Unmanned ship attitude control system and control method
By using INV2 combined navigation positioning module and data mining technology to eliminate outliers in small-scale operation unmanned ships, combining long and short-term memory networks, BP neural networks and two-way long and short-term memory networks for prediction and compensation control, the unmanned ship's navigation trajectory is solved. and the problem of instability of attitude improves prediction accuracy and navigation stability.
Patent Information
- Application Number
- CN202411434498.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2025-02-11
AI Technical Summary
Small-scale unmanned ships are easily affected by undercurrents, wind and waves during navigation, resulting in deviation of navigation trajectory and unstable attitude, which in turn causes problems such as collisions and off-target monitoring targets. The information obtained by the sensor is unstable, affecting the perception function. Accuracy and robustness.
The INV2 combined navigation and positioning module is used to obtain the heading, attitude and track data of the unmanned ship, and the outliers are eliminated through data mining methods to build the unmanned ship's state prediction model and adaptive compensation model, respectively, long-term and short-term memory network, BP neural network and Two-way long and short-term memory networks perform prediction and compensation control.
It improves the accuracy of the status prediction of unmanned ships, enhances the robustness of navigation stability and attitude control, and ensures that unmanned ships can maintain heading and attitude stability when disturbed.
Smart Images
Figure CN119414867A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of unmanned ship attitude control, and in particular relates to an unmanned ship attitude control system and a control method. Background Art
[0002] At present, all parts of the country are actively promoting the construction of the river and lake chief system, implementing hierarchical and segmented water responsibility areas, and conducting in-depth research on the basin, segmental, and seasonal characteristics of water crimes, and striving to form a joint supervision force for water-related industries and a working pattern of "joint management". At the same time, unmanned boats are used to effectively combat various water-related crimes, maintain the stability of water security and the safety of the ecological environment, and serve to ensure the economic development of water areas. In the past, the cleaning of water garbage mostly relied on manual labor, manual identification of garbage locations, and manual driving of boats to garbage points for salvage, which had high labor costs and low efficiency. In recent years, with the continuous progress and maturity of intelligent equipment research, it has become feasible to use unmanned boats to assist or even replace manual labor in performing repetitive and heavy work. Especially for water tasks, the use of unmanned boats instead of manual labor can well avoid the situation where relevant personnel are facing danger due to unstable water conditions, while liberating manpower and greatly ensuring human safety. In addition, many rivers and lakes still use relatively backward collection equipment to collect water quality information. In some waters, there is still a large-scale use of primitive manual observation methods or the establishment of observation stations for observation, and the data measured by these methods are not entirely reasonable. At the same time, the construction of observation stations is expensive and the detection range is small. Mobile water quality monitoring vessels can carry a variety of water quality measurement equipment, which can basically meet the requirements of conventional monitoring and emergency measurement, and greatly improve the ability to early warning and monitoring of water quality deterioration in lakes and rivers. Small workboats are increasingly used in the fields of water security patrol, cleaning, monitoring, etc., and their performance requirements are getting higher and higher.
[0003] However, undercurrents, wind and waves seriously affect its navigation track and posture. Specifically, it manifests in the following two aspects: On the one hand, the deviation of the navigation track makes small working boats prone to collision and monitoring target miss, which in turn causes risks and operational failures in small working boats during navigation. Track control, as the basis for research on unmanned boat regional cruising and target monitoring, is a key link to ensure that unmanned boats complete their tasks safely and efficiently. When using unmanned boats for water surface cleaning and rescue, it is necessary to monitor specific targets in real time to complete the cleaning and rescue tasks. When affected by factors such as wind and waves, the unmanned boat will deviate from the track, causing the monitoring target to miss the target and easily collide. Since the motion system of the unmanned boat is a typical under-actuated system, it has the characteristics of strong interference, model uncertainty and incompleteness, which makes the study of the unmanned boat track control system complex and difficult. On the other hand, small working unmanned boats can be used in a variety of scenarios by carrying different sensors. Equipped with water quality sensors for water sample collection in lakes or reservoirs, or to complete real-time reading of water quality data; equipped with ultrasonic detectors and infrared cameras for reconnaissance or scientific research surveys, equipped with positioning devices and life-saving gear for water rescue, etc. However, due to the changes in the posture of small unmanned operating ships, the information obtained by the visual sensors they carry is unstable, causing a serious reduction in the accuracy and robustness of their perception functions.
[0004] Therefore, it is urgent to predict the navigation trajectory, attitude and other parameters of small-scale unmanned boats, and adopt reasonable compensation control strategies to ensure that the changes in the navigation direction, trajectory, attitude and other quantities of the unmanned boat can meet its operational requirements. Summary of the invention
[0005] The purpose of the present invention is to provide an unmanned ship attitude control system and control method, which can eliminate abnormal values in the unmanned ship heading, attitude and track data obtained by the INV2 combined navigation and positioning module, and adopt different prediction control algorithms for the heading, track and attitude of the small operating unmanned ship. Each algorithm outputs the prediction data with the best accuracy, thereby improving the prediction accuracy of the entire unmanned ship state.
[0006] To achieve the above object, the present invention provides the following technical solutions: An unmanned ship attitude control system and control method, the system comprises an INV2 combined navigation and positioning module, a remote control module, a data transmission module, and a control module, the INV2 combined navigation and positioning module is used to obtain the heading, attitude and track data of the unmanned ship, the INV2 combined navigation and positioning module data is connected to the data transmission module, the data transmission module is used to upload the monitoring data of the INV2 combined navigation and positioning module to the remote control module, the remote control module is connected to the control module by electrical signals, the remote control module is used to send a control signal to the control module according to the uploaded monitoring data, and the control module is electrically connected to the control mechanism of the unmanned ship.
[0007] As a preferred solution of the present invention, the control mechanism includes a propeller drive component and an inertia wheel drive component, and the remote control module controls the propeller drive component and the inertia wheel drive component based on a PLC system.
[0008] The present invention also provides an unmanned ship attitude control method, based on the unmanned ship attitude control system, comprising the following steps: ① Use the INV2 integrated navigation and positioning module to obtain the heading, attitude and track data of the unmanned ship, specifically the heading angle data, attitude placement angle data and track change trajectory relationship of the unmanned ship; ② Use data mining methods to obtain obvious outliers of the heading, attitude, and track data of the unmanned ship; ③ Eliminate obvious outliers; ④ Use the unmanned ship heading, attitude and track data after removing obvious outliers to build an unmanned ship state prediction model; ⑤Construct an adaptive compensation model for unmanned ships.
[0009] As a preferred solution of the present invention, the method for obtaining obvious outliers in step ② is: using a wavelet transform method to extract the lowest frequency component of the original sequence as the running trend of the sequence; removing the running trend of the sequence from the original sequence to obtain a new sequence and calculate the local outlier factor value of the new sequence; using a box plot method to adaptively determine the position of the outlier to obtain the outlier.
[0010] As a preferred solution of the present invention, the long short-term memory network model is used in step ④ to predict the unmanned ship heading data, specifically including: S401. Preprocessing the heading data; S402. Initialize long short-term memory network model parameters; S403. Using the preprocessed heading data to train a long short-term memory network model; S404. Set the initialization parameters of the PSO algorithm; S405. If the trained LSTM network model meets the error requirement or the maximum number of iterations, directly test the LSTM network model and output the predicted value; S406. If the trained LSTM network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S407. Output optimized long short-term memory network model parameters; S408. Initialize the optimized long short-term memory network model parameters and regress and train the long short-term memory network model again.
[0011] As a preferred solution of the present invention, the BP neural network model is used in step ④ to predict the unmanned ship track data, specifically including: S411. Preprocessing the track data; S412. Initialize BP neural network model parameters; S413. Use the preprocessed track data to train the BP neural network model; S414. Set the initialization parameters of the PSO algorithm; S415. If the trained BP neural network model meets the error requirement or the maximum number of iterations, the long short-term memory network model is directly tested and the predicted value is output; S416. If the trained BP neural network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S417. Output optimized BP neural network model parameters; S418. Initialize the optimized BP neural network model parameters and regress and train the BP neural network model again.
[0012] As a preferred solution of the present invention, in step ④, a bidirectional long short-term memory network model is used to predict the unmanned ship posture data, specifically including: S421. Preprocessing the posture data; S422. Initialize bidirectional long short-term memory network model parameters; S423. Using the preprocessed posture data to train a bidirectional long short-term memory network model; S424. Set the initialization parameters of the GA algorithm; S425. If the trained bidirectional long short-term memory network model meets the error requirement or the maximum number of iterations, directly test the long short-term memory network model and output the predicted value; S426. If the trained bidirectional long short-term memory network model does not meet the error requirement or the maximum number of iterations, the set GA algorithm initialization parameters are used to perform a crossover mutation operation to process the bidirectional long short-term memory network model parameters; S427. Output optimized bidirectional long short-term memory network model parameters; S428. Initialize the optimized bidirectional long short-term memory network model parameters and regress and train the bidirectional long short-term memory network model again.
[0013] As a preferred solution of the present invention, the specific adaptive compensation in step ⑤ is: by constructing an unmanned ship state prediction model to obtain the heading prediction data, track prediction data and attitude prediction data of the unmanned ship, the Kalman filter model is used to adaptively compensate the output of the heading prediction data, the fuzzy PID adaptive compensation model is used to adaptively compensate the output of the track prediction data, and the improved PID adaptive compensation model is used to adaptively compensate the output of the attitude prediction data.
[0014] The beneficial effects of the present invention are: The INV2 integrated navigation and positioning module is used to obtain the heading, attitude and track data of the unmanned ship. In view of the fact that INV2 is easily affected by external environmental factors when collecting data, the data mining method is used to process the outliers of the heading, track and attitude data with massive, high frequency and nonlinearity. The wavelet transform method is used to extract the lowest frequency component and use it as the operating trend. Logical operations are used to obtain the new state data sequence and its local outlier factor. The box plot is used to adaptively determine the location of the obvious outliers in the new state data and obtain the obvious outliers to improve the subsequent prediction accuracy. 2. Different prediction control algorithms are used for the heading, track and attitude of small-scale unmanned boats. Each algorithm outputs the prediction data with the best accuracy, thereby improving the prediction accuracy of the entire unmanned boat status; 3. Different compensation algorithms are used to compensate for the disturbed data of the heading and track of the small unmanned boat. The heading and track are controlled by adjusting the speed of the propeller to achieve stable navigation. The compensation control of the small unmanned boat operation adopts an algorithm to adjust the position of the inertia wheel, thereby compensating for the posture changes of the small operating unmanned boat, ensuring that the unmanned boat can maintain a stable posture when it is disturbed by external disturbances. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 It is a schematic diagram of the process provided by the present invention; Figure 2 It is a schematic diagram of the unmanned ship state prediction structure provided by the present invention; Figure 3 It is a schematic diagram of the adaptive compensation control structure provided by the present invention. DETAILED DESCRIPTION
[0016] In order to further understand the content, features and effects of the present invention, the following embodiments are given as examples and described in detail with reference to the accompanying drawings.
[0017] Please also refer to Figures 1 to 3 The unmanned ship attitude control system and control method according to the embodiment of the present invention will be described in detail below with reference to the accompanying drawings.
[0018] like Figure 1 As shown, the unmanned ship attitude control system includes an INV2 combined navigation and positioning module, a remote control module, a data transmission module, and a control module. The INV2 combined navigation and positioning module is used to obtain the heading, attitude and track data of the unmanned ship. The INV2 combined navigation and positioning module data is connected to the data transmission module. The data transmission module is used to upload the monitoring data of the INV2 combined navigation and positioning module to the remote control module. The remote control module is electrically connected to the control module. The remote control module is used to send a control signal to the control module according to the uploaded monitoring data. The control module is electrically connected to the control mechanism of the unmanned ship.
[0019] Furthermore, the INV2 combined navigation and positioning module can be used to digitize the heading, attitude and track data of the unmanned ship, and obtain its heading angle data, attitude placement angle data and track change trajectory relationship quantity.
[0020] The control mechanism includes a propeller drive component and an inertia wheel drive component, and the remote control module controls the propeller drive component and the inertia wheel drive component based on a PLC system.
[0021] The unmanned ship attitude control method, based on the unmanned ship attitude control system, comprises the following steps: ① Use the INV2 integrated navigation and positioning module to obtain the heading, attitude and track data of the unmanned ship, specifically the heading angle data, attitude placement angle data and track change trajectory relationship of the unmanned ship; ② Use data mining methods to obtain obvious outliers of the heading, attitude, and track data of the unmanned ship; ③ Eliminate obvious outliers; Firstly, the lowest frequency component of the original sequence is extracted by using the EWT method as the running trend of the sequence; secondly, the running trend of the sequence is removed from the original sequence to obtain a new sequence and calculate the local outlier factor (LOF) value of the new sequence; finally, the box plot method is used to adaptively determine the position of the outliers, obtain the outliers, and then remove them.
[0022] ④ Use the unmanned ship heading, attitude and track data after removing obvious outliers to build an unmanned ship state prediction model; ⑤Construct an adaptive compensation model for unmanned ships.
[0023] The method for obtaining obvious outliers in step ② is as follows: use the wavelet transform method to extract the lowest frequency component of the original sequence as the running trend of the sequence; remove the running trend of the sequence from the original sequence to obtain a new sequence and calculate the local outlier factor value of the new sequence; use the box plot method to adaptively determine the position of the outlier and obtain the outlier.
[0024] In view of the nonlinear, reciprocating and difficult to accurately predict characteristics of the heading, track and attitude changes of small-scale unmanned ships, this project intends to use the long short-term memory network (LSTM) model to predict the heading of small-scale unmanned ships, and use the particle swarm optimization algorithm (PSO) to optimize the parameters of the LSTM model to improve the prediction accuracy. For the track of unmanned ships, this project intends to use the improved BP neural network for prediction, and use PSO to optimize the parameters in BP to improve its prediction accuracy. For the attitude of unmanned ships, this project intends to use the bidirectional long short-term memory network (Bi-LSTM) prediction model, and use the genetic algorithm (GA) to optimize the parameters in Bi-LSTM.
[0025] In step ④, the long short-term memory network model is used to predict the unmanned ship heading data, specifically including: S401. Preprocessing the heading data; S402. Initialize long short-term memory network model parameters; S403. Using the preprocessed heading data to train a long short-term memory network model; S404. Set the initialization parameters of the PSO algorithm; S405. If the trained LSTM network model meets the error requirement or the maximum number of iterations, directly test the LSTM network model and output the predicted value; S406. If the trained LSTM network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S407. Output optimized long short-term memory network model parameters; S408. Initialize the optimized long short-term memory network model parameters and regress and train the long short-term memory network model again.
[0026] In step ④, the BP neural network model is used to predict the unmanned ship track data, specifically including: S411. Preprocessing the track data; S412. Initialize BP neural network model parameters; S413. Use the preprocessed track data to train the BP neural network model; S414. Set the initialization parameters of the PSO algorithm; S415. If the trained BP neural network model meets the error requirement or the maximum number of iterations, the long short-term memory network model is directly tested and the predicted value is output; S416. If the trained BP neural network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S417. Output optimized BP neural network model parameters; S418. Initialize the optimized BP neural network model parameters and regress and train the BP neural network model again.
[0027] In step ④, a bidirectional long short-term memory network model is used to predict the attitude data of the unmanned ship, specifically including: S421. Preprocessing the posture data; S422. Initialize bidirectional long short-term memory network model parameters; S423. Using the preprocessed posture data to train a bidirectional long short-term memory network model; S424. Set the initialization parameters of the GA algorithm; S425. If the trained bidirectional long short-term memory network model meets the error requirement or the maximum number of iterations, directly test the long short-term memory network model and output the predicted value; S426. If the trained bidirectional long short-term memory network model does not meet the error requirement or the maximum number of iterations, the set GA algorithm initialization parameters are used to perform a crossover mutation operation to process the bidirectional long short-term memory network model parameters; S427. Output optimized bidirectional long short-term memory network model parameters; S428. Initialize the optimized bidirectional long short-term memory network model parameters and regress and train the bidirectional long short-term memory network model again.
[0028] When operating, a small-scale unmanned boat is subject to external interference, which causes the heading, attitude and track of the unmanned boat to change, resulting in the inability to operate normally. In order to make the unmanned boat have higher stability, in step ⑤, the Kalman filter model is used to adaptively compensate the output of the heading prediction data, the fuzzy PID adaptive compensation model is used to adaptively compensate the output of the track prediction data, and the improved PID adaptive compensation model is used to adaptively compensate the output of the attitude prediction data. Specifically, the PID adaptive compensation model collects the heading angle data, attitude placement angle data and track change trajectory relationship of the above-mentioned unmanned boat as sensor information, and performs fuzzy reasoning by determining the difference between the information and the data information of the standard attitude, and finally makes parameters of the fuzzy reasoning concrete. The above-mentioned PID adaptive compensation model is an existing compensation mechanism.
Claims
1. An unmanned ship attitude control system, characterized in that: The unmanned ship attitude control system includes an INV2 combined navigation and positioning module, a remote control module, a data transmission module, and a control module. The INV2 combined navigation and positioning module is used to obtain the unmanned ship's heading, attitude and track data. The INV2 combined navigation and positioning module data is connected to the data transmission module. The data transmission module is used to upload the monitoring data of the INV2 combined navigation and positioning module to the remote control module. The remote control module is electrically connected to the control module. The remote control module is used to send a control signal to the control module according to the uploaded monitoring data. The control module is electrically connected to the control mechanism of the unmanned ship.
2. The unmanned ship attitude control system according to claim 1, characterized in that: The control mechanism includes a propeller drive component and an inertia wheel drive component, and the remote control module controls the propeller drive component and the inertia wheel drive component based on a PLC system.
3. An unmanned ship attitude control method, based on the unmanned ship attitude control system in claim 1, characterized in that: The following steps are involved: ① Use the INV2 integrated navigation and positioning module to obtain the heading, attitude and track data of the unmanned ship, and specifically obtain the heading angle data, attitude placement angle data and track change trajectory relationship of the unmanned ship; ②Using data mining methods to obtain obvious outliers of the heading, attitude and track data of the unmanned ship; ③ Eliminate obvious outliers; ④ Use the unmanned ship heading, attitude and track data after removing obvious outliers to build an unmanned ship state prediction model; ⑤Construct an adaptive compensation model for unmanned ships.
4. The unmanned ship attitude control method according to claim 3, characterized in that: The method for obtaining obvious outliers in step ② is: using the wavelet transform method to extract the lowest frequency component of the original sequence as the running trend of the sequence; removing the running trend of the sequence from the original sequence to obtain a new sequence and calculate the local outlier factor value of the new sequence; using the box plot method to adaptively determine the position of the outlier and obtain the outlier.
5. The unmanned ship attitude control method according to claim 3, characterized in that: In step ④, the long short-term memory network model is used to predict the unmanned ship heading data, specifically including: S401. Preprocessing the heading data; S402. Initialize long short-term memory network model parameters; S403. Using the preprocessed heading data to train a long short-term memory network model; S404. Set the initialization parameters of the PSO algorithm; S405. If the trained LSTM network model meets the error requirement or the maximum number of iterations, directly test the LSTM network model and output the predicted value; S406. If the trained LSTM network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S407. Output optimized long short-term memory network model parameters; S408. Initialize the optimized long short-term memory network model parameters and regress and train the long short-term memory network model again.
6. The unmanned ship attitude control method according to claim 3, characterized in that: In step ④, the BP neural network model is used to predict the unmanned ship track data, specifically including: S411. Preprocessing the track data; S412. Initialize BP neural network model parameters; S413. Use the preprocessed track data to train the BP neural network model; S414. Set the initialization parameters of the PSO algorithm; S415. If the trained BP neural network model meets the error requirement or the maximum number of iterations, the long short-term memory network model is directly tested and the predicted value is output; S416. If the trained BP neural network model does not meet the error requirement or the maximum number of iterations, the speed and position of each particle are updated using the set PSO algorithm initialization parameters; S417. Output optimized BP neural network model parameters; S418. Initialize the optimized BP neural network model parameters and regress and train the BP neural network model again.
7. The unmanned ship attitude control method according to claim 3, characterized in that: In step ④, a bidirectional long short-term memory network model is used to predict the attitude data of the unmanned ship, specifically including: S421. Preprocessing the posture data; S422. Initialize bidirectional long short-term memory network model parameters; S423. Using the preprocessed posture data to train a bidirectional long short-term memory network model; S424. Set the initialization parameters of the GA algorithm; S425. If the trained bidirectional long short-term memory network model meets the error requirement or the maximum number of iterations, directly test the long short-term memory network model and output the predicted value; S426. If the trained bidirectional long short-term memory network model does not meet the error requirement or the maximum number of iterations, the set GA algorithm initialization parameters are used to perform a crossover mutation operation to process the bidirectional long short-term memory network model parameters; S427. Output optimized bidirectional long short-term memory network model parameters; S428. Initialize the optimized bidirectional long short-term memory network model parameters and regress and train the bidirectional long short-term memory network model again.
8. The unmanned ship attitude control method according to any one of claims 3 to 7, characterized in that: The specific adaptive compensation of step ⑤ is: by constructing an unmanned ship state prediction model to obtain the heading prediction data, track prediction data and attitude prediction data of the unmanned ship, the Kalman filter model is used to adaptively compensate the output of the heading prediction data, the fuzzy PID adaptive compensation model is used to adaptively compensate the output of the track prediction data, and the improved PID adaptive compensation model is used to adaptively compensate the output of the attitude prediction data.