CN108615069A - A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization - Google Patents
A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization Download PDFInfo
- Publication number
- CN108615069A CN108615069A CN201810248937.9A CN201810248937A CN108615069A CN 108615069 A CN108615069 A CN 108615069A CN 201810248937 A CN201810248937 A CN 201810248937A CN 108615069 A CN108615069 A CN 108615069A
- Authority
- CN
- China
- Prior art keywords
- particle
- optimization
- inertia weight
- value
- fitness
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The purpose of the present invention is to provide a kind of optimized calculation methods based on improved adaptable quanta particle swarm optimization, the technical solution adopted by the present invention is that establishing suitable evaluation index as needed, the degree of optimization for distinguishing each particle voluntarily distributes different Inertia Weights according to each particle Optimal State.The present invention solves unique control parameter in traditional QPSO algorithms, the single adjustment problem that inertia weight is linearly reduced.By pointedly adjusting the inertia weight of the particle of Different Optimization degree, the ability of algorithm global optimizing can be effectively improved.
Description
Technical field
The present invention relates to a kind of optimized calculation methods solving high-dimensional problem.
Background technology
Due to the good practicability of optimisation technique, generally apply in various engineering fields at present.The purpose of optimization exists
In obtaining preferred plan from a large amount of optimisation strategy, the feature of optimum point of the object function in domain and corresponding solution are determined
Method.It can be defined as using optimization algorithm Solve problems, under the limitation according to certain constraints, be found suitable a series of
Systematic parameter is to ensure that whole system can obtain maximum or minimum output result or performance indicator.
Optimization problem (Optimization Problem) is defined as follows：
In formula, S is solution space, is nonempty set, indicates the search space entirely optimized；F (x) problem of representation models pair
The object function answered, to be defined on the real-valued function in solution space, inside can have multiple variables；Arg min f (x) are indicated
Solve the built-in variable x when f (x) has minimum value1,x2,x3... a series of values, minimum value at this time is equally optimization problem
The target of desired solution.The optimization problem of maximum value can be equivalently represented for solution arg min f (- x).
Under normal circumstances, the solution of optimization problem can be divided into local extremum and the overall situation is most worth, and for above formula, there have to be following fixed
Justice：
A. if there is a point xn *∈ B so thatThere are f (x*)≤f (x), whereinThen claim xn *∈ S are
Local minimum point in the B of domain, f (xn *) it is local minimum；
B. if there is a point xn *∈ S so thatThere are f (x*)≤f (x), whereinThen claim xn *∈ S are whole
The global minima point of a search space, f (xn *) it is global minimum；
Under normal conditions, local minimum is frequently not global minimum.Therefore, optimization algorithm should in searching process
System is avoided to be absorbed in local minimum as possible to miss the global minimum of entire search space.
Population (QPSO) algorithm based on quantum behavior is to cannot achieve freedom for particle in population (PSO) algorithm
Search problem makes a kind of improved method, and this method is newer to particle position using the theoretical method based on quantum physics
Strategy is made that improvement, introduces the concept of potential field and particle is enable to appear in any position in region of search with certain probability.
Successful application in every field shows：This method is fast with the speed of service, optimizing performance is strong, control parameter is easy to adjust less
The advantages that whole.
Currently, what QPSO applied mostly in each field is method that inertia weight linearly reduces.However, this method is neglected
The difference between individual in population is omited, even if in the same period, the quality of each searched out solution of particle is also far from each other.Cause
This, the unified distribution adjustment of Inertia Weight can make some still have the more excellent possible particle of solution of search and enter local optimal searching in advance,
To miss optimal solution.Especially when solving high-dimensional problem, this defect just more highlights.
Invention content
The purpose of the present invention is to provide linearly reduce to bring for inertia weight in traditional QPSO algorithms finally cannot
Seek optimal solution the problem of, and can be directed to each particle difference search condition determine corresponding inertia weight one kind be based on improve
Adaptable quanta particle swarm optimization optimized calculation method.
The object of the present invention is achieved like this：
A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization of the present invention, feature
It is：
(1) object function to be optimized is established；
(2) parameter needed for optimization algorithm, including population scale, problem dimension, maximum iteration, inertia power are initialized
Weight range；
(3) initial solution is generated at random；
(4) entering AQPSO algorithm iteration processes, calculate mbest, mbest is that current particle is averaged optimum position,
Wherein, C (t) indicates that the average value of whole personal best particles, m indicate the number of particle, D problem of representation dimensions；
(5) fitness function value of corresponding object function is calculated；
(6) pbest and gbest is updated, pbest is the optimal location that particle itself lives through, and gbest undergoes for population
The optimal location crossed；
(7) characteristic length for updating i-th of d dimension potential well, as described in following formula：
Li,d(t)=2 ω Cd(t)-Xi,d(t),
Wherein, w is inertia weight, Cd(t) mbest, X of particle are tieed up for di,d(t) position of particle is tieed up for i-th of d；
(8) location information of more new particle；
(9) classify to particle, i-th generation group's average fitness value be calculated,
Wherein, f1 i,f2 i,…,fN iIt indicates each particle fitness value, cost function value is higher thanIndividual make even again
It is classification indicatorsThe optimal f of history in known group in algorithm iterationgbest, calculation optimization level index Δ, such as following formula：
(10) according to classification indicators to particle classifying and one by one determining inertia weight,
When the fitness value of particle n is higher thanWhen, the calculation formula for corresponding to w at this time is
When particle fitness value meetsCorresponding to w calculation formula is
Wherein, wmax、wminSetting minimax inertia weight value is indicated respectively；
When particle n fitness values are less than group's average level of current iteration, the calculation formula of corresponding w is
Wherein, k is the parameter for adjusting the Inertia Weight upper limit；
(11) judge whether to reach maximum iteration or set target, if it is satisfied, otherwise output is as a result, go to step
(4)。
Advantage of the invention is that：
1, unique control parameter in tradition QPSO algorithms, the single adjustment problem that inertia weight is linearly reduced are solved.
2, by pointedly adjusting the inertia weight of the particle of Different Optimization degree, the algorithm overall situation can be effectively improved and sought
Excellent ability.
Description of the drawings
Fig. 1 is UUV path planning track results under the two-dimentional marine environment there are forbidden zone；
Fig. 2 is iterativecurves of the UUV under the two-dimentional marine environment there are forbidden zone obtained by path planning；
Fig. 3 is the path planning track that UUV passes through that recycling is realized on island；
Fig. 4 is the path planning energy expenditure iterativecurve that UUV passes through that recycling is realized on island.
Specific implementation mode
It illustrates below in conjunction with the accompanying drawings and the present invention is described in more detail：
In conjunction with Fig. 1-4, the present invention is that can be directed to the adaptation that each particle difference search condition determines corresponding inertia weight
Property quantum particle swarm (AQPSO) method.The technical solution adopted by the present invention is that establishing suitable evaluation index as needed, distinguish
The degree of optimization of each particle voluntarily distributes different Inertia Weights according to each particle Optimal State, specifically can be according to the following steps
It realizes：
Step 1：According to actual optimization problem, object function to be optimized is established.
Step 2：Initialize all kinds of parameters needed for optimization algorithm, including population scale, problem dimension, greatest iteration time
Number, inertia weight range.
Step 3：It is random to generate initial solution.
Step 4：Into AQPSO algorithm iteration processes, calculates current particle and be averaged optimum position (mbest).
Wherein, C (t) indicates that the average value of whole personal best particles, m indicate the number of particle, D problem of representation dimensions.
Step 5：Calculate the fitness function value of corresponding object function.
Step 6：Update pbest and gbest.
Step 7：Update the characteristic length of i-th of d dimension potential well.As described in following formula,
Li,d(t)=2 ω | Cd(t)-Xi,d(t)|
Wherein, w is inertia weight, Cd(t) mbest of particle is tieed up for d,X i,d(t) it is the position that i-th of d ties up particle.
Step 8：The location information of more new particle
Step 9：Start to classify to particle, i-th generation group's average fitness value be calculated,
Wherein, f1 i,f2 i,…,fN iIndicate each particle fitness value.Cost function value is higher thanIndividual make even again
It is classification indicatorsThe optimal f of history in known group in algorithm iterationgbest, calculation optimization level index Δ, such as following formula：
Step 10：According to classification indicators to particle classifying and one by one determining inertia weight,
When the fitness value of particle n is higher thanWhen, the calculation formula of w is corresponded at this time,
When particle fitness value meetsCorresponding to w calculation formula is,
Wherein, wmax, wminSetting minimax inertia weight value is indicated respectively.
When particle n fitness values are less than group's average level of current iteration, the calculation formula of corresponding w
Wherein, k is the parameter for adjusting the Inertia Weight upper limit, generally takes k=1.5, corresponds to w ∈ (0.5,1) at this time.
Step 11：Judge whether to reach maximum iteration or set target, if it is satisfied, termination algorithm and exporting knot
Otherwise fruit goes to step 4 and continues optimization process.
The present invention can be towards the path planning problem that optimization UUV is optimized based on energy expenditure.In order to test the property of algorithm
Can, consider increase problem complexity and improve problem dimension, separately design there are forbidden zone and pass through island realize recycling two
Tie up two kinds of cases of path planning under marine environment.Through three kinds of algorithms of different, standard PSO, standard QPSO and AQPSO of the present invention
The track that algorithm optimization obtains, as shown in figures 1 and 3.Three tracks are corresponded to respectively with the road of energy consumption target as an optimization in figure
The track that diameter planning is optimized by three kinds of optimization algorithms is as a result, it can be found that being optimized by three kinds of optimization algorithms from figure
To track can effectively utilize ocean circulation and avoid terrain obstruction and forbidden zone to arrive at target location, but three kinds of tracks
Detour is respectively different using the degree of circulation.It can be clearly from the iterativecurve figure of the fitness value of the energy consumption of Fig. 2 and Fig. 4
See that AQPSO algorithm optimizations obtain the corresponding energy expenditure in track and are better than other two kinds of optimization algorithms, wherein standard PSO is calculated
The energy consumption highest that method optimizes, this is because PSO algorithm itself is easy to miss globally optimal solution.And increase in problem dimension
In second case added, in Fig. 3, AQPSO algorithm optimizations obtain the corresponding energy consumption in track and are far below PSO, and are substantially better than
The energy consumption that standard QPSO algorithm optimizations obtain, this fully confirms that the invention is remotely navigated for UUV under the influence of complicated ocean circulation
During the plan optimization of extra large energy consumption path, that is, the complexity and dimension for needing optimization problem are substantially increased, still had excellent
Performance.
Claims (1)
1. a kind of optimized calculation method based on improved adaptable quanta particle swarm optimization, it is characterized in that：
(1) object function to be optimized is established；
(2) parameter needed for optimization algorithm, including population scale, problem dimension, maximum iteration, inertia weight model are initialized
It encloses；
(3) initial solution is generated at random；
(4) entering AQPSO algorithm iteration processes, calculate mbest, mbest is that current particle is averaged optimum position,
Wherein, C (t) indicates that the average value of whole personal best particles, m indicate the number of particle, D problem of representation dimensions；
(5) fitness function value of corresponding object function is calculated；
(6) pbest and gbest is updated, pbest is the optimal location that particle itself lives through, and gbest is what population lived through
Optimal location；
(7) characteristic length for updating i-th of d dimension potential well, as described in following formula：
Li,d(t)=2 ω | Cd(t)-Xi,d(t) |,
Wherein, w is inertia weight, Cd(t) mbest, X of particle are tieed up for di,d(t) position of particle is tieed up for i-th of d；
(8) location information of more new particle；
(9) classify to particle, i-th generation group's average fitness value be calculated,
Wherein, f1 i,f2 i,…,fN iIt indicates each particle fitness value, cost function value is higher thanIndividual be averaged again for point
Class indexThe optimal f of history in known group in algorithm iterationgbest, calculation optimization level index Δ, such as following formula：
(10) according to classification indicators to particle classifying and one by one determining inertia weight,
When the fitness value of particle n is higher thanWhen, the calculation formula for corresponding to w at this time is
When particle fitness value meetsCorresponding to w calculation formula is
Wherein, wmax、wminSetting minimax inertia weight value is indicated respectively；
When particle n fitness values are less than group's average level of current iteration, the calculation formula of corresponding w is
Wherein, k is the parameter for adjusting the Inertia Weight upper limit；
(11) judge whether to reach maximum iteration or set target, if it is satisfied, otherwise output is as a result, go to step (4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810248937.9A CN108615069A (en) | 2018-03-25 | 2018-03-25 | A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810248937.9A CN108615069A (en) | 2018-03-25 | 2018-03-25 | A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108615069A true CN108615069A (en) | 2018-10-02 |
Family
ID=63658691
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810248937.9A Pending CN108615069A (en) | 2018-03-25 | 2018-03-25 | A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108615069A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109866222A (en) * | 2019-02-26 | 2019-06-11 | 杭州电子科技大学 | A kind of manipulator motion planning method based on longicorn palpus optimisation strategy |
| CN110781979A (en) * | 2019-11-08 | 2020-02-11 | 浙江工业大学 | Parameter matching method for plug-in hybrid electric vehicle assembly |
| CN111061285A (en) * | 2019-12-12 | 2020-04-24 | 哈尔滨工程大学 | Ship dynamic positioning thrust distribution method |
| CN116993246A (en) * | 2023-09-26 | 2023-11-03 | 上海伯镭智能科技有限公司 | Intelligent management method and system for unmanned delivery vehicle |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682682A (en) * | 2016-10-20 | 2017-05-17 | 北京工业大学 | Method for optimizing support vector machine based on Particle Swarm Optimization |
-
2018
- 2018-03-25 CN CN201810248937.9A patent/CN108615069A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682682A (en) * | 2016-10-20 | 2017-05-17 | 北京工业大学 | Method for optimizing support vector machine based on Particle Swarm Optimization |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109866222A (en) * | 2019-02-26 | 2019-06-11 | 杭州电子科技大学 | A kind of manipulator motion planning method based on longicorn palpus optimisation strategy |
| CN110781979A (en) * | 2019-11-08 | 2020-02-11 | 浙江工业大学 | Parameter matching method for plug-in hybrid electric vehicle assembly |
| 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 |
| CN116993246A (en) * | 2023-09-26 | 2023-11-03 | 上海伯镭智能科技有限公司 | Intelligent management method and system for unmanned delivery vehicle |
| CN116993246B (en) * | 2023-09-26 | 2023-12-05 | 上海伯镭智能科技有限公司 | Intelligent management method and system for unmanned delivery vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310915B (en) | Data anomaly detection defense method oriented to reinforcement learning | |
| CN108615069A (en) | A kind of optimized calculation method based on improved adaptable quanta particle swarm optimization | |
| CN108108832B (en) | A kind of oil truck oil and gas leakage intelligent monitor system based on wireless sensor network | |
| WO2017035964A1 (en) | Method and system for determining load characteristics of electric power system | |
| CN110750655B (en) | Knowledge base optimization method of intelligent IETM fault maintenance auxiliary system | |
| CN102024179A (en) | Genetic algorithm-self-organization map (GA-SOM) clustering method based on semi-supervised learning | |
| CN110347151A (en) | A kind of robot path planning method merging Bezier optimized Genetic Algorithm | |
| CN104899135A (en) | Software defect prediction method and system | |
| CN111222286A (en) | Parameter optimization method based on power transmission line state estimation | |
| CN107292381A (en) | A kind of method that mixed biologic symbiosis for single object optimization is searched for | |
| CN111310884A (en) | Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm | |
| Ducange et al. | Multi-objective evolutionary fuzzy systems | |
| CN116402002A (en) | Multi-target layered reinforcement learning method for chip layout problem | |
| CN116341605A (en) | Grey wolf algorithm hybrid optimization method based on reverse learning strategy | |
| CN115933693A (en) | Robot path planning method based on adaptive chaotic particle swarm algorithm | |
| Liu et al. | A Hybrid Harmony Search Algorithm with Distribution Estimation for Solving the 0‐1 Knapsack Problem | |
| Lee et al. | Boundary-focused generative adversarial networks for imbalanced and multimodal time series | |
| CN109697531A (en) | A kind of logistics park-hinterland Forecast of Logistics Demand method | |
| CN111967670A (en) | Switch cabinet partial discharge data identification method based on improved wolf algorithm | |
| Su et al. | A cvae-gan-based approach to process imbalanced datasets for intrusion detection in marine meteorological sensor networks | |
| Xu et al. | Reward-based exploration: Adaptive control for deep reinforcement learning | |
| CN108921277A (en) | A kind of multi-modal optimization backbone particle swarm optimization based on improvement species conversation strategy | |
| Li et al. | Fault diagnosis knowledge base optimization based on decision tree | |
| Tu et al. | Transformer fault diagnosis based on improved dung beetle algorithm optimizing SVM | |
| CN107728469A (en) | A kind of laser particle control method |
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: 20181002 |
|
| RJ01 | Rejection of invention patent application after publication |