30 results about "Genetic programming algorithm" patented technology

Disclosed are an unmanned aerial vehicle (UAV) route planning system and a UAV route planning method based on genetic programming. The method comprises the following steps: initial populations of a tree structure are built through a UAV model module; each individual is decoded and the fitness value of each individual is calculated through a genetic programming algorithm module; selecting and breeding operations are performed between the populations, and an optimal population is obtained through a plurality of iteration processes; and finally, an optimal individual is selected from the optimal population and decoded through a UAV task module to obtain an optimal route of genetic programming. According to the invention, initializing, decoding, selecting and breeding steps are performed by use of the tree structure to optimize the route constantly. The optimization process is quick, the method is visual, the performance of the planned route is improved, the computing time is reduced, the degree of fitness is optimized, and the system and the method are of very high feasibility and robustness.

The invention discloses a DFB laser frequency stabilization method based on current control, and the method comprises the steps: carrying out the wavelength-current modeling through employing the current tuning characteristics of a DFB laser and a genetic programming algorithm, so as to set the constant frequency DC work environment of the DFB laser; obtaining a saturated absorption spectrum signal based on the saturated absorption principle; obtaining an odd harmonic differential error signal containing frequency information through employing a phase sensitive detection principle, and carrying out the PID control parameter optimization of the error signal through employing the genetic algorithm. The method achieves the quick and accurate adaptive frequency locking control, gives consideration to the high precision of saturated absorption frequency stabilization, and can guarantee the long-time high precision and stability of the frequency of the laser.

The invention relates to a two-layer genetic integer programming-based complex system DSM (Design Structure Matrix) reconstructing method which can be applied to the industrial fields of aerospace, cars, ships and the like. According to the method, the simultaneous optimization of the element sequence and the clustering scheme in a DSM is realized by adopting a double-segment chromosome coding technique; and layered solving is carried out on a DSM clustering problem by adopting an integer genetic programming algorithm so as to obtain a reconstructed optimal DSM. The method comprises the following steps of firstly building an optimization model through taking DSM-based contact information flow as output and carrying out two-layer optimization on the model; obtaining a preliminary DSM clustering scheme by adopting a genetic integer programming method in the first-layer reconstruction; and carrying out a second search on each cluster in the preliminary scheme by adopting the same algorithm in the second-layer reconstruction so as to obtain a final DSM reconstruction result. Therefore, the method has the advantages of simplifying the design process, shortening the development time and increasing the resource utilization rate.

The invention discloses a large-scale symbol regression method and system based on an adaptive parallel genetic algorithm, and the system comprises a main process module which is used for initializingand calling a CPU thread module and realizing a synchronization barrier and migration operation; a CPU thread module which is used for executing a genetic programming algorithm, realizing EV updatingand calling the GPU adaptive value evaluation module; and a GPU adaptive value evaluation module which comprises a CPU auxiliary thread, a CUDA library function and a CUDA self-defined function and is used for executing adaptive value evaluation. According to the invention, a self-adaptive multi-population evolution mechanism and a parallel computing system of heterogeneous computing resources are introduced into a genetic programming algorithm; effective construction elements are successfully extracted by applying an adaptive multi-population evolution mechanism, so that the performance of agenetic programming algorithm in a complex problem of the multi-construction elements is improved, and by designing a parallel computing system of heterogeneous computing resources, computing resources of a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) are fully utilized, and the searching efficiency is remarkably improved.

The invention belongs to the technical field of image processing, and particularly discloses a graphic image sorting method based on a genetic programming algorithm. The method comprises the steps that (1) image feature sets and training feature sets are constructed; (2) first-stage relevant parameters are set; (3) the image features are extracted; (4) species are initialized and fitness is estimated; (5) individuals are subjected to the survival of the fittest, superior individuals are subjected to crossover and variation, and the fitness is estimated; (6) the individuals are subjected to partial searching; (7) the species are estimated, and if the crossover and variation operation is completed is judged; (8) the good species are selected, and if evolution is over is judged; (9) the species are again initialized according the new features; (10) the better species are selected for crossover and variation; (11) an image matching model is output according to the optimal individual, and an individual tree is decoded to obtain new image features. The training model generated through the method can effectively improve image sorting precision.

The invention discloses a residual life prediction algorithm based on optimal degradation characteristic quantity. The residual life prediction algorithm comprises the steps of extracting the optimal degradation characteristic quantity of equipment and establishing a Weiner process residual life prediction model with a random effect. Compared with the prior art, the residual life prediction algorithm solves the problem of choosing degradation characteristic quantity in a man-made subjective mode and overcomes the bad effect that a single degradation characteristic quantity cannot completely represent equipment states. According to the residual life prediction algorithm, the merits and demerits of a genetic programming algorithm and the Weiner process model with the random effect are combined, and an optimal first prediction time can be chosen; thus, the residual life prediction algorithm based on the optimal degradation characteristic quantity can be established. From simulation results, the prediction accuracy of the algorithm is certainly improved.

A genetic programming algorithm based on local search for dynamic job shop scheduling is applicable to the field of job shop scheduling. According to the technical scheme adopted by the invention, the algorithm comprises the following steps: first, solving the scheduling optimization problem by use of scheduling rules; second, automatically designing scheduling rules by use of a genetic programming method; third, using a heuristic method of local search; fourth, perturbing the current optimal solution of local search; and fifth, introducing a taboo search strategy in the process of local search. The search mechanism of the algorithm achieves balance between development and exploration. Compared with the existing algorithm, compact enough and better scheduling rules can be obtained in a smaller computing time range.

The invention provides an incremental data mining method based on the genetic programming algorithm. Incremental data mining can accomplish incremental model learning task from less to more of data samples; data is input into a model; an input layer performs linear mapping on data, and transmits results into an intermediate layer; the intermediate layer respectively performs nonlinear transformation and space lifting mapping, and outputs results to a voting system; the voting system determines belonging category; a feedback system optimizes network parameters. The learning process is accomplished through multiple iterations. When incremental data is processed, the coupling factor of new and old samples is low, old data samples are not required to be considered when new data is mined, and the method has good succession.

One of the embodiments relates to a feature processing method and a system for machine learning. The method comprises the steps of obtaining a plurality of candidate features from a basic feature setand obtaining a plurality of candidate operators from a basic operator set, and forming a plurality of initial feature combinations by the plurality of candidate features and the plurality of candidate operators; taking the plurality of initial feature combinations as an initial population of a genetic programming algorithm, and performing genetic manipulation on the initial population by adoptingthe genetic programming algorithm to obtain an optimized target population; obtaining a target feature combination based on the optimized target population, wherein the target feature combination isrepresented by calculation results of basic features and basic operators; wherein the basic features belong to a basic feature set, and the basic operators belong to basic operator subsets; and takingthe target feature combination as a feature of machine learning to participate in operation of machine learning.

The invention relates to a ship speed prediction and safe speed control method based on big data mining. The method comprises the following steps: firstly, collecting a sample database of multiple influence factors influencing ship navigation in a plain river network channel; then, constructing a full-connection deep learning neural network model, and optimizing initial parameters by taking prediction precision as a control target; by adopting the optimized full-connection deep learning neural network model, carrying out comparative analysis on the prediction precision of the multiple influence factors under different combinations, and selecting the combination of the optimal influence factors; on the basis of a data sample of the combination of the optimal influence factors, compiling a genetic programming algorithm program by adopting Python, optimizing core parameters by taking prediction precision as a target, and obtaining a parameter combination with the optimal precision; substituting the parameter combination with the optimal precision into an algorithm program to obtain a ship speed prediction model based on a genetic programming method; and finally, combining the ship speed obtained by the ship speed prediction model based on the genetic programming method with the risk coefficient to obtain the safety control ship speed.

The invention provides a genetic programming classification method based on geometric semantics. A training process and a prediction process are separated, so as to complete classification of tested samples; the training process comprises the steps of: solving an optimal individual through the geometric semantics, extracting a classifier formula of the optimal individual, and storing the classifier formula of the optimal individual in a magnetic disk; the prediction process comprises the steps of: invoking the classifier formula, which is stored in the magnetic disk, of the optimal individual, recovering the classifier formula through loading and calculation, and outputting classification results according to the classifier formula, so as to classify a plurality of individuals. The problems of too early convergence, low classification accuracy and the like of an existing genetic programming algorithm are solved, the classification accuracy is high, and the individual formula can be stored.

The invention discloses an improved genetic programming algorithm optimization method for resource-constrained multi-project scheduling. The method comprises the following steps: step 1, initializingparameters, a function set and an attribute set; step 2, collecting project set data under different working conditions, and decomposing the project set data into a training set and a test set; step 3, extracting project information in each working condition project set in the training set as training input, extracting a function set and an attribute set as coding bases, and training populations in the improved genetic programming algorithm; step 4, judging whether the maximum working condition number of the training set is reached or not, if so, outputting an optimal solution set in the population, and if not, returning to the step 3 after converting the project set; step 5, testing the optimal solution set output in the step 4 by adopting a test set and a training set. According to the method, the resource-constrained multi-project scheduling problem under the single / multiple targets can be solved, the defect that traditional genetic programming is prone to falling into local optimumis overcome, and the searching and training capacity of genetic programming is improved.

The invention discloses a method and system for constructing an equivalent circuit model of a lithium ion battery, and equipment. The method comprises the following steps: setting parameter values ofpopulation size, selection rate, crossover rate, mutation rate and maximum evolution algebra; using a genetic programming algorithm to randomly generate initial populations with the same scale and quantity as the populations, with each initial population comprising a plurality of expression trees, each expression tree comprising a random equivalent circuit; according to the selection rate, the crossover rate and the mutation rate, performing selection, crossover and mutation operations on all the initial populations; calculating the fitness of each generated expression tree, arranging the fitness of each expression tree from large to small, taking the expression tree with half of the fitness, performing selection, intersection and mutation operation on the reserved expression trees, and calculating the fitness of the reserved expression trees; and until the reserved fitness of the expression tree meets any one of the maximum iteration number and the fitness function, decoding the expression tree with the maximum fitness to obtain the topological structure of the equivalent circuit model.

The invention relates to the technical field of engineering design, in particular to an engineering design method for conducting automated evolution design on mixing field systems or various different field systems based on a bond graph tool and a genetic programming algorithm. The method includes the steps of assigning an embryo physical structure diagram, conducting user input, setting up an initial population of a genetic programming tree, conducting ontoanalysis, conducting evolution operation and ending judgment. Compared with the prior art, the design method is high in expansibility, wide in application range, high in efficiency, high in normalization and capable of designing a complex project meeting the requirements of a designer through an automated method.

The invention belongs to the technical field of image processing, and particularly discloses a graphic image sorting method based on a genetic programming algorithm. The method comprises the steps that (1) image feature sets and training feature sets are constructed; (2) first-stage relevant parameters are set; (3) the image features are extracted; (4) species are initialized and fitness is estimated; (5) individuals are subjected to the survival of the fittest, superior individuals are subjected to crossover and variation, and the fitness is estimated; (6) the individuals are subjected to partial searching; (7) the species are estimated, and if the crossover and variation operation is completed is judged; (8) the good species are selected, and if evolution is over is judged; (9) the species are again initialized according the new features; (10) the better species are selected for crossover and variation; (11) an image matching model is output according to the optimal individual, and an individual tree is decoded to obtain new image features. The training model generated through the method can effectively improve image sorting precision.

The invention provides an order processing method and device. The method comprises the steps of obtaining a plurality of orders to be processed; according to the plurality of attribute features of the order, constructing an order scoring tree conforming to a genetic programming algorithm, wherein the order scoring tree comprises a logic operation relation among the plurality of attribute features; determining an order score of the order according to a logical operation relationship among multiple attribute features in the order score tree; determining an order sequence of the plurality of orders in combination with the order scores of the orders; based on order sorting and configured material configuration information and evaluation index information, performing simulation operation of material matching and index scoring on the plurality of orders to obtain a comprehensive index score of the plurality of orders; and by taking the optimal comprehensive index score as a target, optimizing the logical operation relationship in the order score tree by adopting a genetic programming algorithm until the comprehensive index score is optimal, and obtaining a target order sequence under the condition of the optimal comprehensive index score. According to the scheme, the optimal order sequence can be accurately and reliably determined.

The invention discloses an optimization method for a heat-engine plant thermal on-line process identification and control algorithm based on a dual-objective parallel ISLAND-HFC mixed model genetic programming algorithm. The steps comprise: 1, establishing a hardware platform; 2, the optimization method being completed by executing foreground interface software and background software by the hardware platform, the background software being formed by field test and data acquisition software, process identification software, and PID controller parameter optimization software. The method makes an ISLAND model and a HFC model organically combine together, anti-prematurity convergence property is good, multi-core CPU resources of an industrial control computer are fully used, multithreading run concurrently, evolution speed is fast, and the method is suitable to solve comprehensive problems. On one hand, dual-objective evolution controls errors between an evolution model and an ideal model, and on the other hand, the structure of an evolution individual is controlled, and finally, optimal individual structure and parameters satisfy requirements. For process identification, a field process model is accurately matched, and for parameter optimization of a PID controller, optimal proportion, integral, and differential parameters are obtained.

The invention discloses a genetic programming-based reconstruction method for axial nonuniform strain of the Bragg gratings and belongs to the field of nonuniform strain reconstruction. The method comprises the following steps: acquiring a structural response signal, randomly generating a Bragg grating axial nonuniform strain distribution expression, calculating a simulated reflection spectrum ofa Bragg grating, calculating a fitness function, optimizing the nonuniform strain distribution expression through the reproduction, intersection and variation operations of the genetic programming, and finally, repeating the last two steps till a preset maximum number of generations is reached. The method uses both a genetic programming algorithm and a modified T-array reflection spectrum formulation to reconstruct the grating axial nonuniform strain distribution expression, expresses a function expression in form of a binary tree without making any assumption concerning the grating axial strain distribution in any form in advance during the random generation of the strain distribution expression and optimizes any individual expression through the genetic manipulation of the binary tree. The method can accelerate convergence and improve calculation efficiency.

A cultivated land parcel classification method based on a genetic programming algorithm is used for collecting cultivated land distribution information in a specified geographic region, and comprises the following steps: step A, collecting land parcel data in satellite image data covering the specified geographic region; step B, on the basis of the segmentation data obtained in the step A, sample selection is carried out, training data and test data are divided according to the proportion of 7: 3, feature data are extracted from the selected samples, and normalization processing is carried out to obtain a feature set; and step C, calculating the feature set data obtained in the step B to obtain an individual with the highest overall precision as a feature for finally classifying all the plots, thereby realizing classification and recognition of the cultivated land in the specified geographic region. The genetic programming algorithm-based cultivated land parcel classification method provided by the invention can provide high-precision cultivated land parcel spatial distribution data, thereby providing a convenient and effective technical means for regional scale cultivated land information monitoring.

The invention discloses an improved genetic programming algorithm optimization method for resource-constrained multi-item scheduling, which includes the following steps: step 1: initializing parameters, function sets and attribute sets; step 2: collecting item set data under different working conditions, and Decompose it into a training set and a test set; step 3: extract the item information of each working condition item set in the training set as the training input, extract the function set and attribute set as the coding basis, and train the population in the improved genetic programming algorithm; step 4: Judging whether the maximum number of working conditions in the training set is reached, if so, output the optimal solution set in the population, if not, return to step 3 after changing the item set; step 5: use the test set and training set to output the optimal solution set in step 4 The solution set is tested; the invention can solve resource-constrained multi-item scheduling problems under single / multiple objectives, improve the defect that traditional genetic programming is easy to fall into local optimum, and improve the search and training capabilities of genetic programming.

The invention provides an incremental data mining method based on a genetic programming algorithm. Incremental data mining can complete the incremental model learning task of increasing data samples from few to many, input data into the model, and the input layer performs linear mapping on the data, and the The result is transmitted to the middle layer; the middle layer performs nonlinear transformation and spatial enhancement mapping respectively, and outputs the result to the voting system; the voting system determines the category; the feedback system optimizes the network parameters. Multiple iterations complete the learning process. When processing incremental data, the coupling degree of old and new data samples is low, and old data samples are no longer considered when mining new data, which has a good inheritance.

Crowd behavior modeling and simulation are technologies having important applications in the fields of public place design and management and the like. The invention applies a genetic programming algorithm to crowd behavior modeling, and relates to the two fields of modeling simulation and intelligent calculation. The method is characterized in that a set of rules capable of reflecting passer-by walking objective rules is automatically extracted and serves as a simulation model, the authenticity of the simulation effect is enhanced, and development of knowledge discovery and other related subjects is promoted. Aiming at the problems that a crowd modeling problem contains a large number of implicit features, effective features are difficult to discriminate and reasonably utilize in the prior art, and manually designed features are seriously limited by human knowledge and experience, the invention provides a novel advanced feature automatic construction technology and a set of auxiliary feature selection technology so as to construct a series of high-performance features and improve the effectiveness of crowd behavior rules.

The invention discloses an urban open channel drainage system control method based on a self-correcting genetic algorithm. According to the method, firstly, modeling is conducted according to an urban open channel system, rainfall prediction amount in the next 24 hours of a city is introduced as system disturbance, and then control over an open channel water pump in the next 24 hours is optimized by means of a genetic programming algorithm to achieve the aims of tracking the target water level and reducing energy consumption. Specially, a state feedback and rolling optimization algorithm is introduced to overcome the defects that a traditional genetic algorithm is large in algorithm model error, poor in disturbance resistance and the like, and robustness and instantaneity of a control system are improved. The method has important scientific meaning and application value in building of an urban open channel system.

The invention discloses an urban rail train network topology optimization method belonging to the technical field of rail transit communication networks. The present invention realizes the optimal design process that simultaneously satisfies the reliability and real-time performance of the communication network under the economic constraints by means of the idea of ​​double-layer planning; the double-layer planning regards the planning of the physical topology of the train network as a discrete network planning , while the logical topology planning is regarded as a shortest time delay planning, and the two-level planning model is solved by using the genetic algorithm; the optimization method of the present invention can effectively find the balance point of cost, reliability and time. The present invention provides a solution result based on genetic genetic algorithm for a given example, indicating that the present invention can be applied to practical engineering problems, and the application of the present invention to the construction of rail transit train network can make the train network operate more economically and efficiently, effectively saving Construction costs, reducing unnecessary waste and redundancy.

The invention discloses an urban open channel drainage system control method based on a self-correcting genetic algorithm. According to the method, firstly, modeling is conducted according to an urban open channel system, rainfall prediction amount in the next 24 hours of a city is introduced as system disturbance, and then control over an open channel water pump in the next 24 hours is optimized by means of a genetic programming algorithm to achieve the aims of tracking the target water level and reducing energy consumption. Specially, a state feedback and rolling optimization algorithm is introduced to overcome the defects that a traditional genetic algorithm is large in algorithm model error, poor in disturbance resistance and the like, and robustness and instantaneity of a control system are improved. The method has important scientific meaning and application value in building of an urban open channel system.

The invention belongs to the technical field of image processing, and specifically discloses a graphic image recognition and matching method based on a genetic programming algorithm of a new coding method, the purpose of which is to obtain more suitable features in image matching through the new algorithm to improve image recognition and matching retrieval Accuracy. The steps include: (1) parameter setting, population initialization and selection of matching methods; (2) calculation of population fitness and genetic operations including crossover, mutation, self-crossover, and self-exchange operations; (3) optimizing individuals after genetic operations, Carry out local search; (4) further optimize the population, and judge whether the evolution is terminated; (5) decode the individual tree, get the way of extracting new features, and obtain new image features; (6) for new image features, according to the set Matching method, the output image matching model. The training model generated by the invention can effectively improve the accuracy of image recognition and matching.

The invention discloses a DFB laser frequency stabilization method based on current control, and the method comprises the steps: carrying out the wavelength-current modeling through employing the current tuning characteristics of a DFB laser and a genetic programming algorithm, so as to set the constant frequency DC work environment of the DFB laser; obtaining a saturated absorption spectrum signal based on the saturated absorption principle; obtaining an odd harmonic differential error signal containing frequency information through employing a phase sensitive detection principle, and carrying out the PID control parameter optimization of the error signal through employing the genetic algorithm. The method achieves the quick and accurate adaptive frequency locking control, gives consideration to the high precision of saturated absorption frequency stabilization, and can guarantee the long-time high precision and stability of the frequency of the laser.

The invention provides an indoor navigation landmark extraction method. The method comprises the following steps: firstly, acquiring landmark saliency attribute data from an indoor scene picture to obtain a landmark saliency attribute measurement value; standardizing the landmark saliency attribute measurement value to obtain final landmark saliency attribute data; meanwhile, obtaining a landmarksaliency value in a questionnaire survey mode; dividing the final landmark saliency attribute data and landmark saliency values into a training set, a verification set and a test set by adopting a five-fold cross validation method, and obtaining a final landmark saliency model by utilizing a genetic programming algorithm according to the training set, the verification set and the test set; and finally, using the final landmark saliency model for calculating the saliency value of each landmark, and extracting the landmark with the highest saliency value. The method has the advantages that the quantitative relation between the landmark significance attribute data and the landmark significance is accurately represented, the prediction accuracy of the most saliency landmark is improved, and the method has practicability.