CN105184001B - A kind of aircushion vehicle secure border decision method - Google Patents
A kind of aircushion vehicle secure border decision method Download PDFInfo
- Publication number
- CN105184001B CN105184001B CN201510593660.XA CN201510593660A CN105184001B CN 105184001 B CN105184001 B CN 105184001B CN 201510593660 A CN201510593660 A CN 201510593660A CN 105184001 B CN105184001 B CN 105184001B
- Authority
- CN
- China
- Prior art keywords
- hovercraft
- state
- neural network
- output
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention is to provide a kind of aircushion vehicle secure border decision method.Based on Lyapunov Theory of Stability and actual hovercraft handling characteristic, the in stable condition Rule of judgment of aircushion vehicle is provided.Then, the state in aircushion vehicle working space is divided into by point of safes and two class of point of instability according to Rule of judgment, stability region judgment models is obtained as sample space training BP neural network.Finally, using dichotomy and BP neural network stability region judgment models come search work space, and then secure border is obtained.During aircushion vehicle controls in real time, secure border can ensure aircushion vehicle safe navigation as the basis for estimation of emergency flight control input in dangerous working condition.Secure border of the invention by calculating the aircushion vehicle under current control law, can reduce aircushion vehicle unstability caused by the erroneous judgement of operator, can also reduce the working strength and mental burden of operator, have great practical value.
Description
Technical Field
The invention relates to a method for judging a safety boundary of a hovercraft, in particular to an off-line calculation method for the safety boundary of the hovercraft based on a neural network.
Background
In the process of navigation and control of the hovercraft, under the disturbance of wind waves or the damage of an apron or the failure of control equipment, improper control of a driver and the like, different risks such as high-speed head burying, low-speed ship turning, high-speed tail flicking and the like can occur. If the driver does not operate properly, the dangerous state of the hovercraft during navigation is represented as follows: as the gyroscopic angular velocity increases, the sideslip angle gradually increases, followed by a rapid increase in the roll and pitch angles. Therefore, in order to ensure the navigation safety, a safety boundary influencing the main variables of the navigation safety needs to be calculated, and whether the hovercraft is subjected to emergency control in the current state is judged.
In foreign publications, the navigational safety limits specified for the U.S. LCAC boat are: the maximum speed pitch angle should generally not be less than 1 deg., when the speed is 50kn and the maximum sideslip angle is 15 deg., the maximum cornering angular velocity is 2 deg/s. The limitation of dangerous working conditions of the Russian hovercraft in navigation is as follows: the rudder can be fully steered at a low speed of 14kn, the limited maximum rotation angular speed is 3.5 degrees/second, and the maximum sideslip angle is 20 degrees; when the navigation speed is 35kn, the rudder angle is limited to 5 degrees, the maximum rotation rate is limited to 1.2 degrees/second, and the maximum sideslip angle is 2 degrees; the rudder cannot be driven for a long time at 40 kn.
In the domestic published literature, according to the actual ship test data, the safety limit of each parameter during navigation is as follows: the normal range of the transverse inclination angle is 0.5-2.5 degrees, and the initial longitudinal inclination angle is larger than the initial longitudinal inclination angle. Under the normal working state, the longitudinal inclination angle of the hovercraft is increased and even can be reached along with the increase of the navigation speed of the hovercraft. At 30kn the gyroscopic angular velocity should not exceed 4 degrees/sec and the sideslip angle should not exceed 20 degrees.
The safety limit is set for manual operation, the influence of a control law is not considered, and when the hovercraft is operated by using the automatic driving system, a safety boundary in a current control law state needs to be provided, so that a basis is provided for whether emergency control is performed.
Disclosure of Invention
The invention aims to provide a hovercraft safety boundary judgment method which provides a basis for whether the hovercraft performs emergency control or not, can reduce hovercraft accidents caused by misjudgment of drivers and can improve navigation stability.
The purpose of the invention is realized by the following steps:
(1) Acquiring a sample space, randomly acquiring a state sequence according to a uniformly distributed principle according to the working ranges of six parameters of a rotation angular velocity, a sideslip angle, a transverse inclination angular velocity, a longitudinal inclination angle and a longitudinal inclination angular velocity of the hovercraft, taking the state sequence as an initial state of the hovercraft, bringing the initial state sequence into a control law to judge the stability of the system after a set time, acquiring a logic value sequence whether the system is stable, and taking the initial state sequence and the corresponding logic value sequence whether the system is stable as the sample space;
(2) Training the BP neural network by using a sample space, so that the initial state in any working range is input, and the BP neural network can obtain a logic value of whether the system is stable;
(3) The states of six parameters in any given working range are input into the trained BP neural network, and the boundary value of the stable state in the working range is searched by adopting a dichotomy according to the output of the neural network.
The stability of the hovercraft in different states and different control laws is different, the method provided by the invention substitutes the current control law of the hovercraft into the model to calculate the logic value of the state stability, and then uses the initial state and the logic value as the training sample of the neural network. In order to enable the output of the BP neural network to accurately determine the stability of the system, an appropriate network structure and a learning function need to be selected. The establishment of the neural network and the dichotomy search of the safety boundary are the core in the invention.
According to the Lyapunov stability theory and the actual hovercraft operating characteristics, the stability judgment condition is designed aiming at six main parameters of the rotation angular velocity, the sideslip angle, the roll angle velocity, the pitch angle and the pitch angle velocity which influence the navigation stability of the hovercraft, the neural network is trained by utilizing the discrete state and the stability of the working space, the judgment of the system stability under the input of any state is realized, and then the safety boundary under the current state of the hovercraft is obtained by utilizing the dichotomy calculation.
Compared with the prior art, the invention has the following advantages and effects: the state of the hovercraft changes rapidly in the navigation process, and if the stability of the system is judged by an operator, on one hand, great working pressure can be caused to a driver, and on the other hand, misjudgment can be easily caused due to time urgency. The hovercraft safety boundary calculation method provided by the invention can calculate the stable boundary of the state according to the current control law and the hovercraft model, so that the pressure of a driver is reduced.
In the real-time control process of the hovercraft, the safety boundary can be used as a judgment basis for emergency control investment in dangerous working conditions, and safe navigation of the hovercraft is guaranteed. By calculating the safety boundary of the hovercraft under the current control law, the invention can reduce the instability of the hovercraft caused by misjudgment of operators, can also reduce the working strength and mental burden of the operators, and has great practical value.
Drawings
Fig. 1 is a flow chart for acquiring a sample space.
FIG. 2 is a diagram showing the training effect of a BP neural network with a hidden layer structure of 10 × 10.
Fig. 3 shows the output curve of the neural network under the stable condition of the hovercraft.
FIG. 4 shows the output curve of the neural network in the unstable hovercraft condition.
FIG. 5 is a flowchart of a hovercraft safety boundary algorithm process based on BP neural network and dichotomy.
FIG. 6 Hovercraft level safety limits.
Detailed Description
The present invention is described in further detail below with reference to the attached drawing figures.
The method mainly comprises the steps of obtaining a sample space, training a BP neural network and calculating a safety boundary by adopting a dichotomy.
With reference to fig. 1, the process of obtaining the sample space includes the following steps:
1. acquiring state information of the hovercraft, and respectively using six parameters of sideslip angle, gyration angular velocity, transverse inclination angle velocity, longitudinal inclination angle and longitudinal inclination angle velocity as variable x 1 ,x 2 ,x 3 ,x 4 ,x 5 ,x 6 Expressed, the vector X = [ X ] 1 ,x 2 ,x 3 ,x 4 ,x 5 ,x 6 ] T The current control law of the hovercraft is recorded as tau, and the model of the hovercraft is recorded asWherein X (0) represents an initial state and t represents time;
2. and selecting an initial state sequence according to the working range of the state. The operating range of the state is represented as follows:
x imin <x i <x imax ,i=1,…,6
x imin ,x imax respectively the lower and upper operating bounds of the state variable,
the number of sequences is determined by the following formula:
wherein epsilon i The maximum deviation of the state data.
And obtaining an initial state sequence (X (0)) according to the principle of uniformly distributing and taking points in a working space.
3. Bringing (X (0), tau) into the hovercraft model, run time T s And then judging whether the system state meets the following conditions:
||x i ||<σ i ,i=1,3,5
wherein sigma i >0,
If the above conditions are satisfied, the air cushion vehicle is considered to be stable under the initial state X (0) and the control law τ, and the corresponding logical value is Y =0; otherwise, the hovercraft is considered to be unstable, and the corresponding logical value is Y =1. And forming a sample space { X (0), Y } by using the initial state sequence { X (0) } and the corresponding logic value sequence { Y }, wherein the sample space { X (0), Y } is used as a training sample of the neural network.
In order to improve the prediction accuracy of the model, according to the characteristics of data of the hovercraft, the neural network selects a BP neural network, the structures of an input layer, a hidden layer and an output layer are selected to be 4 multiplied by 10 multiplied by 1, the activation function of neurons in the hidden layer is selected to be a tansig function, the activation function of neurons in the output layer is selected to be a logsig function, and a Levenberg-Marrdquat algorithm is adopted as a network training function. By off-line training of the sample space, the trained neural network can output a stable logical value V according to the state X (0) of the hovercraft.
With reference to fig. 5, the specific steps of the process for calculating the safety boundary of the hovercraft by using the bisection method and the BP neural network are as follows, wherein X 0 Is any initial state in the input working range; x 1 ,X 2 ,X 3 Is an auxiliary variable; v 0 ,V 1 ,V 3 A logical value output for the neural network; x s Is the state quantity of the safety limit.
Step 1: initializing a state variable X 1 =X 0 ;
And 2, step: state variable X 1 Input into BP network, if output V 0 =0, step 3 is performed; if the output V is 0 =1, perform step 4;
and step 3: state X 1 =X 1 +hX 0 X is to be 1 Input into BP network, if output V 1 =0, repeatedly until output V 1 =1; if output V 1 =1, let X 2 =X 1 ,X 1 =X 1 -hX 0 ;
And 4, step 4: let X 1 =X 1 -hX 0 X is to be 1 Input into BP network, if output V 1 =0, let X 2 =X 1 +hX 0 If outputting V 1 =1, repeat execution;
and 5: by adopting a dichotomy, orderX is to be 3 Input into BP network, if output V 3 =0, judge | | X 3 -X 2 If | ≦ e is true, let X be S =kX 3 Otherwise, let X 1 =X 3 And 5, repeating the step; if output V 3 =1, judge | | | X 3 -X 1 Whether e is less than or equal to | |, if so, making X S =kX 1 Otherwise, make X 2 =X 3 And repeating the step 5;
step 6: let h S =||X 0 ||/||X S L, judge h S &If the result is 1, executing step 7; if not, determining the area as an unsafe area;
and 7: judging |1-h S Whether e is less than or equal to the value, and if so, determining the area as a safe area; if not, determining the area as a safe area boundary;
and step 8: and (6) ending.
Claims (3)
1. A hovercraft safety boundary judgment method is characterized in that:
(1) Obtaining a sample space, specifically comprising the steps of:
(1.1) acquiring state information of the hovercraft, and respectively using six parameters of sideslip angle, gyration angular velocity, transverse inclination angle, transverse inclination angular velocity, longitudinal inclination angle and longitudinal inclination angular velocity as variable x 1 ,x 2 ,x 3 ,x 4 ,x 5 ,x 6 Expressed, the vector X = [ X ] 1 ,x 2 ,x 3 ,x 4 ,x 5 ,x 6 ] T The current control law of the hovercraft is recorded as tau, and the model of the hovercraft is recorded asWherein X (0) represents an initial state and t represents time;
(1.2) selecting an initial state sequence according to the working range of the state, wherein the working range of the state is represented as follows:
x imin <x i <x imax ,i=1,…,6
x imin ,x imax respectively the lower and upper operating bounds of the state variables,
the number of sequences is determined by the following formula:
wherein epsilon i Is the maximum deviation of the state data;
obtaining an initial state sequence (X (0)) according to a principle of uniformly distributing and taking points in a working space;
(1.3) substituting (X (0), tau) into the Hovercraft model, run time T s And then judging whether the system state meets the following conditions:
||x i ||<σ i ,i=1,3,5
wherein σ i >0,
If the above conditions are satisfied, the hovercraft is stable under the initial state X (0) and the control law τ, and the corresponding logical value is Y =0; otherwise, the hovercraft is unstable, the corresponding logical value is Y =1, and a sample space { X (0), Y } is formed by the initial state sequence { X (0) } and the corresponding logical value sequence { Y }, and is used as a training sample of the neural network;
(2) Training the BP neural network by using a sample space to input an initial state in any working range, wherein the BP neural network can obtain a logic value of whether the system is stable;
(3) The states of six parameters in any given working range are input into the trained BP neural network, and the boundary value of the stable state in the working range is searched by adopting a dichotomy according to the output of the neural network.
2. The hovercraft safety margin determination method of claim 1 wherein: the neural network selects a BP neural network, the structure of an input layer, a hidden layer and an output layer is selected to be 4 x 10 x 1, the activation function of neurons of the hidden layer is selected to be a tansig function, the activation function of neurons of the output layer is selected to be a logsig function, and a Levenberg-Marquardt algorithm is adopted as a network training function.
3. The hovercraft safety boundary determination method as defined in claim 2, wherein the hovercraft safety boundary is calculated using a dichotomy and a BP neural network, comprising the steps of: wherein, X 0 Is any initial state within the working range of the input; x 1 ,X 2 ,X 3 Is an auxiliary variable; v 0 ,V 1 ,V 3 A logical value output for the neural network; x s In order to be able to safely limit the state quantities,
step 1: initializing a state variable X 1 =X 0 ;
Step 2: state variable X 1 Input into BP network, if output V 0 =0, step 3 is performed; if the output V is 0 =1, perform step 4;
and step 3: state X 1 =X 1 +hX 0 X is to be 1 Input into BP network, if output V 1 =0, repeat execution until output V 1 =1; if the output V is 1 =1, let X 2 =X 1 ,X 1 =X 1 -hX 0 ;
And 4, step 4: let X 1 =X 1 -hX 0 X is to be 1 Input into BP network, if output V 1 =0, let X 2 =X 1 +hX 0 If outputting V 1 =1, repeatedly executed;
and 5: by adopting a dichotomy, orderX is to be 3 Input into BP network, if output V 3 =0, judge | | | X 3 -X 2 If | | is less than or equal to e, if so, let X S =kX 3 Otherwise, let X 1 =X 3 And repeating the step 5; if the output V is 3 =1, judge | | | X 3 -X 1 Whether e is less than or equal to | |, if so, making X S =kX 1 Otherwise, order X 2 =X 3 And repeating the step 5;
step 6: let h S =||X 0 ||/||X S L, judge h S &If the result is 1, executing a step 7; if not, determining the area as an unsafe area;
and 7: judgment of |1-h S Whether e is equal to or smaller than l, and if so, determining the area as a safety area; if the area is not the safe area boundary, determining the area as the safe area boundary;
and 8: and (6) ending.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510593660.XA CN105184001B (en) | 2015-09-17 | 2015-09-17 | A kind of aircushion vehicle secure border decision method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510593660.XA CN105184001B (en) | 2015-09-17 | 2015-09-17 | A kind of aircushion vehicle secure border decision method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105184001A CN105184001A (en) | 2015-12-23 |
CN105184001B true CN105184001B (en) | 2018-04-17 |
Family
ID=54906080
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510593660.XA Active CN105184001B (en) | 2015-09-17 | 2015-09-17 | A kind of aircushion vehicle secure border decision method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105184001B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110943453B (en) * | 2019-12-23 | 2022-12-02 | 北京交通大学 | Power system fault sample generation and model construction method facing transfer learning |
CN111666633B (en) * | 2020-06-10 | 2022-06-28 | 中国船舶工业集团公司第七0八研究所 | Method for calculating three-dimensional apron responsivity of hovercraft |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103407443A (en) * | 2013-07-11 | 2013-11-27 | 哈尔滨工程大学 | Hovercraft navigation safety coordination and control method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8074939B2 (en) * | 2006-12-01 | 2011-12-13 | The Invention Science Fund I, Llc | Active control of surface drag |
-
2015
- 2015-09-17 CN CN201510593660.XA patent/CN105184001B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103407443A (en) * | 2013-07-11 | 2013-11-27 | 哈尔滨工程大学 | Hovercraft navigation safety coordination and control method |
Non-Patent Citations (1)
Title |
---|
An Idea of Heading Control for Air Cushion Vehicles;Chenglong Wang 等;《Proceedings of the 2007 IEEE International Conference on Integration Technology》;20070324;第334-336页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105184001A (en) | 2015-12-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111186549B (en) | Course autopilot control system with ship collision avoidance function | |
CN111775949B (en) | Personalized driver steering behavior auxiliary method of man-machine co-driving control system | |
CN109283842B (en) | Unmanned ship track tracking intelligent learning control method | |
CN111985614B (en) | Method, system and medium for constructing automatic driving decision system | |
CN108820157A (en) | A kind of Ship Intelligent Collision Avoidance method based on intensified learning | |
US10112607B2 (en) | Driving support device that performs driving support in accordance with characteristics and skills of driver | |
CN111332362B (en) | Intelligent steer-by-wire control method integrating individual character of driver | |
CN108983774B (en) | Single-jet-pump-propelled unmanned surface vehicle self-adaptive course control method based on fuzzy state observer | |
CN110525428B (en) | Automatic parking method based on fuzzy depth reinforcement learning | |
CN112034711A (en) | Unmanned ship sea wave interference resistance control method based on deep reinforcement learning | |
JP4640760B2 (en) | Waveform prediction method of incoming waves and driving control method of navigation body in waves | |
CN105184001B (en) | A kind of aircushion vehicle secure border decision method | |
CN112947494A (en) | Fuzzy PID (proportion integration differentiation) -based automatic ship berthing control method | |
Ercan et al. | An adaptive and predictive controller design for lateral control of an autonomous vehicle | |
CN112987771A (en) | Motor sailing ship path tracking control method based on multi-port output error triggering mechanism | |
CN104590530B (en) | Turn into latitude of emulsion display device and its display methods, auto-pilot | |
US20220274603A1 (en) | Method of Modeling Human Driving Behavior to Train Neural Network Based Motion Controllers | |
CN116750013A (en) | Method, device, equipment and medium for monitoring driving school learning vehicle | |
US10427682B2 (en) | Adaptive method for controlling a vehicle speed and adaptive device for using the same | |
CN116841298A (en) | Finite time tracking control method suitable for four-wheel mobile robot lane change early warning | |
Casado et al. | Identification of nonlinear ship model parameters based on the turning circle test | |
CN116702320A (en) | Unmanned ship response model parameter identification method based on improved particle swarm algorithm | |
CN110103968A (en) | Unmanned vehicle autonomous overtaking track planning system based on three-dimensional laser radar | |
CN108549367A (en) | A kind of man-machine method for handover control based on prediction safety | |
CN110928278A (en) | Autonomous safe sailing method for unmanned ship and unmanned ship |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |