CN108572550A - A kind of online real-time thrust distribution method based on machine learning - Google Patents
A kind of online real-time thrust distribution method based on machine learning Download PDFInfo
- Publication number
- CN108572550A CN108572550A CN201810218580.XA CN201810218580A CN108572550A CN 108572550 A CN108572550 A CN 108572550A CN 201810218580 A CN201810218580 A CN 201810218580A CN 108572550 A CN108572550 A CN 108572550A
- Authority
- CN
- China
- Prior art keywords
- thrust
- propeller
- particle
- desired positions
- machine learning
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Automation & Control Theory (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Aviation & Aerospace Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Feedback Control In General (AREA)
Abstract
The present invention relates to a kind of online real-time thrust distribution method based on machine learning, including：Step S1：It determines that marine propulsion is laid out, is loaded into the parameter of each propeller, wherein the parameter of propeller includes：Propeller thrust direction-agile range, thrust size variable range, thrust direction change rate range, thrust size variation rate range；Step S2：After receiving control signal, target is obtained according to control signal and closes thrust, wherein the conjunction thrust by longitudinal thrust and lateral thrust and with yawing torque and form；Step S3：By the propeller thrust direction-agile range, thrust size variable range, thrust direction change rate range of each propeller, thrust size variation rate range establishes object function as constraints：Step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, obtains Best Thrust distribution.Compared with prior art, the present invention solves the slow defect of particle cluster algorithm convergence rate.
Description
Technical field
The present invention relates to ship power fields, more particularly, to a kind of online real-time thrust distribution side based on machine learning
Method.
Background technology
The rapid development height of modern industry relies on mineral resources.In recent years, with landing field multiple resources exploit and increasingly
In reduction, the theme for being developed into various countries' concern of marine resources, the marine engineering equipments such as mining dredger and ocean platform obtain greatly
Amount application.
In the ocean engineerings operation such as mining dredger and ocean platform, due to being limited by operating environment, anchoring positioning is
Through that cannot meet the requirements, and dynamic positioning technology becomes the hot spot of exploitation because not limited by the depth of water.Dynamic positioning system is
Refer to a kind of power of dependence itself to resist the interference of marine environment load, such as the disturbance of sea wind, wave and wave, ship is made to tie up
Hold in defined position with bow to, or along desired trajectory advance closed-loop system.
Dynamic positioning system generally comprises a series of subsystems peculiar to vessel, including measuring system, push system and control system
Deng the invention mainly relates to the designs of push system.
Push system is the executing agency of dynamic positioning system, it generates required thrust according to the signal from controller
And moment of thrust.Ideal thruster can generate arbitrarily large thrust on 360 degree of directions, but in practice, propeller pushes away
It can be all restricted into angle and thrust size, there are very strong constraints.
The thrust mechanism that dynamic positioning thrust system uses includes open auger paddle, ducted propeller, all-direction propeller
And PODDED PROPULSOR etc..The dynamic positioning system of early stage mainly uses the immutable propulsion device of thrust direction, such as open auger
Paddle and ducted propeller.In the push system of not redundancy, thrust allocation plan mainly by solve system of linear equations come
It realizes.With the appearance of all-direction propeller and PODDED PROPULSOR, thrust direction can also change in real time, this gives dynamic positioning
System brings very big challenge, it is meant that thrust distribution control algolithm has to online real-time under the premise of considering various constraints
Ground calculates the propulsion angle and thrust size of each propeller.Theoretically, which it is excellent can be described as a multi-target non-linear
Change problem, object function need to consider energy expenditure, thrust (square) error, it is unusual the problems such as, constraints is then related to
The equilibrium equation of power and torque, the change rate of thrust amplitude and direction, thruster abrasion etc..
Sequential quadratic programming (SQP) method widely used in early days is optimal using being designed to Nonlinear programming Model
Point, constraint can be considered in this method, such as the problems such as energy-optimised, singular value, mechanical wear, but it is very strong to initial value dependence,
It is difficult to meet global optimum.Hereafter grow up based on L2 norm pseudo inverse methods because simply, it is reliable the features such as, in engineering
Apply in practice relatively broad, but it can not limit thrust size and Orientation change rate and angle forbidden zone.With intelligence
The continuous development of energy algorithm, many intelligent algorithms are attempted for solving thrust assignment problem, overcome L2 norm pseudo inverses method and sequence
The deficiency of row quadratic programming, but research is not mature enough, and is faced with unsecured global convergence and algorithm the convergence speed is slow
Etc. general sex chromosome mosaicism, it is still in the exploratory stage.
Invention content
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind being based on machine learning
Online real-time thrust distribution method.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of online real-time thrust distribution method based on machine learning, including：
Step S1：It determines that marine propulsion is laid out, is loaded into the parameter of each propeller, wherein the parameter of propeller includes：
Propeller thrust direction-agile range, thrust size variable range, thrust direction change rate range, thrust size variation rate model
It encloses；
Step S2：Receive control signal after, according to control signal obtain target close thrust, wherein the conjunction thrust by
Longitudinal thrust and lateral thrust and with yawing torque and composition；
Step S3：The propeller thrust direction-agile range, thrust size variable range, thrust direction of each propeller are become
Rate range, thrust size variation rate range establish object function as constraints：
Wherein：J (u, α, s) is object function, and u is the big minor matrix of thrust of each propeller, and α is thrust direction matrix, and W is
Power entry weighting coefficient, Q are that error weights positive definite matrix, and k is one degree weighting coefficient, and m is propeller number, and i is to promote
Device serial number, c are thrust power weighting matrix, uiFor the thrust size of i-th of propeller, s is error matrix, and ε is the nothing more than 0
Small real number is limited, λ is singular values of a matrix, and B (α) is thrust direction coefficient matrix, ()TFor the transposition of matrix, τ*It is pushed away for practical conjunction
Power；
Step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, is obtained best
Thrust is distributed.
The step S2 includes：
Step S21：After receiving control signal, target is obtained according to control signal and closes thrust；
Step S22：The thrust distribution whether being stored with for the state of target conjunction thrust and current each propeller searched
Scheme, if it has, then the thrust allocation plan using the storage carries out thrust distribution, if it has not, thening follow the steps S3.
In the step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, is obtained
After being distributed to Best Thrust, Best Thrust distribution is stored.
The step S4 is specifically included：
Step S41：Definition corresponds to the current location of propeller according to particle by the molecular group of 2m grain, initializes
The initial position of each particle and individual desired positions；
Step S42：The average desired positions of group are calculated according to the current desired positions of all particles；
Step S43：To each particle in group, its desired positions is updated respectively.
The process for updating its desired positions in the step S43 to each particle specifically includes：
Step S431：The desired positions of particle are updated；
Step S432：If the best values of particle are better than current global desired positions, with updated particle desired positions
Replace the desired positions of the current particle；
Step S433：Judge whether update times reach given threshold, if it has, then particle update terminates, if it has not,
Then return to step S431.
The average desired positions of the group are：
Wherein：C is the average desired positions of group, and M is the number of particle in group, PiFor i-th particle individual most
Good position.
The overall situation desired positions are in current iteration, desired positions best one particle in all particles it is best
Position.
Compared with prior art, the invention has the advantages that：
1) the multi-target non-linear nonsmooth optimization of power-positioning control system thrust allocation optimization problems is established
Model is optimized using the particle cluster algorithm in machine learning.
2) defect of the basic unsecured global convergence of PSO algorithms is overcome by using quantum-behaved particle swarm optimization.
3) the slow defect of particle cluster algorithm convergence rate is solved by the heuristic strategies based on machine learning databases.
Description of the drawings
Fig. 1 is the block diagram of single thruster system；
Fig. 2 is that multi-thruster power and torque synthesize schematic diagram；
Fig. 3 is thrust distribution principle figure；
Fig. 4 is mining dredger thruster arrangement in embodiment；
Fig. 5 is that thrust distributes logical schematic.
Specific implementation mode
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to
Following embodiments.
A kind of online real-time thrust distribution method based on machine learning, in the dynamic positioning system that the application considers,
Measuring system uses Wide-area differential GPS, while coordinating wind speed wind direction sensor, gyro compass and vertical reference unit.Control system
The calculating that control algolithm is then completed by computer manipulates instrument equipped with console and uninterruptible power supply box portable remote.Execution machine
Structure is then made of thruster main body, motor, hydraulic system and actuator interface (containing propeller and steering engine interface etc.).It is single only to push away
Block diagram such as Fig. 1 of power device system.Thrust mechanism (thruster main body) generates thrust, and form is varied, including open auger paddle,
Ducted propeller, all-around propeller and PODDED PROPULSOR.Push system is additionally provided with thrust control system other than thrust mechanism
System, effect are that thrust mechanism is enabled to execute the instruction from controller, and response is sufficiently fast and accurate.Prime mover (motor) is by power
It is supplied to thrust mechanism, under the distribution of thrust control system, according to the signal from power-positioning control system, power is turned
Turn to the thrust of each thruster.
Specifically, this method includes：
Step S1：It determines that marine propulsion is laid out, is loaded into the parameter of each propeller, wherein the parameter of propeller includes：
Propeller thrust direction-agile range, thrust size variable range, thrust direction change rate range, thrust size variation rate model
It encloses；
Step S2：After receiving control signal, target is obtained according to control signal and closes thrust, wherein closes thrust by longitudinal direction
Thrust and lateral thrust and with yawing torque and composition；It specifically includes：
Step S21：After receiving control signal, target is obtained according to control signal and closes thrust；
Step S22：The thrust distribution whether being stored with for the state of target conjunction thrust and current each propeller searched
Scheme, if it has, then the thrust allocation plan using the storage carries out thrust distribution, if it has not, thening follow the steps S3.
It determines that marine propulsion is laid out, determines system input and relevant parameter.One is considered equipped with m thrust device
Ship, a certain moment t, controller according to marine environment and position, bow to deviation, calculate at this time needed for three degree of freedom
Thus the thrust wanted or torque constitute vector：
Its three elements indicate longitudinal force τ respectivelyx, cross force τyAnd yawing torque τz.Thrust allocation algorithm will basis
The vector at the momentDerive the state of m thrust device at this time.There are two parameters for each thrust device
It needs to calculate, is the size u of the thrust of PODDED PROPULSOR respectivelyiWith the direction α of thrusti.Provide αiFor from thrust direction to hull
The angle (clockwise for just) that coordinate system positive direction of the y-axis is constituted, the thrust that gross thrust can be expressed as x and y both directions are closed
At：
If the position that i-th of propulsion device is loaded onto ship in hull coordinate system is (xi, yi), then the thruster makes hull generate
The yawing torque (set clockwise for positive) about ship barycenter be
MZ, i=-xiFY, i+yiFX, i (2)
To same thruster, there are following relationship (Fig. 2)：
αi=arctan (Fx,i/y,i) (4)
Mz,i=yiTicosαi±xiTisinαi (5)
For the system of m thrust device composition, a total of 2m variable can be organized into following state vector：
WhereinThis m propeller combination and
At push system, the total thrust and torque composition of vector that can be generated is：
By synthesizing for power and torque, have
Error can be expressed as caused by thrust allocation algorithm
S=τ-τ* (7)
From (6) and (7) simultaneous：
Above equation group is arranged, is obtained
τ=B (α) u (9)
Wherein,I-th row it is as follows：
Thrust allocation algorithm study the problem of be the rolling how to obtain thruster, pitching, three directions of yawing power
Or torque, it is converted into the size and direction of this m propeller thrust, design scheme needs the power of guarantee thruster as far as possible
It is small, error is as small as possible, but want processing system simultaneously there are various limitations (Fig. 3) such as unusual, change rate.
Can usually there be singular problem in dynamic positioning system, to avoid occurring in thrust assigning process unusual appearance, one
As way be to be quantized into function by unusual, be added in object function, such as be added in object function
It is weight of this in object function that wherein ε ＞ 0, which are used for avoiding numerical problem, k,.The present invention improves above-mentioned function
It is as follows
Wherein the setting of parameter k is determined according to the degree of this influence object function.
The thrust and torque that each thruster of dynamic positioning thrust system generates always exist with controller specified value certain
Deviation, be
S=τ-B (α) x (13)
In order to ensure dynamic positioning required precision, to fully ensure that s is sufficiently close to 0.Error can in constraints
To be handled by object function, i.e., increases a punishment instruction thrust in object function and the practical generation of push system is closed
The error s of reason, this penalty term are taken asWherein weight matrix Q positive definites and characteristic value is sufficiently large.
The change rate of all-direction propeller thrust angle is not above 1deg/s, so thrust direction variation is just like lower limit
System：
Δαmin≤α-α0≤Δαmax (14)
α in formula0Indicate the thrust direction of previous moment thruster, Δ αminIndicate variation in the thrust direction unit interval
The amplitude upper limit, Δ αminIndicate the lower limit of variation, usually negative.
Propeller motor is typically that can all generate certain time lag, therefore thrust size by its mechanical structure and control algolithm
Change rate should limit：
ΔTmin≤u-u0≤ΔTmax (15)
Due to there is interaction between propeller, there is also interactions between propeller and hull, should also be arranged one
A dead zone is set according to the layout of the parameter of ship and propeller：
The maximum thrust that propeller generates then is limited by its power and lower two aspects of cavitation phenomena of high speed, is had
0≤T≤Tmax (17)
According to the relationship of Propeller, power and the thrust of i-th of propeller have following relationship：
So the general power of push system is
This is that one nonlinear equation thus obtains.
Step S3：The propeller thrust direction-agile range, thrust size variable range, thrust direction of each propeller are become
Rate range, thrust size variation rate range establish object function as constraints：
Constraints：
S=B (α) u- τ*
αi∈Γ
|αi-αi0|∈[Δαimin,Δαimax]
u∈[Timin,Timax]
|ui-ui0|∈[ΔTimin,ΔTimax]
Wherein：J (u, α, s) is object function, and u is the big minor matrix of thrust of each propeller, and α is thrust direction matrix, and W is
Power entry weighting coefficient,Q is that error weights positive definite matrix, and k is one degree weighting coefficient, and m is propeller number,
I is propeller serial number, and c is thrust power weighting matrix, uiFor the thrust size of i-th of propeller, s is error matrix, and ε is big
In 0 infinitesimal real number, λ is singular values of a matrix, and B (α) is thrust direction coefficient matrix, ()TFor the transposition of matrix, τ * are real
Close thrust, α in borderiFor the thrust direction of i-th of propeller, Γ is propeller thrust direction-agile range, ai0For i-th of propeller
The thrust direction of eve, ui0For the thrust size of i-th of propeller eve.
In the object function non-convex function.In addition, last is but also function is rough, so the optimization problem
It is a multi-target non-linear nonsmooth optimization, solution procedure is complex.
Step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, is obtained best
Thrust is distributed
Thrust assignment problem is optimized using the particle cluster algorithm (PSO) in machine learning algorithm.Basic PSO is calculated
The defect of method is exactly unsecured global convergence, and quantum behavior particle group optimizing (QPSO) algorithm can overcome this limitation.
Consider N-dimensional object function, it is corresponding with thrust assignment problem, by M：=2m particle composition group X (t)={ X1
(t), X2(t) ..., XM(t) }, its i-th of particle position is
Xi(t)=[XI, l(t), XI, 2..., XI, N] i=1,2 ..., M
Individual desired positions are Pi(t)=[PI, 1(t), PI, 2(t) ..., PI, N(t)], the corresponding desired positions of group can
To be write as G (t)=[G1(t), G2(t) ..., GN(t)], it be in M particle position best one.
To the optimization problem (20), it is as small as possible that we always want to J (u, α, s).It can obtain the best position of particle
It is as follows：
The global best position of particle is as follows：
G (t)=Pg(t)
The thinking of global optimization is to be compared current individual position with the adaptive value of global desired positions, if more
Good, then G (t) can be updated.It enables
It is as follows that the evolution equation of particle at this time can then be write out：
XI, j(t+1)=pI, j(t)±α·|Cj(t) XI, j(t)|·ln[1/uI, j(t)]uI, j(t)~U (0,1)
WhereinFor converging diverging coefficient.
Later, it needs to store Best Thrust distribution,
Step S4 is specifically included：
Step S41：Definition corresponds to the current location of propeller according to particle by the molecular group of 2m grain, initializes
The initial position of each particle and individual desired positions；
Step S42：The average desired positions of group are calculated according to the current desired positions of all particles, group is averaged most
Good position is：
Wherein：C is the average desired positions of group, and M is the number of particle in group, PiFor i-th particle individual most
Good position.
Step S43：To each particle in group, its desired positions is updated respectively, wherein is updated to each particle
The process of its desired positions specifically includes：
Step S431：The desired positions of particle are updated；
Step S432：If the best values of particle are better than current global desired positions, with updated particle desired positions
Replace the desired positions of the current particle；
Step S433：Judge whether update times reach given threshold, if it has, then particle update terminates, if it has not,
Then return to step S431.
Establish the heuristic strategies based on machine learning databases.Although QPSO algorithms solve the problems, such as global convergence,
It is that the algorithm the convergence speed is slower, and dynamic positioning system requires thrust allocation algorithm that can make real-time response.In order to solve this
A problem, the present invention propose the heuristic strategies based on machine learning databases.Front derived the thrust of each sampling
The multi-C vector that size and Orientation is constituted depends on the thrust of the control signal and eve of the controller of sampling instant, and previous
The thrust size and Orientation value at moment is all within stringent lower limit on it.The calculated swaying of controller, surging power and
Yawing torque is both less than according to the calculated average load of operation sea situation within most times, we are just with this average load
The upper limit of the lotus as its all directions.Parameters all in this way have its value bound.According to every one-dimensional actual conditions into
Row quantization, the in this way value range in each dimension become limited from unlimited.Pass through work station or mass computing
Machine calculates the corresponding allocation plan of each value condition, establishes one database according to the algorithm.If in implementation procedure
Instruction from controller meets the range of the database, and thrust allocation plan just can be obtained by tabling look-up.
The present invention illustrates the thrust allocation algorithm proposed for configuring the mining dredger of 4 all-direction propellers, but calculates
Method is not limited to the ship type.The arrangement of mining dredger thruster is as shown in Figure 4：
Use maximum iteration for 2000 times in QPSO algorithms, population size is taken as 20, and specific implementation flow is as follows：
Step 1. enables t=0, initializes in population, the initial position of all particles is denoted as Xi(0), while initialization is a
The desired positions of body, are expressed as Pi(0)=Xi(0)。
Step 2. calculates the average desired positions of population,
Step 3. is executed each particle in population by step 4~7.
Current location X of the step 4. to particle iiIt is calculated, the desired positions of particle is updated.
Step 5. compares the best values of particle i and global desired positions, we are replaced it if in better words.
Step 6. carries out calculating random point current location, updates particle position according to particle evolution equation.
The case where whether step 7. reaches the condition of EP (end of program) to algorithm above is judged, if it is not, t=t+1 is then set, and
Return to step 2；If so, terminator.Object function in problem is to solve minimum problems for one,
Vector is constituted for the thrust size and Orientation of the configuration of 4 all-direction propellers, each samplingControl signal depending on sampling instant controllerWith the thrust T of eve.The thrust relations of distribution of sampling instant in this way are calculated by 11 parameters
Go out, the thrust size and Orientation value of previous moment is all within stringent lower limit on it.It is each using average load as its
The upper limit in direction, then 11 dimension parameters just have its value bound.It is quantified according to every one-dimensional actual conditions,
Then value range becomes limited from unlimited in each dimension, can establish thrust allocation database accordingly, for online real
When provide thrust allocation plan, specific implementation flow it is as shown in Figure 5.
Claims (7)
1. a kind of online real-time thrust distribution method based on machine learning, which is characterized in that including：
Step S1：It determines that marine propulsion is laid out, is loaded into the parameter of each propeller, wherein the parameter of propeller includes：It promotes
Device thrust direction variable range, thrust size variable range, thrust direction change rate range, thrust size variation rate range；
Step S2：After receiving control signal, target is obtained according to control signal and closes thrust, wherein the conjunction thrust is by longitudinal direction
Thrust and lateral thrust and with yawing torque and composition；
Step S3：By the propeller thrust direction-agile range, thrust size variable range, thrust direction change rate of each propeller
Range, thrust size variation rate range establish object function as constraints：
S=B (α) u- τ*
Wherein：J (u, α, s) is object function, and u is the big minor matrix of thrust of each propeller, and α is thrust direction matrix, and W is power
Item weighting coefficient, Q are that error weights positive definite matrix, and k is one degree weighting coefficient, and m is propeller number, and i is propeller sequence
Number, c is thrust power weighting matrix, uiFor the thrust size of i-th of propeller, s is error matrix, and ε is infinitely small more than 0
Real number, λ are singular values of a matrix, and B (α) is thrust direction coefficient matrix, ()TFor the transposition of matrix, τ*Thrust is closed to be practical；
Step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, obtains Best Thrust
Distribution.
2. a kind of online real-time thrust distribution method based on machine learning according to claim 1, which is characterized in that institute
Stating step S2 includes：
Step S21：After receiving control signal, target is obtained according to control signal and closes thrust；
Step S22：The thrust allocation plan whether being stored with for the state of target conjunction thrust and current each propeller is searched,
If it has, then the thrust allocation plan using the storage carries out thrust distribution, if it has not, thening follow the steps S3.
3. a kind of online real-time thrust distribution method based on machine learning according to claim 2, which is characterized in that institute
It states in step S4：Thrust assignment problem is optimized using the particle cluster algorithm in machine learning algorithm, obtains Best Thrust
After distribution, Best Thrust distribution is stored.
4. a kind of online real-time thrust distribution method based on machine learning according to claim 1, which is characterized in that institute
Step S4 is stated to specifically include：
Step S41：Definition is corresponded to the current location of propeller according to particle, is initialized each grain by the molecular group of 2m grain
The initial position of son and individual desired positions；
Step S42：The average desired positions of group are calculated according to the current desired positions of all particles；
Step S43：To each particle in group, its desired positions is updated respectively.
5. a kind of online real-time thrust distribution method based on machine learning according to claim 4, which is characterized in that institute
It states and updates the process of its desired positions in step S43 to each particle and specifically include：
Step S431：The desired positions of particle are updated；
Step S432：If the best values of particle are replaced better than current global desired positions with updated particle desired positions
The desired positions of the current particle；
Step S433：Judge whether update times reach given threshold, if it has, then particle update terminates, if it has not, then returning
Return step S431.
6. a kind of online real-time thrust distribution method based on machine learning according to claim 4, which is characterized in that institute
The average desired positions for stating group are：
Wherein：C is the average desired positions of group, and M is the number of particle in group, PiFor the best position of individual of i-th of particle
It sets.
7. a kind of online real-time thrust distribution method based on machine learning according to claim 4, which is characterized in that institute
It is the desired positions of desired positions best one particle in all particles in current iteration to state global desired positions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810218580.XA CN108572550A (en) | 2018-03-16 | 2018-03-16 | A kind of online real-time thrust distribution method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810218580.XA CN108572550A (en) | 2018-03-16 | 2018-03-16 | A kind of online real-time thrust distribution method based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108572550A true CN108572550A (en) | 2018-09-25 |
Family
ID=63574348
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810218580.XA Pending CN108572550A (en) | 2018-03-16 | 2018-03-16 | A kind of online real-time thrust distribution method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108572550A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709970A (en) * | 2018-12-25 | 2019-05-03 | 哈尔滨工程大学 | A kind of underwater robot six degree of freedom thrust distribution optimization method |
| CN110245740A (en) * | 2019-05-10 | 2019-09-17 | 华中科技大学 | A kind of particle group optimizing method based on sequence near-optimal |
| CN111061285A (en) * | 2019-12-12 | 2020-04-24 | 哈尔滨工程大学 | Ship dynamic positioning thrust distribution method |
| CN111812976A (en) * | 2020-06-06 | 2020-10-23 | 智慧航海(青岛)智能***工程有限公司 | Ship thrust distribution system and thrust distribution method |
| CN112224359A (en) * | 2020-08-05 | 2021-01-15 | 智慧航海（青岛）科技有限公司 | Ship power distribution method capable of being used in different navigational speed modes |
| CN114379744A (en) * | 2020-10-16 | 2022-04-22 | 川崎重工业株式会社 | Ship control system and ship |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH092393A (en) * | 1995-04-17 | 1997-01-07 | Mitsubishi Heavy Ind Ltd | Controlled thrust distributing device of navigating vessel |
| US6450112B1 (en) * | 1999-04-02 | 2002-09-17 | Nautronix, Inc. | Vessel control force allocation optimization |
| CN1710499A (en) * | 2005-07-07 | 2005-12-21 | 上海交通大学 | Ship power-positioning control system based on fuzzy self-adaption algorithm |
| CN102385665A (en) * | 2011-08-30 | 2012-03-21 | 无锡中讯科技有限公司 | Thrust force distribution method of power location system of ship adopting genetic algorithm |
| CN102841970A (en) * | 2012-09-21 | 2012-12-26 | 上海交通大学 | Dynamic positioning thrust distributing device adopting dynamic prohibiting angle and distributing method thereof |
| CN103092077A (en) * | 2013-01-06 | 2013-05-08 | 中国海洋石油总公司 | Thrust distributing method of dynamic positioning system |
| CN103678816A (en) * | 2013-12-19 | 2014-03-26 | 上海交通大学 | Intelligent push force distributing method of power positioning push force system |
| CN103823372A (en) * | 2014-02-24 | 2014-05-28 | 中国船舶重工集团公司第七○二研究所 | Method for distributing thrust of multiple thrusters of ocean engineering equipment dynamic positioning system |
| CN105301963A (en) * | 2015-11-17 | 2016-02-03 | 江苏科技大学 | Thrust optimal distribution method based on ship power management system |
| JP2016184344A (en) * | 2015-03-26 | 2016-10-20 | 三菱重工業株式会社 | Arithmetic unit, moving body system, arithmetic method and program |
| CN106773741A (en) * | 2017-03-02 | 2017-05-31 | 华南理工大学 | A kind of unmanned boat dynamic positioning system and method |
| CN106773722A (en) * | 2017-02-17 | 2017-05-31 | 哈尔滨工程大学 | A kind of thrust force distribution method of power location system of ship based on artificial fish-swarm algorithm |
-
2018
- 2018-03-16 CN CN201810218580.XA patent/CN108572550A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH092393A (en) * | 1995-04-17 | 1997-01-07 | Mitsubishi Heavy Ind Ltd | Controlled thrust distributing device of navigating vessel |
| US6450112B1 (en) * | 1999-04-02 | 2002-09-17 | Nautronix, Inc. | Vessel control force allocation optimization |
| CN1710499A (en) * | 2005-07-07 | 2005-12-21 | 上海交通大学 | Ship power-positioning control system based on fuzzy self-adaption algorithm |
| CN102385665A (en) * | 2011-08-30 | 2012-03-21 | 无锡中讯科技有限公司 | Thrust force distribution method of power location system of ship adopting genetic algorithm |
| CN102841970A (en) * | 2012-09-21 | 2012-12-26 | 上海交通大学 | Dynamic positioning thrust distributing device adopting dynamic prohibiting angle and distributing method thereof |
| CN103092077A (en) * | 2013-01-06 | 2013-05-08 | 中国海洋石油总公司 | Thrust distributing method of dynamic positioning system |
| CN103678816A (en) * | 2013-12-19 | 2014-03-26 | 上海交通大学 | Intelligent push force distributing method of power positioning push force system |
| CN103823372A (en) * | 2014-02-24 | 2014-05-28 | 中国船舶重工集团公司第七○二研究所 | Method for distributing thrust of multiple thrusters of ocean engineering equipment dynamic positioning system |
| JP2016184344A (en) * | 2015-03-26 | 2016-10-20 | 三菱重工業株式会社 | Arithmetic unit, moving body system, arithmetic method and program |
| CN105301963A (en) * | 2015-11-17 | 2016-02-03 | 江苏科技大学 | Thrust optimal distribution method based on ship power management system |
| CN106773722A (en) * | 2017-02-17 | 2017-05-31 | 哈尔滨工程大学 | A kind of thrust force distribution method of power location system of ship based on artificial fish-swarm algorithm |
| CN106773741A (en) * | 2017-03-02 | 2017-05-31 | 华南理工大学 | A kind of unmanned boat dynamic positioning system and method |
Non-Patent Citations (1)
| Title |
|---|
| MING JI 等: "The optimal thrust allocation based on QPSO algorithm for dynamic positioning vessels", 《IEEE XPLORE》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709970A (en) * | 2018-12-25 | 2019-05-03 | 哈尔滨工程大学 | A kind of underwater robot six degree of freedom thrust distribution optimization method |
| CN109709970B (en) * | 2018-12-25 | 2022-01-14 | 哈尔滨工程大学 | Underwater robot six-degree-of-freedom thrust distribution optimization method |
| CN110245740A (en) * | 2019-05-10 | 2019-09-17 | 华中科技大学 | A kind of particle group optimizing method based on sequence near-optimal |
| CN111061285A (en) * | 2019-12-12 | 2020-04-24 | 哈尔滨工程大学 | Ship dynamic positioning thrust distribution method |
| CN111061285B (en) * | 2019-12-12 | 2022-08-02 | 哈尔滨工程大学 | Ship dynamic positioning thrust distribution method |
| CN111812976A (en) * | 2020-06-06 | 2020-10-23 | 智慧航海(青岛)智能***工程有限公司 | Ship thrust distribution system and thrust distribution method |
| CN112224359A (en) * | 2020-08-05 | 2021-01-15 | 智慧航海（青岛）科技有限公司 | Ship power distribution method capable of being used in different navigational speed modes |
| CN112224359B (en) * | 2020-08-05 | 2022-05-06 | 智慧航海（青岛）科技有限公司 | Ship power distribution method capable of being used in different navigational speed modes |
| CN114379744A (en) * | 2020-10-16 | 2022-04-22 | 川崎重工业株式会社 | Ship control system and ship |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108572550A (en) | A kind of online real-time thrust distribution method based on machine learning | |
| CN111966118B (en) | ROV thrust distribution and reinforcement learning-based motion control method | |
| CN112558612B (en) | Heterogeneous intelligent agent formation control method based on cloud model quantum genetic algorithm | |
| CN108445892A (en) | A kind of drive lacking unmanned boat formation control device structure and design method | |
| CN108388250B (en) | Water surface unmanned ship path planning method based on self-adaptive cuckoo search algorithm | |
| Wu et al. | An energy optimal thrust allocation method for the marine dynamic positioning system based on adaptive hybrid artificial bee colony algorithm | |
| CN109683479B (en) | Dynamic positioning thrust distribution device and method based on artificial neural network | |
| CN108563130A (en) | A kind of automatic berthing control method of underactuated surface vessel adaptive neural network, equipment and medium | |
| CN112965371B (en) | Water surface unmanned ship track rapid tracking control method based on fixed time observer | |
| CN108594651A (en) | A kind of dynamic positioning of vessels thrust distribution intelligent optimization method | |
| CN107544258B (en) | Self-adaptive inversion control method for autonomous underwater vehicle | |
| CN111930141A (en) | Three-dimensional path visual tracking method for underwater robot | |
| Cai et al. | A meta-heuristic assisted underwater glider path planning method | |
| Chen et al. | Path planning of AUV during diving process based on behavioral decision-making | |
| Gao et al. | Broad learning system-based adaptive optimal control design for dynamic positioning of marine vessels | |
| Gao et al. | Optimal thrust allocation strategy of electric propulsion ship based on improved non-dominated sorting genetic algorithm II | |
| Liu et al. | A hierarchical disturbance rejection depth tracking control of underactuated AUV with experimental verification | |
| Ahani et al. | Alternative approach for dynamic-positioning thrust allocation using linear pseudo-inverse model | |
| CN114114920A (en) | Thrust distribution method of ship dynamic positioning system based on artificial bee colony improved algorithm | |
| Zhang et al. | Hybrid threshold event-triggered control for sail-assisted USV via the nonlinear modified LVS guidance | |
| CN109901622A (en) | A kind of autonomous underwater robot prediction face S control method based on mechanism model | |
| Liu et al. | The tactics of ship collision avoidance based on quantum‐behaved wolf pack algorithm | |
| CN116541951A (en) | Ship thrust distribution method based on improved aigrette algorithm | |
| CN115903820A (en) | Multi-unmanned-boat pursuit and escape game control method | |
| CN114943168A (en) | Overwater floating bridge combination method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180925 |
|
| RJ01 | Rejection of invention patent application after publication |