Unmanned aerial vehicle multi-target path planning method based on multi-target sheep flock migration algorithm

Abstract

The invention provides an unmanned aerial vehicle multi-target path planning method based on a multi-target sheep flock migration algorithm. The method comprises the following steps: 1, constructing a three-dimensional model of a task environment: 1.1, constructing a grid-based three-dimensional task space model; 1.2, determining collision probability distribution; 2, initializing algorithm parameters and populations; 2.1, initializing multi-target sheep flock migration algorithm parameters; 2.2, randomly generating an initial population in the task area range; 3, performing non-dominated sorting on the population, and selecting head sheep; 3.1, calculating an objective function value of each sheep; 3.2, determining a Pareto optimal solution set by adopting non-dominated sorting; 3.3, selecting the sheep with the largest crowding distance as the head sheep; 4, executing a grazing operator; 5, judging and executing a migration operator or a contraction operator; 5.1, if the optimal solution of the group is updated, executing the migration operator; 5.2, if the optimal solution of the group is not updated, executing a contraction operator; and 5, judging whether the maximum number of iterations is reached or not, if so, outputting an optimal path planning solution, and otherwise, returning to the step 3.

Description

(I) Technical field

[0001] The present invention belongs to the field of unmanned aerial vehicle (UAV) path planning, and in particular relates to a multi-objective UAV path planning method. (II) Background technology

[0002] As related technologies mature, drones are gradually taking on more and more tasks in modern society, such as urban transportation, industrial transportation, route inspection, etc. In the process of performing tasks, drones need to fly back and forth in complex environments, avoiding collisions with obstacles while ensuring the efficiency of performing tasks, which puts forward requirements for the path planning capabilities of drones.

[0003] Most existing path planning methods focus on avoiding collisions, but rarely consider efficiency-related factors such as flight distance and flight time. This path planning method can maximize flight safety, but in some application scenarios with high mission time requirements (such as medical scenarios, combat scenarios, etc.), it may lead to time waste and reduced efficiency, and even mission failure. Therefore, developing a multi-objective path planning method for drones that comprehensively considers obstacle avoidance and mission efficiency has become the key to further improving the mission effectiveness of drones.

[0004] Due to the constraints of the nonlinear dynamic characteristics of the UAV itself and the spatiotemporal complexity of the mission environment, the multi-objective path planning problem of UAVs usually has the characteristics of high dimensionality, nonlinearity, and multi-modality. The use of traditional mathematical programming methods to solve them has problems such as complex calculations and high time costs. At present, various heuristic optimization algorithms are the mainstream methods for solving UAV path planning. They have been widely used because of their simple principles and controllable optimization time. Among them, the flock migration algorithm is a recently proposed swarm intelligence heuristic optimization algorithm with the advantages of good high-dimensional performance and strong multi-modal optimization capabilities. It has successful application precedents in the field of UAVs and has excellent potential for solving multi-objective path optimization problems of UAVs.

[0005] Based on this, this patent comprehensively considers the flight distance and collision risk of UAVs, proposes a multi-objective path planning model for UAVs, and introduces a non-dominated sorting mechanism on the basis of the traditional flock migration algorithm, and proposes a multi-objective improved flock migration optimization algorithm. The algorithm is used to solve the proposed multi-objective path planning model, which is expected to provide a new and efficient solution for UAV path planning under the conditions of comprehensive consideration of mission efficiency and safety. (III) Summary of the invention

[0006] This paper combines the non-dominated sorting idea with the flock migration optimization algorithm, proposes a multi-objective flock migration algorithm, and applies it to the multi-objective path planning problem of unmanned aerial vehicles. It proposes a joint optimization path planning method for unmanned aerial vehicle flight distance and collision risk based on the multi-objective flock migration algorithm. This method can overcome the current defects of unmanned aerial vehicle path planning, such as single consideration of factors, insufficient response to potential obstacles, and slow strategy generation, and provides a new solution for the path planning problem of unmanned aerial vehicles under the condition of considering multiple optimization objectives.

[0007] The specific steps are as follows:

[0008] Step 1: Build a 3D model of the mission environment, including:

[0009] Step 1.1 Use LiDAR to collect mission environment information and use it to build a grid-based three-dimensional mission space model;

[0010] Step 1.2: Determine the high collision risk points in the environment based on prior information, and obtain the probability distribution of collision events through the mixed Gaussian distribution model;

[0011] Step 2: Initialize the parameters of the multi-target flock migration algorithm and randomly generate the initial population, including:

[0012] Step 2.1 Initialize the parameters of the multi-target flock migration algorithm, including the maximum number of iterations, population size, maximum number of searches, benchmark search step length, number of navigation points, etc.;

[0013] Step 2.2 randomly generates an initial population within the mission area, and each sheep represents a path planning scheme;

[0014] Step 3: Perform non-dominated sorting on the flock and select the leader;

[0015] Step 3.1 calculates the path length function and path safety function values ​​corresponding to each individual sheep respectively;

[0016] Step 3.2 uses the non-dominated sorting method to calculate the Pareto optimal solution set of the flock;

[0017] Step 3.3 Calculate the crowding distance Dis of each sheep in the Pareto optimal frontier, and select the sheep with the largest crowding distance as the leader;

[0018] Step 4: The flock executes the herding operator and performs random search;

[0019] Step 5: The flock determines whether to execute the migration operator or the contraction operator based on the update of the group's optimal solution, including:

[0020] Step 5.1 If the optimal solution of the group is updated, the flock executes the migration operator to conduct overall exploration;

[0021] Step 5.2 If the group optimal solution is not updated, the flock executes the contraction operator to further search;

[0022] Step 6 determines whether to end the calculation based on whether the maximum number of iterations has been reached. If so, output the path planning solution corresponding to the current group optimal solution and end the process; otherwise, return to step 3; (IV) Description of the drawings

[0023] Figure 1 This is a schematic flow chart of a method for multi-objective path planning of an unmanned aerial vehicle based on a multi-objective flock migration algorithm proposed in the present invention. (V) Specific implementation methods

[0024] The exemplary implementation of the present invention is described in detail below in conjunction with the accompanying drawings. The following description includes specific details to assist understanding, but these specific details should be shown as exemplary only. Therefore, it will be appreciated by those of ordinary skill in the art that various changes and modifications may be made to the various examples described herein without departing from the scope and spirit of the present disclosure. In addition, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted.

[0025] The terms and words used in the following description and claims are not limited to their literal meanings, but are only used by the inventor to implement a clear and consistent understanding of the present invention. Therefore, it should be clear to those skilled in the art that the following description of various example implementations of the present invention is provided only for illustrative purposes and is not intended to limit the present invention defined by the attached claims and their equivalents.

[0026] Step 1: Build a 3D model of the mission environment, including:

[0027] Step 1.1: Use LiDAR to collect mission environment information and build a grid-based three-dimensional mission space model: First, use LiDAR to collect mission space environment information, including obstacle distribution; then divide the three-dimensional mission space into a certain number of cubic grids according to a certain grid density; then select the initial position of the drone as the coordinate origin, establish a three-dimensional coordinate system, and use the center position coordinates of each grid cube as the coordinates of the grid to build a three-dimensional environment model of the mission space.

[0028] Step 1.2: Determine the high collision risk points in the environment based on prior information, and obtain the probability distribution of collision events through the mixed Gaussian distribution model: According to the prior information of the task environment space, determine the locations of all possible obstacles in the environment, set the grids where they are located as risk points, with a total number of m. According to the distribution of risk points, determine the collision risk probability of the grid with coordinates (x, y, z) in the task space through the mixed Gaussian distribution model:

[0029]

[0030] Among them, (x j ,y j ,z j ) is the coordinate of the jth risk point, and the standard deviation σ j represents the risk diffusion degree of the jth risk point. The collision risk probability of the grid covered by the fixed obstacle in the task space is set to 1.

[0031] Example 1, taking the drone delivery mission as an example. According to step 1, first use the lidar to collect the environmental information of the mission space, and know that the mission space is a 20m×30m×5m rectangular area. Then divide the mission space into 24,000 cubic grids according to the grid density of 0.5m. Establish a three-dimensional coordinate system with the initial position of the drone as the coordinate origin, and assign coordinates to each cubic grid. According to prior information, there are 3 high collision risk points in the mission environment, with coordinates of (5,10,1), (7,16,4) and (19,6,1), and the corresponding standard deviation of risk diffusion is σ 1 =3,σ 2 =4,σ 3 =3, based on which, the collision risk probability of the grid with coordinates (x, y, z) in space can be calculated as:

[0032]

[0033] Step 2: Initialize the parameters of the multi-target flock migration algorithm and randomly generate the initial population, including:

[0034] Step 2.1: Initialize the parameters of the multi-target flock migration algorithm, including: maximum number of iterations T, population size N, maximum number of searches T GOmax , Benchmark search step a / a leader , the number of navigation points n;

[0035] Step 2.2: Randomly generate an initial population within the mission area. Each sheep represents a path planning scheme. Specifically, the path planning of the drone adopts the navigation point method. By selecting n navigation points in the mission space, its coordinates are (x i ,y i ,z i ), i = 1, 2, 3.., n, to guide the flight trajectory of the drone. Therefore, the coordinates of all navigation points are the decision variables of the path planning problem, and the decision variables are written as X = (x 1 ,y 1 ,z 1 ,x 2 ,y 2 ,z 2 ,...,x n ,yn ,z n ), where all coordinates are within the task space.

[0036] Therefore, population initialization is to randomly generate N decision variables X within the task area. i , i=1,2,...,N, forming the initial population Ω={X 1 ,X 2 ,...,X N}.

[0037] Example 2, continued from Example 1.

[0038] Take the drone cargo delivery task as an example. According to step 2, initialize the multi-target flock migration algorithm parameters: T = 200, N = 100, T GOmax =10, a=5, a leader =9, n = 10. Next, 100 path planning solutions X are randomly generated within the task space coordinate range. i =[x 1 ,y 1 ,z 1 ,x 2 ,y 2 ,z 2 ,...,x 10 ,y 10 ,z 10 ], i = 1, 2, 3, ..., 100 as the initial population Ω.

[0039] Step 3: Perform non-dominated sorting on the population and select the leader, including:

[0040] Step 3.1: Calculate the path length function and path safety function values ​​for each individual sheep, specifically: path planning scheme X i The calculation formula of the path length function value is:

[0041]

[0042] Among them, (x n+1 ,y n+1 ,z n+1 ) is the coordinate of the mission target point. Path planning solution X i The calculation formula of the path security function is:

[0043]

[0044] Among them, pr(x i ,y i ,z i ) represents the navigation point (x i ,y i ,zi )’s collision risk probability value.

[0045] Step 3.2: Use the non-dominated sorting method to calculate the Pareto optimal solution set of the flock, specifically: based on the path length function value f of each sheep len and the path security function value f safe , calculate the Pareto optimal solution set of the entire population Ω, denoted as Will Subtract it from the entire population Ω, and continue to calculate the Pareto optimal solution set of the remaining population, recorded as Continue this process until there are no remaining individuals in the population, generating M sets in the process

[0046] Step 3.3: Calculate the crowding distance Dis of each sheep in the Pareto optimal solution set, and select the sheep with the largest crowding distance as the leader. Specifically: For the Pareto optimal frontier of the entire population Calculate the crowding distance value of each sheep, the calculation formula is:

[0047]

[0048] in, Represents the Pareto optimal frontier The number of individuals in and Respectively represent The maximum and minimum values ​​of the path length function corresponding to all individuals in , and Respectively represent The maximum and minimum values ​​of the path safety function for all individuals in . After calculating the crowding distance of all sheep, the one with the largest crowding distance is selected and set as the head sheep.

[0049] Example 3, continued from Example 2.

[0050] Take the drone cargo delivery task as an example. According to step 3, first calculate the path length function value f of each sheep representing the solution in the initial population Ω len (X i ) and the path security function value f safe (X i ), and then use the non-dominated sorting method to calculate all the undominated solutions in the population to form the Pareto optimal frontier Will Subtract it from Ω, and then calculate the non-dominated solution set of the remaining population, and repeat this process until 7 sets are generated. The crowding distance of all sheep in the calculation results shows that Dis(X 2 ) is the largest, so we choose X in 2 For the head sheep.

[0051] Step 4: The flock executes the herding operator and performs a random search. Specifically, after selecting the leader, the solution represented by the leader is set as the optimal solution of the flock X gbest , the entire flock performs random search at a certain step length, and the search process of the leader is:

[0052] X lead (t+1)=X lead (t)+μ lead ×e

[0053] The search process for other sheep is:

[0054] X i (t+1)=X i (t)+μ follow ×e

[0055] Where, e=[e 1 ,e 2 ,...e 3n ] takes random values ​​in the range [-1,1]. lead and μ follow are the benchmark search steps for the head sheep and other sheep respectively, and their values ​​are adjusted dynamically with the number of iterations. During the search process, the path length function value and path safety function value of the newly searched solution are calculated, and it is determined whether the new solution is the dominant solution of the original solution. If so, the individual optimal solution is updated.

[0056] Example 4, continued from Example 3.

[0057] Take the drone cargo delivery mission as an example. According to step 4, 2 Select as the leader and set as the group's optimal solution X gbest The head sheep and other sheep search according to the formula in step 4 respectively, and update their own individual optimal solutions based on the search results.

[0058] Step 5: According to the update of the optimal solution of the group, determine whether to execute the migration operator or the contraction operator, including:

[0059] Step 5.1: If the optimal solution of the group is updated, the flock executes the migration operator. Specifically, after step 4, the flock calculates the optimal solution of the group at this time. If there is an update compared to the beginning of step 4, the flock migrates, and the head sheep moves according to the following formula:

[0060]

[0061] Among them, μ migr The migratory pace of the flock is set in advance; is a random number between [0,1]; Represents the individual optimal solution of the leader. After the leader completes the move, the other sheep move in turn according to the following formula:

[0062]

[0063] in:

[0064]

[0065] Among them, c α ,c β ,c γ are pre-set parameters, and are three unit direction vectors, whose values ​​are:

[0066]

[0067]

[0068] When all sheep have finished moving, step 5 ends.

[0069] Step 5.2: If the optimal solution of the group is not updated, the flock executes the contraction operator, specifically: after step 4, the flock calculates the optimal solution of the group at this time. If it is not updated compared to the beginning of step 4, the flock shrinks, the position of the head sheep remains unchanged, and the positions of other sheep are updated according to the following formula:

[0070]

[0071] Among them, δ 1 and δ 2 is a pre-set shrinkage coefficient, satisfying δ 1 +δ 2 = 1. When all sheep have finished moving, step 5 ends.

[0072] Example 5, continued from Example 4.

[0073] Take the drone cargo delivery task as an example. After step 4, the optimal solution of the group is calculated and compared with the optimal solution of the group at the beginning of step 4. If an update occurs, the flock executes step 5.1 and sets c α =0.25,c β =0.5,c γ = 0.25, the entire flock migrates according to the formula, and after the migration is completed, step 5 ends; if no update occurs, the flock executes step 5.2 and sets δ 1 =0.3,δ 2 =0.7, the flock shrinks according to the formula, and step 5 ends after the movement is completed.

[0074] Step 6: Determine whether to end the calculation based on whether the current number of iterations t reaches the maximum number of iterations T. If t=T, output the path planning solution corresponding to the current group optimal solution and end the method process; if t<T, return to step 3.

[0075] Example 6, continued from Example 5.

[0076] Take the drone cargo delivery task as an example. After 200 iterations, the end condition is reached and the path planning solution corresponding to the current group optimal solution is output: the coordinates of the 10 relay navigation points (x 1 ,y 1 ,z 1 ),(x 2 ,y 2 ,z 2 ),...,(x 10 ,y 10 ,z 10 ) to the drone, and the drone determines its own path based on the obtained coordinates.

[0077] It can be seen that the multi-target path planning of the UAV can be realized according to the steps described in the present invention.

[0078] This paper combines the non-dominated sorting idea with the flock migration optimization algorithm, proposes a multi-objective flock migration algorithm, and applies it to the multi-objective path planning problem of unmanned aerial vehicles. It proposes a joint optimization path planning method for unmanned aerial vehicle flight distance and collision risk based on the multi-objective flock migration algorithm. This method can overcome the current shortcomings of unmanned aerial vehicle path planning, such as single consideration of factors, insufficient response to potential obstacles, and slow strategy generation.

[0079] The examples described above have provided detailed explanations of the implementation methods of various parts of the present invention. The specific implementation forms of the present invention are not limited thereto. For those skilled in the art, various obvious changes to the method of the present invention without departing from the spirit of the method of the present invention and the scope of the claims are within the protection scope of the present invention.

Claims

1. A multi-target path planning method for unmanned aerial vehicles based on a multi-target flock migration algorithm, characterized in that: The following steps are involved: Step 1: Build a 3D model of the mission environment. Specifically, use LiDAR to collect mission environment information and build a grid-based 3D mission space model. Determine high-collision risk points in the environment based on prior information and obtain the probability distribution of collision events through a mixed Gaussian distribution model. Step 2: Initialize the parameters of the multi-target flock migration algorithm and randomly generate the initial population. Specifically, initialize the parameters of the multi-target flock migration algorithm, including the maximum number of iterations T, the number of populations N, the number of navigation points n, etc.; randomly generate the decision variable X in the task area. i ,i=1,2,...,N as the initial population Ω; Step 3: Perform non-dominated sorting on the flock and select the leader. Specifically, calculate the path length function and path safety function values ​​corresponding to each individual sheep respectively; then use the non-dominated sorting method to calculate the Pareto optimal solution set of the flock; calculate the crowding distance Dis of each sheep in the Pareto optimal solution set, and select the sheep with the largest crowding distance as the leader; Step 4: The flock executes the herding operator and performs a random search, specifically: the solution represented by the head sheep is set as the group's optimal solution X gbest , the entire flock searches at a certain step length. If the new solution found is the dominant solution of the original solution, the optimal solution of the individual is updated; Step 5: According to the update of the optimal solution of the group, determine and execute the migration operator or the contraction operator. Specifically, if the optimal solution of the group is updated, the migration operator is executed, and the sheep migrate in turn according to the given formula; if the optimal solution of the group is not updated, the contraction operator is executed, the position of the head sheep remains unchanged, and the other sheep shrink towards the position of the head sheep according to the given formula; Step 6: Determine whether to end the calculation based on whether the current number of iterations t reaches the maximum number of iterations T: if t<T, return to execute step 3; if t=T, output the path planning solution corresponding to the current group optimal solution and end the process.

Images