Formation obstacle avoidance method based on escape algorithm

Abstract

The invention discloses a formation obstacle avoidance method based on an escape algorithm, which is used for solving the problem of formation cooperative obstacle avoidance in a complex environment. According to the method, a population is initialized and elite individuals are retained, the individuals are divided into three types to be differentially updated according to a panic index in an iteration initial stage so as to guarantee search diversity, after the number of iterations reaches a set threshold value, the method is switched to a development stage to deeply excavate an optimal solution, and finally a formation obstacle avoidance path scheme is output. According to the method, efficient collaborative path planning can be realized, and the method is adaptive to multiple types of formation obstacle avoidance scenes.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control and path planning technology, and relates to multi-agent formation cooperative obstacle avoidance technology, and particularly to a formation obstacle avoidance method based on an escape algorithm. Background Technology

[0002] With the increasing demand for multi-agent systems in logistics inspection, military reconnaissance, disaster relief and other fields, formation cooperative obstacle avoidance has become a core scenario for multi-agent systems to perform complex tasks, which puts forward higher requirements for the efficiency, coordination and robustness of path planning.

[0003] Existing formation obstacle avoidance methods are mostly based on traditional swarm intelligence algorithms or geometric obstacle avoidance strategies. These methods suffer from insufficient diversity in the search phase and are prone to getting stuck in local optima during the development phase, making it difficult to meet the real-time collaborative obstacle avoidance requirements of formations in complex dynamic environments.

[0004] This invention proposes a formation obstacle avoidance method based on an escape algorithm. This method effectively solves the problems of weak coordination and low path planning accuracy of traditional methods through population differential updates and deep development guided by an elite pool, providing a reliable solution for the safe and efficient operation of multi-agent formations in complex environments. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a formation obstacle avoidance method based on an escape algorithm, offering the following technical solution:

[0006] S1. Initialize the population to a set size. Individuals are described by multi-dimensional vectors with values ​​limited to corresponding upper and lower bounds. After evaluating fitness, sort them in ascending order and store the best individual in the elite pool. S2. When the number of iterations does not exceed the threshold, individuals are divided into three categories and a panic index is calculated. Calm individuals update rationally, compliant individuals follow, and panicked individuals explore irregularly, achieving diversified search. S3. After the number of iterations exceeds the threshold, enter the development stage. All individuals are considered calm. Combine the elite pool with the random individual positions to optimize and update in order to deeply explore the optimal solution region. S4. After the iteration termination condition is met, extract the optimal individual parameters from the elite pool, transform them into a formation obstacle avoidance path scheme, and output the guidance for the formation to cooperate in obstacle avoidance in complex environments.

[0007] Preferably, the specific content of S1 is as follows:

[0008] Initialize a population of size N, where each individual is represented by a D-dimensional vector X. i =(X i，1 X i，2 , ..., X i，D The description states that the value of the i-th individual in the j-th dimension is...

[0009] X i，j =LBj +r i，j ×(UB j -LB j )

[0010] Among them, UB j and LB j These are the upper and lower bounds of the j-th dimension, respectively, ensuring that the initial position of each individual is randomly distributed within the feasible space, r. i，j It is a random variable that is uniformly distributed between 0 and 1.

[0011] After initializing the population, the fitness of each individual is evaluated, and then the population is sorted in ascending order according to fitness. The individual with the best fitness is stored in the elite pool E, which represents the number of potential safe exits discovered by the population.

[0012] Preferably, the specific content of S2 is as follows:

[0013] When the number of iterations does not exceed T / 2, individuals are categorized into three classes: calm, compliant, or panicked. These categories correspond to different behavioral responses during the evacuation. At the beginning of each iteration, the panic index is calculated using the following formula.

[0014]

[0015] Where t is the current iteration number and T is the total number of iterations.

[0016] Individuals in the calm group behaved rationally, and the position update formula was:

[0017]

[0018] Among them, X i，j Let m1 be the value of the i-th individual in the j-th dimension, m1 be a binary variable, ω1 be an adaptive Levy weight, and C be the value of the j-th individual. j P(t) is the mean of the calm group on the j-th dimension, P(t) is the panic index, and the vector v is the mean of the calm group on the j-th dimension. c，j Defined as follows

[0019] v c，j =R c，j -X i，j +ε j

[0020] Among them, R c，j X is a randomly generated position within this dimension of the calm group. i，j It is the value of the i-th individual in the j-th dimension, ε j It is a slight adjustment in individual movement.

[0021] Individuals in the compliant group will follow the behavior of the calm and panicked groups; the position update formula is as follows.

[0022]

[0023] Among them, X i，j X is the value of the i-th individual in the j-th dimension, m1 and m2 are binary variables, ω1 and ω2 are adaptive Levy weights, and X... p，j Individuals are randomly selected from the panic group, P(t) is the panic index, and vector v is the panic index. h，j Defined as follows

[0024] v h，j =R h，j -X i，j +ε j

[0025] Among them, R h，j It conforms to the randomly generated position within that dimension of the group, X i，j It is the value of the i-th individual in the j-th dimension, ε j It is a slight adjustment in individual movement.

[0026] Panic-driven individuals exhibit more irregular exploration in the solution space, with the position update formula being:

[0027]

[0028] Among them, X i，j Let m1 and m2 be the value of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the value of the j-th individual. j X is an individual randomly selected from the elite pool. rand，j Individuals are randomly selected from the crowd, P(t) is the panic index, and the vector v is... h，j Defined as follows

[0029] v p，j =R p，j -X i，j +ε j

[0030] Among them, X i，j R is the value of the i-th individual in the j-th dimension. p,j It is a randomly generated position within this dimension of the panicked group, ε j It is a slight adjustment in individual movement.

[0031] Preferably, the specific content of S3 is as follows:

[0032] As the number of iterations exceeds T / 2, the algorithm enters the development phase. At this stage, all individuals are considered to be in a state of calm. The position information update formula for this phase is:

[0033]

[0034] Among them, X i，j Let m1 and m2 be the position of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the position of the i-th individual in the j-th dimension. j It is a position in the elite pool, X rand,j It is the position of an individual randomly selected from the crowd.

[0035] Preferably, the specific content of S4 is as follows:

[0036] Determine whether the iteration has reached the preset total number of iterations or met the convergence condition. If so, terminate the iteration. Extract the parameter information corresponding to the best individual from the elite pool and transform it into a specific path planning scheme for formation obstacle avoidance. Clarify the movement trajectory and obstacle avoidance nodes of each member of the formation and output the scheme to guide the formation to coordinate obstacle avoidance in complex environments, ensuring movement safety and efficiency.

[0037] Implementing the embodiments of the present invention will have the following beneficial effects: This method relies on the differentiated update mechanism of the escape algorithm, and through population classification and panic index regulation, it takes into account both global search diversity and local development depth, effectively avoiding the drawbacks of traditional formation obstacle avoidance algorithms that are prone to getting trapped in local optima. It can flexibly adapt to multi-agent formation scenarios, achieve efficient collaborative obstacle avoidance in complex dynamic environments, reduce the collision risk when the formation performs tasks, and provide reliable technical support for the engineering application of multi-agent systems. Attached Figure Description

[0038] To more clearly understand the embodiments of the present invention or the solutions of the prior art, the relevant drawings will be briefly described below. These drawings illustrate specific implementations of the present invention and are intended to provide necessary information to those skilled in the art, enabling them to obtain the relevant content without additional creative thinking. The drawings visually illustrate the structure, composition, and workflow of the present invention, helping to deeply understand the detailed aspects of the embodiments. The specific content is as follows:

[0039] Figure 1 This is a flowchart of a formation obstacle avoidance method based on an escape algorithm;

[0040] Figure 2 This is a simulation result diagram of the formation obstacle avoidance described in one embodiment. Detailed Implementation

[0041] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings of the embodiments. It should be emphasized that the following embodiments are merely specific exemplary cases of the present invention and do not cover all feasible implementations of the present invention. All other embodiments derived by those skilled in the art based on the embodiments of the present invention without creative effort should be included within the protection scope of the present invention. The specific method is shown below; see details. Figures 1-2 The method includes:

[0042] S1. Initialize the population to a set size. Individuals are described by multi-dimensional vectors with values ​​limited to corresponding upper and lower bounds. After evaluating fitness, sort them in ascending order and store the best individual in the elite pool. S2. When the number of iterations does not exceed the threshold, individuals are divided into three categories and a panic index is calculated. Calm individuals update rationally, compliant individuals follow, and panicked individuals explore irregularly, achieving diversified search. S3. After the number of iterations exceeds the threshold, enter the development stage. All individuals are considered calm. Combine the elite pool with the random individual positions to optimize and update in order to deeply explore the optimal solution region. S4. After the iteration termination condition is met, extract the optimal individual parameters from the elite pool, transform them into a formation obstacle avoidance path scheme, and output the guidance for the formation to cooperate in obstacle avoidance in complex environments.

[0043] In some specific embodiments, the specific content of S1 is as follows:

[0044] Initialize a population of size N, where each individual is represented by a D-dimensional vector X. i =(X i，1 X i，2 ,…,X i，D The description states that the value of the i-th individual in the j-th dimension is...

[0045] X i，j =LB j +r i，j ×(UB j -LB j )

[0046] Among them, UB j and LB j These are the upper and lower bounds of the j-th dimension, respectively, ensuring that the initial position of each individual is randomly distributed within the feasible space, r. i，j It is a random variable that is uniformly distributed between 0 and 1.

[0047] After initializing the population, the fitness of each individual is evaluated, and then the population is sorted in ascending order according to fitness. The individual with the best fitness is stored in the elite pool E, which represents the number of potential safe exits discovered by the population.

[0048] In some specific embodiments, the specific content of S2 is as follows:

[0049] When the number of iterations does not exceed T / 2, individuals are categorized into three classes: calm, compliant, or panicked. These categories correspond to different behavioral responses during the evacuation. At the beginning of each iteration, the panic index is calculated using the following formula.

[0050]

[0051] Where t is the current iteration number and T is the total number of iterations.

[0052] Individuals in the calm group behaved rationally, and the position update formula was:

[0053]

[0054] Among them, X i，j Let m1 be the value of the i-th individual in the j-th dimension, m1 be a binary variable, ω1 be an adaptive Levy weight, and C be the value of the j-th individual. j P(t) is the mean of the calm group on the j-th dimension, P(t) is the panic index, and the vector v is the mean of the calm group on the j-th dimension. c，j Defined as follows

[0055] v c，j =R c，j -X i，j +ε j

[0056] Among them, R c，j X is a randomly generated position within this dimension of the calm group. i，j It is the value of the i-th individual in the j-th dimension, ε j It is a slight adjustment in individual movement.

[0057] Individuals in the compliant group will follow the behavior of the calm and panicked groups; the position update formula is as follows.

[0058]

[0059] Among them, X i，j Let X be the value of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and X be the value of the j-th individual. p，j Individuals are randomly selected from the panic group, where P(t) is the panic index and the vector v is the panic index. h，j Defined as follows

[0060] v h，j =R h，j -X i，j +ε j

[0061] Among them, R h，j It conforms to the randomly generated position within that dimension of the group, X i，j It is the value of the i-th individual in the j-th dimension, ε i It is a slight adjustment in individual movement.

[0062] Panic-driven individuals exhibit more irregular exploration in the solution space, with the position update formula being:

[0063]

[0064] Among them, Xi，j Let m1 and m2 be the value of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the value of the i-th individual in the j-th dimension. j X is an individual randomly selected from the elite pool. rand，j Individuals are randomly selected from the crowd, P(t) is the panic index, and the vector v is... h，j Defined as follows

[0065] v p，j =R p，j -X i，j +ε j

[0066] Among them, X i，j R is the value of the i-th individual in the j-th dimension. p，j It is a randomly generated position within this dimension of the panicked group, ε j It is a slight adjustment in individual movement.

[0067] In some specific embodiments, the specific content of S3 is as follows:

[0068] As the number of iterations exceeds T / 2, the algorithm enters the development phase. At this stage, all individuals are considered to be in a state of calm. The position information update formula for this phase is:

[0069]

[0070] Among them, X i，j Let m1 and m2 be the position of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the position of the i-th individual in the j-th dimension. j It is a position in the elite pool, X rand，j It is the position of an individual randomly selected from the crowd.

[0071] In some specific embodiments, the specific content of S4 is as follows:

[0072] Determine whether the iteration has reached the preset total number of iterations or met the convergence condition. If so, terminate the iteration. Extract the parameter information corresponding to the best individual from the elite pool and transform it into a specific path planning scheme for formation obstacle avoidance. Clarify the movement trajectory and obstacle avoidance nodes of each member of the formation and output the scheme to guide the formation to coordinate obstacle avoidance in complex environments, ensuring movement safety and efficiency.

[0073] Figure 2The image shows a simulation result of the formation obstacle avoidance described in one embodiment. Based on the motion characteristics of the unmanned surface vessel and the requirements of the optimization algorithm, the key parameters, initial state, and simulation conditions of the system are set as follows: The formation consists of one center vessel and two follower vessels, with an initial position of (0, 0) and a target position of (100, 100). There are four static obstacles at positions (45, 10), (60, 35), (20, 70), and (80, 60). There are three dynamic obstacles: obstacle 1 starts at (10, 20) and ends at (70, 20); obstacle 2 starts at (10, 40) and ends at (80, 80); obstacle 3 starts at (70, 70) and ends at (30, 70). The maximum simulation duration is 100 seconds, and the time step is 1 second.

[0074] This invention addresses the problems of insufficient diversity in multi-agent formation obstacle avoidance search under complex dynamic environments, susceptibility to local optima, and low collaborative efficiency. It provides a formation obstacle avoidance method based on an escape algorithm. This method initializes a large-scale population and constructs an elite pool to retain the best individuals. By combining a panic index, individuals are divided into three categories for differentiated updates to ensure global search. After iteration exceeds the limit, the method enters a development phase to further explore the optimal solution region, ultimately outputting a highly efficient collaborative formation obstacle avoidance path scheme.

[0075] The above description is merely a preferred embodiment, intended to clarify the technical concept and not to limit the scope of application. The scope of protection of this invention covers various implementation forms and technical solutions. The technical solutions described herein are only specific examples of the core principles and not the only implementation path. Without departing from the core principles of this invention, improvements to existing solutions, parameter adjustments, expansion of subdivided scenarios, and other implementation methods that conform to the principles are all included in the scope of protection of this invention. All modifications, alterations, and derivative solutions that do not depart from the core idea fall within the technical system of this invention.

Claims

1. A formation obstacle avoidance method based on an escape algorithm, characterized in that... It includes the following steps: S1. Initialize the population of a set size. Individuals are described by multi-dimensional vectors with values ​​limited to corresponding upper and lower bounds. After evaluating fitness, sort them in ascending order and store the best individuals in the elite pool. S2. When the number of iterations does not exceed the threshold, individuals are divided into three categories and the panic index is calculated. Calm individuals update rationally, compliant individuals follow, and panicked individuals explore irregularly to achieve diversified search. S3. After the number of iterations exceeds the threshold, the development phase begins. All individuals are considered to be in a calm state. The optimization and update are combined with the elite pool and the random individual positions to deeply explore the optimal solution region. S4. After the iteration termination condition is met, the optimal individual parameters are extracted from the elite pool, transformed into a formation obstacle avoidance path scheme, and output to guide the formation to collaboratively avoid obstacles in complex environments.

2. The formation obstacle avoidance method based on an escape algorithm according to claim 1, characterized in that: The specific content of S1 is as follows: Initialize a population of size N, where each individual is represented by a D-dimensional vector X. i =(X i，1 X i,2 ,…,X i，D The description states that the value of the i-th individual in the j-th dimension is X. i，j =LB j +r i，j ×(UB j -LB j ) Among them, UB j and LB j These are the upper and lower bounds of the j-th dimension, respectively, ensuring that the initial position of each individual is randomly distributed within the feasible space, r. i，j It is a random variable that is uniformly distributed between 0 and 1; After initializing the population, the fitness of each individual is evaluated, and then the population is sorted in ascending order according to fitness. The individual with the best fitness is stored in the elite pool E, which represents the number of potential safe exits discovered by the population.

3. The formation obstacle avoidance method based on an escape algorithm according to claim 1, characterized in that: The specific content of S2 is as follows: When the number of iterations does not exceed T / 2, individuals are categorized into three classes: calm, compliant, or panicked. These categories correspond to different behavioral responses during the evacuation. At the beginning of each iteration, the panic index is calculated using the following formula. Where t is the current iteration number and T is the total number of iterations; Individuals in the calm group behaved rationally, and the position update formula was: Among them, X i，j Let m1 be the value of the i-th individual in the j-th dimension, m1 be a binary variable, ω1 be an adaptive Levy weight, and C be the value of the j-th individual. j P(t) is the mean of the calm group on the j-th dimension, P(t) is the panic index, and the vector v is the mean of the calm group on the j-th dimension. c，j Defined as follows v c，j =R c，j -X i，j +ε j Among them, R c，j X is a randomly generated position within this dimension of the calm group. i，j It is the value of the i-th individual in the j-th dimension, ε j It is a slight adjustment in individual movement; Individuals in the compliant group will follow the behavior of the calm and panicked groups; the position update formula is as follows. Among them, X i，j Let X be the value of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and X be the value of the j-th individual. p，j Individuals are randomly selected from the panic group, where P(t) is the panic index and the vector v is the panic index. h，j Defined as follows v h，j =R h，j -X i，j +ε j Among them, R h，j It conforms to the randomly generated position within that dimension of the group, X i，j It is the value of the i-th individual in the j-th dimension, ε j It is a slight adjustment in individual movement; Panic-driven individuals exhibit more irregular exploration in the solution space, with the position update formula being: Among them, X i，j Let m1 and m2 be the value of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the value of the i-th individual in the j-th dimension. j X is an individual randomly selected from the elite pool. rand，j Individuals are randomly selected from the crowd, P(t) is the panic index, and the vector v is... h，j Defined as follows v p，j =R p，j -X i，j +ε j Among them, X i，j R is the value of the i-th individual in the j-th dimension. p，j It is a randomly generated position within this dimension of the panicked group, ε j It is a slight adjustment in individual movement.

4. The formation obstacle avoidance method based on an escape algorithm according to claim 1, characterized in that: The specific content of S3 is as follows: As the number of iterations exceeds T / 2, the algorithm enters the development phase. At this stage, all individuals are considered to be in a state of calm. The position information update formula for this phase is: Among them, X i，j Let m1 and m2 be the position of the i-th individual in the j-th dimension, m1 and m2 be binary variables, ω1 and ω2 be adaptive Levy weights, and E be the position of the i-th individual in the j-th dimension. j It is a position in the elite pool, X rand,j It is the position of an individual randomly selected from the crowd.

5. The formation obstacle avoidance method based on an escape algorithm according to claim 1, characterized in that: The specific content of S4 is as follows: Determine whether the iteration has reached the preset total number of iterations or met the convergence condition. If so, terminate the iteration. Extract the parameter information corresponding to the best individual from the elite pool and transform it into a specific path planning scheme for formation obstacle avoidance. Clarify the movement trajectory and obstacle avoidance nodes of each member of the formation and output the scheme to guide the formation to coordinate obstacle avoidance in complex environments, ensuring movement safety and efficiency.

Images