Path planning method based on improved fish optimization algorithm

Abstract

The invention belongs to the field of agricultural intellectualization, and particularly relates to a path planning method based on an improved fish optimization algorithm, which has the capability of performing path planning in an orchard which is a relatively complex environment, so that the spatial distance of a path is shorter, and the efficiency of path planning is improved. The method comprises the steps that an orchard environment grid map is generated according to orchard environment data which are surveyed and mapped in advance, a moving task of an orchard robot from a starting point to a target point is designed, Logistic-Tent chaotic mapping and a crisscross strategy are introduced to improve a fish optimization algorithm, path planning is carried out based on the improved fish optimization algorithm, and a path track is obtained. According to the method, the spatial distance of the planned path can be shortened in the complex orchard environment, the path planning efficiency is improved, and therefore the working efficiency of the orchard robot is improved.

Description

Technical Field

[0001] This invention belongs to the field of agricultural intelligence, specifically involving a path planning method based on an improved remora optimization algorithm, applied to intelligent robot path planning in orchard environments.

[0002] Key Project of Jiangsu Provincial Key Research and Development Program (Industrial Foresight and Key Core Technologies) (Project No.: BE2021016). Background Technology

[0003] With the advancement of agricultural modernization, orchard robots are playing an increasingly important role in fruit picking and inspection. However, traditional path planning methods often face problems such as slow search speed and easy getting trapped in local optima. Existing research mainly focuses on the application of single algorithms and lacks effective handling of the complex environment of orchards. Therefore, there is an urgent need for a new path planning algorithm to improve the working efficiency and path planning accuracy of orchard robots.

[0004] Currently, path planning algorithms can be broadly classified into two categories: global path planning and local path planning. Global path planning typically relies on a pre-obtained environment map and uses methods such as graph search and optimization to find the optimal path.

[0005] Patent CN202411844884.9 proposes a robot global path planning method and system based on probability interval division. Based on the obstacle area and total obstacle area within each equally divided interval, it calculates the sampling probability within the interval based on obstacles, determines the positions of the start and end points and the equally divided intervals containing the start and end points, sets the sampling probability within the intervals based on position between the equally divided intervals containing the start and end points, combines the sampling probabilities within the intervals based on obstacles and positions to obtain the sampling probability within the region, selects nodes based on the RRT* algorithm, samples the equally divided intervals using a variable step size method, iteratively generates an initial path, optimizes the generated initial path, and obtains the robot's global path. This improves the speed of generating the initial path from random numbers and enhances the algorithm's purposefulness and directionality.

[0006] Patent CN202410663985.X proposes a dynamic planning method for transport vehicles. By constructing a weighted directed graph, it generates global path planning using key nodes, predicts potential traffic congestion points, and adjusts driving strategies in a timely manner to improve the flexibility of planning. However, traditional global path planning algorithms often cannot provide sufficient adaptability when facing complex orchard environments and require a lot of computing resources. In recent years, path planning methods based on intelligent optimization algorithms have gradually attracted attention. Many intelligent optimization algorithms have been applied to the field of path planning. They search for the optimal path by simulating the optimization mechanism in nature.

[0007] Patent CN202410336368.9 proposes a path planning method for wheeled plant protection robots. First, orchard environmental information is acquired through lidar. The point cloud density is reduced by applying a voxel grid method. The ground point cloud is segmented using a grid method, and the robot's inter-row passage area is extracted by clustering. Then, combined with the kinematic model of the plant protection robot and the constraints of the operation specifications, a series of candidate trajectory sets are generated by using a model prediction method. This improves the traditional ant colony algorithm-based path planning by incorporating the robot's passage cost into the node search, thereby effectively improving the algorithm's operating efficiency. However, these methods have some shortcomings in practical applications, such as being prone to getting trapped in local optima, slow convergence speed, and low path trajectory accuracy.

[0008] The existing technology proposes a remora optimization algorithm, which mainly simulates the remora attaching to hosts of different sizes to complete the foraging process. This algorithm covers the sailfish optimization algorithm and the whale optimization algorithm. By simulating the remora's attachment to the host, dynamic switching of the host and predation strategy, it achieves a balance between global and local search capabilities, thereby effectively solving the path planning problem in complex environments.

[0009] To address the shortcomings of existing algorithms in orchard path planning, this invention proposes a path planning method based on an improved remora optimization algorithm. By improving the remora optimization algorithm, this invention can effectively avoid getting trapped in local optima, improve global search capabilities, and thus plan a shorter path, thereby improving the working efficiency of orchard robots. Summary of the Invention

[0010] To improve the working efficiency of orchard robots, this invention proposes a path planning method based on an improved harbinger optimization algorithm. This algorithm combines Logistic-Tent chaotic mapping and cross-path strategy to optimize the path planning process of orchard robots.

[0011] The technical solution of the present invention is:

[0012] A path planning method based on an improved remora optimization algorithm includes the following steps:

[0013] S1. Modeling is done using real orchard scene data. LiDAR and camera sensors are used to acquire orchard terrain and obstacle information. The collected environmental data is transformed into a model that can be used for path planning. A two-dimensional map containing fruit trees, obstacles and passable areas is created. The environment is processed using the grid method, dividing the real scene into several grids of the same size to facilitate the application of the path planning algorithm. Obstacle areas are marked as black grids and passable areas are marked as white grids.

[0014] S2. In the modeled orchard environment, the user uses a mobile device to set the robot's starting position in the orchard. Usually, an easily accessible area is selected, such as the orchard entrance or the robot's parking position. The coordinate information of the starting point is recorded. The target position of the robot in the orchard is set. Usually, an area where the operation needs to be carried out is selected, such as a specific fruit tree or the other end of the orchard. At the same time, it is ensured that the target point is located in a passable area and that the point on an obstacle is avoided. The coordinate information of the target point is recorded.

[0015] S3. Design the cost function to ensure that the path is as short as possible without encountering obstacles;

[0016] The cost function for the orchard path can then be defined as follows:

[0017]

[0018] Where i is the iteration number, and L(i) is the straight-line distance moved in each iteration, as shown in the formula:

[0019]

[0020] Where, x i ,y i x represents the current location. i-1 ,y i-1 Given the position before iteration and g as the grid precision, the path planning problem is transformed into finding the minimum cost function, as shown in the formula:

[0021] min f.

[0022] S4. Path search is performed using an improved remora optimization algorithm. Logistic-Tent chaotic mapping is introduced to initialize the population and cross-path strategy into the traditional remora optimization algorithm. In each iteration, the fitness of the current path is evaluated and the path length and obstacle avoidance ability are calculated.

[0023] S4.1. Logistic-Tent chaotic mapping is used to improve population diversity and enhance the algorithm's global search capability. The expression for the Logistic-Tent chaotic mapping sequence is shown in the formula:

[0024]

[0025] S4.2 Improve the traditional remora optimization algorithm by using a cross-sectional strategy, including randomly selecting two dimensions for cross-sectional operations to enhance the algorithm's global search capability;

[0026] In a complex environment like an orchard, robot path planning needs to consider not only the terrain's complexity but also the tree distribution and the specific requirements of the task. The cross-sectional strategy includes both lateral and longitudinal cross-sectional approaches. Using a lateral cross-sectional strategy for population search can effectively reduce blind spots and enhance the algorithm's global search performance. To employ this strategy, individuals in the population are first randomly paired, and then the paired individuals are crossed laterally. Assuming SM... i1 and SM i2 It is the previous generation of individuals after pairing, and its offspring are and The resulting formula is shown below:

[0027]

[0028] Among them SM i1j and SM i2j They represent SM respectively i1 and SM i2 The j-th dimension, j = 1, 2, ..., D, and They are and In the j-th dimension of the offspring generated by the horizontal crossover at the j-th position, r1 and r2 are random numbers uniformly distributed in the range (0,1). Then the generated offspring are compared with the previous generation, and individuals with smaller objective function values ​​are retained.

[0029] Traditional remora optimization algorithms may converge prematurely due to stagnation in certain dimensions of the population. Vertical crossover effectively addresses this issue, helping the algorithm escape local optima. The vertical crossover strategy involves crossover operations on all individuals across two different dimensions. In the later stages of the remora optimization algorithm, it is prone to getting trapped in local optima, largely because some individuals in the population reach local optima in specific dimensions, causing premature convergence. Analysis reveals that the remora optimization algorithm lacks a necessary mutation mechanism, failing to effectively intervene in individuals already in local optima, thus hindering further exploration of the global optimum. Therefore, after performing horizontal operations, vertical crossover should be introduced to effectively prevent the algorithm from getting trapped in local optima prematurely and enhance its global search capability. When performing vertical operations, two dimensions j are randomly selected. 1 and j 2 His descendants The jth 1 The dimensions are shown in the formula below; other dimensions are the same as the previous generation SM. i Keep it the same:

[0030]

[0031] Where r is a random number uniformly distributed in the range (0,1), the generated offspring are compared with their previous generation, and individuals with smaller objective function values ​​are retained;

[0032] During the iteration process, if an individual's dimension successfully escapes the local optimum through vertical crossover, that dimension will immediately be shared with other individuals in the population through horizontal crossover. This not only updates the dimension but may also provide opportunities for other dimensions that are still in the local optimum to escape the local optimum. Such crossover operations can effectively promote the widespread dissemination of information in the population, thereby enhancing the algorithm's global search capability. The combination of horizontal and vertical crossover operations can significantly improve the algorithm's convergence accuracy and convergence efficiency.

[0033] S5. Based on the algorithm results, generate the final path planning scheme and output the path coordinates and path length.

[0034] Beneficial effects:

[0035] This method addresses the challenges of complex orchard environments and the low efficiency of traditional path planning algorithms. By introducing Logistic-Tent chaotic mapping and a cross-sectional strategy, the IROA algorithm is improved, significantly enhancing its global search capability. Compared to traditional algorithms, the improved IROA demonstrates superior performance in orchard robot path planning, exhibiting faster convergence speed and higher planning accuracy.

[0036] The core process is as follows: First, generate an orchard environment raster map based on pre-surveyed orchard environmental data; second, design the orchard robot's movement task from the starting point to the target point; and then, integrate the application...

[0037] The improved remora optimization algorithm based on Logistic-Tent chaotic mapping and cross-sectional strategy is used for path search. Finally, the optimized path trajectory is output. In summary, the improved remora optimization algorithm path planning method proposed in this invention can effectively shorten the spatial distance of the planned path in the complex environment of orchards, significantly improve the efficiency of path planning, and thus improve the overall working efficiency of orchard robots. Attached Figure Description

[0038] Figure 1 This is a flowchart of a path planning method based on an improved remora optimization algorithm;

[0039] Figure 2 This is an example of an orchard environment raster map based on a path planning method using an improved remora optimization algorithm;

[0040] Figure 3 This is a schematic diagram of the robot's movable directions based on a path planning method using an improved remora optimization algorithm.

[0041] Figure 4 This is a path planning result in a modeled orchard environment, based on an improved remora optimization algorithm. Detailed Implementation

[0042] To enhance understanding of the present invention, the invention will be further described in detail below with reference to embodiments and accompanying drawings. These embodiments are only for explaining the invention and do not constitute a limitation on the scope of protection of the invention.

[0043] The specific implementation of the present invention mainly includes steps such as environment modeling, setting the starting point and the target point, designing the cost function, and implementing the path search algorithm.

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, such as... Figure 1 , 2 As shown in points 3 and 4, the details are as follows:

[0045] S1. Environmental Modeling: Environmental data of the orchard, including obstacle information, is collected using sensor devices such as LiDAR and cameras. The raw data acquired by the sensors is preprocessed to remove noise and unnecessary information to ensure data accuracy. The processed data is then converted into a two-dimensional raster map. The environment is divided into several raster of the same size using a raster method. Each raster is marked as a passable area (0) or an impassable area (1), forming a raster map array, such as... Figure 2 As shown;

[0046] S2. Setting the starting and target points: Select the robot's starting position in the orchard, ensuring this position is within a passable area. Record the coordinates of the starting point. Determine the robot's target position, again ensuring this position is within a passable area. Record the coordinates of the target point. In this embodiment, (1,1) is selected as the starting point and (20,20) as the target point. The orchard robot's movable directions are as follows: Figure 3 As shown;

[0047] S3. Design the cost function: The path should be as short as possible without touching obstacles. Therefore, the cost function for the orchard path can be defined as follows:

[0048]

[0049] Where i is the iteration number, and L(i) is the straight-line distance moved in each iteration, as shown in the formula:

[0050]

[0051] Where, x i ,y i x represents the current location. i-1 ,y i-1Given the position before iteration and g as the grid precision, the path planning problem is transformed into finding the minimum cost function, as shown in the formula:

[0052] min f.

[0053] S4. Path Search Algorithm Implementation: An improved remora optimization algorithm, incorporating Logistic-Tent chaotic mapping for population initialization and a cross-path optimization strategy, is used to implement path planning for the orchard robot. In each iteration, the fitness of the current path is evaluated, and the path length is calculated. In this embodiment, for the improved remora optimization algorithm, the initial population size is set to 30, and the maximum number of iterations is set to 500. The path planning results are as follows: Figure 4 As shown;

[0054] S5. Output Results: Based on the algorithm results, the final path planning scheme is generated, and the path length is output. In this embodiment, the path trajectory scheme is as follows: Figure 4 As shown, the path length is 28.5312.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A path planning method based on an improved remora optimization algorithm, characterized in that, Includes the following steps: S1: Modeling is done using real orchard scene data. LiDAR and camera sensors are used to acquire information about the orchard environment and obstacles. The collected environmental data is transformed into a model that can be used for path planning, creating a two-dimensional map that includes fruit trees, obstacles and passable areas. S2: In the modeled orchard environment, the user uses a mobile device to determine the robot's starting position in the orchard and records the coordinate information of the starting point. The user then determines the robot's target position in the orchard and records the coordinate information of the target point. S3: Design a cost function to ensure that the path is as short as possible without touching obstacles; S4: Use the improved remora optimization algorithm to search for paths, and in each iteration, evaluate the fitness value of the current path and calculate the path length and obstacle avoidance capability. S5: Based on the algorithm results, generate the final path planning scheme and output the path coordinates and path length.

2. The path planning method based on the improved remora optimization algorithm according to claim 1, characterized in that, In step S3, the cost function is designed to minimize the path length and avoid obstacles to ensure the feasibility of the path. Based on the description of the orchard path planning problem, the designed cost function should satisfy the requirement that the path is as short as possible without encountering obstacles. Therefore, the cost function for the orchard path can be defined as follows: Where i is the iteration number, and L(i) is the straight-line distance moved in each iteration, as shown in the formula: Where, x i ,y i x represents the current location. i-1 ,y i-1 Given the position before iteration and g as the grid precision, the path planning problem is transformed into finding the minimum cost function, as shown in the formula: min f.

3. The path planning method based on the improved remora optimization algorithm according to claim 1, characterized in that, In step S4, the traditional remora optimization algorithm is improved by using Logistic-Tent chaotic mapping to initialize the population and a cross-cutting strategy. Specifically: S4.

1. Logistic-Tent chaotic mapping is used to improve population diversity and enhance the algorithm's global search capability. The expression for the Logistic-Tent chaotic mapping sequence is shown in the formula: S4.2 Improve the traditional remora optimization algorithm by using a cross-sectional strategy, including randomly selecting two dimensions for cross-sectional operations to enhance the algorithm's global search capability; In a complex environment like an orchard, robot path planning needs to consider not only the terrain's complexity but also the tree distribution and the specific requirements of the task. The cross-sectional strategy includes both lateral and longitudinal cross-sectional approaches. Using a lateral cross-sectional strategy for population search can effectively reduce blind spots and enhance the algorithm's global search performance. To employ this strategy, individuals in the population are first randomly paired, and then the paired individuals are crossed laterally. Assuming SM... i1 and SM i2 It is the previous generation of individuals after pairing, and its offspring are and The resulting formula is shown below: Among them SM i1j and SM i2j They represent SM respectively i1 and SM i2 The j-th dimension, j = 1, 2, ..., D, and They are and In the j-th dimension of the offspring generated by the horizontal crossover at the j-th position, r1 and r2 are random numbers uniformly distributed in the range (0,1). Then the generated offspring are compared with the previous generation, and individuals with smaller objective function values ​​are retained. Traditional remora optimization algorithms may converge prematurely due to stagnation in certain dimensions of the population. Vertical crossover effectively addresses this issue, helping the algorithm escape local optima. The vertical crossover strategy involves crossover operations on all individuals across two different dimensions. In the later stages of the remora optimization algorithm, it is prone to getting trapped in local optima, largely because some individuals in the population reach local optima in specific dimensions, causing premature convergence. Analysis reveals that the remora optimization algorithm lacks a necessary mutation mechanism, failing to effectively intervene in individuals already in local optima, thus hindering further exploration of the global optimum. Therefore, after performing horizontal operations, vertical crossover should be introduced to effectively prevent the algorithm from getting trapped in local optima prematurely and enhance its global search capability. When performing vertical operations, two dimensions j are randomly selected. 1 and j 2 His descendants The jth 1 The dimensions are shown in the formula below; other dimensions are the same as the previous generation SM. i Keep it the same: Where r is a random number uniformly distributed in the range (0,1), the generated offspring are compared with their previous generation, and individuals with smaller objective function values ​​are retained; During the iteration process, if an individual's dimension successfully escapes the local optimum through vertical crossover, that dimension will immediately be shared with other individuals in the population through horizontal crossover. This not only updates the dimension but may also provide opportunities for other dimensions that are still in a local optimum to escape the local optimum. Such crossover operations can effectively promote the widespread dissemination of information in the population, thereby enhancing the algorithm's global search capability. The combination of horizontal and vertical crossover operations can significantly improve the algorithm's convergence accuracy and convergence efficiency.

Images