CN-121977555-A - Unmanned aerial vehicle cluster dynamic path planning method for complex mountain terrain

CN121977555ACN 121977555 ACN121977555 ACN 121977555ACN-121977555-A

Abstract

The invention belongs to the technical field of unmanned plane path planning. The invention provides an unmanned aerial vehicle cluster dynamic path planning method facing a dynamic complex mountain area environment. The method solves the optimization problem by constructing a cost function covering energy consumption, obstacle avoidance, flying height control, terrain collision avoidance and inter-machine safety, converting path planning into an optimization problem of path cost minimization, and providing an enhanced arithmetic optimization algorithm (EGAOA). The EGAOA algorithm integrates elite pool t distribution variation and Gaussian fusion search strategies, and uses reinforcement learning to update algorithm population candidate solutions in a most suitable mode in each optimization iteration process, so that the balance between a dynamic path planning algorithm exploration stage and a development stage is realized, and the capability of the algorithm for finding high-quality paths is improved. The EGAOA algorithm can effectively plan a safe and efficient flight path for the unmanned aerial vehicle cluster in a complex dynamic mountain area environment, and provides powerful support for the unmanned aerial vehicle cluster to execute tasks in the mountain area.

Inventors

TIAN JUN
WANG WENTAO

Assignees

南开大学

Dates

Publication Date: 20260505
Application Date: 20260109

Claims (7)

1. The unmanned aerial vehicle cluster dynamic path planning method for the complex mountain terrain is characterized by comprising the following steps of: Establishing a simulation scene and a cost function, establishing a simulation mountain area scene according to mountain area terrain environment elevation map data, and establishing an optimized target cost function comprising energy consumption, obstacle avoidance, flight height control, terrain collision avoidance and inter-machine safety; performing EGAOA algorithm initialization and selection of candidate solution updating stages, wherein each path node of the unmanned aerial vehicle cluster is encoded into one candidate solution in EGAOA algorithm population, and a Q-learning algorithm is used for selecting a proper updating stage for updating according to each candidate solution state; The method for solving the EGAOA algorithm comprises an exploration stage, a transition stage, a development stage and a calculation stage, wherein the exploration stage uses elite pool t distribution variation strategy to update algorithm population candidate solutions on the basis of arithmetic optimization algorithm multiplication and division operators; decoding path nodes and drawing paths after the EGAOA algorithm iteration updating is finished, screening out optimal candidate solutions in the population after the EGAOA algorithm completes single iteration updating, decoding the optimal candidate solutions into path nodes of the unmanned aerial vehicle cluster in the next step, circulating the process until all unmanned aerial vehicles reach the end point, and finally splicing the nodes generated in each step to form a complete unmanned aerial vehicle cluster flight track.
2. The method of claim 1, wherein establishing a simulation scenario and cost function, establishing a simulation mountain scenario based on mountain terrain environment elevation map data, and establishing an optimization objective cost function comprising energy consumption, obstacle avoidance, fly height control, terrain collision avoidance, and inter-machine safety, further comprises: Establishing a simulated terrain scene according to the elevation map data of the target mountain terrain, and defining a starting point and an ending point of the unmanned aerial vehicle cluster; The updating mode and path modeling of the unmanned aerial vehicle cluster are as follows: In the proposed path planning model, it is assumed that the drone cluster consists of N drones, the next position of each drone being based on its current position Speed and velocity of Yaw angle And pitch angle Dynamic update, wherein Representing the current space coordinates and parameters of the ith unmanned aerial vehicle 、 And Is constrained within a specific range: , And The spatial coordinate updating process of the ith unmanned aerial vehicle in the cluster is shown by the following formula: Searching the most suitable parameters in the search space by adopting EGAOA algorithm 、 And The 3 formulas are used for assisting the unmanned aerial vehicle cluster to continuously update the position coordinates thereof so as to avoid the obstacle and finally reach the destination, wherein the flight path of the unmanned aerial vehicle cluster The method is formed by combining a starting point of an unmanned aerial vehicle cluster, updated position nodes in each step and destination points, and the flight path of an ith unmanned aerial vehicle is expressed as Which is a set of M path nodes, Representing spatial coordinates of an ith unmanned aerial vehicle at a jth path node, wherein , ; The unmanned aerial vehicle cluster path planning cost function specifically comprises the following sub-cost functions: Energy consumption cost function: For the ith unmanned aerial vehicle, adjacent nodes And The Euclidean distance between them is recorded as The total path length of each unmanned aerial vehicle is obtained by summing the distances between all adjacent nodes on the path, and the energy consumption cost function is as follows: Wherein the method comprises the steps of Representing the current node of the ith unmanned aerial vehicle To the destination Is used for the remaining distance of (a), A weight coefficient representing the parameter; Is punishment term coefficient for measuring the cost of flying steps Representing the total number of steps required by the unmanned aerial vehicle cluster to reach the destination; Value of sum Default settings are 5 and 100, or flexibly adjusted according to specific task requirements; obstacle avoidance cost function: There are Q static obstacles and K dynamic obstacles in the environment, each modeled as a cylinder, for the kth dynamic obstacle The origin coordinates of which are defined as The current coordinates are defined as The endpoint coordinates are defined as Updating the mathematical model of its next step location is shown in the following formula: Wherein, the And Respectively representing the speed and the movement direction angle of the kth dynamic obstacle, and assuming that the ith unmanned aerial vehicle in the cluster is from a path node Move to And the (i+1) th unmanned aerial vehicle slave Move to Both trajectories encounter static and dynamic obstacles in the vicinity thereof to Move to For example, let And Respectively represent the center and the radius of the q-th static obstacle, wherein , And Representing the center and radius of the kth dynamic barrier respectively, And Respectively represent the slaves of the unmanned plane To the point of The safe distance between the flight segment of the (c) and the collision zone of the static obstacle and the dynamic obstacle is kept, W represents the radius of the unmanned plane, And Respectively represent slave flight segments To the point of And The mathematical model of the obstacle avoidance cost function is as follows: Wherein the method comprises the steps of And Respectively representing the cost of unmanned aerial vehicle clusters for avoiding dynamic barriers and static barriers; the mathematical model of (2) is shown in the following formula: Wherein, the And Respectively representing the current position and the next position of the kth dynamic obstacle to the flight section of the unmanned plane The threat cost of the composition; is calculated by the formula of (2) The same, if the unmanned aerial vehicle collides with the obstacle, the cost function value is set to infinity, which indicates that the corresponding path is not feasible, the cost function of avoiding the static obstacle The formula of (2) is as follows: Wherein the method comprises the steps of Representing the q-th static obstacle to the flight section of the unmanned plane The threat cost caused; fly height control cost function: The flying height range of the unmanned aerial vehicle group is set as The range is adjusted according to the actual terrain condition, and the path node is The flight height of the ith unmanned aerial vehicle is The mathematical model of the unmanned aerial vehicle group flight altitude control cost function is as follows: Avoiding terrain collision cost function: Taking an ith unmanned aerial vehicle as an example, generating a group of finer interpolation points including U interpolation points for a path segment between two adjacent path nodes by adopting cubic spline interpolation The spatial coordinates of the inner u-th interpolation point are Wherein The terrain height of the unmanned aerial vehicle at the XOY plane coordinate is recorded as Then, the height of each interpolation point on the flight path is matched with the height of the corresponding terrain Comparing if less than The path is considered to collide with the terrain, the cost value is set to be infinity, and the path is marked as an infeasible path; the mathematical model of the cost function for avoiding the terrain collision is shown as follows: Unmanned aerial vehicle safety cost function: For any two unmanned aerial vehicles selected randomly in the cluster, namely unmanned aerial vehicle a and unmanned aerial vehicle b, at time j, the positions of unmanned aerial vehicle a and unmanned aerial vehicle b are respectively expressed as And Their Euclidean distance is expressed as If the distance is smaller than If the distance is zero, collision is considered to occur, the planned unmanned aerial vehicle cluster flight path is not feasible, the cost function value is set to infinity, and the mathematical model of the safety cost function among unmanned aerial vehicles is as follows: Wherein the symbol penalty represents a penalty term for forcing a safe distance between unmanned aerial vehicles; total cost function: The total cost function integrates various factors including energy consumption, obstacle avoidance, flight altitude limitation, terrain collision threat and safety cost among unmanned aerial vehicles, and the total path planning cost function of unmanned aerial vehicle clusters in the dynamic mountain area environment is recorded as The following is shown: Wherein, the Representing the weight of each sub-cost function in the total cost function.
3. The method of claim 2, wherein performing EGAOA algorithm initialization and selection of a candidate solution update stage, specifically comprises: algorithm candidate decoding and population initialization: each candidate solution in the EGAOA algorithm population is encoded as a parameter, i.e., speed, required by the next set of path nodes of the drone cluster Yaw angle And pitch angle ; The initialization process of EAAOA algorithm is performed using the following formula: Wherein, the And The minimum and maximum values of the search space are defined respectively, Is an in-interval A number that is randomly generated in the memory, Representing the i-th candidate solution in the population, Representing the dimensions of which, Is the number of algorithm populations; selection of a candidate solution update phase: The Q-learning algorithm is used for selecting a proper updating stage for updating according to each candidate solution state, and the method comprises the following steps of initializing a Q table containing Np rows and 3 columns, wherein Np is the number of candidate solutions, and the 3 columns respectively correspond to Q values of the exploring, transitional and developing stages according to a formula Updating the Q value in the Q table, wherein And EGAOA updating through Q-learning according to the self-adaptive selection exploration, transition and development stages of each candidate solution.
4. A method according to claim 3, wherein performing a solution process of EGAOA algorithm, in particular comprises: The exploration and development stage of constructing a basic arithmetic optimization algorithm AOA: in the exploration stage of AOA, the algorithm updates the candidate solution by using multiplication and division operators, and the update mathematical expression of the candidate solution in the stage is shown as follows: Wherein, the Representing the i-th candidate solution generated in the t +1 th iteration, Representing the best candidate solution searched so far, constant Is a very small positive number, parameter And Respectively used for adjusting the dynamic characteristics and the search sensitivity parameters of the optimization process; In the development stage of AOA, the candidate position is updated by addition and subtraction operation, and in this stage, the updating formula of the candidate solution is shown as the following formula: the exploration, transition and development phases of the build EGAOA algorithm: In the exploration stage of EGAOA algorithm, after finishing the preliminary update of the candidate solution by utilizing multiplication and division operators in the AOA exploration stage in EGAOA algorithm, introducing a t distribution exploration strategy based on elite pool to carry out secondary update on the candidate solution, constructing an elite pool containing 5% of high-quality solutions before each generation so as to enhance the diversity of the search direction, then utilizing a t distribution variation disturbance mechanism to carry out disturbance on the updated candidate solution under the guidance of an elite pool individual, and a EGAOA exploration stage mathematical model is as follows: Wherein the method comprises the steps of Is the maximum number of iterations, t is the current number of iterations, And Two random solutions in elite pool, and at the end of EGAOA, the new candidate solution is obtained Comparing the cost function value with the candidate solution obtained in the original AOA algorithm exploration stage, and selecting the solution with the small cost function as a new candidate solution to enter the next iteration; The transition stage of EGAOA algorithm, aiming at the problem that the original AOA algorithm only comprises two stages of exploration and development and lacks an intermediate transition mechanism, introduces a Gaussian fusion search strategy for constructing a new transition stage, wherein the Gaussian fusion search strategy consists of a Gaussian center disturbance operator and a multi-direction differential mutation operator, and selects probability parameters A ranking decision of the cost function value of each candidate solution is used for selecting one of two operators to update the current candidate solution if the current candidate solution And when the number is larger than the random number r, updating the current candidate solution by using a Gaussian center disturbance operator, otherwise, updating the current candidate solution by using a multi-direction differential mutation operator, wherein the mathematical model updated by the candidate solution at the stage is as follows: Wherein, the A cost function value ranking representing the i-th candidate solution, Representing a gaussian perturbation operator to generate new candidate solutions In the process of (a), Is the optimal candidate solution in the current population, ~ The method comprises the steps of selecting five different candidate solutions randomly from a population, using the five different candidate solutions to form a multi-directional differential mutation operator to update a current candidate solution, comparing the cost function value of the new candidate solution and the current solution by adopting a greedy strategy, and reserving a solution with a better cost function value as an iteration result; The development phase of EGAOA algorithm is the same as the AOA algorithm development phase, and the algorithm population candidate solution is updated by using the following formula: 。
5. Unmanned aerial vehicle cluster dynamic path planning device towards complicated mountain area topography, characterized in that includes: the first module is used for establishing a simulation scene and a cost function, establishing a simulation mountain area scene according to mountain area terrain environment elevation map data, and establishing an optimized target cost function comprising energy consumption, obstacle avoidance, flight height control, terrain collision avoidance and inter-machine safety; the second module is used for executing EGAOA algorithm initialization and selection of candidate solution updating stages, and comprises the steps of encoding each path node of the unmanned aerial vehicle cluster into one candidate solution in EGAOA algorithm population, and selecting a proper updating stage for updating by using a Q-learning algorithm according to each candidate solution state; The third module is used for executing the solving process of EGAOA algorithm, and comprises an exploration stage, a transition stage, a development stage and a calculation stage, wherein the exploration stage uses elite pool t distribution variation strategy to update algorithm population candidate solutions on the basis of arithmetic optimization algorithm multiplication and division operators; And the fourth module is used for decoding path nodes and drawing paths after the EGAOA algorithm iteration updating is finished, screening out optimal candidate solutions in the population after the EGAOA algorithm completes single iteration updating, decoding the optimal candidate solutions into path nodes of the unmanned aerial vehicle cluster in the next step, circulating the process until all unmanned aerial vehicles reach the end point, and finally splicing the nodes generated in each step to form a complete unmanned aerial vehicle cluster flight track.
6. An electronic device comprising a processor and a memory communicatively coupled to the processor; the memory stores computer-executable instructions; The processor executes computer-executable instructions stored in the memory to implement the method of any one of claims 1-4.
7. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-4.

Description

Unmanned aerial vehicle cluster dynamic path planning method for complex mountain terrain Technical Field The invention relates to the technical field of unmanned aerial vehicle planning, in particular to an unmanned aerial vehicle cluster dynamic path planning method for complex mountain terrain. Background Unmanned aerial vehicle cluster path planning is used as a key technology of an intelligent unmanned aerial vehicle system and is widely applied to the fields of military reconnaissance, disaster relief, mountain area logistics and the like. With the development of three-dimensional terrain modeling and dynamic obstacle perception technologies, a path planning algorithm based on group intelligence gradually becomes a hotspot of an embodiment of the invention. In the prior art, an Arithmetic Optimization Algorithm (AOA) constructs a collaborative optimization system comprising exploration and development stages by simulating four arithmetic logics, and a core flow of the collaborative optimization system covers key links such as initial solution generation, neighborhood search, global convergence and the like. Specifically, the AOA updates the candidate solution using four arithmetic operations, and guides the search direction through the globally optimal solution, where the diversity of the exploration phase remains with the locally optimal capability of the development phase to form two major struts of the algorithm performance. However, existing AOA methods have systematic drawbacks in dynamic complex terrain scenarios. Specifically, the exploration stage is updated only by relying on the current optimal solution, and a multi-level solution set guiding mechanism is not constructed, so that the searching direction is single. After the algorithm is iterated, the overall optimal solution discovery rate of the algorithm is low, the algorithm is easy to sink into a local optimal trap, and path quality is poor for unmanned aerial vehicle planning. In the development stage, neighborhood information is not effectively utilized, solutions are updated only through simple addition and subtraction operation, and the depth mining capability of a local optimal region is lacked. In addition, the traditional static variation strategy cannot adapt to the search requirements of different iteration stages, and the defects of insufficient early exploration and poor later convergence precision are particularly remarkable. These defects make it difficult for the algorithm to generate an optimal path that combines safety and efficiency under the multi-dimensional constraint conditions such as three-dimensional terrain constraint, dynamic obstacle avoidance, cluster collaborative optimization, and the like. Disclosure of Invention The invention aims to provide a dynamic path planning method of an unmanned aerial vehicle cluster facing to complex mountain terrain, which effectively solves the path planning problem of the unmanned aerial vehicle cluster in the complex dynamic mountain environment through a constructed enhanced arithmetic optimization algorithm EGAOA. Therefore, a first object of the present invention is to provide a method for planning a dynamic path of an unmanned aerial vehicle cluster facing to a complex mountain terrain. The second aim of the invention is to provide an unmanned aerial vehicle cluster dynamic path planning device facing complex mountain terrain. A third object of the present invention is to propose an electronic device. A fourth object of the present invention is to propose a computer readable storage medium. A fifth object of the invention is to propose a computer programme product. In order to achieve the above objective, an embodiment of a first aspect of the present invention provides a method for planning a dynamic path of an unmanned aerial vehicle cluster facing a complex mountain terrain, including: Establishing a simulation scene and a cost function, establishing a simulation mountain area scene according to mountain area terrain environment elevation map data, and establishing an optimized target cost function comprising energy consumption, obstacle avoidance, flight height control, terrain collision avoidance and inter-machine safety; performing EGAOA algorithm initialization and selection of candidate solution updating stages, wherein each path node of the unmanned aerial vehicle cluster is encoded into one candidate solution in EGAOA algorithm population, and a Q-learning algorithm is used for selecting a proper updating stage for updating according to each candidate solution state; The method for solving the EGAOA algorithm comprises an exploration stage, a transition stage, a development stage and a calculation stage, wherein the exploration stage uses elite pool t distribution variation strategy to update algorithm population candidate solutions on the basis of arithmetic optimization algorithm multiplication and division operators; decoding path nodes and drawing paths afte