CN109188907A - A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform - Google Patents
A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform Download PDFInfo
- Publication number
- CN109188907A CN109188907A CN201811074461.8A CN201811074461A CN109188907A CN 109188907 A CN109188907 A CN 109188907A CN 201811074461 A CN201811074461 A CN 201811074461A CN 109188907 A CN109188907 A CN 109188907A
- Authority
- CN
- China
- Prior art keywords
- particle
- extreme value
- algorithm
- current
- individual extreme
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
A kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform, belongs to a kind of Control System of Stable Platform algorithm.The present invention is directed to existing defect, provide a kind of control effect is good, response rapidly, strong antijamming capability, precision are high, overshoot is small, fast convergence rate control algolithm.In the present invention, initial solution group is randomly generated, each particle position solution is brought into ADRC algorithm, the corresponding fitness value of each particle and global extremum are determined in the form of fitness function of the present invention；Particle rapidity and position are updated, aforesaid operations are repeated, updates current individual extreme value and preceding global extremum；The difference E of the fitness of the given new and old position of particle is calculated, difference receives new explanation less than 0；Difference is greater than 0, then receives new explanation so that exp (- E/T) > rand (0,1) is probability, until reaching equilibrium state, exports optimal global extremum to get automatic disturbance rejection controller optimized parameter is arrived.Present invention is mainly used for the particle group optimizings of stabilized platform.
Description
Technical field
The invention belongs to a kind of Control System of Stable Platform and its algorithms, and in particular to one kind is applied to Stable Platform System
Control algolithm.
Background technique
Currently, the mechanical structure of well known stabilized platform is broadly divided into uniaxial, twin shaft and three-axle steady platform, each axis composition
Structure is identical.The axial quantity for including in platform is corresponding with the direction in space varied number of platform.Stabilized platform uses three axis
Mechanical structure to guarantee the disturbance that is subject to of carrier isolation three direction in spaces of x, y, z, to realize that carrier is stablized.Processing at present
The algorithm of feedback quantity mainly has pid control algorithm, FUZZY ALGORITHMS FOR CONTROL, hereditary control algolithm, Sliding mode variable structure control.It is conventional
Pid algorithm is also PID control algorithms, is current industrial most widely used control algolithm, in applying for it
Due to the limitation of itself in journey, the contradiction of system stability and accuracy will necessarily be encountered, is finally often derived from body three parts control
The compromise of production, it is difficult to obtain optimum efficiency；FUZZY ALGORITHMS FOR CONTROL is by a large amount of real data induction and conclusions, it is being controlled
In limited by preset parameter and membership function, the lasting variation being unable to during suitable solution；Sliding mode variable structure control is due to cunning
Dynamic model state is a kind of nonlinear Control, therefore unrelated with image parameter and disturbance, this, which allows for it, has quick response, to parameter
Change and disturb it is insensitive, debate knowledge online without system, the shortcomings that simple advantage of physics realization, the algorithm, is when state rail
After mark reaches sliding-mode surface, it is difficult to it strictly slides along sliding-mode surface towards equalization point, but is backed across in sliding-mode surface two sides, from
And generate trembling, that is, buffet problem.
In elementary particle colony optimization algorithm, since particle rapidity and the randomness of position can cause convergence rate slowly and influence
Control effect.
Automatic disturbance rejection controller (active disturbance rejection control, ADRC) is that Mr. Han Jingqing exists
Develop the technical spirit-" error is eliminated based on error " of PID control, and draw modern control theory achievement on the basis of, be
One kind can solve, and there is a wide range of and labyrinth (non-linear, time-varying, coupling, buffeting) uncertain system control problem to have
Efficacious prescriptions method.Its core concept is using simple " integrator tandem type " as the standard type of feedback system, different in system dynamic
It is considered as " total disturbance " (including interior disturb and disturb outside) in the part of standard type, " total disturbance " is estimated, and Active Compensation " is always disturbed
It is dynamic " influence to system, thus standard type is linearly turned to full of disturbance, uncertain and nonlinear controlled device, so that
The design of control system is from complexity to simple, intuitive from being abstracted into.But its parameter is more with respect to for PID, and without too
More ready-made theories and rule can be followed.Therefore the parameter tuning problem of automatic disturbance rejection controller, is its efficient application in practical work
The root problem that must be faced when industry object.
ADRC parameter tuning be in the case where its form or structure have determined, by adjusting ADRC each section parameter with
Reach the control requirement to target, therefore its core is exactly multi-objective optimization question.ADRC controller just has more than ten parameters to need
Processing, this makes the adjusting work difficulty of parameter very big.Generally during actual parameter tuning nonlinear function characteristic
Parameter has corresponding reference value, but still needs to adjust there are five parameter, they are SEF output quantity u respectively0Adjustable parameter k1、
k2；The pre- appraisal z of ESO1, z2, z3Adjustable parameter b01、b02、b03, and this five parameters influence each other.For an active disturbance rejection
For control system, how five parameters are effectively coordinated to combine and obtain good control effect and be only key.
Genetic algorithm main thought is to simulate the chromosome replication occurred in biological natural selection and genetic process, is intersected,
The phenomenon that variation.Procreation evolution generation upon generation of, finally converges to the individual that a group most adapts to environment since any population, is a kind of
Global efficient optimizing algorithm, but it is easy to converge to locally optimal solution and cause " precocity " phenomenon and then influence control effect.
Annealing algorithm is the similitude in analog physical between the annealing process of solid matter and general combinatorial optimization problem
And generate, main thought is to be traversed entire search space stage by stage in the form of high temperature to low temperature and sought solution；And it not only connects
It is solved by optimization, inferior solution is also received with certain probability, it can probabilistic can jumped out from locally optimal solution and finally tend to be global
Optimal solution；But the disadvantage is that largely meaningless using not enough, often being done when solving to a certain range to the feedback information in system
Redundancy iteration, refinement really solves low efficiency.
Particle swarm algorithm (partical swarm optimization, PSO) belongs to a kind of evolutionary computation technique algorithm,
It is one " particle " regarded as in search space that its main thought, which is each initial solution, all particles have one by
The adaptive value that optimised function determines, each particle determine the direction and distance that they move there are one speed, then grain
Son just follow current optimal particle and searched in solution space.PSO is similar with genetic algorithm, is all by system initialization a group
Then random particles find optimal solution by iteration；In each iteration, particle is updated certainly by tracking two " extreme value "
Oneself；One is optimal solution that particle itself is found, this solution is called individual extreme value；Another extreme value is that entire population is looked at present
The optimal solution arrived, this extreme value are global extremums.But in algorithm early stage, that there are precision is low, easily diverging the shortcomings that.
To sum up, any single control method cannot all reach good control effect.It is therefore desirable to a kind of control effect
Good, response is rapidly, strong antijamming capability, precision are high, overshoot is small, fast convergence rate control algolithm.
Summary of the invention
The present invention for existing control algolithm control effect is poor, low-response, poor anti jamming capability, precision is low, overshoot is big,
The slow defect of convergence rate, provide a kind of control effect is good, response rapidly, strong antijamming capability, precision are high, overshoot is small, receive
Hold back fireballing control algolithm.
A kind of skill of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to the present invention
Art scheme is as follows:
A kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to the present invention, it is described
Genetic Annealing Particle Swarm Mixed Algorithm the following steps are included:
Step 1: initial solution group is randomly generated, the initial velocity sp0, initial of each particle in initial solution group is randomly generated
Position x0, setting annealing initial temperature T0, annealing speed a, the number of iterations m, coefficient of variation cm, interaction coefficent cp, weight w1、w2、
w3And the adjustable parameter of ADRC algorithm；
Step 2: by the initial position x of each particle0It is brought into ADRC algorithm with adjustable parameter, passes through fitness function
The corresponding fitness value of each particle is determined, it is specified that the fitness value of current particle is individual extreme value；According to each particle institute
Corresponding individual extreme value determines optimum individual extreme value, and optimum individual extreme value is set to global extremum；
Step 3: by particle present speed spkWith current location xkBring formula (1) and formula (2), more new particle into respectively
Present speed spk+1With current location xk+1；
spk+1=spkcm+cp(pbk-xk)+cp(gbk-xk) (1)
Wherein, wherein pbkFor individual extreme value；gbkFor global extremum；
xk+1=xk+spk+1 (2)
Step 4: by updated position xk+1It brings into ADRC algorithm with adjustable parameter, is obtained respectively by fitness function
A corresponding current individual extreme value of particle and current global extremum；By the current individual extreme value of each particle and former individual extreme value phase
Comparison, if current individual extreme value updates current individual extreme value better than former individual extreme value；By current global extremum and former global pole
Value compares, if current global extremum updates current global extremum better than former individual extreme value；
Step 5: bringing particle present speed and current location into formula (1) and formula (2) respectively, more new particle is worked as
Preceding speed spk+1With current location xk+1；And present speed is no more than maximum speed, i.e. spk+1< spmax；
Step 6: calculate current particle position fitness and update the difference E of preceding particle position fitness, if E < 0 or exp (-
E/T) > rand (0,1) then the new position of more new particle and repeats step 2, until reaching equilibrium state, executes step 7；
Step 7: judge whether to meet termination condition, if satisfied, then stop calculating and export optimal global extremum to get
To automatic disturbance rejection controller optimized parameter；If not satisfied, then cool down, i.e. T=aT0, and repeat step 2.
Further: in step 1, the adjustable parameter includes 5 undetermined parameters；Respectively state error feedback control
Rate SEF output quantity u processed0Adjustable parameter k1、k2；The pre- appraisal z of extended state observer ESO1, z2, z3Adjustable parameter b01、
b02、b03。
Further: in step 2 and step 4, the fitness function formula are as follows:
Wherein, e (t) is feedback error, and e (t)=v1 (k)-y, v1 are tracking input signal, and y is output quantity；T0 is control
Stablize the time, t0=1/r, r are the parameter for adjusting transient process speed；Mp is the overshoot controlled in transient process, Mp=u-
Y, u are control amount.
Further: in step 7, whether the termination condition is whether to reach the number of iterations m or reach to terminate temperature
Degree.
Further: the Control System of Stable Platform include main control unit, power amplifier, torque motor, gyroscope,
A/D converter, angle measuring system and stabilized platform, the stabilized platform is by the controlled quantity of state of feedback by gyroscope and angle measuring system
It is sent to main control unit by A/D converter, the ADRC of genetic Annealing Particle Swarm Mixed Algorithm parameter tuning is embedded in main control unit
In, main control unit carries out processing to feedback quantity and control amount is sent to torque motor by power amplifier, it is driven to make phase
It should act.
A kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to the present invention has
Beneficial effect is:
The present invention has that strong antijamming capability, precision are high, overshoot is small, convergence is fast independent of accurate object model
Fast feature is spent, in the minds of the master control core of this platform control system.It organically combines genetic algorithm to intersect, the bionical spy of variation
Point, simulated annealing probabilistic the characteristics of jumping out locally optimal solution and particle swarm algorithm ability of searching optimum solve ADRC
Parameter tuning problem, improve ADRC control precision.ADRC is set to search comparatively ideal parameter quickly and will control rapidly
Signal is sent into torque motor, improves stabilized platform control effect, and platform stable control precision can be improved, and carrier obtains more on platform
Add stable inertial space position.
Detailed description of the invention
Fig. 1 is genetic Annealing Particle Swarm Mixed Algorithm flow chart；
Fig. 2 is Control System of Stable Platform block diagram；
Fig. 3 is ADRC basic block diagram；
Fig. 4 is genetic Annealing Particle Swarm Mixed Algorithm ADRC parameter tuning simulation architecture figure；
Fig. 5 is genetic Annealing Particle Swarm Mixed Algorithm parameter tuning ADRC step signal figure；
Fig. 6 is typical ADRC step signal figure；
Fig. 7 is genetic Annealing Particle Swarm Mixed Algorithm parameter tuning ADRC Disturbance Rejection response diagram；
Fig. 8 is typical ADRC Disturbance Rejection response diagram.
Specific embodiment
Below with reference to embodiment, the following further describes the technical solution of the present invention, and however, it is not limited to this, all right
Technical solution of the present invention is modified or replaced equivalently, and without departing from the spirit and scope of the technical solution of the present invention, should all be contained
Lid is within the protection scope of the present invention.
Embodiment 1
Illustrate the present embodiment in conjunction with Fig. 1-Fig. 8, it is in the present embodiment, according to the present invention a kind of applied to stabilized platform
The genetic Annealing Particle Swarm Mixed Algorithm of control system, the genetic Annealing Particle Swarm Mixed Algorithm the following steps are included:
Step 1: initial solution group is randomly generated, the initial velocity sp0, initial of each particle in initial solution group is randomly generated
Position x0, setting annealing initial temperature T0, annealing speed a, the number of iterations m, coefficient of variation cm, interaction coefficent cp, weight w1、w2、
w3And the adjustable parameter of ADRC algorithm；
Step 2: by the initial position x of each particle0It is brought into ADRC algorithm with adjustable parameter, passes through fitness function
The corresponding fitness value of each particle is determined, it is specified that the fitness value of current particle is individual extreme value；According to each particle institute
Corresponding individual extreme value determines optimum individual extreme value, and optimum individual extreme value is set to global extremum；
Step 3: by particle present speed spkWith current location xkBring formula (1) and formula (2), more new particle into respectively
Present speed spk+1With current location xk+1；
spk+1=spkcm+cp(pbk-xk)+cp(gbk-xk) (1)
Wherein, wherein pbkFor individual extreme value；gbkFor global extremum；
xk+1=xk+spk+1 (2)
Step 4: by updated position xk+1It brings into ADRC algorithm with adjustable parameter, is obtained respectively by fitness function
A corresponding current individual extreme value of particle and current global extremum；By the current individual extreme value of each particle and former individual extreme value phase
Comparison, if current individual extreme value updates current individual extreme value better than former individual extreme value；By current global extremum and former global pole
Value compares, if current global extremum updates current global extremum better than former individual extreme value；
Step 5: bringing particle present speed and current location into formula (1) and formula (2) respectively, more new particle is worked as
Preceding speed spk+1With current location xk+1；And present speed is no more than maximum speed, i.e. spk+1< spmax；
Step 6: calculate current particle position fitness and update the difference E of preceding particle position fitness, if E < 0 or exp (-
E/T) > rand (0,1) then the new position of more new particle and repeats step 2, until reaching equilibrium state, executes step 7；
Step 7: judge whether to meet termination condition, if satisfied, then stop calculating and export optimal global extremum to get
To automatic disturbance rejection controller optimized parameter；If not satisfied, then cool down, i.e. T=aT0, and repeat step 2.
More specifically: in step 1, the adjustable parameter includes 5 undetermined parameters；Respectively state error is fed back
Control rate SEF output quantity u0Adjustable parameter k1、k2；The pre- appraisal z of extended state observer ESO1, z2, z3Adjustable parameter b01、
b02、b03。
More specifically: in step 2 and step 4, the fitness function formula are as follows:
Wherein, e (t) is feedback error, and e (t)=v1 (k)-y, v1 are tracking input signal, and y is output quantity；T0 is control
Stablize the time, t0=1/r, r are the parameter for adjusting transient process speed；Mp is the overshoot controlled in transient process, Mp=u-
Y, u are control amount.
More specifically: in step 7, whether the termination condition is whether to reach the number of iterations m or reach to terminate temperature
Degree.
More specifically: the Control System of Stable Platform includes main control unit, power amplifier, torque motor, gyro
Instrument, A/D converter, angle measuring system and stabilized platform, the stabilized platform is by the controlled quantity of state of feedback by gyroscope and angle measurement
System is sent to main control unit by A/D converter, and the ADRC of genetic Annealing Particle Swarm Mixed Algorithm parameter tuning is embedded in master control
In unit, main control unit carries out processing to feedback quantity and control amount is sent to torque motor by power amplifier, it is driven to do
Corresponding actions out.
It is a kind of genetic Annealing Particle Swarm Mixed Algorithm used in the present invention.Using particle swarm algorithm thought as basic framework,
Particle is guided using the self-information (location information and velocity information) of particle, individual extreme value information and three category information of global extremum
Next step iterative position；It is low for algorithm early stage precision, the shortcomings that easily diverging, the bionical thought of genetic algorithm is introduced, with current
Solution and the crossover operation during individual extreme value and the difference of global extremum this two mimic biology chromosomal inheritances, interaction coefficent
cp, replace the Studying factors in general particle swarm algorithm, and then avoid leading to low precision because of empirical value deficiency in algorithm early stage,
It is not easy the problem of restraining；Search speed item spkSimulate mutation operation, coefficient of variation cm；As shown in formula 1.And each particle fortune
Scanning frequency degree is restricted to spmax, act predominantly on for command deployment precision, prevent particle from flying over outstanding region easily；Finally
Resulting new explanation not necessarily excellent solution after fitness function screens uses thought --- the probability of simulated annealing here
Property receive inferior solution, avoid algorithm from falling into the predicament of locally optimal solution, and then constitute the algorithm flow of complete set.
Using each particle in the initial solution group generated at random self-information (self-information of particle include position believe
Breath and velocity information), individual extreme value information and global extremum information guide the iterative position of particle next step, current solution respectively with
Individual extreme value and global extremum make poor, the crossover operation during two difference mimic biology chromosomal inheritances, interaction coefficent cp,
Replace the Studying factors in general particle swarm algorithm；Search speed item spkSimulate mutation operation, coefficient of variation cm；Each particle fortune
Scanning frequency degree is restricted to spmax, it is used for command deployment precision, prevents particle from flying over outstanding region easily；
spk+1=spkcm+cp(pbk-xk)+cp(gbk-xk) (1)
xk+1=xk+spk+1 (2)
The invention patent uses following fitness function:
Wherein, e (t) is feedback error, is v1(k)-y；t0Stablize the time to control, takes t0=1/r, MpIn control process
Overshoot, be u-y；w1, w2, w3For weight.It is desirable that fitness function is the smaller the better.
The following are the specific operating procedures of algorithm:
1. a n initial solution (particle) is randomly generated, each solution is to include 5 undetermined parameters；It is initial that particle is randomly generated
Speed and position, maximum speed are limited to spmax；Anneal initial temperature T0, annealing speed a and the number of iterations, cm, cpParameter value；
2. the corresponding fitness value of each particle is determined using automatic disturbance rejection controller simulation architecture figure is built, it is current to adapt to
Angle value is individual extreme value；Global extremum is found out according to the individual extreme value of each particle；
3. generating the new position of given particle according to the particle current location of formula (1) and speed；
4. calculating the fitness value of each new position of particle according to fitness function formula (3)；For giving particle, if its
Better than original individual extreme value current adaptive value is arranged then as individual extreme value in adaptive value；If its adaptive value is better than the original overall situation
Extreme value, then it is global extremum that current adaptive value, which is arranged,；
5. updating particle rapidity, and it is limited in maximum speed sp according to formula (1)maxIt is interior；According to formula (2), more
The current position of new particle；
6. calculating the difference E of fitness caused by two positions.If E<0 or exp (- E/T)>rand (0,1), then receive new value
Return step 2 is until reaching equilibrium state；
7. judging whether to meet termination condition (generally the number of iterations and final temperature).If satisfied, then stopping calculating simultaneously
Optimal solution is exported to get automatic disturbance rejection controller optimized parameter is arrived；Otherwise, cool down T=aT0And return step 2.
In Fig. 1, a kind of genetic Annealing population mixing calculation applied to Control System of Stable Platform is described with flow chart
Platform stable precision can be improved in method, the inertial space position for keeping carrier acquisition thereon more stable.The algorithm is stored in master control core
In, and and power amplifier, torque motor, A/D conversion, gyro is connected with angle measuring system sequence constitutes a space drive shaft.
It organically combines genetic algorithm to intersect, the bionical feature of variation, simulated annealing is probabilistic jumps out local optimum
The characteristics of solution and particle swarm algorithm global iterative search capability, and applied the front end TD and controlled device in ADRC structure
ADRC parameter is adjusted between output.
In Fig. 2, in the ADRC insertion main control unit through genetic Annealing Particle Swarm Mixed Algorithm parameter tuning, main control unit is logical
It crosses this algorithm and the attitude angle converted through A/D and angular speed feedback quantity is handled and sent out control amount, sent by power amplification
To torque motor, it is driven to make corresponding actions to guarantee that carrier keeps stablizing in inertial space.
In Fig. 3, ADRC is mainly by TD (Nonlinear Tracking Differentiator), and (expansion state is seen by SEF (state error feedback rate control), ESO
Survey device) composition.Wherein: TD is usually used in inputting parameter v0Transition process arranging provides tracking input signal v as early as possible1, and extract
Differential signal v2；ESO main detection and unknown disturbance is estimated, provides pre- appraisal z1, z2, z3And by its feedback compensation to controller
In；SEF is by nonlinear function the tracking signal v generated by TD1With differential signal v2With the ESO object model provided
State estimator z1, z2The error e of formation1、e2It is combined into linearisation and generates a control amount u0, and z is estimated in advance with ESO3And benefit
Repay factor b0Ratio formed control amount u pass to controlled device.
3 provide the specific implementation of ADRC each section algorithm with reference to the accompanying drawings.Using second order ADRC as main study subject, it is assumed that its
Are as follows:
Y=v0(k)
Wherein b is uncertainty coefficient, and w (k) is uncertain disturbance, b b0Rough estimate, be adjustable parameter.
TD algorithm is realized:
e0=v1(k)-v0(k)
v1(k+1)=v1(k)+hv2(k)
v2(k+1)=v2(k)+h·fh
Fh=fhan (e0,v2k,r,h0)
v1It (k) is input signal v0Tracking signal, v2It (k) is its differential signal, fhan is the comprehensive letter of high sp eed and optimal control
Number, r are the parameter for adjusting transient process speed, h0For filtering factor, h is integration step, this three is adjustable parameter.
ESO algorithm is realized:
E=z1-y
Wherein,For the differential signal of zi (i=1,2,3).Reasonably select appropriate b01, b02, b03Expansion state can be made to observe
The good tracking system original variable v of device variable0(k)。
SEF algorithm is realized
e1=v1-z1
e2=v2-z2
u0=k1fal(e1,a,δ)+k2fal(e2,a,δ)
U=u0-z3/b0
As 0 < a < 1, function fal has small error, large gain；It is on the contrary then have the characteristics of big error, small gain.Rationally
Each parameter of nonlinear Control amount fal is selected, to realize to integral tandem type object nonlinear Control, and with u pairs of practical control amount
Unknown disturbance and dynamic characteristic compensate.
In Fig. 4, handle is added in ADRC model through genetic Annealing Particle Swarm Mixed Algorithm in MATLAB emulation.
In Fig. 5 and Fig. 6, in the TD input terminal through genetic Annealing Particle Swarm Mixed Algorithm setting parameter ADRC and typical case ADRC
The step signal that given amplitude is " 1 " is placed, the response wave shape figure of system is obtained by oscillograph.Two figures comparison it is found that using
The dynamic property of the Control System of Stable Platform of genetic Annealing Particle Swarm Mixed Algorithm setting parameter ADRC is greatly improved,
Regulating time can original 500ms shorten to 50ms, overshoot significantly reduces.
In Fig. 7 and Fig. 8, in order to test through genetic Annealing Particle Swarm Mixed Algorithm parameter tuning ADRC and typical case ADRC to disturbing
Dynamic rejection ability, herein in system simulation model be added amplitude be 10 °, frequency cycle be 5Hz random waveform noise into
The experiment of row Disturbance Rejection, obtains the Disturbance Rejection response curve under the perturbation action of Fig. 7 and Fig. 8.It is compared by two figures it is found that disturbing
In the identical situation of dynamic frequency, amplitude the latter of the spike of corresponding perturbation response curve is larger, and peak-peak is about 3.5 °.
ADRC i.e. through genetic Annealing Particle Swarm Mixed Algorithm parameter tuning is compared to the stabilized platform control strategy that typical ADRC is used as
It can preferably disturbance suppression act on, guarantee the lasting accuracy of platform.
In the present invention, initial solution group is randomly generated, each particle position solution is brought into ADRC algorithm, is adapted to the present invention
Degree functional form determines the corresponding fitness value of each particle and global extremum；Particle rapidity and position are updated, above-mentioned behaviour is repeated
Make, updates current individual extreme value and preceding global extremum；Calculate the difference E of the fitness of the given new and old position of particle, difference less than 0,
Receive new explanation；Difference is greater than 0, then receives new explanation so that exp (- E/T) > rand (0,1) (T is current annealing temperature) is probability, directly
To equilibrium state is reached, optimal global extremum is exported to get automatic disturbance rejection controller optimized parameter is arrived.
Claims (5)
1. a kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform, which is characterized in that the heredity
Annealed Particle group hybrid algorithm the following steps are included:
Step 1: initial solution group is randomly generated, initial velocity sp0, the initial position of each particle in initial solution group is randomly generated
x0, setting annealing initial temperature T0, annealing speed a, the number of iterations m, coefficient of variation cm, interaction coefficent cp, weight w1、w2、w3And
The adjustable parameter of ADRC algorithm；
Step 2: by the initial position x of each particle0It brings into ADRC algorithm, is determined by fitness function each with adjustable parameter
The corresponding fitness value of a particle is, it is specified that the fitness value of current particle is individual extreme value；According to corresponding to each particle
Individual extreme value determines optimum individual extreme value, and optimum individual extreme value is set to global extremum；
Step 3: by particle present speed spkWith current location xkBring formula (1) and formula (2) into respectively, more new particle is worked as
Preceding speed spk+1With current location xk+1；
spk+1=spkcm+cp(pbk-xk)+cp(gbk-xk) (1)
Wherein, wherein pbkFor individual extreme value；gbkFor global extremum；
xk+1=xk+spk+1 (2)
Step 4: by updated position xk+1It is brought into ADRC algorithm with adjustable parameter, each grain is obtained by fitness function
Sub corresponding current individual extreme value and current global extremum；The current individual extreme value of each particle is opposite with former individual extreme value
Than if current individual extreme value updates current individual extreme value better than former individual extreme value；By current global extremum and former global extremum
It compares, if current global extremum updates current global extremum better than former individual extreme value；
Step 5: bringing particle present speed and current location into formula (1) and formula (2), the current speed of more new particle respectively
Spend spk+1With current location xk+1；And present speed is no more than maximum speed, i.e. spk+1< spmax；
Step 6: calculating current particle position fitness and updating the difference E of preceding particle position fitness, if E < 0 or exp (- E/T)
> rand (0,1) then the new position of more new particle and repeats step 2, until reaching equilibrium state, executes step 7；
Step 7: judging whether to meet termination condition, if satisfied, stopping calculating and exporting optimal global extremum then to get to certainly
Disturbance rejection control device optimized parameter；If not satisfied, then cool down, i.e. T=aT0, and repeat step 2.
2. a kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to claim 1,
It is characterized in that, the adjustable parameter includes 5 undetermined parameters in step 1；Respectively state error feedback rate control SEF
Output quantity u0Adjustable parameter k1、k2；The pre- appraisal z of extended state observer ESO1, z2, z3Adjustable parameter b01、b02、b03。
3. a kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to claim 1,
It is characterized in that, in step 2 and step 4, the fitness function formula are as follows:
Wherein, e (t) is feedback error, and e (t)=v1 (k)-y, v1 are tracking input signal, and y is output quantity；T0 is that control is stablized
Time, t0=1/r, r are the parameter for adjusting transient process speed；Mp is the overshoot controlled in transient process, and Mp=u-y, u are
Control amount.
4. a kind of genetic Annealing Particle Swarm Mixed Algorithm applied to Control System of Stable Platform according to claim 1,
It is characterized in that, whether the termination condition is to reach the number of iterations m or whether reach final temperature in step 7.
5. the Control System of Stable Platform of genetic Annealing Particle Swarm Mixed Algorithm described in -4, feature exist according to claim 1
In the Control System of Stable Platform includes main control unit, power amplifier, torque motor, gyroscope, A/D converter, angle measurement
The controlled quantity of state of feedback is passed through A/D converter by gyroscope and angle measuring system by system and stabilized platform, the stabilized platform
It is sent to main control unit, the ADRC of genetic Annealing Particle Swarm Mixed Algorithm parameter tuning is embedded in main control unit, main control unit pair
Feedback quantity carries out processing and control amount is sent to torque motor by power amplifier, it is driven to make corresponding actions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811074461.8A CN109188907A (en) | 2018-09-14 | 2018-09-14 | A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811074461.8A CN109188907A (en) | 2018-09-14 | 2018-09-14 | A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109188907A true CN109188907A (en) | 2019-01-11 |
Family
ID=64911181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811074461.8A Pending CN109188907A (en) | 2018-09-14 | 2018-09-14 | A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109188907A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109856976A (en) * | 2019-03-12 | 2019-06-07 | 哈尔滨工程大学 | It is a kind of that Auto-disturbance-rejection Control is tracked based on the adaptive track laying air cushion vehicle for intersecting particle group optimizing |
| CN110472839A (en) * | 2019-07-25 | 2019-11-19 | 上海电力大学 | Thermal power plant's control system Information Security Evaluation system based on SA-PSO-AHP |
| CN111045328A (en) * | 2019-12-20 | 2020-04-21 | 中国科学院光电技术研究所 | Sliding mode frequency domain parameter identification method based on simulated annealing particle swarm for photoelectric tracking platform |
| CN111092602A (en) * | 2019-12-27 | 2020-05-01 | 京信通信系统（中国）有限公司 | Modeling method and device of power amplifier, computer equipment and storage medium |
| CN115441478A (en) * | 2022-10-24 | 2022-12-06 | 盛世华通(山东)电气工程有限公司 | Photovoltaic power smoothing method based on SA-PSO (SA-particle swarm optimization) boundary |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408589A (en) * | 2014-10-24 | 2015-03-11 | 陕西科技大学 | AGV optimization scheduling method based on mixed particle swarm optimization |
| CN106502092A (en) * | 2016-10-21 | 2017-03-15 | 东南大学 | A kind of thermal process model parameter identification method using improvement Hybrid Particle Swarm |
| CN107272403A (en) * | 2017-06-14 | 2017-10-20 | 浙江师范大学 | A kind of PID controller parameter setting algorithm based on improvement particle cluster algorithm |
-
2018
- 2018-09-14 CN CN201811074461.8A patent/CN109188907A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408589A (en) * | 2014-10-24 | 2015-03-11 | 陕西科技大学 | AGV optimization scheduling method based on mixed particle swarm optimization |
| CN106502092A (en) * | 2016-10-21 | 2017-03-15 | 东南大学 | A kind of thermal process model parameter identification method using improvement Hybrid Particle Swarm |
| CN107272403A (en) * | 2017-06-14 | 2017-10-20 | 浙江师范大学 | A kind of PID controller parameter setting algorithm based on improvement particle cluster algorithm |
Non-Patent Citations (3)
| Title |
|---|
| WANG XIAOBIN等: "On ADRC for Photoelectrical Stabilized Platform", 《2011 SECOND INTERNATIONAL CONFERENCE ON MECHANIC AUTOMATION AND CONTROL ENGINEERING》 * |
| 朱丽玲等: "基于遗传算法的ADRC参数整定及其应用", 《仪器仪表用户》 * |
| 瞿永飞: "基于遗传算法的陀螺稳定平台速度环的自抗扰控制", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109856976A (en) * | 2019-03-12 | 2019-06-07 | 哈尔滨工程大学 | It is a kind of that Auto-disturbance-rejection Control is tracked based on the adaptive track laying air cushion vehicle for intersecting particle group optimizing |
| CN110472839A (en) * | 2019-07-25 | 2019-11-19 | 上海电力大学 | Thermal power plant's control system Information Security Evaluation system based on SA-PSO-AHP |
| CN111045328A (en) * | 2019-12-20 | 2020-04-21 | 中国科学院光电技术研究所 | Sliding mode frequency domain parameter identification method based on simulated annealing particle swarm for photoelectric tracking platform |
| CN111045328B (en) * | 2019-12-20 | 2022-09-20 | 中国科学院光电技术研究所 | Sliding mode frequency domain parameter identification method based on simulated annealing particle swarm and aiming at photoelectric tracking platform |
| CN111092602A (en) * | 2019-12-27 | 2020-05-01 | 京信通信系统（中国）有限公司 | Modeling method and device of power amplifier, computer equipment and storage medium |
| CN111092602B (en) * | 2019-12-27 | 2023-10-20 | 京信网络系统股份有限公司 | Modeling method, modeling device, computer equipment and storage medium of power amplifier |
| CN115441478A (en) * | 2022-10-24 | 2022-12-06 | 盛世华通(山东)电气工程有限公司 | Photovoltaic power smoothing method based on SA-PSO (SA-particle swarm optimization) boundary |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109188907A (en) | A kind of genetic Annealing Particle Swarm Mixed Algorithm and its Control System of Stable Platform applied to Control System of Stable Platform | |
| Mandlekar et al. | Iris: Implicit reinforcement without interaction at scale for learning control from offline robot manipulation data | |
| Li et al. | Integral reinforcement learning for linear continuous-time zero-sum games with completely unknown dynamics | |
| Reynoso-Meza et al. | Controller tuning using evolutionary multi-objective optimisation: current trends and applications | |
| CN108133258A (en) | A kind of mixing global optimization method | |
| Quirynen | Numerical simulation methods for embedded optimization | |
| Kubalík et al. | Symbolic regression methods for reinforcement learning | |
| Deepa et al. | Model order formulation of a multivariable discrete system using a modified particle swarm optimization approach | |
| Song et al. | A new self-learning optimal control laws for a class of discrete-time nonlinear systems based on ESN architecture | |
| CN112462611A (en) | Sliding friction modeling method for precise electromechanical system | |
| Izci et al. | A novel modified opposition‐based hunger games search algorithm to design fractional order proportional‐integral‐derivative controller for magnetic ball suspension system | |
| Jagannathan et al. | Neural-network-based state feedback control of a nonlinear discrete-time system in nonstrict feedback form | |
| Hein et al. | Generating interpretable fuzzy controllers using particle swarm optimization and genetic programming | |
| Li et al. | Hybrid reinforcement learning for optimal control of non-linear switching system | |
| Wang et al. | Advanced policy learning near-optimal regulation | |
| Zhu et al. | Baidu apollo auto-calibration system-an industry-level data-driven and learning based vehicle longitude dynamic calibrating algorithm | |
| Priyambada et al. | Fuzzy-PID controller optimized TLBO approach on automatic voltage regulator | |
| Arshad et al. | Deep Deterministic Policy Gradient to Regulate Feedback Control Systems Using Reinforcement Learning. | |
| Wang et al. | Survey of transient performance control | |
| Reid et al. | Mutual Q-learning | |
| Guan et al. | Robust adaptive recurrent cerebellar model neural network for non-linear system based on GPSO | |
| Li et al. | Morphing Strategy Design for UAV based on Prioritized Sweeping Reinforcement Learning | |
| Tsourdos et al. | Fuzzy multi-objective design for a lateral missile autopilot | |
| CN109657778B (en) | Improved multi-swarm global optimal-based adaptive pigeon swarm optimization method | |
| Wang et al. | A general adaptive dynamic programming approach with experience replay |
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: 20190111 |
|
| RJ01 | Rejection of invention patent application after publication |