Multi-target path planning method and device based on improved mantis optimization algorithm

Abstract

The invention relates to the field of robot target path planning, and discloses a multi-target path planning method and device based on an improved mantis optimization algorithm, and the method comprises the steps: improving the mantis optimization algorithm, and building a grid environment map and a multi-target fitness function, an obstacle avoidance function and a smooth path function corresponding to path planning; the population position optimization corresponds to path nodes in the grid environment map, iteration is performed on all individuals of the population, and fitness values are formed; calculating comprehensive fitness by using a multi-target fitness function to obtain an updated global optimal position and a corresponding optimal fitness value; completing path planning of a multi-target fitness function; and observing the convergence curve graph and the shortest path graph, and verifying the feasibility of multi-target path planning. According to the method, the convergence speed of the algorithm is improved, the global and local search capabilities are balanced, the shortest path length is effectively reduced, path redundant nodes and path inflection points are reduced, and the path quality and efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the field of robot target path planning, and specifically relates to a multi-target path planning method and device based on an improved mantis optimization algorithm. Background Art

[0002] With the rapid development of artificial intelligence and robotics, the application of mobile robots in industrial automation, intelligent manufacturing and unmanned driving has been expanding. However, path planning in complex environments is still a key issue in mobile robot research. The quality of its planning is directly related to the work efficiency and safety of the robot. To address this problem, meta-heuristic algorithms based on swarm intelligence are proposed, among which the Dung Beetle Optimization (DBO) algorithm has attracted much attention due to its simplicity and efficiency. DBO simulates the behaviors of mantises such as rolling, dancing, and reproduction to solve complex optimization problems, and shows potential in path planning. However, the DBO algorithm has deficiencies in global search capabilities and path planning stability, and is particularly prone to falling into local optimality, which limits its application in complex environments. Therefore, it is of great significance to study and improve the DBO algorithm to enhance its global search and path planning capabilities.

[0003] Traditional path planning algorithms such as A* algorithm, Dijkstra algorithm, and RRT algorithm work well in static environments, but only consider the shortest path length and lack consideration of many constraints, which makes them inadequate in dynamic and complex environments. Experts have found that meta-heuristic algorithms are more suitable for solving this type of multi-constraint problem. These algorithms are inspired by group behavior in nature and solve complex optimization problems by simulating group behavior in nature. In recent years, various meta-heuristic algorithms have been born, such as Black Hawk Optimization Algorithm (BEO), Black Kite Optimization Algorithm (BKA), Pelican Optimization Algorithm (POA), Zebra Optimization Algorithm (ZOA), and Ivy Optimization Algorithm (IVYA). These algorithms are highly flexible and robust. Although they are often widely used in various fields to reduce the risk of local optimality based on the improvement of existing population intelligent optimization algorithms, their convergence accuracy and stability in complex environments still need to be improved. At present, the improvement of intelligent optimization algorithms has achieved initial results and has been initially applied in algorithm performance and engineering applications, but they still face many challenges in terms of accuracy and convergence speed. In view of the shortcomings of the DBO algorithm, such as unstable global search and poor path planning effect, Figure 1 As shown, firstly, since the population initialization stage is randomly generated, this may lead to uneven population distribution, poor ergodicity, and slow convergence; secondly, the algorithm's global search ability and local development ability are insufficient, and it is easy to fall into local optimality, affecting the convergence performance of the algorithm. Summary of the invention

[0004] In order to solve the problems of the prior art, the present invention is based on the improvement strategy of the metaheuristic algorithm, and applies it in the robot simulation experiment, and proposes a multi-objective path planning method and device based on the improved mantis optimization algorithm, which enhances the global search and path planning capabilities. The simulation experiment and rank sum test, as well as the time complexity analysis, verify the effectiveness and feasibility of the IDBO algorithm.

[0005] The present invention is implemented by adopting the following technical scheme: a multi-objective path planning method based on an improved mantis optimization algorithm, comprising:

[0006] The original mantis algorithm is optimized for path length to obtain the improved mantis optimization algorithm;

[0007] Establish a grid environment map based on the working environment and obstacle distribution information;

[0008] Establish multi-objective fitness function, obstacle avoidance function and smooth route function corresponding to path planning;

[0009] The mantis population position optimization of the improved mantis optimization algorithm corresponds to the path nodes in the grid environment map, where the individual coordinates in the mantis population correspond to the node coordinates of the path planning;

[0010] According to the improved mantis optimization algorithm, all individuals in the population are iterated, and the comprehensive fitness of each iteration is calculated using the multi-objective fitness function to obtain the updated global optimal position and the corresponding optimal fitness value, and record and compare them;

[0011] The global optimal position of the individual is used as the node position of the grid map path planning, and the node starting point, each node, and the node end point of the path planning are connected in sequence. The path length is obtained by calculating the Euclidean distance between each adjacent path node to complete the path planning of the multi-objective fitness function.

[0012] Observe the convergence curves and shortest path graphs of each algorithm to verify the feasibility of multi-objective path planning.

[0013] Preferably, the improved mantis optimization algorithm comprises: using a refraction reverse learning mechanism to construct an initialization population, setting a maximum number of iterations and related parameters of the population, selecting an elite selection strategy to sort the initialization population, and screening to obtain a high-quality population;

[0014] The Osprey optimization algorithm is selected and an adaptive weight mechanism is added to calculate the individual fitness value and the global optimal position, output the global optimal position and the corresponding individual fitness value of the rolling ball, and couple the position update formula as the global optimal solution; update the individual position information of the rolling ball according to the obstacle mode;

[0015] The dynamic selection strategy and adaptive t distribution are used, the number of iterations is used as the degree of freedom parameter for perturbation, the path is iteratively optimized through the mantis algorithm, and the position information of the breeding individuals is updated;

[0016] According to the corresponding formula, the Jacobian curve and the composite fitness function are selected to find the local optimal solution in the grid environment map, improve the smoothness of the foraging individual path movement in the search space, and update the foraging individual position information;

[0017] According to the original mantis algorithm, the location information of all stealing individuals is updated to obtain the location of the stealing population;

[0018] Perform a greedy selection strategy update and evaluation on the fitness values ​​of all individuals in the current population, and the position of the mantis increases by one after iteration;

[0019] Repeat the above steps 1 to 5 until the number of iterations is equal to the maximum number of iterations and converges to the optimal value of the path comprehensive score; wherein the composite fitness function is a fitness value that is weighted by combining the results of the multi-objective fitness functions.

[0020] Preferably, the elite selection strategy is used to sort and screen the initialized population to obtain a high-quality population, including: first randomly initializing the mantis population A, setting the maximum number of iterations and related parameters of the population, and then generating the population B through refraction reverse learning, merging and sorting the fitness values ​​of the two groups of populations, and finally selecting the top N best individuals to form a new population X.

[0021] Preferably, the selection of dynamic selection strategy and adaptive t distribution includes: adjusting the reproduction strategy by calculating the probability according to the current population characteristics, and completing the generation method and distribution pattern adjustment of new reproduction individuals.

[0022] Preferably, the establishing of the grid environment map includes: establishing a simple grid environment map and a complex grid environment, initializing relevant parameters of the grid map path planning, and specifying the starting point and the ending point of the environment grid map.

[0023] Preferably, the establishment of a multi-objective fitness function corresponding to the path planning includes:

[0024] Path safety function: The path safety is determined by calculating the minimum distance between two adjacent path segments and obstacles and setting distance constraints;

[0025] Path smoothness function: Use the cosine theorem to calculate the turning angles (cosine values) between adjacent path nodes, and make the path closer to a straight line by minimizing the angle of each turning point;

[0026] Path length function: The path length is minimized by summing the Euclidean distances between adjacent nodes in the path, and the shortest path length of all nodes is used to reduce the moving time.

[0027] Preferably, the multi-objective fitness function corresponding to the path planning is established, and a final comprehensive fitness value is formed by weighted calculation of the results of the path safety function, the smoothness function and the path length function.

[0028] In a second aspect, an embodiment of the present invention proposes a multi-objective path planning device based on an improved mantis optimization algorithm, comprising: an improvement module for optimizing the path length of the original mantis algorithm to obtain an improved mantis optimization algorithm;

[0029] The grid environment module is used to establish a grid environment map based on the working environment and obstacle distribution information.

[0030] Function module, used to establish multi-objective fitness function, obstacle avoidance function and smooth route function corresponding to path planning;

[0031] The node correspondence module is used to correspond the mantis population position optimization in the improved mantis optimization algorithm to the path nodes in the grid environment map, where the individual coordinates in the mantis population correspond to the node coordinates of the path planning;

[0032] The calculation module is used to iterate all individuals in the population according to the improved mantis optimization algorithm, calculate the comprehensive fitness of each iteration using the multi-objective fitness function, obtain the updated global optimal position and the corresponding optimal fitness value, and record and compare;

[0033] The node connection module is used to use the individual global optimal position as the node position of the grid map path planning, connect the node starting point of the path planning, each node, and the node end point of the path planning in sequence, and calculate the Euclidean distance between each adjacent path node to obtain the path length, thus completing the path planning of the multi-objective fitness function;

[0034] The simulation verification module is used to observe the convergence curve graph and shortest path graph of each algorithm and verify the feasibility of multi-objective path planning.

[0035] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a multi-objective path planning method based on an improved mantis optimization algorithm as provided in any embodiment of the present invention.

[0036] The present invention is applied to multi-objective path planning of mobile robots. Compared with the prior art, it can improve the algorithm convergence speed, balance the global and local search capabilities, effectively reduce the shortest path length, reduce the redundant nodes of the path, reduce the inflection points of the path, and improve the path quality and efficiency. A hybrid strategy is added to the original mantis algorithm, and the refraction direction learning is added to construct the initial population in the initialization stage, and the elite selection strategy is adopted to screen and optimize the population to improve the population diversity and algorithm convergence speed; the osprey optimization algorithm is added to the mantis rolling ball stage and coupled with the optimal solution to enhance the algorithm's global search capability, and a dynamic selection strategy and an adaptive t distribution are introduced in the mantis breeding stage to balance the algorithm's global exploration and local development capabilities. The Jacobian curve is introduced in the mantis foraging stage to enhance the algorithm's ability to jump out of the local optimum and avoid falling into the local optimum. The feasibility of the algorithm is verified by comparing the optimized algorithm with the original algorithm and other swarm intelligence optimization algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flowchart of the original Mantis optimization algorithm.

[0038] Figure 2 The present invention is a flow chart of a multi-objective path planning method based on an improved mantis optimization algorithm.

[0039] Figure 3 The invention discloses a simple grid environment map for a multi-objective path planning method based on an improved mantis optimization algorithm.

[0040] Figure 4 The invention discloses a complex grid environment map for a multi-objective path planning method based on an improved mantis optimization algorithm.

[0041] Figure 5 This is an example diagram of the results of simple grid environment map path planning based on the multi-objective path planning method of the present invention based on the improved mantis optimization algorithm and other algorithms.

[0042] Figure 6 This is an example diagram of the results of complex grid environment map path planning using the multi-objective path planning method based on the improved mantis optimization algorithm of the present invention and other algorithms.

[0043] Figure 7 This is an algorithm iteration comparison diagram of a simple grid map of the multi-objective path planning method based on the improved mantis optimization algorithm of the present invention.

[0044] Figure 8 This is a comparison diagram of algorithm iterations of complex grid maps of the multi-objective path planning method based on the improved mantis optimization algorithm of the present invention.

[0045] Fig. 9 2 is a schematic diagram of the structure of a multi-objective path planning device of an improved mantis optimization algorithm according to an embodiment of the present invention;

[0046] Fig.10 It is a structural schematic diagram of a multi-objective path planning device of an improved mantis optimization algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0048] Embodiment 1 The present invention proposes an improved mantis optimization algorithm, which is divided into population initialization, rolling mantis stage, breeding mantis stage, foraging mantis stage, and stealing mantis stage. It includes:

[0049] Step 1: Constructing the initial population X includes, in the initialization stage, using the refraction reverse learning mechanism to construct the initialization population, setting the maximum number of iterations and related parameters of the population, selecting the elite selection strategy to sort and screen the initialization population to obtain a high-quality population; enhancing the global search capability to improve the convergence accuracy and speed.

[0050] Refraction reverse learning mechanism: used in the population initialization stage to improve the diversity and randomness of the initial population, avoid early convergence problems, and thus more effectively explore paths in complex map environments. Refraction reverse learning (Refracted Opposition-Based Learning, referred to as ROBL) is a new algorithm improvement mechanism that comes from the refraction phenomenon of light in a medium. On the basis of reverse learning, the path of light in the medium is used to guide the search process to find the optimal solution. It has good global search capabilities and fast convergence. In the optimization problem, the search space acts as a medium, and the optimal solution in the solution space is the path of light in the medium.

[0051] Preferably, selecting an elite selection strategy to sort and screen the initialized population to obtain a high-quality population includes: first randomly initializing a mantis population A, setting the maximum number of iterations and related parameters of the population, then generating a population B through refraction reverse learning, merging and sorting the fitness values ​​of the two groups of populations, and finally selecting the top N best individuals to form a new population X. The optimization problem is solved by simulating the individual selection and evolution process in biological evolution, the candidate solution is represented as a particle swarm, and the fitness of each particle is evaluated by the fitness function.

[0052] Step 2: Select the Osprey optimization algorithm and add an adaptive weight mechanism to calculate the individual fitness value and the global optimal position, output the global optimal position and the corresponding individual fitness value of the rolling ball, and couple the position update formula as the global optimal solution; update the individual position information of the rolling ball according to the obstacle mode; prevent the algorithm from falling into the local optimum, and balance the global exploration and local development capabilities of the algorithm.

[0053] Step 3: Select the dynamic selection strategy and adaptive t distribution, use the number of iterations as the degree of freedom parameter for perturbation, iteratively optimize the path through the mantis algorithm, and update the position information of the breeding individuals; in each iteration, randomly generate the path, dynamically update the comprehensive fitness value of the breeding individuals, dynamically update the intermediate nodes of the path, smooth the obstacles of the breeding individuals, and draw the adaptive planning path;

[0054] Among them, the adaptive weight mechanism dynamically adjusts the search strategy weight value according to the search status of the current environmental conditions or the fitness of the individual in the iteration stage to complete the intensity adjustment of the search strategy. The search strategy weight includes the weight of the global search and the weight of the local search. It is used to dynamically adjust the search direction and focus on the trade-off of operations. It allows the search strategy to be automatically adjusted at different iteration stages and under different environmental conditions to enhance the global search capability and the ability to cope with complex obstacles. For example, when the algorithm detects that the search effect of the current solution is not good, the weight of the global search can be increased, and vice versa, the weight of the local search can be increased to balance the ability of exploration and development.

[0055] In view of the fact that the Mantis Optimization Algorithm relies on the worst support and lacks communication with other populations, the Osprey Optimization Algorithm (OOA) is added to the rolling ball stage to couple with the optimal solution and add an adaptive weight mechanism. When dealing with complex obstacles, the position update formula is changed to couple with the global optimal solution. The global search capability is enhanced. Its purpose is to enhance the adaptability of handling obstacles and complex paths during the path planning process.

[0056] Preferably, the dynamic selection strategy and adaptive t distribution are selected to include: adjusting the reproduction strategy by calculating the probability according to the current population characteristics, and completing the generation mode and distribution morphology adjustment of the new individuals. The reproduction strategy is dynamically adjusted by calculating the probability during the mantis reproduction stage. The adaptive t distribution is to adjust the generation mode and distribution characteristics of the new individuals. It is used for the diversity and adaptability of individuals. The adaptive weight mechanism and the adaptive t distribution can be used in combination to improve the flexibility and performance of the algorithm.

[0057] In the reproduction phase, dynamic selection strategies and adaptive t-distribution are added to enhance global search capabilities and avoid falling into local optimality. This ensures that the algorithm can take into account the optimality of the path and the needs of multiple objectives when planning the path. This prevents the algorithm from falling into local optimality and balances global exploration and local development capabilities to ensure that the algorithm can take into account the optimality of the path and the needs of multiple objectives when planning the path.

[0058] The new individuals generated during operation will dynamically adjust their distribution pattern according to the current population characteristics. Adaptability enables the population to flexibly adapt to different search environments, increase the diversity among individuals, and help avoid early convergence.

[0059] Step 4: Select the Jacobian curve and composite fitness function according to the corresponding formula, find the local optimal solution in the grid environment map, improve the smoothness of the path movement of the foraging individual in the search space, update the location information of the foraging individual; enhance the ability to jump out of the local optimum. Improve the overall smoothness and security of the path, especially suitable for path planning in complex environments. Jacobian curves usually refer to elliptic curves under Jacobian coordinates used in elliptic curve cryptography. In path planning problems, the composite fitness value is usually the lower the better, indicating that the path is shorter, safer and smoother, thus meeting the optimization target requirements.

[0060] Step 5: In the stealing phase, the location information of all stealing individuals is updated according to the original mantis algorithm to obtain the location of the stealing population;

[0061] Step 6: Perform a greedy selection strategy update and evaluation on the fitness values ​​of all individuals in the current population. After iteration, the position of the i-th mantis increases once. After all steps (i.e. rolling, reproduction, foraging, stealing) are executed, perform a global greedy selection strategy update and evaluation on the fitness values ​​of all individuals in the current population. Evaluate and update the fitness values ​​of all individuals in the current population to ensure that the global optimal fitness value of each generation is the latest. Reduce the frequency of greedy selection and avoid performing it at every step, thereby reducing the computational burden. At the same time, ensure that the selection of the best individual in the current round is the result of comprehensive all update steps.

[0062] Step 7: Repeat the above steps 1 to 5 until the number of iterations is equal to the maximum number of iterations and converges to the theoretical optimal value of the path comprehensive score. The composite fitness function is a fitness value that combines the results of the multi-objective fitness function by weighting.

[0063] The optimal convergence values ​​include: shortest path length: for example, in a 20×20 grid map, IDBO's shortest path is 27.841, which is 2.57% shorter than the original DBO algorithm, and the path length is shortened by 15.35% in a 40×40 grid map. Convergence speed: The IDBO algorithm converges faster and has a stronger ability to jump out of local optimality. By observing the stability and convergence rate of the convergence curve, verify whether the algorithm reaches the optimal solution within a small number of iterations. Path smoothness: In addition to path length, attention should also be paid to the continuity of the path and the number of turning points. The higher the smoothness, the smoother the robot's movement in a complex environment.

[0064] The definition of the Mantis Optimization Algorithm converging to the vicinity of the optimal value depends on the experimental scenario. On the CEC2005 benchmark function, the optimal value refers to the known theoretical optimal value of each benchmark function, and the convergence effect of the algorithm is evaluated by the degree of proximity to the theoretical optimal value. In the path planning simulation experiment, the optimal value is defined as the comprehensive score of the path, which includes the minimization of the path length, the optimization of the path smoothness, and the obstacle avoidance effect. Specifically, the optimal value evaluation standard of the path planning experiment can be the solution that makes the comprehensive score the highest or the error within an acceptable range.

[0065] The optimized Mantis algorithm is applied in engineering practice, such as wind power generation prediction parameters, welding beam fixed parameters, optimization of neural network prediction models, etc., which verifies its superior optimization ability and engineering practicality. The purpose of mobile robot path planning is to generate the optimal safe path. However, the real-time performance of the mobile robot in local path planning is poor, and the path is long in global path planning. In order to improve the accuracy of real-time obstacle avoidance prediction in local path planning, shorten the path length of global path planning, and reduce the path planning time, so as to obtain a better safe path, the present invention proposes an improved Mantis optimization algorithm (IDBO) based on a hybrid strategy, and applies it to the environmental map created by the grid method, and conducts simulation experiments.

[0066] Figure 2 The flowchart of the multi-objective path planning method based on the improved mantis optimization algorithm provided by the present invention is applicable to the case of multi-objective path planning based on the improved mantis optimization algorithm of the hybrid strategy, mainly applied to multi-objective path planning of robots on a grid map. The method can be executed by a multi-objective path planning device based on the improved mantis optimization algorithm, which can be implemented by hardware and / or software and can generally be integrated in an electronic device. The method specifically includes:

[0067] Step S1, performing path length optimization on the original mantis algorithm to obtain an improved mantis optimization algorithm;

[0068] Step S2: Establishing a grid environment map based on the working environment and obstacle distribution information;

[0069] like Figure 3 , Figure 4 As shown, the grid environment map establishment includes: establishing a simple grid environment map G and a complex grid environment map G', initializing the relevant parameters of the grid map path planning, and specifying the starting point and the end point of the grid environment map.

[0070] Step S3: Establish a multi-objective fitness function, obstacle avoidance function and smooth route function corresponding to the path planning; make the shortest path have the least turning points and the smallest redundancy.

[0071] Preferably, establishing a multi-objective fitness function corresponding to path planning includes:

[0072] Path safety function (i.e. obstacle avoidance function): determines path safety by calculating the minimum distance between two adjacent path segments and obstacles and setting distance constraints; used to ensure that the path avoids obstacles as much as possible and improves safety; evaluates the distance to obstacles in the path to ensure that the robot path avoids obstacles as much as possible to improve travel safety.

[0073] Path smoothness function: Use the cosine theorem to calculate the turning angles (cosine values) between adjacent path nodes, and optimize the path smoothness by minimizing the angle of each turning point and reducing the turning angles. This reduces sharp turns in the path, makes the path closer to a straight line, and makes the robot move more smoothly with fewer sharp turns. This reduces sharp turns in the path and improves the smoothness of the path.

[0074] Path length function: The minimum path length is obtained by accumulating the Euclidean distances between adjacent nodes in the path. The shortest path length of all nodes is used to reduce the robot's moving time; it is used to evaluate the quality of each path and ensure that the robot can reach the end point by the shortest path.

[0075] The multi-objective fitness function forms the final comprehensive fitness value by weighted calculation of the results of the path safety function, smoothness function and path length function. The multi-objective fitness function is a comprehensive evaluation indicator. The multi-objective fitness function refers to multiple independent fitness functions established for different goals in path planning (such as path length, safety, smoothness, etc.). The performance of each goal is evaluated separately in order to balance these goal requirements in path planning so that the algorithm can optimize the path quality in multiple aspects. The multi-objective fitness function will calculate and output multiple fitness values ​​separately.

[0076] Step S4, the mantis population position optimization in the improved mantis optimization algorithm corresponds to the path nodes in the grid environment map, wherein the individual coordinates in the mantis population correspond to the node coordinates of the path planning, that is, the individual position corresponds to the node position; the rolling mantis, breeding mantis, foraging mantis, and stealing mantis population position optimization correspond to the path nodes in the grid map, and the observation results are run independently 30 times and compared with other algorithms.

[0077] Step S5, according to the improved mantis optimization algorithm, all individuals in the population are iterated, and the comprehensive fitness of each iteration is calculated using the multi-objective fitness function to obtain the updated global optimal position (Best_pos) and the optimal fitness value (Best_score), and record and compare them, and find the optimal position by comparing the fitness of each position; and use the global optimal position as the node position of the grid map path planning;

[0078] Iterate all individuals: In each iteration, the position of the individual is updated according to the fitness value, and all individuals in the population are evaluated and updated, with the goal of finding a better solution.

[0079] Output the updated global optimal position and the corresponding individual fitness value of the rolling ball: After each iteration, compare the fitness value of the current individual with the current global optimal fitness value. If a better fitness value is found, update the current global optimal position and optimal fitness value. Through multiple iterations, gradually converge to the global optimal solution. The final global optimal position and optimal fitness value are obtained through multiple iterations and comprehensive fitness calculations, and are reflected in the final path planning results as the shortest path, the least turning points, and the highest path smoothness.

[0080] Obtain the final global optimal position and optimal fitness value: In each iteration, calculate the fitness value of each individual in the population, and determine the current global optimal position and fitness value in each generation, and select the individual with the best fitness from all individuals as the new global optimal position and fitness value.

[0081] Step S6: Using the individual global optimal position as the node position of the grid map path planning, connecting the node starting point of the path planning, each node, and the node end point of the path planning in sequence, and obtaining the path length by calculating the Euclidean distance between each adjacent path node to complete the path planning of the multi-objective fitness function;

[0082] Step S7, observe the convergence curve graph and shortest path graph of each algorithm in the simulation experiment to verify the feasibility of multi-objective path planning; various algorithms include improved mantis optimization algorithm, original algorithm and other swarm intelligence optimization algorithms. Preferably, the simulation experiment based on the benchmark test function and the Wilcoxon rank sum test, as well as the time complexity analysis, verify the effectiveness and feasibility of the IDBO algorithm.

[0083] By applying the improved Mantis optimization algorithm to the map environment created by the grid method, the shortest path length, the number of turning points from the starting point to the end point, and the time spent in path planning of different algorithms are observed to verify the effectiveness of the IDBO algorithm.

[0084] Combine the following Figure 1 and Figure 2 The present invention further describes the multi-objective path planning method based on the improved mantis optimization algorithm: the implementation process of the improved mantis optimization algorithm includes:

[0085] In the first step, the maximum number of iterations of the population is set to T, T = 500, and the refractive index n is derived according to the refractive reverse learning diagram. n is obtained using formula (1):

[0086]

[0087] In formula (1), in the refraction reverse learning diagram, the search space of the solution on the x-axis is [a, b], the incident angle and the refraction angle are α and β, and the incident light length and the refracted light length are l and l * , let k = l / l * , substitute into formula (2) and expand it to multi-dimensional space, and obtain the refraction direction solution

[0088]

[0089] In the formula, x i,j represents the position of the i-th mantis in the j-dimension, and its reverse solution is a j and b j are the upper and lower bounds of the search space, respectively. By adjusting the parameter k, the search range of the early iteration is expanded to enhance the global search capability. In order to improve the global search capability and quality of the initial population, an elite selection strategy is adopted. First, the mantis population A is randomly initialized, and then the population B is generated through refraction reverse learning; the fitness values ​​of the two groups of populations are merged and sorted, and finally the top N best individuals are selected to form a new population X, and its fitness value is calculated. The function expression of the population is as follows:

[0090]

[0091] Where X is the population matrix after initialization, x N,D is the individual position in the population, N is the number of screening individuals set by the algorithm, and Dim is the dimension.

[0092] In the second step, the Osprey Optimization Algorithm (OOA) is introduced in the rolling ball stage and an adaptive weight mechanism is added. In addition, when dealing with obstacles, the global optimal solution is changed to anti-disturbance to improve the search quality and robustness. After the introduction of the Osprey algorithm, let ε = 0.9, r∈(0,1) be a random number, and when r<0.9, there is no obstacle. At this time, the rolling ball position update formula is shown in formula (4):

[0093]

[0094] Where t is the position of the ith mantis after the tth iteration, T is the maximum number of iterations, ω is a parameter that changes with the number of iterations, r1∈[0,1] is a random number, and x best represents the optimal position, and I is a random number between the set {1,2}. When r≥0.9, the mantis encounters an obstacle, and the update formula is shown in formula (5):

[0095]

[0096] Among them, θ is the deflection angle. If θ=0 or θ=π / 2, the individual position is not updated. The osprey optimization algorithm is a new biological heuristic algorithm that imitates the osprey's predation behavior and finds the optimal solution by simulating the osprey's predation process. It has the characteristics of strong optimization ability and fast convergence speed.

[0097] The third step is to introduce the adaptive t-distribution dynamic adjustment strategy in the breeding stage, and dynamically adjust the breeding strategy in the breeding process by calculating the probability p. The update formula of probability p and the population update formula after the introduction of adaptive t-distribution are as follows:

[0098]

[0099] Where ω1 and ω2 are dynamic selection probability parameters, t_iter represents the adaptive t distribution parameter, and r is a constant between [0,1]. It is the position after adaptive t-distribution mutation disturbance, b1 and b2 represent the generation of a random number matrix of size 1×Dim (dimension).

[0100] The fourth step is to add Jacobian curves in the foraging phase to guide mantis individuals to move along a smooth path in the search space, which helps to reduce unnecessary jumps and large turns, avoid falling into local optimality, and thus improve the efficiency of global search. The population update formula is as follows:

[0101] x i (t+1)=r2×x i (t)+(e θ -2×x best× sinθ) / (sinθ-cosθ) (8)

[0102] Where r2 is a random number between (0,1) and θ∈(0,π). Different from the original algorithm, the individual update strategy is determined by the random number rand. If rand < 0.5, the update is performed using formula (8), otherwise the position is updated using formula (9).

[0103] X i (t+1)=X b +S×g×(|X i (t)-X * |+|X i (t)-X b |) (9)

[0104] Where g is a vector that follows a normal distribution and S is a constant.

[0105] Step 5: In the stealing phase, the population is updated according to the original algorithm steps. The update formula is as follows:

[0106] X i (t+1)=X b +S×g×(|X i (t)-X * |+|X i (t)-X b |) (10)

[0107] Among them, X i (t) is the position of the i-th mantis, g is a vector that follows a normal distribution, and S is a constant.

[0108] Repeat steps 1 to 5 until the global optimal position and optimal fitness value converge to the theoretical optimal value of the path comprehensive score.

[0109] The optimized Mantis algorithm can be applied in the fields of microgrids, wind farms, and drone path planning. The present invention will conduct simulation experiments on robot multi-objective path planning by improving the Mantis optimization algorithm based on a hybrid strategy, so that it can reach the shortest path from the starting point to the end point at a faster speed.

[0110] The multi-objective path planning process of the grid map includes the following steps: The first step is to initialize the relevant parameters of the grid map path planning. Specifically, the two-dimensional continuous coordinate system required for path planning is rasterized to obtain the grid map G. Each obstacle and non-obstacle unit in the map is replaced by binary numbers "0" and "1". "0" represents the passable moving area, and "1" represents the obstacle as a black single grid. The black circle represents the starting point coordinate (x start ,y start ), the black triangle represents the end point coordinates (x end,y end ), set the maximum number of iterations of path planning to T. Each individual (particle) represents a possible path, and the size of the population is the number of multiple possible paths generated. Each path is composed of a series of intermediate nodes from the starting point to the end point.

[0111] The second step is to define the fitness function to evaluate the quality of each path. The fitness function takes into account factors such as path safety, smoothness, and path length. Its mathematical model (11) is as follows:

[0112] Formula (11) represents minimizing path safety and ensuring that the path avoids obstacles as much as possible. f1 represents the minimum distance between two adjacent path segments and obstacles:

[0113]

[0114] Formula (12) represents minimizing the path smoothness. The cosine theorem is used to calculate the turning angle. By minimizing the angle of each turning point, the path is made closer to a straight line, ensuring that the path turning is as smooth as possible and reducing large-angle turns in the robot path.

[0115]

[0116] Among them, the smoothing path function, parameter: δ is the corner of two adjacent sections, and the turning angle (cosine value) of the adjacent path segments is used to measure the smoothness. A smaller turning angle (or closer to a straight line) indicates a better path smoothness. d(pi,pi+1) is the distance between two adjacent path nodes. It is used to ensure that the path is as smooth as possible, reduce sharp turns, avoid frequent direction changes, and optimize the stability of the robot's movement.

[0117]

[0118] Formula (14) represents minimizing the path length, which is based on the accumulation of the Euclidean distances between adjacent points in the path to obtain the shortest path of all nodes, thereby reducing the total time of the robot movement.

[0119]

[0120] The three parameters of the above fitness function, obstacle avoidance function, parameters: P is the path, pi is the i-th node of the path, and its coordinates are (x i ,y i ), O j is the jth obstacle, For the road section between node i and node i+1, the results of the three functions are weighted and combined to obtain the final fitness value f x , d(p i ,p i+1) is the distance between two adjacent path nodes. D is the minimum distance parameter: the minimum distance between each point in the path and the nearest obstacle.

[0121] In the third step, the path length f3(P) is obtained by calculating the Euclidean distance between each adjacent path node, which is used to evaluate the path distance. Formula (14) provides the formula for calculating the length of the path. The total length of the path is the cumulative value of the Euclidean distance between all adjacent points. By accumulating and minimizing the path length, the robot's movement time in the task can be effectively reduced, thereby improving the efficiency and energy saving of the system.

[0122] The fourth step is to iteratively optimize the path through the Mantis Optimization Algorithm. In each generation, the path is randomly generated, and the intermediate nodes of the path are dynamically updated in combination with the Mantis Optimization Algorithm, so that the path can effectively avoid obstacles, the path is smoother and as short as possible.

[0123] Step 5: Smoothing and drawing the path: Smooth the final optimized path, draw the final path plan and display the results.

[0124] Smoothing is calculated using equations (12) and (13). It measures the smoothness of the path by calculating the angle difference (turning angle) between three adjacent points. By minimizing the angle of each turning point, the path is closer to a straight line and sharp turns are reduced. Smoothing and minimizing the path length complement each other. By reducing sharp turns in the path, not only can the smoothness of the path be improved, but it also helps to reduce the total length of the path, making the robot walk more efficiently.

[0125] Figure 3 and Figure 4 As shown in the figure, the improved mantis optimization algorithm (IDBO) based on hybrid strategy of the present invention, the classic particle swarm algorithm (PSO), and the original mantis optimization algorithm (DBO) were respectively run 30 times in the same environment to conduct a simulation experiment of grid map path planning. The maximum number of iterations of the paths of the two maps was set to 300, and the grid map path planning experiment was conducted 10 times for each comparison algorithm. The optimal value, average value and standard deviation of the 10 experiments were taken respectively, and the convergence curve was drawn. Figure 7 , Figure 8 shown.

[0126] In the 20*20 simple grid map, the shortest path of the IDBO algorithm is 27.841, which is the best among the three algorithms. The path length is 2.57% shorter than that of the DBO algorithm. The average path length and standard deviation of the IDBO algorithm are also the lowest among the three algorithms, indicating the stability of the IDBO algorithm in path planning problems. In the 40*40 complex grid map, the shortest path of the IDBO algorithm is 57.098, which is also better than other algorithms. The path length is 15.35% shorter than the traditional DBO algorithm. The average and standard deviation indicators are the best, indicating that IDBO has better stability and optimization performance when solving complex problems. Figure 5 , Figure 6 This is a comparison chart of path planning under two maps. The IDBO algorithm not only has the shortest path, but also has the fewest turning points. Compared with other algorithms, it has better continuity and better optimization performance.

[0127] like Figure 7 In the 20*20 simple grid map, the IDBO algorithm continuously jumps out of the local optimum and iterates to the optimal path length first. The Jacobian curve proposed in this invention helps the algorithm to jump out of the local optimum, and the refractive reverse learning improves the population diversity and improves the convergence accuracy and speed. Figure 8 As shown in the figure, in the 40*40 grid map, the IDBO algorithm is superior to other algorithms in terms of convergence speed and shortest path, which proves its feasibility in solving complex problems.

[0128] Fig. 9 The schematic diagram of the structure of the multi-objective path planning device based on the improved mantis optimization algorithm provided in the second embodiment includes: an improvement module 210, which is used to optimize the path length of the original mantis algorithm to obtain an improved mantis optimization algorithm module;

[0129] The grid environment module 220 is used to establish a grid environment map module according to the working environment and obstacle distribution information;

[0130] Function module 230, used to establish a multi-objective fitness function module, an obstacle avoidance function module and a smooth route function module corresponding to path planning;

[0131] The node correspondence module 240 is used to correspond the population position optimization in the improved mantis optimization algorithm to the path nodes in the grid environment map, wherein the individual coordinates in the mantis population correspond to the node coordinates of the path planning, that is, the individual position corresponds to the node position;

[0132] The calculation module 250 is used to iterate all individuals in the population according to the improved mantis optimization algorithm, calculate the comprehensive fitness of each iteration using the multi-objective fitness function, obtain the updated global optimal position and the corresponding optimal fitness value, and record and compare;

[0133] The node connection module 260 is used to use the individual global optimal position as the node position of the grid map path planning, connect the node starting point of the path planning, each node, and the node end point of the path planning in sequence, calculate the Euclidean distance between each adjacent path node to obtain the path length, and complete the path planning of the multi-objective fitness function;

[0134] The simulation verification module 270 is used to observe the convergence curve graph and the shortest path graph of each algorithm in the simulation experiment to verify the feasibility of the multi-objective path planning.

[0135] Preferably, the improved mantis optimization algorithm module is obtained, including the following: in the initialization stage, the refraction reverse learning mechanism is used to construct the initialization population, the maximum number of iterations and related parameters of the population are set, and the elite selection strategy is selected to sort the population and screen the high-quality population module;

[0136] The Osprey optimization algorithm is selected and an adaptive weight mechanism is added to calculate the individual fitness value and the global optimal position, the global optimal position and the corresponding individual fitness value of the rolling ball are output, and the position update formula is coupled as the global optimal solution; the individual position information of the rolling ball is updated according to the obstacle mode;

[0137] The dynamic selection strategy and adaptive t distribution module are selected, the number of iterations is used as the degree of freedom parameter for perturbation, the position information of the breeding individuals is updated, and the path is optimized through iteration; in each iteration, the path is randomly generated, the comprehensive fitness value is updated, the intermediate nodes of the path are dynamically updated, the obstacles are smoothed, and the adaptive planning path is drawn;

[0138] According to the corresponding formula, the Jacobian curve and the composite fitness function are selected to find the local optimal solution in the grid environment map, update the position information of the foraging individual, and improve the smoothness of the individual path movement in the search space.

[0139] In the stealing phase, the location information of all stealing individuals is updated according to the original mantis algorithm steps to obtain the location of the stealing population;

[0140] Perform a greedy selection strategy update and evaluation on the fitness values ​​of all individuals in the current population. After the tth iteration, the position of the i-th mantis increases once;

[0141] Repeat the above steps 1 to 5 until the number of iterations is equal to the maximum number of iterations and converges to the theoretical optimal value.

[0142] Optionally, the module for selecting an elite selection strategy to sort the population and screen high-quality populations includes: first randomly initializing a mantis population A, setting the maximum number of iterations and related parameters of the population, then generating a population B through refractive reverse learning, merging and sorting the fitness values ​​of the two groups of populations, and finally selecting the top N best individuals to form a new population X.

[0143] Optionally, the selection of dynamic selection strategy and adaptive t distribution module includes: dynamically adjusting the generation method and distribution morphological characteristics of new rolling ball individuals according to the current population characteristics; dynamically adjusting the reproduction strategy by calculating the probability during the mantis reproduction stage.

[0144] Optionally, the grid environment map establishment module includes: establishing a simple grid environment map and a complex grid environment map, initializing relevant parameters of grid map path planning, and specifying the starting point and the ending point of the grid environment map.

[0145] Optionally, a multi-objective fitness function module corresponding to path planning is established, and the comprehensive fitness value is calculated by weighting and combining the results of multiple objective functions such as path safety, smoothness and path length. Including: Path safety function module: By calculating the minimum distance between two adjacent path segments and obstacles in the path, setting distance constraints to determine path safety; used to ensure that the path avoids obstacles as much as possible;

[0146] Path smoothness function module: Use the cosine theorem to calculate the turning angle (cosine value) between adjacent path nodes, and make the path closer to a straight line by minimizing the angle of each turning point; thereby reducing sharp turns in the path and improving the smoothness of the path; Path length function module: Obtain the total path length based on the Euclidean distance between adjacent nodes in the path, and the shortest path length of all nodes to reduce the moving time; Weight the results of minimizing path safety, smoothness and path length to obtain the fitness value, and obtain a multi-objective fitness function to evaluate the quality of each path.

[0147] The multi-objective path planning device based on the improved mantis optimization algorithm of the present invention can execute the multi-objective path planning method based on the improved mantis optimization algorithm provided by any embodiment, and has the corresponding functional modules and beneficial effects of the execution method.

[0148] Fig.10 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention is shown in FIG. Fig.10 As shown, the electronic device includes a processor 310, a memory 320, an input device 330 and an output device 340; the number of the processor 310 in the electronic device can be one or more. Fig.10A processor 310 is taken as an example; the processor 310, the memory 320, the input device 330 and the output device 340 in the electronic device can be connected via a bus or other means. Fig.10 The example of connecting via a bus is taken. The memory 320, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the multi-objective path planning method based on the improved mantis optimization algorithm in the embodiment (for example, the grid environment map module and the node corresponding module in the multi-objective path planning device based on the improved mantis optimization algorithm). The processor 310 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 320, that is, realizes the above-mentioned multi-objective path planning method based on the improved mantis optimization algorithm.

[0149] The memory 320 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 320 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some instances, the memory 320 may further include a memory remotely arranged relative to the processor 310, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The input device 330 may be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. The output device 340 may include a display device such as a display screen.

[0150] In the description of the present invention, it is necessary to understand that the indicated orientation or positional relationship is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0151] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-objective path planning method based on an improved mantis optimization algorithm, characterized in that: include: The original mantis algorithm is optimized for path length to obtain the improved mantis optimization algorithm; Establish a grid environment map based on the working environment and obstacle distribution information; Establish multi-objective fitness function, obstacle avoidance function and smooth route function corresponding to path planning; The mantis population position optimization of the improved mantis optimization algorithm corresponds to the path nodes in the grid environment map, where the individual coordinates in the mantis population correspond to the node coordinates of the path planning; According to the improved mantis optimization algorithm, all individuals in the population are iterated, and the comprehensive fitness of each iteration is calculated using the multi-objective fitness function to obtain the updated global optimal position and the corresponding optimal fitness value, and record and compare them; The global optimal position of the individual is used as the node position of the path planning of the grid environment map. The node starting point, each node, and the node end point of the path planning are connected in sequence. The path length is obtained by calculating the Euclidean distance between each adjacent path node to complete the path planning of the multi-objective fitness function. Observe the convergence curves and shortest path graphs of each algorithm to verify the feasibility of multi-objective path planning.

2. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 1 is characterized in that: The improved mantis optimization algorithm includes: using a refraction reverse learning mechanism to construct an initialization population, setting a maximum number of iterations and related parameters of the population, and selecting an elite selection strategy to sort the initialization population and screen to obtain a high-quality population; The Osprey optimization algorithm is selected and an adaptive weight mechanism is added to calculate the individual fitness value and the global optimal position, output the global optimal position and the corresponding individual fitness value of the rolling ball, and couple the position update formula as the global optimal solution; update the individual position information of the rolling ball according to the obstacle mode; The dynamic selection strategy and adaptive t-distribution are used, the number of iterations is used as the degree of freedom parameter for perturbation, the path is iteratively optimized through the mantis algorithm, and the position information of the breeding individuals is updated; According to the corresponding formula, the Jacobian curve and the composite fitness function are selected to find the local optimal solution in the grid environment map, improve the smoothness of the foraging individual path movement in the search space, and update the foraging individual position information; According to the original mantis algorithm, the location information of all stealing individuals is updated to obtain the location of the stealing population; Perform a greedy selection strategy update and evaluation on the fitness values ​​of all individuals in the current population, and the position of the mantis increases by one after iteration; Repeat the above steps 1 to 5 until the number of iterations is equal to the maximum number of iterations and converges to the optimal value of the path comprehensive score; wherein the composite fitness function is a fitness value that is weighted by combining the results of the multi-objective fitness functions.

3. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 2 is characterized in that: The elite selection strategy is used to sort and screen the initialized population to obtain a high-quality population, including: first randomly initializing the mantis population A, setting the maximum number of iterations and related parameters of the population, then generating the population B through refraction reverse learning, merging and sorting the fitness values ​​of the two groups of populations, and finally selecting the top N best individuals to form a new population X.

4. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 2 is characterized in that: The selection of dynamic selection strategy and adaptive t distribution includes: adjusting the reproduction strategy by calculating the probability according to the current population characteristics, completing the generation method and distribution pattern adjustment of new reproduction individuals.

5. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 1 is characterized in that: The establishment of the grid environment map includes: establishing a simple grid environment and a complex grid environment, initializing relevant parameters of the path planning of the grid environment map, and specifying the starting point and the ending point of the grid environment map.

6. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 1 is characterized in that: The multi-objective fitness function corresponding to the path planning is established as follows: Path safety function: The path safety is determined by calculating the minimum distance between two adjacent path segments and obstacles and setting distance constraints; Path smoothness function: Use the cosine theorem to calculate the turning angles between adjacent path nodes, and optimize the path smoothness through the turning angles; Path length function: Minimize the path length by summing the Euclidean distances between adjacent nodes in the path.

7. The multi-objective path planning method based on the improved mantis optimization algorithm according to claim 6 is characterized in that: The multi-objective fitness function corresponding to the path planning is established, and a final comprehensive fitness value is formed by weighted calculation through the results of the path safety function, the smoothness function and the path length function.

8. A multi-objective path planning device based on an improved mantis optimization algorithm, characterized in that: include: The improvement module is used to optimize the path length of the original mantis algorithm to obtain the improved mantis optimization algorithm. The grid environment module is used to establish a grid environment map based on the working environment and obstacle distribution information. Function module, used to establish multi-objective fitness function, obstacle avoidance function and smooth route function corresponding to path planning; The node correspondence module is used to correspond the mantis population position optimization in the improved mantis optimization algorithm to the path nodes in the grid environment map, where the individual coordinates in the mantis population correspond to the node coordinates of the path planning; The calculation module is used to iterate all individuals in the population according to the improved mantis optimization algorithm, calculate the comprehensive fitness of each iteration using the multi-objective fitness function, obtain the updated global optimal position and the corresponding optimal fitness value, and record and compare; The node connection module is used to use the individual global optimal position as the node position of the grid map path planning, connect the node starting point of the path planning, each node, and the node end point of the path planning in sequence, and calculate the Euclidean distance between each adjacent path node to obtain the path length, thus completing the path planning of the multi-objective fitness function; The simulation verification module is used to observe the convergence curve graph and shortest path graph of each algorithm and verify the feasibility of multi-objective path planning.

9. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the multi-objective path planning method based on the improved mantis optimization algorithm as described in any one of claims 1-7.

Images