An Optimization Method for S-Surface Control Parameters of Autonomous Underwater Vehicles Based on IAFSA Algorithm

Abstract

This invention provides a method for optimizing S-plane control parameters of an autonomous underwater vehicle (AUV) based on the IAFSA algorithm. The method includes: improving the classic artificial fish swarm algorithm by employing feeding behavior, adaptive step size, and a field of view with attenuation factors to form the IAFSA algorithm; designing an S-plane heading controller and an S-plane forward velocity controller for the AUV; and optimizing and tuning five control parameters of the S-plane forward velocity controller and the S-plane heading controller using the IAFSA algorithm to form an S-plane path tracking controller based on the IAFSA algorithm. Simulation and experimental analysis show that the improved fish swarm algorithm has a faster convergence speed and a significantly enhanced ability to escape local optima. The S-plane path tracking controller performance using the tuned parameters is 96% lower than before tuning. Underwater experiments have verified the feasibility and effectiveness of applying the IAFSA algorithm in tuning the two-dimensional path tracking control parameters of an AUV.

Description

Technical Field

[0001] This invention relates to optimization algorithms, specifically a method for optimizing the S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm. Background Technology

[0002] Autonomous underwater vehicles (AUVs) have become an important tool for underwater exploration and marine resource development, currently serving multiple fields such as marine military and marine economy. Due to the highly nonlinear and strongly coupled dynamic characteristics of AUVs, as well as the uncertainty of the underwater environment, the design of their underwater motion controllers is difficult to refer to relatively mature unmanned systems such as unmanned vehicles and drones, which has become a fundamental bottleneck for other important functions.

[0003] Reference [1] combined the concepts of fuzzy controller and PD controller to design an S-face controller for underwater robots, which has the characteristics of fewer input parameters, high fitting degree to underwater systems and strong practicality, and its effectiveness has been proven in subsequent field tests. Reference [2] used other nonlinear functions with better convergence to replace the Sigmoid function in the S-face controller, forming a generalized S-face underwater robot controller, which can achieve better control effect. However, the control parameters in the traditional S-face controller are often fixed, lack adaptive ability, and need to be manually set and adjusted. This kind of approximation method makes the AUV still have a large tracking deviation during underwater movement, which does not meet the actual application. Therefore, applying intelligent optimization algorithms to the tuning of control parameters has become an effective solution (references [3-7]). In recent years, some scholars have studied the application of intelligent algorithms in the tuning of S-face controller parameters, including particle swarm optimization (references [8-11]), neural networks (references [12-13]), simulated annealing algorithm (reference

[14] ), etc. Among the many optimization algorithms, the artificial fish swarm algorithm was proposed by Li Xiaolei in 2002 (reference

[15] ). It has global optimization capabilities, fast convergence speed, and good fit to uncertain models. It is suitable for complex nonlinear mathematical models such as AUV path tracking control. However, the artificial fish swarm algorithm proposed by Li Xiaolei has problems such as blindness in the later stage of iteration, slow convergence speed, and difficulty in escaping once trapped in a local optimum.

[0004] References:

[0005] [1] Liu Xuemin, Xu Yuru. S-surface control method for underwater robot motion [J]. Ocean Engineering, 2001, 19(3):81-84.

[0006] [2] Wang Jianguo, Wu Gongxing, Wan Lei, et al. Design of underwater robot controller with generalized S-plane [J]. Journal of Electrical Machines and Control, 2009, 13(A01): 144-148.

[0007] [3] Zhou Zexing. A three-dimensional trajectory tracking method for AUV based on DRNN-S control [J]. Ship Science and Technology, 2021, 43(21): 96-99, 178. DOI: 10.3404 / j.issn.1672-7649.2021.11.017.

[0008] [4] Yang Qiushi. Research on underactuated ship trajectory tracking control based on S-plane [D]. Liaoning: Dalian Maritime University, 2021.

[0009] [5] Northwestern Polytechnical University. A cooperative control method for quadrotor UAVs based on expert S-surface control: CN202110190552.3[P]. 2021-07-16.

[0010] [6] Shandong University. AUV path tracking method and system based on S-surface control and TD3: CN202110239801.3[P]. 2021-06-29.

[0011] [7] Pan Wei. Research on three-dimensional path tracking control for recovery of fully driven autonomous underwater robot [D]. Jiangsu: Jiangsu University of Science and Technology, 2021.

[0012] [8] Lü Chong, Pang Yongjie, Wang Bo, et al. Improved S-surface control and hardware-in-the-loop simulation of underwater robot [J]. Journal of Shanghai Jiaotong University, 2010, 44(7): 957-961, 967.

[0013] [9] Tang Xudong, Pang Yongjie, Wan Lei. Application of improved PSO algorithm in tuning of S-face motion control parameters of underwater robot [J]. Journal of Applied Basic and Engineering Sciences, 2009, 17(1):153-160.

[0014]

[10] Li Hongyu, Wang Ying, Lu Zhen, et al. Coordinated Attitude Control of Underwater Robots Based on PSO-GA Algorithm and Neural Network [J / OL]. China Test: 1-7 [2022-07-11]. https: / / kns-cnki-net.webvpn.jmu.edu.cn / kcms / detail / 51.1714.TB.20220322.1838.048.html

[0015]

[11] Liu Sheng, Ren Dong, Li Bing. S-face control of submersible based on SA-PSO algorithm [J]. Control Engineering, 2011, 18(05):710-714. DOI:10.14107 / j.cnki.kzgc.2011.05.043.

[0016]

[12] Tang Xudong, Pang Yongjie, Wang Jianguo. Adaptive motion control of underwater robot S-face based on single neuron [J]. Computer Applications, 2007, 27(12):2899-2901.

[0017]

[13] Wan Lei, Tang Wenzheng, Li Yueming. BP neural network S-surface control of intelligent underwater robot [J]. Industrial Instrumentation and Automation Devices, 2019(2):13-17.

[0018]

[14] Sun Yushan, Li Yueming, Zhang Yinghao, et al. Application of improved simulated annealing algorithm in the optimization of motion control parameters of S-face of underwater robot [J]. Acta Ordnance et al., 2013, 34(11):1418-1423.

[0019]

[15] Li Xiaolei. A novel intelligent optimization method - artificial fish swarm algorithm [D]. Zhejiang University, 2003.

[0020] This patent application was supported by the National Natural Science Foundation of China (42122025, 41974005). Summary of the Invention

[0021] The purpose of this invention is to provide a method for optimizing the S-plane control parameters of an unmanned underwater vehicle based on the IAFSA algorithm, which can reduce the deviation of AUV two-dimensional path tracking to a certain extent, thereby achieving AUV path tracking control with smaller errors and faster convergence speed.

[0022] A method for optimizing S-plane control parameters of an unmanned underwater vehicle based on the IAFSA algorithm includes the following steps:

[0023] The classic artificial fish swarm algorithm is improved by adopting feeding behavior, adaptive step size, and field of view with decay factor to form the IAFSA algorithm;

[0024] Design of an S-plane heading controller and an S-plane forward velocity controller for an autonomous underwater vehicle based on an S-plane controller;

[0025] The IAFSA algorithm is used to optimize and tune five control parameters of the S-plane forward velocity controller and the S-plane heading controller to form an S-plane path tracking controller based on the IAFSA algorithm.

[0026] Furthermore, the specific steps for designing the S-plane heading controller for the autonomous underwater vehicle based on the S-plane controller are as follows:

[0027] The control law is generated using the Sigmoid function to control the angle of the vertical servo, thereby controlling the heading of the AUV. The S-plane heading controller is designed as follows:

[0028]

[0029] In the formula, u1 is the forward velocity output of the thrust controller, k1 and k2 are control parameters, and e1 and These represent the differences between the actual forward velocity and the desired value, respectively, with Δu1 representing the incremental output of the heading controller; where...

[0030] e1 = v b -v (2)

[0031] In the formula, v b This represents the expected forward velocity, where v is the current velocity of the AUV.

[0032] Furthermore, the specific steps for designing the S-face forward velocity controller for the autonomous underwater vehicle based on the S-face controller are as follows:

[0033] Using an S-face controller, a control law is generated through the Sigmoid function to control the thrust of the propeller, thereby controlling the forward velocity of the AUV. The design of the S-face forward velocity controller is shown in formula (3):

[0034]

[0035] In the formula, u2 is the forward velocity output of the thrust controller, k3, k4, and k5 are control parameters, and e2 and These represent the difference between the actual forward velocity and the desired value, respectively, with Δu2 representing the incremental output of the forward velocity controller.

[0036] e2=ψ b -(ψ+β) (4)

[0037] Where e2 represents the actual forward velocity, ψ b ψ represents the desired heading of the AUV in the geodetic coordinate system, β represents the actual heading of the AUV, and β represents the drift angle.

[0038] Furthermore, the improvement of the classic artificial fish swarm algorithm by employing feeding behavior, adaptive step size, and a field of view with a decay factor to form the IAFSA algorithm specifically includes:

[0039] The algorithm improves upon the classic artificial fish swarm algorithm by incorporating feeding competition behavior into the basic behaviors of artificial fish. The specific mathematical expression is as follows:

[0040]

[0041] In the formula, Y i Indicates the current position X i The food concentration, when it is relative to a certain location X in the field of vision. V Food concentration Y VWhen the ratio is less than a certain threshold, the artificial fish will engage in predation behavior. This threshold is represented by the product of the probability of engagement β = rand(0,1) and the crowding factor.

[0042] The algorithm employs an adaptive step size to improve the classic artificial fish swarm algorithm. Specifically, a larger Step value is used when the current position is far from an area with higher food concentration, and a smaller Step value is used when the current position is close. The specific mathematical expression is as follows:

[0043] Step = Rand * || X i -X V ‖ (6)

[0044] In the formula, X i Indicates the current position of the artificial fish, X V This indicates a location within its field of vision, when X V If the food concentration at a given location is greater than the concentration at the current location, the artificial fish will swim a distance Step in that direction.

[0045] The algorithm improves the classic artificial fish swarm algorithm by using a field of view with a decay factor. Specifically, when the algorithm starts iterating, a larger Visual value is used to expand the optimization range. As the iteration process progresses, the Visual value is gradually and appropriately decreased to accelerate the convergence speed. Mathematically, this is expressed as:

[0046] Visual k =αVisual k-1 (7)

[0047] Where α∈(0,1) represents the decay rate of the field of view, and k represents the number of iterations.

[0048] Furthermore, the classic artificial fish swarm algorithm is improved by incorporating feeding behavior, adaptive step size, and field of view with a decay factor. The conditions for selecting each behavior are as follows:

[0049] ①Swarming behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions is n. f and the center position of the peer group X c The food concentration at the center is Y. c If equation (8) is satisfied, then it represents the center position X. c If the crowding level is low and the food concentration is high, then the artificial fish i will move towards the center position X. cMove one step in the direction and update the next position according to equation (9); if equation (8) is not satisfied, then perform foraging behavior according to the fixed step size value Step0:

[0050]

[0051]

[0052] ② Tail-end collision behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions n f The location of the companion with the highest food concentration, X j and its food concentration value Y j If equation (10) is satisfied, then it represents X. j If the crowding level is low and the food concentration is high, then the current artificial fish i is heading towards X. j Move one step in the direction and update the next position according to formula (11); if formula (10) is not satisfied, then perform foraging behavior according to the fixed step value Step0;

[0053]

[0054]

[0055] Where X max This is the location with the highest food concentration;

[0056] ③ Food-stealing behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k A certain companion's position X was randomly observed within the room. V The food concentration at this location is Y. V If equation (12) is satisfied, then it represents X. V If the companion at location X has a much higher food concentration than the current artificial fish, then the artificial fish should move directly to location X. V Location; if equation (12) is not satisfied, then foraging behavior is performed according to the fixed step size value Step0:

[0057]

[0058] ④ Foraging behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k Within its field of vision k Randomly select a position X inside j The food concentration at this location is Y. j If Y i <Y j If Y = 0, it means that the food concentration at the selected location is high. In this case, the artificial fish moves one step in that direction and updates the next position according to formula (13). i >Y j If the random behavior is executed, that is, a random direction and step size are selected for movement, and the next position is updated according to formula (14);

[0059]

[0060] X i =X i +Visual*rand(14)

[0061] Furthermore, the optimization and tuning of five control parameters of the forward velocity controller and the heading controller using the IAFSA algorithm to form an S-plane path tracking controller based on the IAFSA algorithm specifically includes:

[0062] Step 1: Initialize the fish swarm parameters. The initial fish swarm parameters include the number of artificial fish N, the initial field of view Visual, the number of attempts try_number, the crowding factor δ, the maximum number of iterations MAXGAN, and the decay factor α.

[0063] Step 2: Initialize the artificial fish swarm within a given range, randomly generate position coordinates within the upper and lower limits of each dimension to form the artificial fish swarm;

[0064] Step 3: Repeat the following steps using MAXGAN 10 times until the optimal food concentration value for the artificial fish population is found:

[0065] b. Each artificial fish selects and executes the appropriate behavior according to the conditions;

[0066] b. Evaluate the objective function value of the artificial fish's position after performing the corresponding action. If the objective function value of the new position is better, update the state; otherwise, retain the state.

[0067] c. Determine if the current iteration count has reached MAXGAN. If it has, exit the loop and record the five-dimensional coordinates (k1, k2, k3, k4, k5) of the optimal position, where k1, k2, k3, k4, k5 are the control parameters of the S-face forward velocity controller and the S-face forward controller. If it has not reached, repeat the above loop.

[0068] Step 4: Substitute the five-dimensional coordinates (k1, k2, k3, k4, k5) of the optimal food concentration position back into the S-face forward velocity controller and S-face forward controller of the AUV to complete the two-dimensional path tracking task and generate a path tracking trajectory map.

[0069] This invention first addresses the application of optimization algorithms in the parameter tuning of two-dimensional path tracking controllers for autonomous underwater vehicles (AUVs). It improves the artificial fish swarm algorithm by employing methods such as feeding behavior, adaptive step size, and field of view with attenuation factors, which accelerates the convergence speed in the later stages of iteration and helps escape local optima, thus forming the IAFSA algorithm. Secondly, based on the S-plane controller, an S-plane forward velocity controller and an S-plane bow controller for AUV two-dimensional path tracking are designed. Then, the IAFSA algorithm is used to optimize and tune five control parameters in the S-plane forward velocity controller and the S-plane bow controller to improve controller performance. Finally, simulation experiments and sea trials were completed. Simulation and experimental analysis show that the improved fish swarm algorithm has a faster convergence speed and a significantly enhanced ability to escape local optima. The S-plane path tracking controller performance using the tuned parameters is 96% lower than before tuning. Underwater trials further verify the feasibility and effectiveness of the IAFSA algorithm in tuning AUV two-dimensional path tracking control parameters. Attached Figure Description

[0070] Figure 1 A schematic diagram of a Sigmoid surface;

[0071] Figure 2 Visual concept image of an artificial fish;

[0072] Figure 3 This is a flowchart illustrating the optimization of the control parameters of the S-face forward velocity controller and the S-face forward controller using the IAFSA algorithm in this invention.

[0073] Figure 4 This is a comparison chart of the index values ​​of the IAFSA algorithm of this invention and the classic fish swarm algorithm;

[0074] Figure 5 The path tracking trajectory before control parameter tuning;

[0075] Figure 6 The path tracking trajectory after the control parameters are tuned;

[0076] Figure 7 This invention relates to an embodiment of the AUV two-dimensional path tracking sea trial trajectory based on an S-plane controller. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] This invention provides a method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm, comprising the following steps:

[0079] Step 1: Improve the classic artificial fish swarm algorithm by adopting feeding behavior, adaptive step size, and field of view with attenuation factor to form the IAFSA algorithm.

[0080] In a body of water, fish often find areas with abundant food, either independently or by following other fish. Therefore, the area with the highest fish population generally corresponds to the area with the highest food concentration. Artificial fish swarm algorithms leverage this characteristic by constructing artificial fish to mimic various survival behaviors of natural fish, thus achieving optimal selection. The artificial fish perceive their environment through vision, and the artificial fish model uses methods such as... Figure 2 The method shown achieves virtual vision for artificial fish.

[0081] Figure 2 In this diagram, X represents the current position of the artificial fish, Step represents the step size, and Visual represents the visual range. If X... V If the food concentration at a location is higher than the current location's concentration, proceed in that direction to X. next Location, X V and X next The relationship with X can be expressed by equations (1-2):

[0082] X V =X + visual*r (1)

[0083]

[0084] Where r is a random number in the interval [-1, 1].

[0085] Artificial fish exhibit the following typical behaviors:

[0086] 1) Foraging behavior: Under normal circumstances, fish swim freely and randomly in the water. When they find food, they will swim quickly towards the direction where the food concentration increases. (Artificial fish X) i Randomly select a location X within the field of vision. j Food concentration Y at this location j If it is better than the current location concentration Y i Then artificial fish X i Towards observation position X j Move one step. The updated position coordinates are represented as follows:

[0087]

[0088] If the food concentration at the next location does not meet the above conditions, the artificial fish randomly selects another location to observe the food concentration. When the number of re-observations reaches the try_number value, the artificial fish performs random behavior.

[0089] 2) Schooling Behavior: Due to their natural instinct to seek advantage and avoid harm, fish will form groups while swimming to increase their chances of survival and avoid predators. This can be achieved by calculating the number of other artificial fish X in the school. j With artificial fish X i The distance can be used to determine X. i Within the visual range, there is the number of artificial fish (n) and the virtual center location (X) formed by other artificial fish. c and its food concentration Y c If Y satisfies c / n>δY i This indicates that the food concentration in the central area is high and not crowded, indicating that the artificial fish X i Towards X c Move one step; otherwise, perform foraging behavior. The position coordinate update is represented as follows:

[0090]

[0091] 3) Tail-chasing behavior: When one or more fish in a school find food, their nearby companions will quickly follow them to the food source. This is determined by calculating the number of other artificial fish (X) in the school. j With artificial fish X i The distance can be used to determine the number of artificial fish, n, within the current artificial fish visual range, and the location X with the highest food concentration can be observed. max and its food concentration Y max If Y max / n>δY i This indicates that the location with the highest food concentration is still not crowded, and the current artificial fish X... i Towards X max Move one step; otherwise, perform foraging behavior. The position update formula is as follows:

[0092]

[0093] 4) Random Behavior: Solitary fish in water typically swim randomly in order to search for food sources or nearby fish over a wider area. The position update formula is:

[0094] X i =X i +Visual*rand (6)

[0095] To address some of the problems existing in the classic fish swarm algorithm, this invention uses the following strategies for improvement:

[0096] 1) Food snatching behavior

[0097] In the real natural world, when the surrounding food concentration drops to a certain level, fish will become aggressive due to hunger. If other fish occupy more food within their field of vision, the hungry fish will attempt to take their place; this is called predation behavior. Therefore, predation behavior can be considered as a basic behavior of artificially created fish. The specific mathematical expression is as follows:

[0098]

[0099] In the above formula, Y i Indicates the current position X i The food concentration, when it is relative to a certain location X in the field of vision. V Food concentration Y V When the ratio is less than a certain value, the artificial fish will engage in predation behavior. This threshold is represented by the product of the probability of engagement β = rand(0,1) and the crowding factor.

[0100] The purpose of incorporating feeding behavior is to strengthen the tendency of artificial fish swarms to cluster toward a better value of the objective function, while introducing randomness makes it easier to escape local optima.

[0101] 2) Adaptive step size

[0102] When artificial fish swim towards areas with high food concentration, they randomly select a position in that direction. If the current position is close to the point of highest food concentration, it will oscillate around the optimal value, leading to longer algorithm execution time. This is also the main reason for the slow convergence speed in the later stages of the algorithm iteration. To address this issue, an adaptive step size approach can be considered. That is, a larger step value is used when the current position is far from the point of higher food concentration, and a smaller step value is used when it is close. The specific mathematical expression is as follows:

[0103] Step = Rand * || X i -X V ‖ (8)

[0104] In the above formula, X i Indicates the current position of the artificial fish, X V This indicates a location within its field of vision, when X V If the food concentration at a given location is greater than the current concentration, the artificial fish will swim a distance Step in that direction. The purpose of adding the Rand() function is to provide randomness and prevent the convergence speed from being too fast and getting stuck in a local optimum. If the artificial fish does not move to a location with a higher food concentration after performing the swarming and tail-chasing behavior, Step will still use a fixed value Step0 to increase randomness and escape local optima.

[0105] 3) Field of view with attenuation factor

[0106] The foraging behavior of fish plays a crucial role in solving discrete optimization problems. While foraging, the fish maintain a constant field of vision. As the artificial fish gradually approaches the optimal solution, its state X... i If only 1-2 dimensions differ from the optimal solution, using the original fixed Visual value will cause the artificial fish to blindly search for the best, significantly increasing the complexity of the algorithm in its later iterations. Furthermore, using a fixed Visual value results in either an excessively large field of view leading to slow convergence, or an excessively small field of view increasing the likelihood of getting trapped in local optima. To avoid this drawback, the following strategies can be used to improve foraging behavior:

[0107] When the algorithm begins iteration, a larger Visual value is used to expand the optimization range. As the iteration progresses, the Visual value is gradually and appropriately decreased to accelerate convergence. Mathematically, this can be expressed as:

[0108] Visual k =aVisual k-1 (9)

[0109] Where α∈(0,1) represents the decay rate of the field of view, and k represents the number of iterations.

[0110] Each time the artificial fish chooses a new action, it calculates the food concentration value using the current coordinates to determine if it's optimal. This calculation process involves completing a path-tracking task to obtain the index value. This step is the key to improving the fish swarm algorithm.

[0111] Based on the classic artificial fish swarm algorithm, improvements were made using the aforementioned feeding behavior, adaptive step size, and field of view with attenuation factor. The conditions for selecting each behavior are as follows:

[0112] ①Swarming behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. iAccording to equation (9), the field of view for the current iteration number is obtained. k According to equation (8), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions is n. f and the center position of the peer group X c The food concentration at the center is Y. c If equation (10) is satisfied, then it represents the center position X. c If the crowding level is low and the food concentration is high, then the artificial fish i will move towards the center position X. c Move one step in the direction and update the next position according to formula (4); if formula (10) is not satisfied, then perform foraging behavior according to the fixed step value Step0;

[0113]

[0114] ② Tail-end collision behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (9), the field of view for the current iteration number is obtained. k According to equation (8), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions n f The location of the companion with the highest food concentration, X j and its food concentration value Y j If equation (11) is satisfied, then it represents X. j If the crowding level is low and the food concentration is high, then the current artificial fish i is heading towards X. j Move one step in the direction and update the next position according to formula (5); if formula (11) is not satisfied, then perform foraging behavior according to the fixed step value Step0;

[0115]

[0116] ③ Food-stealing behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (9), the field of view for the current iteration number is obtained. k According to equation (8), the adaptive step size Step is obtained, and within its field of view Visual k A certain companion's position X was randomly observed within the room. V The food concentration at this location is Y. V If equation (12) is satisfied, then it represents X. V If the companion at location X has a much higher food concentration than the current artificial fish, then the artificial fish should move directly to location X. VLocation; if equation (12) is not satisfied, foraging behavior is performed according to the fixed step value Step0.

[0117]

[0118] ④ Foraging behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (9), the field of view for the current iteration number is obtained. k Within its field of vision k Randomly select a position X inside j The food concentration at this location is Y. j If Y i <Y j If Y = 0, it means that the food concentration at the selected location is high. In this case, the artificial fish moves one step in that direction and updates the next position according to formula (3). i >Y j If the position is selected, random behavior is performed, that is, a random direction and step size are selected for movement, and the next position is updated according to formula (6).

[0119] Step 2: Design the S-face heading controller and S-face forward velocity controller for the autonomous underwater vehicle based on the S-face controller.

[0120] Designing S-surface controllers using Sigmoid surfaces (e.g.) Figure 1 This method offers better control performance for underwater nonlinear control systems such as AUVs. A control law is generated using the Sigmoid function to control the angle of the vertical servo, thereby controlling the AUV's heading.

[0121] The S-plane heading controller is designed as follows:

[0122]

[0123] In the formula, u1 is the forward velocity output of the thrust controller, k1 and k2 are control parameters, and e1 and These represent the differences between the actual forward velocity and the desired value, respectively, with Δu1 representing the incremental output of the heading controller; where...

[0124] e1 = v b -v (14)

[0125] In the formula, v b This represents the expected forward velocity, where v is the current velocity of the AUV;

[0126] An S-face controller is used, and a control law is generated through the Sigmoid function to control the thrust of the propeller, thereby controlling the forward speed of the AUV. In the process of using the S-face forward speed controller, there is often a problem that the steady-state error caused by forward damping cannot be eliminated. Therefore, integral control is introduced into the exponential term of the Sigmoid function to eliminate this type of steady-state error. Specifically, the design of the S-face forward speed controller is shown in formula (15):

[0127]

[0128] In the formula, u2 is the forward velocity output of the thrust controller, k3, k4, and k5 are control parameters, and e2 and These represent the difference between the actual forward velocity and the desired value, respectively, with Δu2 representing the incremental output of the forward velocity controller; where...

[0129] e2=ψ b -(ψ+β) (16)

[0130] In the formula, ψ b ψ represents the desired heading of the AUV in the geodetic coordinate system, β represents the actual heading of the AUV, and β represents the drift angle.

[0131] Step 3: Optimize and tune the five control parameters of the S-plane forward velocity controller and the S-plane heading controller using the IAFSA algorithm to form an S-plane path tracking controller based on the IAFSA algorithm. Step 3 specifically includes:

[0132] Step 1: Initialize fish swarm parameters; In this embodiment, the initialized fish swarm parameters include the number of artificial fish N, initial field of view Visual, number of attempts try_number, crowding factor δ, maximum number of iterations MAXGAN MAXGAN, decay factor α, and fixed step size Step0, as detailed in Table 1:

[0133] Table 1 Initialization parameters for the fish swarm

[0134]

[0135] Step 2: Initialize the artificial fish swarm within a given range, randomly generate position coordinates within the upper and lower limits of each dimension to form the artificial fish swarm;

[0136] Step 3: Repeat the following steps MAXGAN times:

[0137] a. Each artificial fish selects and performs the appropriate behavior according to the conditions;

[0138] b. Evaluate the objective function value of the artificial fish's location after performing the corresponding action. The evaluation process is as follows:

[0139] The parameters to be optimized in the two S-face controllers mentioned above include five: k1, k2, k3, k4, and k5. Since the dynamic model of the AUV has strong coupling and nonlinearity when moving underwater, the coupling between the forward velocity and the bow output cannot be decoupled. Therefore, when using the IAFSA algorithm to optimize the control parameters, the position coordinates of the artificial fish are set as a five-dimensional vector (k1, k2, k3, k4, k5). Substituting the five-dimensional position coordinates of the artificial fish back into the forward velocity controller and the bow controller of the AUV, the sum of the absolute value of the deviation multiplied by the integral of the time term over time in the two S-face controllers is selected as the objective function of the algorithm. This function reflects both the magnitude of the error (control accuracy) and the speed of error convergence, balancing control accuracy and convergence speed. It reflects the control accuracy and speed of the control system. The smaller the value, the better the controller effect. The objective function is expressed using equation (17):

[0140]

[0141] Where Y represents the controller metric (i.e., the objective function of the IAFSA algorithm), and t represents the time taken for the AUV to complete this path tracking task.

[0142] It is worth noting that in the description of the classic artificial fish swarm algorithm, the ultimate goal of the artificial fish is to find the highest food concentration. However, in the AUV two-dimensional path tracking control task, it is necessary to find the control parameter value that minimizes the controller index, that is, to find the lowest food concentration value of the artificial fish (k1,k2,k3,k4,k5). The principle is the same, only the description is different.

[0143] During the process of optimizing controller parameters using the IAFSA algorithm (such as...) Figure 3 Once the artificial fish selects its current behavior, it will track the deviation e and its derivative generated by the complete path tracking task. Obtain the food concentration value, i.e., the controller index Y value.

[0144] If the objective function at the new position is better, then update the state; otherwise, retain the state.

[0145] c. Determine if the current iteration count has reached MAXGAN. If it has, exit the loop and record the optimal position of the artificial fish. If it has not, repeat the above loop.

[0146] Step 4: Substitute the five-dimensional coordinates of the optimal position of the artificial fish (i.e., the five control parameters in the forward velocity controller and the heading controller) back into the S-plane forward velocity controller and the S-plane heading controller to generate the path tracking trajectory as follows. Figure 6 As shown.

[0147] To verify the superiority of the improved IAFSA algorithm over the classic artificial fish swarm algorithm, both were simulated and optimized under the same parameter environment, and the curves of the objective function changing with the number of iterations were obtained (e.g., Figure 4 As shown in the figure, the classic fish swarm algorithm converges significantly slower than the improved fish swarm algorithm within the range of 10-40 iterations. After 40 iterations, the index value stagnates around 334, struggling to escape this local optimum and failing to reach the optimal value of the objective function in later stages. However, after the aforementioned three measures, the improved artificial fish swarm algorithm shows improved convergence speed in the middle of the iterations and escapes the local optimum after 60 iterations, optimizing the index value to 321.4, thus verifying the superiority of the improved fish swarm algorithm.

[0148] like Figure 5 As shown, comparing the path tracking trajectory before and after optimization reveals that the controller index value Y = 8062.7 before optimization, and Y = 321.4 after optimization, representing a 96% reduction. Clearly, the optimized path tracking controller performs better. Furthermore, comparing the two-dimensional path tracking trajectories before and after parameter tuning shows that the improved fish swarm algorithm significantly reduces oscillations and tracking errors during two-dimensional path tracking, validating the rationality and feasibility of applying the improved fish swarm algorithm to the parameter tuning of AUV S-plane path tracking control.

[0149] Finally, to verify the effectiveness of this algorithm in practical engineering, a two-dimensional path tracking field test of an AUV based on an S-plane controller was conducted in the waters near the Tanggu Fort in Tianjin. Similar to the simulation above, control parameter values ​​suitable for the specific time and location were obtained through IAFSA algorithm optimization, and these parameters were loaded into the AUV to complete the two-dimensional path tracking test. To avoid random errors caused by environmental factors and human actions, the same path tracking test was performed three times. After the test, the data was read and the flight path was plotted as follows. Figure 7 As shown in the figure, the control parameters optimized and tuned by the IAFSA algorithm enable the S-plane controller to perform the AUV two-dimensional path tracking task well. The AUV successfully starts from the starting point and reaches the preset endpoint along the planned path, maintaining a small path deviation, thus verifying the feasibility and effectiveness of the present invention.

[0150] It should be understood that modifications or equivalent substitutions can still be made to the specific embodiments of the present invention, and any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm, characterized in that, Includes the following steps: The classic artificial fish swarm algorithm is improved by adopting feeding behavior, adaptive step size, and field of view with decay factor to form the IAFSA algorithm; Design of an S-plane heading controller and an S-plane forward velocity controller for an autonomous underwater vehicle based on an S-plane controller; The IAFSA algorithm is used to optimize and tune five control parameters of the S-face forward velocity controller and the S-face heading controller to form an S-face path tracking controller based on the IAFSA algorithm. The improved classical artificial fish swarm algorithm, which incorporates feeding behavior, adaptive step size, and a field of view with a decay factor, forms the IAFSA algorithm, specifically including: The algorithm improves upon the classic artificial fish swarm algorithm by incorporating feeding competition behavior into the basic behaviors of artificial fish. The specific mathematical expression is as follows: (5)； In the formula, Y i Indicates the current position X i The food concentration, when it is relative to a certain location X in the field of vision. V Food concentration Y V When the ratio is less than a certain threshold, the artificial fish will engage in predation behavior. This threshold is represented by the product of the probability of engagement β=rand(0,1) and the crowding factor. The algorithm employs an adaptive step size to improve the classic artificial fish swarm algorithm. Specifically, a larger Step value is used when the current position is far from an area with higher food concentration, and a smaller Step value is used when the current position is close. The specific mathematical expression is as follows: (6)； In the formula, X i Indicates the current position of the artificial fish, X V This indicates a location within its field of vision, when X V If the food concentration at a given location is greater than the concentration at the current location, the artificial fish will swim a distance Step in that direction. The algorithm improves the classic artificial fish swarm algorithm by using a field of view with a decay factor. Specifically, when the algorithm starts iterating, a larger Visual value is used to expand the optimization range. As the iteration process progresses, the Visual value is gradually and appropriately decreased to accelerate the convergence speed. Mathematically, this is expressed as: (7)； in The decay rate represents the field of view, and k represents the number of iterations.

2. The method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm as described in claim 1, characterized in that, The specific steps for designing the S-plane heading controller for an autonomous underwater vehicle based on the S-plane controller are as follows: The control law is generated using the Sigmoid function to control the angle of the vertical servo, thereby controlling the heading of the AUV. The S-plane heading controller is designed as follows: (1)； In the formula, u1 is the forward velocity output of the thrust controller, k1 and k2 are control parameters, and e1 and These represent the differences between the actual forward velocity and the expected value. This represents the incremental output of the bow controller; where (2)； In the formula, v b This represents the expected forward velocity, where v is the current velocity of the AUV.

3. The method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm as described in claim 1, characterized in that, The specific steps for designing the S-face forward velocity controller for an autonomous underwater vehicle based on an S-face controller are as follows: Using an S-face controller, a control law is generated through the Sigmoid function to control the thrust of the propeller, thereby controlling the forward velocity of the AUV. The design of the S-face forward velocity controller is shown in formula (3): (3)； In the formula, u2 is the forward velocity output of the thrust controller, k3, k4, and k5 are control parameters, and e2 and These represent the differences between the actual forward velocity and the expected value. This indicates the incremental output of the forward speed controller; (4)； Where e2 represents the actual forward velocity, ψ b ψ represents the desired heading of the AUV in the geodetic coordinate system, β represents the actual heading of the AUV, and β represents the drift angle.

4. The method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm as described in claim 1, characterized in that, The improved classic artificial fish swarm algorithm incorporates feeding behavior, adaptive step size, and field of view with attenuation factor. The selection conditions for each behavior are as follows: ①Swarming behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions is n. f and the center position of the peer group X c The food concentration at the center is Y. c If equation (8) is satisfied, then it represents the center position X. c If the crowding level is low and the food concentration is high, then the artificial fish i will move towards the center position X. c Move one step in the direction and update the next position according to equation (9); if equation (8) is not satisfied, then perform foraging behavior according to the fixed step size value Step0: (8)； (9)； ② Tail-end collision behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k The number of observable companions n f The location of the companion with the highest food concentration, X j and its food concentration value Y j If equation (10) is satisfied, then it represents X. j If the crowding level is low and the food concentration is high, then the current artificial fish i is heading towards X. j Move one step in the direction and update the next position according to formula (11); if formula (10) is not satisfied, then perform foraging behavior according to the fixed step value Step0; (10)； (11)； Where X max This is the location with the highest food concentration; ③ Food-stealing behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k According to equation (6), the adaptive step size Step is obtained, and within its field of view Visual k A certain companion's position X was randomly observed within the room. V The food concentration at this location is Y. V If equation (12) is satisfied, then it represents X. V If the companion at location X has a much higher food concentration than the current artificial fish, then the artificial fish should move directly to location X. V Location; if equation (12) is not satisfied, then foraging behavior is performed according to the fixed step size value Step0: (12)； ④ Foraging behavior: Let the current position of artificial fish i be X. i The food concentration at this location is Y. i According to equation (7), the field of view for the current iteration number is obtained. k Within its field of vision k Randomly select a position X inside j The food concentration at this location is Y. j If Y i <Y j If Y = 0, it means that the food concentration at the selected location is high. In this case, the artificial fish moves one step in that direction and updates the next position according to formula (13). i >Y j If the random behavior is executed, that is, a random direction and step size are selected for movement, and the next position is updated according to formula (14); (13)； (14)。 5. The method for optimizing S-plane control parameters of an autonomous underwater vehicle based on the IAFSA algorithm as described in claim 4, characterized in that, The process of optimizing and tuning five control parameters of the S-face forward velocity controller and the S-face heading controller using the IAFSA algorithm to form an S-face path tracking controller based on the IAFSA algorithm specifically includes: Step 1: Initialize the fish swarm parameters. The initial fish swarm parameters include the number of artificial fish N, the initial field of view Visual, the number of attempts try_number, the crowding factor δ, the maximum number of iterations MAXGAN, and the decay factor α. Step 2: Initialize the artificial fish swarm within a given range, randomly generate position coordinates within the upper and lower limits of each dimension to form the artificial fish swarm; Step 3: Repeat the following steps using MAXGAN 10 times until the optimal food concentration value for the artificial fish population is found: a. Each artificial fish selects and performs the appropriate behavior according to the conditions; b. Evaluate the objective function value of the artificial fish's position after performing the corresponding action. If the objective function value of the new position is better, update the state; otherwise, retain the state. c. Determine if the current iteration count has reached MAXGAN. If it has, exit the loop and record the five-dimensional coordinates (k1, k2, k3, k4, k5) of the optimal position, where k1, k2, k3, k4, k5 are the control parameters of the S-face forward velocity controller and the S-face forward controller. If it has not reached, repeat the above loop. Step 4: Substitute the five-dimensional coordinates (k1, k2, k3, k4, k5) of the optimal food concentration position back into the S-face forward velocity controller and S-face forward controller of the AUV to complete the two-dimensional path tracking task and generate a path tracking trajectory map.

Images