Unmanned aerial vehicle three-dimensional path planning method based on neuron chaos wolf pack algorithm

Abstract

The invention relates to the technical field of unmanned aerial vehicle route planning, in particular to an unmanned aerial vehicle three-dimensional route planning method based on a neuron chaos wolf pack algorithm. Constructing a discrete neuron hyper-chaos mapping system, improving a wolf pack algorithm based on the discrete neuron hyper-chaos mapping system in combination with a neuron attention mechanism, and determining a neuron hyper-chaos wolf pack algorithm; initializing individual positions in the wolf pack, correspondingly updating the individual positions of the wolf pack according to a neuron hyper-chaos wolf pack algorithm, and judging whether the updated individual positions of the wolf pack meet a stopping criterion of the neuron hyper-chaos wolf pack algorithm or not through a constraint cost function until the individual positions of the wolf pack are updated; determining optimal flight path points of the unmanned aerial vehicle to form an optimal solution sequence, and generating an optimal flight path of the unmanned aerial vehicle according to the optimal solution sequence; the algorithm is effectively prevented from falling into a local optimal solution, the global optimal solution search precision of the algorithm is improved, and the optimal flight path of the unmanned aerial vehicle is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV path planning, and particularly relates to a three-dimensional path planning method for UAVs based on a neuron chaotic wolf pack algorithm. Background Art

[0002] UAVs have unique characteristics such as being lightweight, flexible, and efficient, and are currently widely used in many fields. For example, high-altitude equipment inspection, aerial photography, disaster prevention and mitigation, crop protection, electronic reconnaissance, etc. With the development of UAV technology, its application scenarios will gradually expand from outdoor open environments to urban spaces. The extension and expansion of application scenarios have led to a more complex working environment for UAVs, and the types and densities of obstacles that need to be avoided will also increase sharply accordingly.

[0003] Common path planning algorithms in the prior art include graph theory methods or methods based on swarm intelligence and bionics. Among them, graph theory methods usually include Dijkstra algorithm, A* algorithm, and Floyd algorithm, etc. Although this method is easy to implement and can be used in directed graphs and negative weight networks, its computational complexity and space complexity are relatively high, and it is not applicable to large-scale scenarios, with relatively limited scenario usage; methods based on swarm intelligence and bionics usually include particle swarm algorithm, genetic algorithm, and various bionic population algorithms, etc. This algorithm can obtain the best flight path of the UAV, but during the search process, it often falls into local optimal solutions, resulting in limited search scope and inability to accurately search for the best flight route. Summary of the Invention

[0004] In order to solve the technical problems that the existing graph theory algorithms have high computational complexity and space complexity and are not applicable to large-scale scenarios, and the swarm intelligence and bionics methods have limited search and low accuracy, the purpose of the present invention is to provide a three-dimensional path planning method for UAVs based on a neuron chaotic wolf pack algorithm, and the specific technical solutions adopted are as follows:

[0005] Collect the waypoints of the UAV, obtain the flight path of the UAV according to the waypoints, and establish a constraint cost function;

[0006] Construct a discrete neuron hyperchaotic mapping system, and based on the discrete neuron hyperchaotic mapping system, combine the neuron attention mechanism to improve the wolf pack algorithm, and determine the neuron hyperchaotic wolf pack algorithm;

[0007] Initialize the individual positions in the wolf pack, update the individual positions of the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and judge whether the updated individual positions of the wolf pack satisfy the stop criterion of the neuron hyperchaotic wolf pack algorithm through the constraint cost function. If not, continue to iterate for position correction until the update of the individual positions of the wolf pack is completed;

[0008] Determine the optimal flight path points of the UAV by completing the updated positions of the wolf pack individuals to form an optimal solution sequence, and generate the best flight path of the UAV according to the optimal solution sequence.

[0009] Preferably, collect the flight path points of the UAV, obtain the flight path of the UAV according to the flight path points, and establish a constraint cost function, including:

[0010] Collect the flight path points of the UAV, and construct a sequence set {x i , i = 1, …, n}, where x i represents the i-th flight path point; n represents the total number of flight path points;

[0011] Calculate the path lengths between pairwise flight path points based on the sequence set, and perform screening to determine the path constraint conditions, and calculate the basic terrain constraint conditions and external threat constraint conditions in turn, and comprehensively form the space constraint conditions;

[0012] Establish a constraint cost function according to the path constraint conditions and the space constraint conditions.

[0013] Preferably, calculate the path lengths between pairwise flight path points based on the sequence set, and perform screening to determine the path constraint conditions, and calculate the basic terrain constraint conditions and external threat constraint conditions in turn, and comprehensively form the space constraint conditions, including:

[0014] Calculate the path lengths between pairwise flight path points based on the sequence set, and the corresponding calculation formula is:

[0015]

[0016] where dist represents the path length; x i represents the i-th flight path point; x i+1 represents the adjacent (i + 1)-th flight path point; n represents the total number of flight path points;

[0017] Determine the path constraint conditions, and the corresponding calculation formula is:

[0018] Path shortest = min(dist i )

[0019] where Path shortest represents the shortest path;

[0020] Define the basic terrain as the distribution of independent mountain peaks above the reference sea level, and calculate the height values of the independent mountain peaks, and the corresponding calculation formula is:

[0021]

[0022] where h terrain represents the height value of the independent mountain peak; w1 represents the constraint weight coefficient of the basic terrain; Hterrain Represents the equivalent elevation value of the base terrain; Represents the spatial constraint range of the base terrain; (x1, y1) represents the spatial coordinate position of the base terrain;

[0023] Determine the base terrain constraint conditions, and the corresponding calculation formula is:

[0024]

[0025] Among them, Crash terrain Represents the collision judgment condition of the base terrain;

[0026] Define the external threat as the continuous mountain peak distribution based on the sea level, calculate the height value of the continuous mountain peaks, and the corresponding calculation formula is:

[0027]

[0028] Among them, h threadn Represents the height value of the continuous mountain peaks; w2 represents the constraint weight coefficient of the external threat; T thread Represents the equivalent height of the external threat; Represents the spatial constraint range of the external threat; (x2, y2) represents the spatial coordinate position of the external threat; (x cti , y cti ) represents the associated parameter; h terrain (x2, y2) represents the equivalent base terrain constraint height where the external threat is located;

[0029] Determine the external threat constraint conditions, and the corresponding calculation formula is:

[0030]

[0031] Among them, Crash thread Represents the collision judgment condition of the external threat.

[0032] Preferably, establish a constraint cost function according to the path constraint conditions and spatial constraint conditions, and the corresponding calculation formula is:

[0033] Path best (x i ) = Path shortet (x i ), Crash terrain (x i ) > 0, Crash thread (x i ) > 0

[0034] Among them, Path best Represents the best path.

[0035] Preferably, a discrete neuron hyperchaotic mapping system is constructed. Based on the discrete neuron hyperchaotic mapping system and combined with the neuron attention mechanism, the wolf pack algorithm is improved to determine the neuron hyperchaotic wolf pack algorithm, including:

[0036] Based on the discrete Rulkov mapping model, a quadratic nonlinear term and a sine curve nonlinear term are introduced to obtain the neuron Rulkov mapping;

[0037] The wolf pack algorithm is combined with the bottom-up perception paradigm to determine the neuron hyperchaotic wolf pack algorithm.

[0038] Preferably, the obtained neuron Rulkov mapping has the corresponding calculation formula:

[0039]

[0040] where x(n) represents the membrane potential state variable of the neuron at time n; y(n) represents the internal state variable of the neuron at time n; δ represents the membrane potential change coefficient; and α represents the neuron excitation coefficient.

[0041] Preferably, the individual positions in the wolf pack are initialized. According to the neuron hyperchaotic wolf pack algorithm, the positions of the wolf pack individuals are updated correspondingly. The constraint cost function is used to determine whether the updated positions of the wolf pack individuals meet the stopping criterion of the neuron hyperchaotic wolf pack algorithm. If not, the position correction is continued iteratively until the update of the wolf pack individual positions is completed, including:

[0042] Initialize the individual positions in the wolf pack;

[0043] Define the group roles of the wolf pack as the head wolf, scout wolf, and fierce wolf. Update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm; update the position of the fierce wolf according to the call of the head wolf and the attack on the prey;

[0044] Use the constraint cost function to determine whether the iteration of the neuron hyperchaotic wolf pack algorithm needs to be terminated for the updated positions of the wolf pack individuals. If not, the neuron hyperchaotic wolf pack algorithm continues to iteratively correct the positions of the wolf pack individuals until the update of the wolf pack individual positions is completed.

[0045] Preferably, the formula for initializing the individual positions in the wolf pack is:

[0046] {x i ,i=1,…,dim}=lb i +rand*(ub i -lb i )

[0047] where x i represents the initial position of the individual in the wolf pack, that is, the initial trajectory point of the UAV; dim represents the solution space dimension of the individual in the wolf pack; lbi denotes the minimum value of the i-th dimension of the solution space; ub i denotes the maximum value of the i-th dimension of the solution space.

[0048] Preferably, update the position of the exploring wolves in the wolf pack according to the neuron hyperchaotic wolf pack algorithm; update the position of the attacking wolves according to the leading wolf's summoning and attacking the prey, including:

[0049] Update the position of the exploring wolves in the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and the corresponding calculation formula is:

[0050]

[0051] where denotes the position of the exploring wolf after searching in the p-th direction; x i denotes the current position of the exploring wolf; p denotes the search direction of the exploring wolf; N denotes the total number of search directions; step_a denotes the search step length of the exploring wolf; levy denotes the Levy flight coefficient; α denotes the neuron excitation coefficient; attention bottom_up denotes the bottom_up attention attraction term; NHCM denotes the discrete neuron hyperchaotic mapping system; δ denotes the membrane potential change coefficient; β denotes the neuron excitation value coefficient; y(n) denotes the internal state variable of the neuron at time n;

[0052] After the leading wolf summons, update the position of the attacking wolves, and the corresponding calculation formula is:

[0053]

[0054] where denotes the position of the attacking wolf after rushing towards the m-th direction of the leading wolf; x j denotes the current position of the attacking wolf; step_b denotes the search step length of the attacking wolf; g d denotes the current position of the leading wolf; λ denotes the attention weight coefficient; sin(8π*j) denotes the attention deflection coefficient;

[0055] When the wolf pack attacks the prey, update the position of the attacking wolves, and the corresponding calculation formula is:

[0056]

[0057] where denotes the position of the attacking wolf after attacking towards the q-th direction; x k denotes the current position of the attacking wolf; step_d denotes the attack step length of the attacking wolf; G d denotes the current position of the leading wolf; exp(-10 -2 *k) denotes the attention attenuation coefficient.

[0058] Preferably, by determining the optimal flight path points of the UAV based on the updated positions of the wolf pack individuals to form an optimal solution sequence, and generating the best flight path of the UAV according to the optimal solution sequence, including:

[0059] Screen the updated positions of the wolf pack individuals according to the constraint cost function to determine the optimal flight path points of the UAV, and form an optimal solution sequence;

[0060] Perform continuous interpolation calculation on the optimal solution sequence using cubic spline interpolation to obtain a continuous curve and get the best flight path of the UAV.

[0061] The present invention has the following beneficial effects:

[0062] The discrete neuron hyperchaotic mapping system constructed in this application is based on the traditional Rulkov chaotic mapping and introduces a rhythm constraint, which not only enriches the dynamic characteristics of the Rulkov neuron but also simulates the high-frequency signal characteristics of the neuron; moreover, this system provides more complex and diverse dynamic behaviors for the bionic optimization algorithm, thereby enhancing its global search ability; based on this system and combined with the neuron attention mechanism, the wolf pack algorithm is improved to determine the neuron hyperchaotic wolf pack algorithm, which can effectively avoid the algorithm falling into a local optimal solution, enhance the global search ability of the algorithm, and further improve the global optimal solution search accuracy of the algorithm to obtain the best flight path of the UAV; subsequently, the effectiveness of the neuron hyperchaotic wolf pack algorithm proposed in this application is verified to ensure its potential and practicality for solving complex optimization problems. Description of the Drawings

[0063] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0064] Figure 1 It is a flowchart of the steps of a UAV three-dimensional path planning method based on the neuron chaotic wolf pack algorithm provided by an embodiment of the present invention;

[0065] Figure 2 It is a comparison diagram of the optimal path fitness curves of the sparse scene and the dense scene of a UAV three-dimensional path planning method based on the neuron chaotic wolf pack algorithm provided by an embodiment of the present invention. Detailed Embodiments

[0066] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details a method for three-dimensional path planning of an unmanned aerial vehicle (UAV) based on a neuron chaotic wolf pack algorithm, including its specific implementation manner, structure, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0068] The following specifically describes the specific solution of a method for three-dimensional path planning of an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm provided by the present invention in conjunction with the accompanying drawings.

[0069] Please refer to Figure 1 , which shows a flowchart of the steps of a method for three-dimensional path planning of an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm provided by an embodiment of the present invention. The method includes:

[0070] Step S1: Collect the waypoints of the UAV, obtain the flight path of the UAV according to the waypoints, and establish a constraint cost function;

[0071] Step S2: Construct a discrete neuron hyperchaotic mapping system. Based on the discrete neuron hyperchaotic mapping system, combined with the neuron attention mechanism, improve the wolf pack algorithm to determine the neuron hyperchaotic wolf pack algorithm;

[0072] Step S3: Initialize the individual positions in the wolf pack, update the individual positions of the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and determine whether the updated individual positions of the wolf pack satisfy the stop criterion of the neuron hyperchaotic wolf pack algorithm through the constraint cost function. If not, continue to iterate for position correction until the update of the individual positions of the wolf pack is completed;

[0073] Step S4: Determine the optimal waypoints of the UAV to form an optimal solution sequence through the updated individual positions of the wolf pack, and generate the best flight path of the UAV according to the optimal solution sequence.

[0074] For better illustration, as the application scenarios of UAVs become increasingly diverse, they need to make different optimal path plans according to the variability of complex urban spatial environments or outdoor open environments to assist UAVs in efficiently and safely completing tasks such as aerial photography, geographical mapping, agricultural monitoring, disaster assessment, and express delivery, and overcome the defects of the existing technology. Therefore, a method for three-dimensional path planning of an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm in the first embodiment of the present application is proposed.

[0075] Furthermore, in step S1, it includes:

[0076] Step S11: Collect the waypoints of the drone and construct a sequence set {x i , i = 1, …, n}, where x i represents the i-th waypoint; n represents the total number of waypoints.

[0077] Specifically, use positioning methods such as the Global Positioning System to obtain the waypoints of the drone to ensure the accuracy of the position data, so as to generate a sequence set; according to the specific requirements of the flight mission, connect the waypoints to form the flight path of the drone.

[0078] Step S12: Calculate the path lengths between every two waypoints based on the sequence set, and perform screening to determine the path constraint conditions, and calculate the basic terrain constraint conditions and external threat constraint conditions in turn, and comprehensively form the space constraint conditions.

[0079] It should be noted that the shortest flight path distance of the drone is used as the path constraint condition for trajectory optimization; the space constraint conditions include two types: basic terrain constraint conditions and external threat constraint conditions; it can be understood that when the drone is performing a flight mission, it is usually restricted by various external conditions. In order to ensure that the drone can complete the established mission while ensuring the safety of its own equipment, when making a path plan, it is necessary to consider the space constraint conditions. Among them, the basic terrain constraint refers to the space and geographical environment within the operation area of the drone, such as mountains, rivers, altitude and other constraint conditions; the external threat constraint refers to the anti-aircraft firepower, detection radar and other constraint conditions within the operation area in a complex urban environment, as well as the impact of aerial obstacles existing in the urban environment, such as urban power grids, aerial billboards and other objects on the flight safety of the drone, and the interference caused by high-rise and super high-rise buildings in the city to its trajectory.

[0080] Furthermore, in step S12, it includes:

[0081] Step S121: Calculate the path lengths between every two waypoints based on the sequence set, and the corresponding calculation formula is:

[0082]

[0083] i where dist represents the path length; x i+1 represents the i-th waypoint; x shortest

[0084] represents the adjacent (i + 1)-th waypoint; n represents the total number of waypoints;

[0085] Determine the path constraint conditions, and the corresponding calculation formula is: shortest Path i = min(dist i)

[0086] Among them, Path shortest represents the shortest path.

[0087] Specifically, the path length between waypoints can be obtained through the Euclidean distance, that is, the straight-line distance between two waypoints. Among them, the Euclidean distance is obtained based on the principle of the shortest distance between two points in geometry.

[0088] Step S122: Define the basic terrain as the distribution of independent peaks on the reference sea level, and calculate the height value of the independent peaks. The corresponding calculation formula is:

[0089]

[0090] Among them, h terrain represents the height value of the independent peak; w1 represents the constraint weight coefficient of the basic terrain; H terrain represents the equivalent altitude value of the basic terrain; represents the spatial constraint range of the basic terrain; (x1, y1) represents the spatial coordinate position of the basic terrain;

[0091] Determine the basic terrain constraint condition. The corresponding calculation formula is:

[0092]

[0093] Among them, Crash terrain represents the collision judgment condition of the basic terrain.

[0094] Make an explanation. Converting the basic terrain into the distribution of independent peaks on the reference sea level is conducive to calculating the basic terrain constraint conditions; for basic terrains such as lakes, rivers, and oceans, although their altitude is low, due to their large area and the limited flight range and communication control distance of the drone, in order to ensure the safety of the drone's own equipment, it is necessary to make the drone detour on the basic terrain. Therefore, a relatively high equivalent altitude value H terrain of the basic terrain is set; at the same time, when planning the drone path, to avoid collisions with the basic terrain constraints, the basic terrain constraint condition, that is, Crash terrain , is determined. When Crash terrain >0, it means that the sequence of waypoints formed by the flight path of the drone at this time will not collide with the basic terrain constraint conditions. Otherwise, a collision occurs, and then the flight path is re-optimized to obtain a new sequence of waypoints.

[0095] Step S123: Define the external threat as the continuous peak distribution based on the sea level, and calculate the height value of the continuous peaks. The corresponding calculation formula is:

[0096]

[0097] Among them, h threadn represents the height value of consecutive mountain peaks; w2 represents the constraint weight coefficient of external threats; T thread represents the equivalent height of external threats; represents the spatial constraint range of external threats; (x2, y2) represents the spatial coordinate position of external threats; (x cti , y cti ) represents the correlation parameter; h terrain (x2, y2) represents the equivalent basic terrain constraint height where external threats are located;

[0098] Determine the external threat constraint conditions, and the corresponding calculation formula is:

[0099]

[0100] Among them, Crash thread represents the collision judgment condition of external threats.

[0101] Make an explanation. Since there are many types of external threat constraints, the calculation of the cost function is complex. For the sake of analysis, convert external threats into a continuous mountain peak distribution based on sea level; it can be understood that there may be a correlation relationship between external threat constraints, that is, the mutual dependence at the technical level, the mutual influence at the strategic level, and how to coordinate in actual operation to achieve the best use effect, etc. For example, there is a strong correlation relationship between air defense fire threat and detection radar threat, and there is also a strong correlation relationship between urban power grid and buildings. Therefore, these correlation relationships between external threats must be considered when performing continuous mountain peak conversion.

[0102] It can be explained that T thread represents the equivalent height of external threats. For different types of external threats, the threshold of the equivalent height is also different; specifically, for air defense fire and detection radar threats, T thread is its effective fire strike radius; for urban power grid and aerial billboard threats, T thread is its maximum height value; for urban high-rise and super high-rise building threats, T thread is the height value of its building vertex position; similarly, when Crash thread > 0, it means that the sequence of waypoints formed by the flight path of the UAV at this time will not collide with the external threat constraint conditions. On the contrary, if there is a collision, re-optimize the flight path.

[0103] Step S13: Establish a constraint cost function according to the path constraint conditions and spatial constraint conditions. That is, by comprehensively considering different constraint conditions, to put forward more reasonable screening conditions for subsequent path planning.

[0104] Further, in step S13, a constraint cost function is established according to the path constraint condition and the space constraint condition, and the corresponding calculation formula is:

[0105] Path best (x i ) = Path shortest (x i ), Crash terrain (x i ) > 0, Crash thread (x i ) > 0

[0106] Among them, Path best represents the optimal path.

[0107] It can be understood that the discrete Rulkov mapping model is a discrete-time neuron model based on a nonlinear dynamic system. It can represent the action characteristics of the membrane potential of brain neurons. By adjusting the firing pattern of neurons, it can reflect various types of neuron activity states. Its defined expression is:

[0108]

[0109] Among them, x(n) and y(n) respectively represent the membrane potential and internal state of the neuron at time n; in the discrete two-dimensional Rulkov mapping, the membrane potential state variable x(n) represents the activation state of the neuron; the internal state variable y(n) represents the excitation value of the neuron; δ represents the membrane potential change coefficient; β represents the neuron excitation value coefficient; γ represents the excitation bias.

[0110] It can be explained that in the discrete Rulkov mapping model, the internal state variable y(n) is a linear formula, while neuron excitation is usually a non-linear state variable. Therefore, in order to better fit the action of the neuron membrane potential and the change of the excitation state, an improvement is made on the existing model to construct a new discrete neuron hyperchaotic mapping system to more accurately simulate the action of the neuron membrane potential and the change of neuron excitation, and then an improvement of the wolf pack algorithm is made through the neuron attention mechanism to determine a new algorithm, that is, the neuron hyperchaotic wolf pack algorithm.

[0111] Further, in step S2, it includes:

[0112] Step S21: Based on the discrete Rulkov mapping model, introduce a quadratic non-linear term and a sine curve non-linear term to obtain the neuron Rulkov mapping.

[0113] Further, in step S21, to obtain the neuron Rulkov mapping, the corresponding calculation formula is:

[0114]

[0115] Among them, x(n) represents the membrane potential state variable of the neuron at time n; y(n) represents the internal state variable of the neuron at time n; δ represents the membrane potential change coefficient; and α represents the neuron excitation coefficient.

[0116] It is explained that in the calculation formula of the neuron Rulkov map, the excitation bias γ is set to -1, and the introduced sine function term is used to simulate the periodic discharge phenomenon of the neuron; the dynamic characteristics of the neuron Rulkov map are analyzed to make it have hyperchaotic characteristics, that is, to make the constructed discrete neuron hyperchaotic map system (NHCM, Neural hyper-chaotic map) diverge exponentially in different directions, enhancing the randomness and unpredictability of path planning to improve the concealment and safety of the UAV during mission execution.

[0117] Step S22: Combine the wolf pack algorithm with the bottom-up perception paradigm to determine the neuron hyperchaotic wolf pack algorithm.

[0118] It can be understood that the wolf pack algorithm is proposed based on simulating the hunting strategy of wolf packs and is used to solve optimization problems. However, when this algorithm optimizes the objective function of complex surfaces, it will fall into a local optimum and cannot search for the global optimum solution. To overcome this defect, a neuron attention mechanism is introduced to search for the target faster and complete the corresponding tasks.

[0119] It is explained that according to different types of attention, there are two different types of perception paradigms: top-down and bottom-up; among them, in the top-down perception paradigm, the organism first determines the target of interest and then searches in the surrounding environment based on the target characteristics. For targets with obvious characteristics, it can enable the organism to quickly capture its location and improve the search efficiency. However, for targets with unclear characteristics or in a more complex environment, the accuracy of searching for the target is low; the bottom-up perception paradigm means that when the organism has not yet consciously perceived the target object, information is transmitted from the external sensory organs to the brain from bottom to top. After the brain receives this information, it starts to mobilize attention resources to search. It can achieve a wide-area search for target objects in the environment and can also accurately perceive the target submerged in the complex environment, but it will consume time and perception resources.

[0120] Specifically, when the cerebral cortex receives external target stimuli, that is, when it is in the bottom-up perception paradigm, the generated bioelectric signal sequence has chaotic properties. Therefore, by selecting and combining based on two different paradigms, the wolf pack algorithm is combined with the bottom-up perception paradigm. That is, the constructed discrete neuron hyperchaotic mapping system NHCM is introduced into the wolf pack algorithm as an external bottom-up signal source, enabling it to take into account both the top-down and bottom-up attention paradigms during target search, effectively conducting a wide-area search of the entire problem domain, and making the neural hyper-chaotic wolf pack algorithm (NHC-WPA, Neural Hyper-chaotic Wolf Pack Algorithm) have better convergence speed and optimization accuracy.

[0121] It can be understood that the working principle of the wolf pack algorithm is based on three basic behaviors in the wolf pack: exploration, wandering, and hunting. Among them, exploration represents the wolf pack's cognition of the surrounding environment and the search for potential prey; the wandering behavior means that the wolf pack randomly searches unexplored areas to find new resources; the hunting behavior is to obtain food, and the wolf pack will launch a collective attack on the prey under the leadership of the alpha wolf. At the same time, the entire wolf pack follows the natural selection law of "survival of the fittest". Strong individuals can obtain more food and resources, while weak individuals are eliminated in the process of natural evolution. In the wolf pack, individuals have different group roles such as alpha wolves, scout wolves, and fierce wolves.

[0122] It can be explained that the alpha wolf represents the current optimal solution to the problem found; the scout wolves represent the group of sub-optimal solutions, which wander around the alpha wolf. Once a scout wolf finds a better problem solution than the current alpha wolf, the scout wolf will challenge the alpha wolf and become the new leader of the wolf pack. When a new alpha wolf is generated, the alpha wolf will immediately issue a summons, and the fierce wolves on the periphery of the wolf pack will immediately move towards the position of the current alpha wolf to complete the position update. Then, the wolf pack analyzes and sorts the problem solutions collected by the scout wolves and fierce wolves. Individuals with sub-optimal solutions will transform into scout wolves and continue to wander, while other individuals will become fierce wolves and be distributed on the periphery of the wolf pack. That is, iterative execution is carried out for the wolf pack to search for the global optimal solution of the problem.

[0123] Furthermore, in step S3, it includes:

[0124] Step S31: Initialize the individual positions in the wolf pack.

[0125] It should be noted that the distribution of individual positions in the wolf pack can affect the hunting effect of the entire wolf pack. That is, the more evenly the individual positions in the wolf pack are distributed in the entire solution space, the better. Based on the global ergodicity characteristics of chaotic mapping, the individuals in the wolf pack can be evenly distributed in the entire solution space, enabling the wolf pack algorithm to have a faster search convergence speed and a higher global optimal solution search accuracy.

[0126] Further, in step S31, the positions of the individuals in the wolf pack are initialized, and the corresponding calculation formula is:

[0127] {x i ,i = 1, …, dim} = lb i + rand * (ub i - lb i )

[0128] where x i represents the initial position of an individual in the wolf pack, i.e., the initial flight path point of the UAV; dim represents the dimension of the solution space of an individual in the wolf pack; lb i represents the minimum value of the i-th dimension of the solution space; ub i represents the maximum value of the i-th dimension of the solution space.

[0129] It should be noted that in this application, the wolf pack algorithm is adopted, that is, the flight path points of the UAVs are analogized to the individuals in the wolf pack; the positions of the individuals in the wolf pack are initialized, and the objective function is determined, that is, the optimization problem is solved centered on the objective function; where rand represents a random number between [0, 1]; the dimension of the solution space refers to a measure of the possible range and complexity of the wolf pack searching for the optimal solution, that is, the selectable space, as well as various dimensions such as the influencing factors of the selection.

[0130] Step S32: Define the group roles of the wolf pack as the lead wolf, the scout wolf, and the fierce wolf, update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm; update the position of the fierce wolf according to the lead wolf's summons and attack on the prey.

[0131] It should be explained that before the wolf pack algorithm starts to iterate, it will select the individual with the highest fitness as the lead wolf in the wolf pack. When the fitness value of other individuals in the wolf pack is better than the current lead wolf, this individual challenges the current lead wolf and then becomes the new lead wolf; specifically, the individual with the best fitness value in the initialization stage is set as the lead wolf, and at the same time, according to the role assignment rules of the wolf pack algorithm, the scout wolf and the fierce wolf are set proportionally.

[0132] Further, in step S32, it includes:

[0133] Step S321: Update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and the corresponding calculation formula is:

[0134]

[0135] where represents the position of the scout wolf after searching in the p-th direction; x irepresents the current position of the scout wolf; p represents the search direction of the scout wolf; N represents the total number of search directions; step_a represents the search step length of the scout wolf; Levy represents the Levy flight coefficient; α represents the neuron excitation coefficient; attention bottom_up represents the bottom_up attention attraction term; NHCM represents the discrete neuron hyperchaotic mapping system; δ represents the membrane potential change coefficient; β represents the neuron excitation value coefficient; y(n) represents the internal state variable of the neuron at time n.

[0136] It is explained that the scout wolf searches in N directions, then moves towards the direction with a better fitness value, and compares with the fitness value of the leading wolf again. If it is better than the fitness value of the leading wolf, the scout wolf will challenge the leading wolf and become the new leading wolf; after becoming the new leading wolf, the scout wolf will immediately issue a summons, and at this time all scout wolves and fierce wolves will also re-go to the position of the new leading wolf; if the fitness value of the current scout wolf is worse than that of the current leading wolf, it will continue to wander.

[0137] Specifically, introducing the bottom_up attention attraction term can improve the exploration accuracy of the global optimal solution of the scout wolf individuals; among them, Levy represents the Levy flight coefficient, which uses the occasional long-distance flight of the Levy function to simulate the jumping action of the scout wolf.

[0138] As an optional implementation manner, in this embodiment, a random number rand ∈ [0, length(NHCM)-1], δ = 6, β = 10 -3 .

[0139] It can be understood that after the leading wolf issues a summons, all scout wolves and fierce wolves search towards the position of the leading wolf. When the scout wolves and fierce wolves move, they have different search step lengths. Since the scout wolves are closer to the leading wolf, in order to avoid missing the position of the leading wolf due to too fast moving speed, their step lengths are set smaller; while the fierce wolves are at the edge of the wolf pack, and in order to reach the position of the leading wolf faster, they need to have a larger search step length.

[0140] Step S322: After the leading wolf issues a summons, update the position of the fierce wolf, and the corresponding calculation formula is:

[0141]

[0142] Among them, represents the position after the fierce wolf rushes towards the m-th direction of the leading wolf; x j represents the current position of the fierce wolf; step_b represents the search step length of the fierce wolf; g d represents the current position of the leading wolf; λ represents the attention weight coefficient; sin(8π*j) represents the attention deflection coefficient.

[0143] As an alternative implementation, in this embodiment, the periodic property of the sine function is used to achieve a 360° global search.

[0144] It can be understood that when the distance between an individual in the wolf pack and the lead wolf is less than the attack distance d, the entire wolf pack starts to attack the prey, and when the wolf pack starts to attack, the positions of all individuals in the entire wolf pack will be updated; the calculation formula for the attack distance d is:

[0145]

[0146] where D represents the dimension of the solution space; represents the distance weight coefficient; M dim represents the maximum value of the dim-th dimension of the solution space; m dim represents the minimum value of the dim-th dimension of the solution space.

[0147] Step S323: When the wolf pack attacks the prey, update the position of the fierce wolf, and the corresponding calculation formula is:

[0148]

[0149] where represents the position of the fierce wolf after attacking in the q-th direction; x k represents the current position of the fierce wolf; step_d represents the attack step length of the fierce wolf; G d represents the current position of the lead wolf; exp(-10 -2 *k) represents the attention attenuation coefficient.

[0150] It should be noted that exp(-10 -2 *k) represents the attention attenuation coefficient. When the fierce wolf individual is gradually approaching the prey, the attention will be fully concentrated on the current prey and will not be interfered by the outside world. And as the number of iterations increases, the attention attenuation coefficient will gradually decay to 0.

[0151] Step S33: Determine whether to terminate the iteration of the neuron hyperchaotic wolf pack algorithm through the constraint cost function for the updated positions of the wolf pack individuals. If not, the neuron hyperchaotic wolf pack algorithm continues to iterate and correct the positions of the wolf pack individuals until the positions of the wolf pack individuals are updated.

[0152] Understandably, based on the aforementioned logic, the position of each individual in the wolf pack is updated. Even if the fitness values of all individuals in the wolf pack are updated and re-sorted, according to the rules of the wolf pack algorithm, the corresponding lead wolves, scout wolves, and fierce wolves are re-designated. Then, based on the constraint cost function, it is determined whether the updated positions of the wolf pack individuals need to terminate the iteration of the neural hyper-chaotic wolf pack algorithm, that is, to determine whether the updated positions of the wolf pack individuals satisfy all the constraint conditions included in the constraint cost function. If so, the iteration process is stopped. If not, the neural hyper-chaotic wolf pack algorithm continues to iterate and correct the positions of the wolf pack individuals to perform the correction of the positions of the wolf pack individuals until the update of the positions of the wolf pack individuals is completed, which can more accurately realize the flight path planning of the UAV.

[0153] Furthermore, in step S4, it includes:

[0154] Step S41: Screen according to the constraint cost function through the updated positions of the wolf pack individuals to determine the optimal flight path points of the UAV, and form an optimal solution sequence;

[0155] Step S42: Use cubic spline interpolation to perform continuous interpolation calculation on the optimal solution sequence to obtain a continuous curve and get the best flight path of the UAV.

[0156] Specifically, based on the wolf pack individuals whose positions are updated after step S3, the UAV flight path points that satisfy the constraint cost function are screened out and determined as the optimal flight path points, and the optimal flight path points are integrated to form an optimal solution sequence; cubic spline interpolation is used to perform continuous interpolation calculation on the optimal solution sequence, that is, the discrete optimal flight path points are connected to form a continuous curve as the best flight path of the UAV. Among them, through cubic spline interpolation, it is ensured that the curve is not only continuous between each point, but also has continuous first-order and second-order derivatives, ensuring the smoothness of the flight path.

[0157] For better illustration, the present application conducts a simulation experiment to verify the feasibility of the proposed neural hyper-chaotic wolf pack algorithm (NHC-WPA, Neural Hyper-chaotic Wolf Pack Algorithm) for optimization problems, and uses the wolf pack algorithm (WPA, Wolf Pack Algorithm), grey wolf algorithm (GWO, Grey Wolf Optimizer), and giant trevally algorithm (GTO, Giant Trevally Optimizer) as comparison algorithms to test the three-dimensional path planning performance for scenarios with different obstacle densities, that is, for sparse scenarios and dense scenarios respectively, according to the neural hyper-chaotic wolf pack algorithm and the comparison algorithms, perform global optimal flight path planning, and obtain the fitness curves of each different algorithm in the two scenarios.

[0158] Please refer to Figure 2, which shows the comparison graph of the optimal path fitness curves of the sparse scene and the dense scene of a three-dimensional path planning method for unmanned aerial vehicles based on the neuron chaotic wolf pack algorithm provided by an embodiment of the present invention. Among them, the left figure (a) is the optimal flight path fitness curve of the sparse scene; the right figure (b) is the optimal flight path fitness curve of the dense scene.

[0159] It is explained that in the sparse scene and the dense scene with different obstacle densities, the NHC-WPA algorithm has a relatively fast convergence speed. Compared with other algorithms, it can quickly search for a better global optimal solution, while the comparison algorithms are all trapped in local optima to varying degrees, indicating that the NHC-WPA algorithm has a better convergence speed and global optimal search ability for both simple and complex scenes, and has stronger practicability.

[0160] It should be noted that the above-mentioned sequence of embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0161] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A three-dimensional path planning method for unmanned aerial vehicles based on a neuron chaotic wolf pack algorithm, characterized in that the flight path points of the unmanned aerial vehicle are collected, the flight path of the unmanned aerial vehicle is obtained according to the flight path points, and a constraint cost function is established; a discrete neuron hyperchaotic mapping system is constructed, and based on the discrete neuron hyperchaotic mapping system, combined with the neuron attention mechanism, the wolf pack algorithm is improved to determine the neuron hyperchaotic wolf pack algorithm; the individual positions in the wolf pack are initialized, the individual positions of the wolf pack are updated correspondingly according to the neuron hyperchaotic wolf pack algorithm, and it is judged whether the updated individual positions of the wolf pack meet the stopping criterion of the neuron hyperchaotic wolf pack algorithm through the constraint cost function. If not, continue to iterate for position correction until the update of the individual positions of the wolf pack is completed; the optimal flight path points of the unmanned aerial vehicle are determined through the updated individual positions of the wolf pack to form an optimal solution sequence, and the best flight path of the unmanned aerial vehicle is generated according to the optimal solution sequence.

2. The method for three-dimensional path planning of an unmanned aerial vehicle based on the neuron chaotic wolf pack algorithm according to claim 1, wherein, The flight path points of the unmanned aerial vehicle are collected, the flight path of the unmanned aerial vehicle is obtained according to the flight path points, and a constraint cost function is established, including: Collect the waypoint of the UAV and construct a sequence set {x i , i = 1, …, n}, where x i represents the i-th waypoint; n represents the total number of waypoints; calculating the path lengths between two flight path points based on the sequence set, and screening to determine the path constraint conditions, calculating the basic terrain constraint conditions and external threat constraint conditions in turn, and comprehensively forming the space constraint conditions; establishing a constraint cost function according to the path constraint conditions and the space constraint conditions.

3. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 2, characterized in that, Calculating the path lengths between two flight path points based on the sequence set, and screening to determine the path constraint conditions, calculating the basic terrain constraint conditions and external threat constraint conditions in turn, and comprehensively forming the space constraint conditions, including: Calculating the path lengths between two flight path points based on the sequence set, and the corresponding calculation formula is: Among them, dist represents the path length; x i represents the i-th track point; x i+1 represents the adjacent (i + 1)-th track point; n represents the total number of track points; Determining the path constraint conditions, and the corresponding calculation formula is: Path shortest = min(dist i ) Among them, Path shortest represents the shortest path; Defining the basic terrain as the distribution of independent mountain peaks above the reference sea level, calculating the height values of the independent mountain peaks, and the corresponding calculation formula is: Among them, h terrain represents the height value of an independent peak; w1 represents the constraint weight coefficient of the basic terrain; H terrain represents the equivalent altitude value of the basic terrain; represents the spatial constraint range of the basic terrain; (x1, y1) represents the spatial coordinate position of the basic terrain; Determining the basic terrain constraint conditions, and the corresponding calculation formula is: Among them, Crash terrain represents the collision judgment condition of the basic terrain; Defining the external threat as the distribution of continuous mountain peaks based on the sea level, calculating the height values of the continuous mountain peaks, and the corresponding calculation formula is: Among them, h threadn represents the height value of consecutive peaks; w2 represents the constraint weight coefficient of external threats; T thread represents the equivalent height of external threats; represents the spatial constraint range of external threats; (x2, y2) represents the spatial coordinate position of external threats; (x cti , y cti ) represents the correlation parameter; h terrain (x2, y2) represents the equivalent base terrain constraint height where the external threat is located; Determining the external threat constraint conditions, and the corresponding calculation formula is: Among them, Crash thread represents the collision judgment condition for external threats.

4. The three-dimensional path planning method for an unmanned aerial vehicle based on the neuron chaotic wolf pack algorithm according to claim 3, wherein, Establishing a constraint cost function according to the path constraint conditions and the space constraint conditions, and the corresponding calculation formula is: Path best (x i ) = Path shortest (x i ), Crash terrain (x i ) > 0, Crash thread (x i ) > 0 Among them, Path best represents the optimal path.

5. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 1, characterized in that Constructing a discrete neuron hyperchaotic mapping system, and based on the discrete neuron hyperchaotic mapping system, combined with the neuron attention mechanism, improving the wolf pack algorithm to determine the neuron hyperchaotic wolf pack algorithm, including: Based on the discrete Rulkov mapping model, introducing a quadratic nonlinear term and a sine curve nonlinear term to obtain the neuron Rulkov mapping; Combining the wolf pack algorithm with the bottom-up perception paradigm to determine the neuron hyperchaotic wolf pack algorithm.

6. The method for three-dimensional path planning of an unmanned aerial vehicle based on the neuron chaotic wolf pack algorithm according to claim 5, wherein Obtaining the neuron Rulkov mapping, and the corresponding calculation formula is: wherein, x(n) represents the membrane potential state variable of the neuron at time n; y(n) represents the internal state variable of the neuron at time n; δ represents the membrane potential change coefficient; α represents the neuron excitation coefficient.

7. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 1, wherein, Initialize the individual positions in the wolf pack, update the individual positions of the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and judge whether the updated individual positions of the wolf pack satisfy the stopping criterion of the neuron hyperchaotic wolf pack algorithm through the constraint cost function. If not, continue to iterate for position correction until the update of the individual positions of the wolf pack is completed, including: Initialize the individual positions in the wolf pack; Define the group roles of the wolf pack as the alpha wolf, the scout wolf, and the aggressive wolf. Update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm; update the position of the aggressive wolf according to the alpha wolf's call and attack on the prey; Judge whether the updated individual positions of the wolf pack need to terminate the iteration of the neuron hyperchaotic wolf pack algorithm through the constraint cost function. If not, the neuron hyperchaotic wolf pack algorithm continues to iterate to correct the individual positions of the wolf pack until the update of the individual positions of the wolf pack is completed.

8. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 7, characterized in that Initialize the individual positions in the wolf pack, and the corresponding calculation formula is: {x i , i = 1, …, dim} = lb i + rand * (ub i - lb i ) Among them, x i represents the initial position of an individual in the wolf pack, that is, the initial flight path point of the UAV; dim represents the dimension of the solution space of an individual in the wolf pack; lb i represents the minimum value of the i-th dimension of the solution space; ub i represents the maximum value of the i-th dimension of the solution space.

9. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 8, wherein Update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm; Update the position of the aggressive wolf according to the alpha wolf's call and attack on the prey, including: Update the position of the scout wolf in the wolf pack according to the neuron hyperchaotic wolf pack algorithm, and the corresponding calculation formula is: Among them, represents the position of the exploration wolf after searching in the p-th direction; x i represents the current position of the exploration wolf; p represents the search direction of the exploration wolf; N represents the total number of directions of the search direction; step_a represents the search step size of the exploration wolf; Levy represents the Levy flight coefficient; α represents the neuron excitation coefficient; attention bottom_up represents the bottom_up attention attraction term; NHCM represents the discrete neuron hyperchaotic mapping system; δ represents the membrane potential change coefficient; β represents the neuron excitation value coefficient; y(n) represents the internal state variable of the neuron at time n; After the alpha wolf makes a call, update the position of the aggressive wolf, and the corresponding calculation formula is: Among them, represents the position of the fierce wolf after raiding towards the m-th direction of the alpha wolf; x j represents the current position of the fierce wolf; step_b represents the search step size of the fierce wolf; g d represents the current position of the alpha wolf; λ represents the attention weight coefficient; sin(8π*j) represents the attention deflection coefficient; When the wolf pack attacks the prey, update the position of the aggressive wolf, and the corresponding calculation formula is: Among them, represents the position of the fierce wolf after attacking in the q-th direction; x k represents the current position of the fierce wolf; step_d represents the attack step length of the fierce wolf; G d represents the current position of the lead wolf; exp(-10 -2 *k) represents the attention attenuation coefficient.

10. A three-dimensional path planning method for an unmanned aerial vehicle based on a neuron chaotic wolf pack algorithm according to claim 1, characterized in that, Determine the optimal trajectory points of the UAV from the updated individual positions of the wolf pack to form an optimal solution sequence, and generate the best flight path of the UAV according to the optimal solution sequence, including: Screen according to the constraint cost function from the updated individual positions of the wolf pack to determine the optimal trajectory points of the UAV, and form an optimal solution sequence; Use cubic spline interpolation to perform continuous interpolation calculation on the optimal solution sequence to obtain a continuous curve and get the best flight path of the UAV.

Images