collision risk of each grid point, the fitness of each particle is calculated through a fitness function;

[0021] Based on each fitness level and the historical best fitness, update the individual's historical best position and the global best position;

[0022] The adaptive parameters are updated by using a linear decay strategy, and the population learning factor is dynamically adjusted based on the maximum value of each fitness, resulting in the updated adaptive parameters and the updated population learning factor.

[0023] Based on the updated individual historical best position, the updated global best position, the updated adaptive parameters, and the updated group learning factor, the velocity vector and position of each particle are updated to obtain the global best position of the current iteration.

[0024] Determine if the current iteration meets the iteration conditions; if so, use the current iteration's global best position as the global virtual navigation path; otherwise, re-enter the iteration and update the iteration count.

[0025] Furthermore, the calculation of the fitness of each particle includes:

[0026] For each candidate path, dense sampling is performed, and based on the drone collision risk of each grid point, the total collision risk value of all sampled points in different particles is generated;

[0027] Based on the position information of each particle and the scope of the no-fly zone, compliance penalty items are generated;

[0028] Calculate the path length of each particle based on its position information;

[0029] Calculate the straight-line distance of the drone's flight based on its starting and ending points;

[0030] The fitness of each particle is generated using a fitness function based on the straight-line distance of flight, the total collision risk value of each particle, and the path length.

[0031] In the above process, this method quantifies the quality of particles by integrating the fitness function of total collision risk value, compliance penalty term and path length, and then updates the inertia weight with a linear decay strategy. This solves the technical problems of traditional particle swarm optimization algorithm, which is prone to insufficient global search or convergence stagnation due to fixed parameters, path evaluation focuses only on a single indicator and ignores compliance constraints and risk accumulation in airport scenarios, and global path generation lacks accurate risk data support and is prone to falling into high-risk areas. In addition, the method solves the local minimum problem of traditional APF (Artificial Potential Field Algorithm) by generating virtual navigation paths through adaptive particle swarm optimization algorithm, so that the improved artificial potential field algorithm can achieve local dynamic obstacle avoidance, taking into account both global optimum and real-time performance.

[0032] Furthermore, the processing steps of the improved artificial potential field algorithm include:

[0033] Based on the guiding constraints, set the guiding node, gravity coefficient, and basic repulsion coefficient for the current iteration; based on the route risk and the basic repulsion coefficient, calculate the manned aircraft route repulsion coefficient for the current iteration.

[0034] Based on the drone's starting point, ending point, and guiding node, the drone's path is simulated. Combining the gravity coefficient and the manned aircraft's repulsion coefficient, the drone's dual gravity, total repulsion, and unit motion direction are calculated in the current iteration.

[0035] Based on the current unit motion direction, update the drone's position information and calculate the corresponding drone collision risk;

[0036] Based on the updated location information, the collision risk of the drone is screened and iterated until the drone reaches the destination, generating an initial optimized path for global virtual navigation.

[0037] The initial optimized path for global virtual navigation is verified for path smoothness and executability, and a global virtual navigation optimized path is generated.

[0038] Furthermore, the calculation of the dual gravitational force, total repulsive force, and unit motion direction of the UAV in the current iteration includes:

[0039] Based on the current iteration's guiding node, the UAV's location information, and its endpoint, combined with the gravity coefficient, the UAV's dual gravity is calculated.

[0040] Based on the airport integrated operation environment model, risk source priorities are set, and based on the repulsion coefficient of manned aircraft routes, the repulsion of different risk sources under the current iteration is calculated through the repulsion potential function and negative gradient.

[0041] Define the moment of encounter between the UAV and the manned aircraft, and combine the velocity vectors of the UAV and the manned aircraft with the repulsive force factor between the computers;

[0042] Based on the inter-machine repulsion factor, the repulsion coefficient of the manned aircraft route is updated and combined with the repulsion from different risk sources to calculate the total repulsion of the UAV;

[0043] Calculate the resultant force of the drone based on its dual gravitational and total repulsive forces; then calculate the unit direction of motion based on the resultant force.

[0044] In the above process, firstly, this method generates a complete process of "guiding constraints to determine direction—dual gravity to stabilize the overall situation—tiered repulsion to avoid risks—inter-machine factors to adapt to dynamic conflicts—smooth verification to ensure execution" through core features such as dual gravity, risk source priority hierarchical repulsion, and dynamic adjustment of inter-machine repulsion factors. This solves the technical bottleneck of traditional artificial potential field algorithms that rely solely on the gravity of the target point and are prone to getting trapped in local minima, ensuring that local optimization always fits the global safety framework. Secondly, based on the repulsion coefficient of manned aircraft flight paths, risk source repulsion divided by priority, and inter-machine repulsion factors that integrate relative speed and encounter time, it breaks through the limitations of traditional repulsion that relies solely on distance and is slow to respond to dynamic conflicts between manned and aircraft, significantly improving the obstacle avoidance accuracy in high closed-loop speed and short-time encounter scenarios. Thirdly, path smoothing and executability verification make up for the defects of some algorithms that generate paths with sharp corners and cannot be directly adapted to UAV flight control systems. Finally, the "global optimal framework" is transformed into a "locally executable path", forming a closed loop of "environment modeling - risk quantification - global planning - local optimization", which solves the technical problems of blindness in local obstacle avoidance, insufficient dynamic conflict avoidance, and poor path practicality in airport integrated operation scenarios.

[0045] Furthermore, the generation of the initial optimized path for global virtual navigation includes:

[0046] Determine whether the drone collision risk of the updated location information exceeds the threshold; if so, remove the updated location information, update the step size or repulsion weight, and recalculate the repulsion force; otherwise, retain the updated location information as the drone's location information for the next iteration until the number of iterations is met, and generate the initial optimized path for global virtual navigation.

[0047] Furthermore, the verification of path smoothing and executability of the initial optimized path for global virtual navigation includes:

[0048] The initial optimized path of the global virtual navigation is deredundant and key points are retained to generate the initial optimized path.

[0049] A smooth path is generated by fitting spline curves to the discrete points in the initial optimized path.

[0050] Dense sampling is performed on the smooth path, and the minimum safe distance and the risk of drone collision are checked point by point. The sampling points are then filtered in combination with the curvature / turning radius constraints to generate a global virtual navigation optimization path.

[0051] In the above process, this method uses a collision risk threshold judgment mechanism, combined with dynamic adjustment of step size or repulsion weight, and collision risk data to eliminate high-risk points that exceed the safety threshold, thus preventing the local optimized path from falling into dangerous areas. This solves the problem that traditional artificial potential field algorithms are prone to path safety hazards due to delayed risk judgment when avoiding obstacles locally. At the same time, by removing redundancy, retaining key points, and fitting spline curves, redundant nodes and sharp corners in the initial optimized path are eliminated. Then, through dense sampling and point-by-point verification of minimum safe distance and other constraints, the method makes up for the shortcomings of some algorithms in generating paths with insufficient smoothness and inability to be directly adapted to UAV flight control systems. This solves the problem of hidden safety vulnerabilities or non-compliance with executability standards after path optimization.

[0052] A UAV path planning system based on an improved artificial potential field algorithm includes:

[0053] The data acquisition module is used to collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and to construct an airport integrated operation environment model through a uniform grid partitioning algorithm;

[0054] The risk calculation module is used to calculate the drone collision risk at each grid point based on the airport integrated operation environment model and the Reich model.

[0055] The path generation module is used to set the start and end points of the drones and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm.

[0056] The path optimization module is used to optimize the global virtual navigation path based on the collision risk of each UAV and through an improved artificial potential field algorithm, thereby generating a global virtual navigation optimized path.

[0057] In the aforementioned process, this system centers on "end-to-end collaboration and multi-objective fusion," with four modules forming a closed loop of "modeling-risk quantification-global planning-local optimization." Data is interconnected and logic is integrated, breaking through the limitations of traditional isolated modules. It coordinates static constraints and dynamic variables to achieve a unified approach to multiple objectives: safety and compliance, high efficiency and scalability, engineering practicality, and dynamic adaptation. It can respond in real-time to changes in airport airspace, and the generated paths can be directly imported into the flight control system. This precisely addresses the core challenge of balancing "obstacle avoidance, risk, and real-time performance" in integrated airport operations, demonstrating the comprehensive advantages of an integrated solution. Attached Figure Description

[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0060] Figure 2 This is a system structure diagram of Embodiment 1 of the present invention;

[0061] Figure 3 This is a schematic diagram of the global virtual navigation path in Embodiment 2 of the present invention;

[0062] Figure 4 This is a comparison diagram of the path between the method of this invention and the traditional artificial potential field algorithm in Embodiment 2 of the present invention;

[0063] Figure 5 This is a comparison chart of the path risk assessment effects generated by the method of this invention and the traditional artificial potential field algorithm in Embodiment 2 of the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0065] Please see Figure 1This embodiment provides a UAV path planning method based on an improved artificial potential field algorithm. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0066] A method for UAV path planning based on an improved artificial potential field algorithm, comprising:

[0067] S1. Collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and construct an airport integrated operation environment model through a uniform grid partitioning algorithm;

[0068] Airport geographic data refers to the airport's geographic boundaries. In this embodiment, the airport's geographic boundaries are X∈[50, 950]m and Y∈[50, 950]m. Fixed obstacle parameters include information such as the location, size, and type of the terminal / radar station. For example, the dimensions of a rectangular terminal are (500m, 300m), and its semi-major axis is 50m. The no-fly zone includes the main runway, apron, etc. Manned aircraft route information includes origin and destination coordinates, flight frequency (e.g., 0.8 flights / hour for takeoff routes), and flight direction, which are updated in real-time at 1-minute intervals via the airport air traffic control interface.

[0069] The airspace is divided into 1m×1m grids, a unified coordinate system is defined, and the airport center (500m, 500m) is used as the reference origin. The coordinate constraints of the UAV's starting point (100m, 100m) and ending point (900m, 900m) are defined, and a three-dimensional model with coordinate information and environmental information is constructed, namely the airport integrated operation environment model.

[0070] S2. Based on the airport integrated operation environment model, the collision risk of drones at each grid point is calculated using the Reich model; the drone collision risk includes distance risk, route risk, and no-fly risk.

[0071] S2 includes:

[0072] S2-1. Based on the coordinate information of fixed obstacles and the position information of UAVs at each grid point in the airport integrated operation environment model, the distance risk between the UAV and the fixed obstacles is calculated using the Reich model. For example, let the position information of the UAV be... , No. The area occupied by each fixed obstacle is (It can be of any shape). The geometric center of the obstacle is taken as... And its equivalent radius is defined as the radius of the circumcircle. :

[0073] ;

[0074] The minimum distance from the drone to the obstacle. Defined as:

[0075] ;

[0076] The distance risk of the obstacle The corresponding formula is:

[0077] ;

[0078] Distance risk of this grid point Take the maximum value of the distance risk from all fixed obstacles, that is:

[0079] ;

[0080] in, This represents an exponential function with base e. Indicates the occupied area is Spatial points within (or on the boundary of) the geometric center are used to find points on the boundary of the geometric center. The farthest point; Represents the geometric center With the occupied area as Inner (or boundary) distance from geometric center The distance between the farthest points, This represents the smallest positive number that prevents the denominator from being zero.

[0081] For example, when the fixed obstacle is a circular obstacle, then... The formula corresponding to its minimum distance is:

[0082] ;

[0083] in, Represents the distance function (distance between points). Represents the maximum value function. Indicates the radius of the circular obstacle. , These represent the two-dimensional coordinates of the UAV.

[0084] S2-2. Based on the route information in the airport integrated operation environment model, calculate the route risk of each grid point using the Reich model. Assume the human-machine route is... Minimum distance from the drone to the flight path for:

[0085] ;

[0086] in, For the route Any point in it.

[0087] Route risk and flight frequency Regarding this, we first need to truncate and normalize the flight frequencies: , This is a truncation function.

[0088] When a drone approaches within 100 meters of the flight path, the risk decreases with distance; this is defined as flight path risk. for:

[0089] ;

[0090] Among them, distance The value is 100m. When a drone intrudes into the flight path... It is 0 at this time. .

[0091] S2-3. Based on the no-fly zone range and UAV location information in the airport integrated operation environment model, set the no-fly risk for each grid point. When a UAV is within the no-fly zone... At that time, the risk of no-fly zone Direct assignment, i.e. .

[0092] S2-4. Based on the distance risk, flight path risk, and no-fly zone risk of each grid point, calculate the UAV collision risk of each grid point using a weighted summation minimum value function. The corresponding formula is:

[0093] ;

[0094] in, Describes the minimum value function. , , These represent the distance risk weight, route risk weight, and no-fly risk weight, respectively. The sum of the distance risk weight, route risk weight, and no-fly risk weight is 1.

[0095] Drone collision risk Within the range of [0, 1], when the risk of drone collision... If the risk exceeds the safety threshold (0.1), the grid point is considered to have exceeded the risk limit.

[0096] S3. Set the starting point and ending point of the drones, and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm;

[0097] Traditional artificial potential field algorithms rely entirely on the instantaneous force field calculation of the UAV's current position. When the gravitational force at the target point and the repulsive force of surrounding obstacles reach equilibrium at a non-target point, the resultant force becomes zero, causing the UAV to stagnate or oscillate repeatedly, ultimately failing to reach its destination and resulting in planning failure. Therefore, directly applying traditional artificial potential field algorithms to path planning in complex environments easily leads to getting trapped in local minima. To address this fundamental problem, this embodiment introduces an Adaptive Particle Swarm Optimization (APSO) algorithm as the front-end global planner. APSO simulates swarm intelligence to search for a virtual navigation path from the starting point to the destination with the lowest overall risk in the global solution space. This path is essentially a "safe corridor" framework composed of key waypoints, independent of local force fields, thus completely avoiding the trap of local minima. This assists the subsequent improved artificial potential field method in performing refined local planning under the guidance of this global framework.

[0098] S3 includes:

[0099] S3-1. Based on the start and end points of the UAV, generate multiple candidate paths; in this embodiment, the number of candidate paths is 30, and the corresponding dimension is the number of intermediate nodes × 2.

[0100] S3-2. Treat each candidate path as a particle, initialize the initial position and initial velocity vector of each particle, and set adaptive parameters; the adaptive parameters include inertia weight and group learning factor;

[0101] S3-3. Based on the airport fusion operation environment model and the UAV collision risk of each grid point, the fitness of each particle is calculated through the fitness function;

[0102] S3-3 includes:

[0103] S3-3-1. Perform dense sampling on each candidate path, and generate the total collision risk value of all sampled points in different particles based on the UAV collision risk of each grid point.

[0104] S3-3-2. Based on the position information of each particle and the no-fly zone range, generate compliance penalty items; when the path intrudes into the no-fly zone or manned aircraft flight path, the value of the compliance penalty item is 1000; otherwise, the value of the compliance penalty item is 0.

[0105] S3-3-3. Calculate the path length of each particle based on its position information;

[0106] S3-3-4. Calculate the straight-line flight distance of the drone based on its starting and ending points;

[0107] S3-3-5. Based on the straight-line distance of flight, the total collision risk value of each particle, and the path length, the fitness of each particle is generated through a fitness function.

[0108] Taking any particle as an example, calculate the ratio of the particle's path length to its straight-line distance, and use this as the normalized value of the path length; then, according to the formula:

[0109] ;

[0110] Calculate fitness .in, This represents the total collision risk value. This indicates compliance penalties. This represents the normalized value of the path length.

[0111] S3-4. Based on each fitness and the historical best fitness, update the individual's historical best position and the global best position;

[0112] S3-5. Update the adaptive parameters using a linear decay strategy, and dynamically adjust the population learning factor based on the maximum value of each fitness to obtain the updated adaptive parameters and the updated population learning factor.

[0113] The formula corresponding to the linear decay strategy is:

[0114] ;

[0115] in, This represents the updated adaptive parameters. , These represent the maximum and minimum values ​​of the adaptive parameters, respectively. Indicates the current iteration number. This indicates the maximum number of iterations. In this embodiment, , The values ​​are 0.8 and 0.3, respectively.

[0116] In this embodiment, the group learning factor is adjusted within the range of [1.5, 2.5]. In each iteration, the change in the current global optimal fitness compared to the optimal value in the previous several iterations is calculated. If this change is less than 1 × 10 for three consecutive iterations... -5 If the optimal fitness is not significantly improved, the group learning factor is increased by 0.1 to enhance the particle's tendency to learn from the globally optimal individual, thereby improving the population's ability to explore the world beyond local optima. Conversely, if the optimal fitness is still significantly improving, the group learning factor is decreased by 0.1 to appropriately reduce the weight of social learning, which is conducive to the particle's more refined local development.

[0117] S3-6. Based on the updated individual historical best position, the updated global best position, the updated adaptive parameters, and the updated swarm learning factor, update the velocity vector and position of each particle to obtain the global best position for the current iteration. In this embodiment, the update methods involved in S3-5 and S3-6 are the same as those in existing adaptive particle swarm optimization algorithms.

[0118] S3-7. Determine if the current iteration meets the iteration condition; if so, use the global best position of the current iteration as the global virtual navigation path; otherwise, re-enter the iteration and update the iteration count. The iteration condition is that the number of iterations reaches 50.

[0119] In addition, ensure that the risk of drone collision at each node of the generated global virtual navigation path is less than the safety threshold.

[0120] S4. Based on the collision risk of each UAV, the global virtual navigation path is optimized by an improved artificial potential field algorithm to generate a global virtual navigation optimized path.

[0121] VNP nodes are extracted from the global virtual navigation path, integrated into a node sequence, and used as global framework constraints. This embodiment uses a combined force of "dual gravitational field + risk-driven hierarchical repulsive field" to iteratively generate discrete waypoints, and performs smoothing and executability constraint checks on the discrete waypoints, outputting an optimized global virtual navigation path. The expression for the node sequence is: , This represents the total number of VNP nodes in the global virtual navigation path. , , These represent the first, second, and last VNP nodes in the global virtual navigation path, respectively.

[0122] The improved artificial potential field algorithm uses a global virtual navigation path as a guiding framework to construct a dual gravitational field (the resultant gravitational force is a weighted synthesis of the gravitational force of the target point and the gravitational force of the virtual navigation path nodes) and a risk-driven hierarchical repulsive field. Through iterative generation of the potential field force, an executable smooth trajectory that satisfies safety constraints is generated. Therefore, the processing steps of the improved artificial potential field algorithm include:

[0123] S4-1. Based on the guiding constraints, set the guiding node, gravity coefficient, and basic repulsion coefficient for the current iteration; based on the route risk and the basic repulsion coefficient, calculate the manned aircraft route repulsion coefficient for the current iteration.

[0124] The initial setup for the bootstrap node is the first VNP node in the global virtual navigation path, with the corresponding node index being: If the current iteration's guiding node does not satisfy the guiding constraints, i.e., it does not meet the following:

[0125] ;

[0126] Then choose the first one. VNP nodes As a bootstrap node. This indicates the location information of the drone in the current iteration. Indicates the first There are several VNP nodes. Before each iteration, it is necessary to determine whether the guiding node of the current iteration satisfies the guiding constraints, ensuring that the local search always progresses along the global direction given by the VNP, until... Or a drone will reach the destination.

[0127] To avoid unnecessary impact on long-distance flight paths, the preferred trigger condition is set: when the distance between the UAV and the manned aircraft flight path is less than 150 meters, the flight path repulsion calculation is initiated, i.e., the manned aircraft flight path repulsion coefficient. The corresponding formula is:

[0128] ;

[0129] in, This indicates the location information of the drone in the current iteration. Indicates the basic repulsion coefficient. , Both represent risk amplification parameters. This indicates route risk. >0, ≥1.

[0130] If the distance between the UAV and the manned aircraft's flight path is not less than 150 meters, then the repulsion coefficient of the manned aircraft's flight path is... It is 0.

[0131] S4-2. Based on the UAV's starting point, ending point, and guiding node, perform path simulation for the UAV. Combine the gravity coefficient and the manned aircraft's repulsion coefficient to calculate the UAV's dual gravity, total repulsion, and unit motion direction under the current iteration.

[0132] Specifically, S4-2 includes:

[0133] S4-2-1. Based on the current iteration's guiding node, the UAV's position information, and the endpoint, and combined with the gravity coefficient, calculate the UAV's dual gravity.

[0134] Calculating only a single gravitational force exacerbates the problem of traditional artificial potential field algorithms getting trapped in local minima. A dual gravitational field, however, adds a gravitational force from a guiding node on the virtual path, effectively setting multiple "midway traction points." When the gravitational force at the target point "fails" due to obstacles, the gravitational force from the guiding node provides a pull force in another direction, breaking the original force balance. Simultaneously, the guiding node dynamically switches as the drone moves forward, forming a continuous forward traction that ensures the drone always travels along a globally safe corridor, effectively avoiding being trapped in a local location. Therefore, this embodiment introduces a dual gravitational field into the improved artificial potential field algorithm and calculates the dual gravitational force on the drone under this field.

[0135] The formula corresponding to S4-2-1 is:

[0136] ;

[0137] ;

[0138] in, This represents the dual gravity in the current iteration. This represents the gravitational coefficient (valued at 5.0 in this embodiment). , These represent the first gravitational weight and the second gravitational weight, respectively. Indicates the destination of the drone. Represents the first gravitational force. This represents the second gravitational force. It increases as the drone moves away from the endpoint. To align with the VNP global framework, improve the drone's performance as it approaches the finish line. This is to ensure convergence to the target point.

[0139] S4-2-2. Based on the airport integrated operation environment model, risk source priorities are set, and based on the repulsion coefficient of manned aircraft routes, the repulsion of different risk sources under the current iteration is calculated through the repulsion potential function and negative gradient.

[0140] Specifically, because traditional APF (Advanced Persistent Force Field) methods determine repulsive force intensity solely by distance, the magnitude of the repulsive force cannot accurately reflect the actual level of danger. This can lead to insufficient obstacles causing collisions or excessive obstacle avoidance sacrificing efficiency and feasibility. Therefore, this embodiment employs a risk-driven, graded repulsive field.

[0141] Risk sources are categorized by priority into no-fly zones (highest priority), manned flight paths (medium priority), and fixed obstacles (lowest priority). For any risk source, the minimum safe distance from the drone to that risk source is defined as... The effective distance threshold is .

[0142] when When the risk source does not generate a repulsive force, the corresponding repulsive force is 0; otherwise, it is based on:

[0143] ;

[0144] Calculate the repulsive potential function of the risk source. .

[0145] Based on:

[0146] ;

[0147] Calculate the repulsive force of this risk source. and minimum safe distance . This represents the repulsion coefficient corresponding to the risk source.

[0148] Furthermore, when the source of risk is a point obstacle, it is advisable to... When the risk source is a line segment / polygonal region, the outward normal unit vector corresponding to the nearest boundary point can be taken. Indicates the first The location of each risk source Indicates the negative gradient. Indicates the minimum safe distance. This indicates the direction of the external law unit away from the source of risk.

[0149] S4-2-3. Define the moment of encounter between the UAV and the manned aircraft, and combine the velocity vectors of the UAV and the manned aircraft with the repulsive force factor between the computers;

[0150] It should be explained that, in order to enhance the avoidance sensitivity to conflict scenarios of "high closing velocity and short encounter time", an inter-machine repulsion factor is added to the manned-machine related repulsion force. Let the velocity vector of the UAV be... The velocity vector and position information of the manned aircraft are respectively , The relative speed of drones and manned aircraft and relative position They are respectively:

[0151] ;

[0152] ;

[0153] Define the moment of encounter between drones and manned aircraft. for:

[0154] ;

[0155] in, Indicates the moment of encounter between the drone and the manned aircraft. The upper limit of the predicted time window, This means truncating the value within the interval. Inside.

[0156] Inter-machine repulsion factor The corresponding formula is:

[0157] ;

[0158] in, , These represent the speed weight and time weight, respectively (both of which are greater than 0). Indicates speed reference value, Indicates a time reference value. This indicates protection against small quantities.

[0159] S4-2-4. Based on the inter-aircraft repulsion factor, update the repulsion coefficient of the manned aircraft route and combine it with the repulsion of different risk sources to calculate the total repulsion of the UAV.

[0160] According to the formula:

[0161] ;

[0162] Update the manned aircraft flight path repulsion coefficient to obtain the updated manned aircraft flight path repulsion coefficient. .

[0163] When a drone is close to a manned aircraft / fixed obstacle / no-fly zone, the greater the relative speed or the closer the encounter, the stronger the repulsive force, thus increasing the collision avoidance safety margin. The repulsive force on the manned aircraft / fixed obstacle / no-fly zone is calculated using the same method as in S4-2-2; and the various repulsive forces are superimposed according to the priority of "no-fly zone > manned aircraft flight path > fixed obstacle" to obtain the total repulsive force. ,Right now:

[0164] ;

[0165] in, This indicates the repulsive force corresponding to manned and machine-generated objects. This represents the repulsive force corresponding to a fixed obstacle. This represents the repulsive force corresponding to the no-fly zone.

[0166] S4-2-5. Based on the dual attraction and total repulsion of the drone, according to the formula:

[0167] ;

[0168] Calculate the resultant force of the drone ;

[0169] Based on the resultant force, according to the formula:

[0170] ;

[0171] Calculate the direction of unit motion If it appears If the value is less than the preset threshold, it indicates that the current drone may be in a local minimum or oscillation. It is preferable to switch to the next VNP guiding node, adjust the weight, or shorten the step size to get out of trouble, so as to keep the path advancing along the VNP global framework until the target point is reached.

[0172] S4-3. Based on the unit motion direction of the current iteration, update the drone's position information and calculate the corresponding drone collision risk;

[0173] The formula for updating the drone's location information is:

[0174] ;

[0175] in, This indicates the updated location information of the drone. This represents the step size (which can be a constant or adaptive to the minimum risk distance).

[0176] Determine whether the drone collision risk of the updated location information exceeds the threshold; if so, remove the updated location information, update the step size or repulsion weight, and recalculate the repulsion force; otherwise, retain the updated location information as the drone's location information for the next iteration until the number of iterations is met, and generate the initial optimized path for global virtual navigation.

[0177] S4-5. Based on the updated location information, the VNP nodes are screened and iterated until the drone reaches the destination, generating the initial optimized path for global virtual navigation.

[0178] S4-6. Verify the smoothness and executability of the initial optimized path for global virtual navigation, and generate the optimized path for global virtual navigation.

[0179] S4-6 includes:

[0180] S4-6-1. Perform redundancy removal and key point retention operations on the initial optimized path of the global virtual navigation to generate the initial optimized path;

[0181] S4-6-2. Fit the discrete points in the initial optimized path with spline curves to generate a smooth path;

[0182] S4-6-3. Perform dense sampling on the smooth path, check the minimum safe distance and the risk of drone collision point by point, and filter the sampling points in combination with the curvature / turning radius constraints to generate a global virtual navigation optimization path.

[0183] In summary, this invention accurately constructs an airport integrated operation environment model through uniform grid partitioning, and comprehensively quantifies distance risk, route risk, and no-fly zone risk using the Reich model, achieving refined assessment of collision risk. This solves the problems of one-sided risk assessment and coarse environmental description in traditional path planning. Furthermore, it utilizes an adaptive particle swarm optimization algorithm to generate a global virtual navigation path, providing global guidance for the improved artificial potential field algorithm. This effectively avoids the technical bottleneck of traditional artificial potential field algorithms easily getting trapped in local minima. Simultaneously, it performs local optimization of the global virtual navigation path with collision risk as the core, ensuring that the path meets the stringent safety constraints of airports while also considering real-time responsiveness in dynamic environments. This solves the core challenge of simultaneously achieving "obstacle avoidance reliability, risk controllability, and real-time adaptability" in path planning within the integrated operation scenario of airport airspace UAVs and transport aviation. It also improves the smoothness and executability of the path, adapting to the actual application needs of UAV flight control systems, and further alleviates the problem of path correction lag in dynamic environments (such as sudden changes in manned aircraft flight path frequency or the addition of obstacles).

[0184] like Figure 2 As shown, a UAV path planning system based on an improved artificial potential field algorithm includes:

[0185] The data acquisition module is used to collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and to construct an airport integrated operation environment model through a uniform grid partitioning algorithm;

[0186] The risk calculation module is used to calculate the drone collision risk at each grid point based on the airport integrated operation environment model and the Reich model.

[0187] The path generation module is used to set the start and end points of the drones and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm.

[0188] The path optimization module is used to optimize the global virtual navigation path based on the collision risk of each UAV and through an improved artificial potential field algorithm, thereby generating a global virtual navigation optimized path.

[0189] A UAV path planning system based on an improved artificial potential field algorithm also includes a real-time path correction module for real-time correction of the generated global virtual navigation optimized path. When the manned aircraft route information or fixed obstacle parameters in the airport fusion operation environment model change, steps 2-4 are re-executed, and the path correction response time does not exceed 5 seconds, with the corrected path having an overlap of no less than 80% with the original path.

[0190] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0191] Example 2:

[0192] This embodiment is based on Embodiment 1.

[0193] A global virtual navigation path is generated using an adaptive particle swarm optimization algorithm, such as... Figure 3 As shown. The current intermediate nodes are (250, 220), (380, 350), (520, 480), (650, 610), (800, 780), then the global virtual navigation path is: starting point → (250, 220) → (380, 350) → (520, 480) → (650, 610) → (800, 780) → ending point.

[0194] The improved artificial potential field algorithm from Example 1 and the traditional artificial potential field algorithm are then compared experimentally. Figure 4 As shown, traditional artificial potential field algorithms cannot assess the risks of a drone's path when entering the takeoff route. The improved artificial potential field algorithm produces a smoother and more direct path, avoiding various no-fly zones, high-risk areas, and obstacles while shortening the travel distance and improving the efficiency and safety of path planning. Figure 5 As shown, the improved artificial potential field algorithm can calculate the actual risk value, while the path generated by the traditional artificial potential field algorithm will enter the high-risk area due to the underestimation of the risk value.

[0195] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0196] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A UAV path planning method based on an improved artificial potential field algorithm, characterized in that, include: Collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and construct an airport integrated operation environment model through a uniform grid partitioning algorithm; Based on the airport integrated operation environment model, the collision risk of drones at each grid point is calculated using the Reich model; the drone collision risk includes distance risk, route risk, and no-fly risk. Set the start and end points of the drones, and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm; Based on the collision risk of each UAV, the global virtual navigation path is optimized by an improved artificial potential field algorithm to generate an optimized global virtual navigation path. The calculation of drone collision risk at each grid point includes: Based on the coordinate information of fixed obstacles and the position information of UAVs at each grid point in the airport integrated operation environment model, the distance risk between the UAV and the fixed obstacles is calculated by the Reich model. Based on the route information in the airport integrated operation environment model, the route risk of each grid point is calculated using the Reich model. Based on the no-fly zone range and drone location information in the airport integrated operation environment model, the no-fly risk of each grid point is set; Based on the distance risk, flight route risk, and no-fly risk of each grid point, the drone collision risk of each grid point is calculated by a weighted summation minimum value function. The processing steps of the improved artificial potential field algorithm include: Based on the guiding constraints, set the guiding node, gravity coefficient, and basic repulsion coefficient for the current iteration; based on the route risk and the basic repulsion coefficient, calculate the manned aircraft route repulsion coefficient for the current iteration. Based on the drone's starting point, ending point, and guiding node, the drone's path is simulated. Combining the gravity coefficient and the manned aircraft's repulsion coefficient, the drone's dual gravity, total repulsion, and unit motion direction are calculated in the current iteration. Based on the current unit motion direction, update the drone's position information and calculate the corresponding drone collision risk; Based on the updated location information, the collision risk of the drone is screened and iterated until the drone reaches the destination, generating an initial optimized path for global virtual navigation. The initial optimized path for global virtual navigation is verified for path smoothness and executability, and a global virtual navigation optimized path is generated.

2. The UAV path planning method based on an improved artificial potential field algorithm according to claim 1, characterized in that, The generation of the global virtual navigation path includes: Based on the drone's origin and destination, multiple candidate paths are generated; Each candidate path is treated as a particle, and the initial position and initial velocity vector of each particle are initialized, and adaptive parameters are set; the adaptive parameters include inertia weight and group learning factor; Based on the airport integrated operation environment model and the UAV collision risk of each grid point, the fitness of each particle is calculated through a fitness function; Based on each fitness level and the historical best fitness, update the individual's historical best position and the global best position; The adaptive parameters are updated by using a linear decay strategy, and the population learning factor is dynamically adjusted based on the maximum value of each fitness, resulting in the updated adaptive parameters and the updated population learning factor. Based on the updated individual historical best position, the updated global best position, the updated adaptive parameters, and the updated group learning factor, the velocity vector and position of each particle are updated to obtain the global best position of the current iteration. Determine if the current iteration meets the iteration conditions; if so, use the current iteration's global best position as the global virtual navigation path; otherwise, re-enter the iteration and update the iteration count.

3. The UAV path planning method based on an improved artificial potential field algorithm according to claim 2, characterized in that, The calculation of the fitness of each particle includes: For each candidate path, dense sampling is performed, and based on the drone collision risk of each grid point, the total collision risk value of all sampled points in different particles is generated; Based on the position information of each particle and the scope of the no-fly zone, compliance penalty items are generated; Calculate the path length of each particle based on its position information; Calculate the straight-line distance of the drone's flight based on its starting and ending points; The fitness of each particle is generated using a fitness function based on the straight-line distance of flight, the total collision risk value of each particle, and the path length.

4. The UAV path planning method based on an improved artificial potential field algorithm according to claim 1, characterized in that, The calculation of the dual gravitational force, total repulsive force, and unit motion direction of the UAV in the current iteration includes: Based on the current iteration's guiding node, the UAV's location information, and its endpoint, combined with the gravity coefficient, the UAV's dual gravity is calculated. Based on the airport integrated operation environment model, risk source priorities are set, and based on the repulsion coefficient of manned aircraft routes, the repulsion of different risk sources under the current iteration is calculated through the repulsion potential function and negative gradient. Define the moment of encounter between the UAV and the manned aircraft, and combine the velocity vectors of the UAV and the manned aircraft with the repulsive force factor between the computers; Based on the inter-machine repulsion factor, the repulsion coefficient of the manned aircraft route is updated and combined with the repulsion from different risk sources to calculate the total repulsion of the UAV; Calculate the resultant force of the drone based on its dual gravitational and total repulsive forces; then calculate the unit direction of motion based on the resultant force.

5. The UAV path planning method based on an improved artificial potential field algorithm according to claim 1, characterized in that, The generation of the initial optimized path for global virtual navigation includes: Determine whether the drone collision risk of the updated location information exceeds the threshold; if so, remove the updated location information, update the step size or repulsion weight, and recalculate the repulsion force; otherwise, retain the updated location information as the drone's location information for the next iteration until the number of iterations is met, and generate the initial optimized path for global virtual navigation.

6. The UAV path planning method based on an improved artificial potential field algorithm according to claim 1, characterized in that, The verification of path smoothing and executability of the initial optimized path for global virtual navigation includes: The initial optimized path of the global virtual navigation is deredundant and key points are retained to generate the initial optimized path. A smooth path is generated by fitting spline curves to the discrete points in the initial optimized path. Dense sampling is performed on the smooth path, and the minimum safe distance and the risk of drone collision are checked point by point. The sampling points are then filtered in combination with the curvature / turning radius constraints to generate a global virtual navigation optimization path.

7. A UAV path planning system based on an improved artificial potential field algorithm, used to implement the UAV path planning method based on an improved artificial potential field algorithm as described in any one of claims 1 to 6, characterized in that, include: The data acquisition module is used to collect airport geographic data, fixed obstacle parameters, manned aircraft route information and no-fly zone range, and to construct an airport integrated operation environment model through a uniform grid partitioning algorithm; The risk calculation module is used to calculate the drone collision risk at each grid point based on the airport integrated operation environment model and the Reich model. The path generation module is used to set the start and end points of the drones and generate a global virtual navigation path based on the collision risk of each drone using an adaptive particle swarm optimization algorithm. The path optimization module is used to optimize the global virtual navigation path based on the collision risk of each UAV and through an improved artificial potential field algorithm, thereby generating a global virtual navigation optimized path.

Images