Search

CN-122018281-A - Submarine navigation device pose control method based on BP neural network PID control

CN122018281ACN 122018281 ACN122018281 ACN 122018281ACN-122018281-A

Abstract

The invention provides a submarine navigation attitude control method based on BP neural network PID control, which aims at a submarine navigation 'main pushing and auxiliary pushing' propeller system, realizes self-tuning of PID control parameters and is beneficial to improving the precision and efficiency of depth setting and directional control under the condition of disturbance. The invention adopts the characteristics of mutual matching and mutual restriction of three parameters of PID control, the conventional PID controller has fixed parameters, no self-adaptive capacity, and the optimal solution of various nonlinear combinations of the three parameters in the PID controller can be obtained through self-learning by means of the mapping capacity of the BP neural network to the nonlinear function, thereby ensuring the depth fixing, directional control precision and dynamic performance of the submarine under complex sea conditions.

Inventors

  • WANG YAJUN
  • GUO XIAOWEI
  • ZHANG JIARUI
  • XUE HONG
  • LIU DAN

Assignees

  • 中国电子科技集团公司第二十研究所

Dates

Publication Date
20260512
Application Date
20251016

Claims (8)

  1. 1. The submarine vehicle pose control method based on BP neural network PID control is characterized by comprising the following steps of: s1, determining the structure of a BP neural network, wherein the structure comprises the number of layers and the number of nodes of each layer, selecting an initial value of a connection weight between each layer, and selecting an activation function; s2, forward network calculation; S3, selecting a performance index function E (k) as the quadratic power of the output error: (10) Wherein r (k) is a set heading angle, y (k) is an actual heading angle at the current moment, namely, a control law u (k) output by the PID controller acts on a controlled object and then is measured by a compass, and an incremental PID control law is as follows: (11) Wherein e (K) is an output error at the current moment, namely an error between a set heading angle and an actual heading angle, namely r (K) -y (K), and the control parameter K P 、K I 、K D is the output O 1 (3) (k) 、O 2 (3) (k) 、O 3 (3) (K) of the BP neural network; s4, if the performance index function E (k) is smaller than or equal to a threshold value, ending the training stage, otherwise, reversely transferring errors, and modifying the weight layer by layer according to a gradient descent method, so that the network weight coefficient is modified according to the learning rate eta until the performance index E (k) is smaller than or equal to the threshold value; s5, adding 1 to k, returning the updated weight value in the S4 to the S2 to restart calculation, and if the performance index function E (k) is smaller than or equal to a threshold value, indicating that the BP neural network at the moment meets the requirement, fixing network parameters, namely connecting weight values between the neurons of two adjacent layers, ending a training stage and entering a working stage; S6, in the working stage, a steering mode is needed to be considered when calculating the error e (k), wherein the steering mode is that the submarine is selected to turn left or right when the heading is adjusted, and in order to save energy and improve response speed, the minimum path steering is selected, namely, the path with the absolute value of the difference value within 180 degrees is selected to steer, so that the error e (k) is converted into a range of-180 degrees to 180 degrees, the error is right, and the error is left when the error is negative; And S7, the work of the propeller is controlled by a proportional valve and two left and right electromagnetic valves, the proportional valve controls the rotating speed of the propeller, the output of the electromagnetic valve is the switching value to control whether the propeller works, the bow side pushing proportional valve controls the bow side pushing rotating speed, the bow side pushing left electromagnetic valve is opened, the bow side pushing right electromagnetic valve is closed to jointly control the bow part to move leftwards, and simultaneously, the stern side pushing proportional valve controls the stern side pushing rotating speed, the stern side pushing left electromagnetic valve is closed, the stern side pushing right electromagnetic valve is opened to jointly control the stern part to move rightwards, and the bow and the stern cooperatively steer the boat body leftwards.
  2. 2. The submarine pose control method based on BP neural network PID control according to claim 1, wherein the method is characterized by comprising the following steps: In the S1, the BP neural network is used for optimizing PID control parameters, so that the node number of an output layer is 3, the input of the BP neural network respectively corresponds to a proportional coefficient K P , an integral coefficient K I and a differential coefficient K D of PID control, the input of the BP neural network comprises setting a heading angle, an actual output value, an error and a bias term, the bias term takes 1, the node number of the input layer is 4, the hidden layer number of the BP neural network is 1, the node number is 5, and the finally formed BP neural network structure is that the input layer comprises 4 nodes which are respectively a set value, an actual value, an error and a bias term, the hidden layer comprises 5 nodes, and the output layer comprises 3 nodes including the proportional coefficient, the integral coefficient and the differential coefficient; The connection weight initial values of the hidden layer and the output layer are random values on [ -1,1 ]; The activation function of the hidden layer selects the hyperbolic tangent function: (1) e is the base of the natural logarithm, and the derivative of the activation function of the hidden layer is: (2) The output layer nodes are K P 、K I and K D ,K P 、K I and K D respectively, which cannot be numbers smaller than zero, so the non-negative hyperbolic tangent function is selected as the activation function of the output layer: (3) the derivative of the activation function of the output layer is: (4)。
  3. 3. the submarine pose control method based on BP neural network PID control according to claim 2, wherein the method is characterized by comprising the following steps: In the step S2, the output of the input layer of the BP neural network is: (5) wherein x i (k) is the input of the neural network input layer at the current moment, and the upper corner mark (1) represents the input layer; The input of the hidden layer is: (6) Wherein ω ji (2) (k) represents the connection weight between the hidden layer neuron j and the input layer neuron i, and (1) of the upper corner mark represents the hidden layer; The output of the hidden layer is: (7) Wherein, f () is a hyperbolic tangent function shown in the formula (1); the input of the output layer is: (8) Wherein ω lj (3) (k) represents the connection weight between the output layer neuron l and the hidden layer neuron j; the output of the output layer is: (9) In the formula (9), g () is a non-negative hyperbolic tangent function shown in the formula (1.3).
  4. 4. The submarine pose control method based on BP neural network PID control according to claim 3, wherein the method is characterized by comprising the following steps: In the step S4, the learning rate η is 0.3, and the weight updating calculation formula of the output layer is as follows: (12) In the formula, (13) G' () is the derivative of the non-negative hyperbolic tangent function in equation (4); The weight updating calculation formula of the hidden layer is as follows: (14) In the formula, (15) F' () is the derivative of the hyperbolic tangent function of equation (2).
  5. 5. The submarine pose control method based on BP neural network PID control according to claim 1, wherein the method is characterized by comprising the following steps: In the step S6, the specific calculation method is divided into 4 cases: 1. If it is And (2) and Then ; 2. If it is And (2) and Then ; 3. If it is And (2) and Then ; 4. If it is And (2) and Then 。
  6. 6. The submarine pose control method based on BP neural network PID control according to claim 1, wherein the method is characterized by comprising the following steps: in the step S7, in order to prevent the propeller from being frequently started, a dead zone of +/-0.5 degrees is set according to the automatic orientation precision requirement, the error is in the dead zone, the orientation target is considered to be reached, all the directional electromagnetic valves are closed, and the propeller does not work at the moment; When the error is outside the dead zone, steering of the boat body is controlled through opening/closing of electromagnetic valves in different directions, and the rotating speed of the propeller is controlled through a proportional valve, wherein the control quantity of the proportional valve is obtained by a PID control law based on a BP neural network, the BP neural network entering a stable working stage continuously adjusts three control parameter proportional coefficients K P , integral coefficients K I and differential coefficients K D of the PID control law, and the output of the PID controller is distributed by thrust to obtain the control quantity of the proportional valve for controlling the bow and stern side pushing rotating speeds.
  7. 7. An electronic device, comprising: one or more processors; a memory; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method of any of claims 1-6.
  8. 8. A computer readable storage medium, characterized in that the computer readable storage medium stores a program code, which is callable by a processor for performing the method according to any one of claims 1-6.

Description

Submarine navigation device pose control method based on BP neural network PID control Technical Field The invention relates to the field of neural networks, in particular to a control method based on the neural network. Background The current navigation control system of the submarine is widely applied to PID control algorithm, the control parameters of the control system are fixed and cannot be automatically adjusted along with the change of conditions, and the accuracy and efficiency of positioning and attitude determination can be reduced under disturbance. The invention aims to design a PID control method capable of realizing parameter self-tuning optimization, which is suitable for automatic navigation control of a submarine craft system so as to improve the performance of positioning and attitude determination control. Various types of submarines are widely applied to the fields of submarine resource exploration, ocean science discovery, rescue of a failure submarine, global climate change research, ocean archaeology, submarine military target detection and the like at present, and have important strategic positions in the construction of ocean countries. In the working process of the submarine, complex sea conditions such as complex ocean current disturbance which is difficult to measure, rugged and changeable submarine topography and the like can be faced, and how to ensure autonomous, safe and reliable navigation control is the basis for deep sea exploration of the submarine. Taking a rescue scene of the accident submarine as an example, the rescue boat can sail to the sea area where the accident submarine is located along with the supporting mother ship, after the mother ship is used for laying the rescue boat, the rescue boat needs to sail and submerge to the position where the accident submarine is located, the posture of the rescue boat is adjusted, and the rescue boat is butted with the submarine by utilizing the butt joint skirt so as to transfer trapped personnel. The whole process can be subjected to a plurality of interference factors such as ocean current disturbance, complex terrain, target movement and the like. Therefore, the depth setting and the directional control performance of the lifeboat are extremely important. At present, a great deal of research results are available for navigation control systems of the submarine, such as 201811580167.4 patent applied by China university of ocean, an ESO-based underwater robot PID motion control method is designed, 202310921070.X patent applied by Nanjing university of post is designed, an fuzzy PID-based underwater robot depth control method is designed, 201811013607.8 patent applied by China university of science and technology is designed, and a bow-stern combined steering control strategy is designed for depth control of the submarine. The disadvantages of the prior art are: The current navigation control algorithm is mainly classical PID control and variants thereof, parameters are fixed, the optimal solution of the parameters of the controller is continuously changed under the interference of ocean current disturbance, target drift and the like, and the conventional PID controller has poor robustness and is easy to generate control deviation because the parameters cannot be self-optimized. In order to achieve both mobility during target entering and stability during wave-resisting and current-resisting, the submarine is required to be provided with a plurality of pairs of propellers, bow rudder and a stern rudder, the steering engine works in a cruising state with higher navigational speed, and fine positioning and attitude determination control is mainly realized through a propeller system. In the current research on the navigation control method of the submarine, there are few specific depth and orientation control methods for the 'main pushing and auxiliary pushing' propeller system. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides a submarine pose control method based on BP neural network PID control. Aiming at a main pushing and auxiliary pushing propeller system of the submarine, the invention researches a navigation control method, and provides a submarine pose control method based on BP (Back Propagation) neural network PID control, which realizes self-tuning of PID control parameters and is beneficial to improving the precision and efficiency of depth setting and directional control under the condition of disturbance. The technical scheme adopted for solving the technical problems is as follows: s1, determining the structure of a BP neural network, wherein the structure comprises the number of layers and the number of nodes of each layer, selecting an initial value of a connection weight between each layer, and selecting an activation function; s2, forward network calculation; S3, selecting a performance index function E (k) as the quadratic power of the output error: (10) Wher