299 results about "Error function" patented technology

A computer-implemented method for evaluating and optimizing a large language model (LLM) includes receiving input data for a large language model; transforming the input data into a standardized format compatible with a multi-metric evaluation framework; applying a set of evaluation metrics to the transformed data, wherein the set of metrics comprises a hallucination metric, a groundedness metric, a relevance metric, a recall metric, a precision metric, a consistency metric, and a Coherence metric; generating an error function that quantifies an overall performance of the LLM, wherein the error function takes as input parameters of the LLM and incorporates the set of evaluation metrics to produce an error score; minimizing the error function wherein a lower error score indicates better performance of the LLM; and outputting an improved LLM with improved parameters.

The invention discloses a settlement prediction system based on soft soil creep, which relates to the technical field of settlement prediction, and dynamically calculates stress increment, total stress increment and pore water pressure by collecting the loading rate and underground water level change of a foundation in real time, and further calculates vertical effective stress based on the intermediate variables. Therefore, an accurate void ratio calculation formula is constructed. The systematized process not only improves the precision of settlement prediction, but also enhances the adaptability of the model to the change of actual working conditions. Specifically, through collection and processing of real-time data, the method can quickly respond to changes of foundation conditions, and it is ensured that model parameters are always in the optimal state. Calculating the void ratio by using the initial model parameters and evaluating the difference between a predicted value and an actual observed value in combination with an error function, so as to realize iterative optimization of the initial model parameters and finally obtain a corrected empirical model. The robustness and reliability of the model are remarkably improved, and settlement prediction is more accurate and reliable.

The invention discloses a three-level inverter dual-tracking model prediction control method and an inverter, and the method comprises the steps: constructing an integral sliding mode observer, and obtaining a current prediction value and a lumped disturbance estimation value through the integral sliding mode observer; calculating a voltage error of the inverter, and substituting the voltage error into the PI controller to obtain a voltage tracking control item; based on the voltage tracking control term, a switching state that minimizes the cost function is selected to control the inverter. The integral sliding mode observer is configured to calculate a current prediction error between a current prediction value of a current control period and a system output current, introduce an integrator to track the current prediction error, and generate an error function based on an output result of the integrator and the current prediction error. And feeding back a current predicted value and a lumped disturbance estimated value of the next control period through an error function. According to the invention, the method can achieve the quick estimation of the operation state of the inverter and the robust compensation of disturbance, reduces the steady-state tracking error, and reduces the total harmonic distortion of the inverter.

The invention discloses an ambiguity determination method for satellite positioning, a terminal and a medium in the field of satellite navigation and positioning, and the method comprises the steps: obtaining observation data of a base station and a mobile station, eliminating errors through employing a double-difference technology, and carrying out the cycle slip correction, and obtaining the corrected observation data; preliminarily estimating the speed and position of the mobile station; based on the corrected observation data, extracting an optimal ambiguity subset through satellite-level and ambiguity-level screening; constructing a factor graph model by using the optimal ambiguity subset, taking the preliminarily estimated speed and position of the mobile station as initial values, minimizing an error function in a factor graph, and optimizing to obtain a floating solution of ambiguity; obtaining an integer ambiguity solution based on the floating point solution; and verifying the stability of an ambiguity resolving result through consistency, and finally obtaining an ambiguity fixed solution. According to the method provided by the invention, high precision, high efficiency and high stability of ambiguity calculation are realized, and the method is suitable for a high-precision satellite positioning system in a complex environment.

The invention provides an unmanned vehicle autonomous driving method and system based on multi-sensor data fusion, and relates to the technical field of data fusion, and the method comprises the steps: obtaining the heterogeneous multi-sensor data of an unmanned vehicle, constructing the graph representation of the heterogeneous multi-sensor data, extracting the features of the data of each sensor through a deep neural network, and obtaining the data of each sensor; respectively mapping the extracted features of the sensor data to a shared hidden space through a mapping function to obtain respective low-dimensional feature vectors in the shared hidden space; establishing a joint optimization target, including minimizing an error function of a constraint edge and minimizing a mapping error of heterogeneous multi-sensor data in a shared hidden space, and minimizing the joint optimization target by adopting an alternate optimization strategy to obtain an optimal fusion result; and utilizing a pre-trained prediction distribution model, taking the uncertainty distribution diagram of the optimal fusion result as input, and combining the current state and the navigation target of the unmanned vehicle to generate an optimal behavior decision and path planning, so as to realize autonomous driving of the unmanned vehicle.

The invention provides an external parameter registration and inter-frame loading method based on a laser radar and a camera, and the method comprises the steps: obtaining point cloud data collected by the laser radar and image data collected by the camera, and carrying out the fitting and optimization of the point cloud data, obtaining a feature point coordinate value of the calibration object in the laser radar coordinate system and a feature linear equation of the calibration object in the laser radar coordinate system; determining feature point coordinate values of a calibration object in the image data under a camera coordinate system; determining a characteristic linear equation of the calibration object in a camera coordinate system; determining an error function of intersection point matching constraint based on the feature point coordinate values under the laser radar coordinate system and the camera coordinate system; based on the characteristic linear equations under the laser radar coordinate system and the camera coordinate system, determining an error function of the reprojection constraint; and determining an external parameter registration result of the laser radar and the camera through an iterative nearest point registration algorithm based on the error function of the intersection point matching constraint and the error function of the reprojection constraint.

A method for performing a computing task includes: extracting one or more features from a user content; converting the features to a floating point query vector; quantizing the floating point query vector; obtaining a database vector including one or more floating point feature vectors; determining a compensation vector based on a data distribution of the floating point query vector; quantizing the floating point feature vectors; determining an error function based on a difference between data distributions of i) the quantized query vector compensated with the compensation vector, and ii) the floating point query vector; determining, based on the error function, values of the compensation vector corresponding to the quantized feature vectors; combining the quantized query vectors and the values of the compensation vector to obtain one or more compensated query vectors; and performing the computing task using the compensated query vectors and the quantized feature vectors to obtain an output.

The invention relates to the technical field of robot control, and discloses a mechanical arm trajectory planning control method and system based on a BAFARNN model. The method comprises the steps that a mechanical arm kinematics model is established, and a trajectory tracking problem is converted into a time-varying equation; designing a bounded adaptive function to activate a recurrent neural network model, defining an error function and constructing a dynamic equation; designing a piecewise adaptive coefficient function, and dynamically adjusting the gain according to an error norm and time; setting a Lissajous curve as an expected trajectory, and initializing a simulation environment; the joint speed is solved in real time through an ODE numerical method, and the mechanical arm is driven to move; actual motion data is collected and compared with an instruction, and closed-loop feedback control is triggered when the actual motion data exceed a threshold value. According to the method, rapid convergence is achieved through the piecewise adaptive coefficient function, the bounded activation function and the negative feedback mechanism are adopted to suppress noise, and high-precision and real-time trajectory tracking of the mechanical arm in the dynamic environment is achieved.

The invention provides a motor parameter online identification method based on an improved golden section method, and belongs to the technical field of motor parameter identification, and the method comprises the steps: S1, obtaining a residual function of a recursive least square method with a forgetting factor when parameter identification is not carried out on a motor model for the first time; s2, detecting whether the residual function exceeds a preset threshold value or not, and executing a step S3 when the residual function exceeds the preset threshold value; s3, taking the residual function as a target function, and adjusting the forgetting factor by using a golden section method to obtain an optimized forgetting factor; and S4, carrying out motor parameter online identification by using a recursive least square method with the optimized forgetting factor. The method has the advantages that the forgetting factor is dynamically adjusted through the golden section method, the residual function serves as the target function, the optimal solution of the forgetting factor under the corresponding working condition can be found when the system changes rapidly, the tracking speed is increased, the steady-state precision is further improved, and therefore the overall performance of parameter identification is improved.

A quality prediction and adaptive compensation method and apparatus for curved surface assembly are provided. The method includes: inputting a geometric error function and a thermal error function into a spatial error model, to obtain a machining error prediction model; superimposing obtained machining errors on a theoretical surface of an assembly surface, to obtain a predicted machining surface; calculating, according to an assembly median plane determined based on the assembly surface, shape errors and assembly gap errors, and predicting curved surface assembly quality of a part by using the shape errors and the assembly gap errors; calculating an adaptive compensation amount of each assembly plane of the assembly surface based on the shape errors, the assembly gap errors, and the machining errors, when the curved surface assembly quality does not meet a preset assembly quality requirement, and compensating the corresponding assembly plane by using the adaptive compensation amount.

The invention belongs to the technical field of flow instruments, and provides a dynamic compensation and error correction system of a high-precision flow instrument, which comprises a multi-dimensional data acquisition and preprocessing module, a module capable of being additionally provided with a sensor, a module for establishing a data stream with a timestamp and processing data, a module for fluid network modeling and priori knowledge construction, and a data processing module. A distributed collaborative optimization and parameter calibration module capable of constructing a directed graph model and generating initial compensation parameters; a network association reasoning and dynamic weight distribution module capable of adding a global penalty term to optimize parameters on the basis of a traditional error function; a working condition adaptation calibration and cross-domain parameter migration module capable of constructing an association graph, detecting abnormity and distributing weights; real-object-free calibration and parameter migration can be realized; according to the system, through multi-module cooperation, the problem that traditional single-table compensation is local and is not global is solved, and high precision and real-time performance of industrial-grade cooperative metering are supported.

The invention relates to the technical field of power grid data prediction, and discloses an energy storage scheduling decision-oriented source load matching rate prediction method, system and device, a medium and a product, and the method comprises the steps: constructing an initial source load matching rate prediction model based on a deep learning algorithm, and forming a mixed loss function through a prediction error function and a decision evaluation function; and parameter optimization is carried out on the initial source load matching rate prediction model based on a mixed gradient descent learning method to obtain a source load matching rate prediction model, so that the source load matching rate is accurately predicted, and the decision precision under the condition of predicting the source load matching rate is improved.

The invention discloses a fractal mask bilinear interpolation-based checkerboard angular point positioning method, which comprises the following steps of S1, acquiring a checkerboard image by using image acquisition equipment, and extracting candidate angular points in the checkerboard image; s2, determining a direction angle of the edge of the local area by using an edge detection algorithm and a polar coordinate transformation and clustering algorithm; s3, based on the direction angle of the edge of the local area, using a gray integral method and a least square method to preliminarily position the coordinates of candidate angular points; s4, in the neighborhood of the preliminarily positioned candidate angular point coordinates, obtaining a sub-pixel-level gray value by using fractal mask bilinear interpolation, and calculating local gradient distribution; s5, dynamically generating a weight mask based on a fractal theory, constructing a central point minimum error function according to a corner local gradient consistency principle, iteratively optimizing candidate corner coordinates, and outputting a sub-pixel level position; according to the method, the sub-pixel-level accurate positioning of the checkerboard angular points is realized by fusing the edge detection, the gray integration and the iterative optimization of the fractal theory.

The present invention belongs to the field of lifting equipment control technology, and specifically relates to a dynamic modeling method and control method for double-bridge lifting equipment. The dynamic equations of a five-degree-of-freedom double-bridge crane are constructed using the Lagrangian method, and the geometric constraint relationships between trolley I and trolley II in the direction parallel to the bridge and the vertical direction are constructed. The dynamic model of the double-bridge crane is constructed by combining the dynamic equations and the geometric constraint relationships, and the dynamic model is transformed into a general model. Based on the general model, a first error function is constructed based on the difference between the actual position and the expected position, and a virtual controller is constructed based on the first error function. A second error function is constructed based on the virtual controller, and an actual controller is constructed based on the first error function, the virtual controller, and the second error function. This technical solution models the system as a whole through the constraint relationship between the two trolleys, taking the mass of the hook into consideration during the modeling process, and solving the singularity problem.

The invention discloses a self-adaptive cooperative control strategy based on a variable speed reaching law and a neural network, belongs to the field of intelligent control, and mainly aims at estimating and compensating uncertainty, external disturbance and actuator saturation existing in a queue in queue control and improving the stability, robustness and control performance of the whole queue. The method comprises the following steps: establishing a queue system model with uncertainty, external disturbance and actuator saturation; establishing a Gaussian error function approximation strategy; establishing a tracking error; establishing a variable speed reaching law; establishing a queue integral sliding mode error; establishing a queue coupling sliding mode error; establishing an adaptive parameter estimation mechanism for external disturbance and Gaussian function approximation errors in the queue; establishing a neural network estimation mechanism for uncertainty in the queue; and establishing a queue integral sliding mode control strategy based on a variable speed reaching law, a Gaussian error function approximation strategy, an adaptive parameter estimation mechanism and a neural network estimation mechanism. The method is used for queue cooperative intelligent control.

A system and method include generating synthetic data by generating a first set of hyperparameters for a first trained machine learning model and a second set of hyperparameters for a second trained machine learning model, generating a plurality of synthetic data vectors using the first and second trained machine learning models, computing an error function for the first and second set of hyperparameters using a third machine learning model, computing an objective function value, responsive to determining that the objective function value is not an optimal value, updating the first set of hyperparameters and the second set of hyperparameters or responsive to determining that the objective function value is an optimal value outputting the plurality of synthetic data vectors as a set of synthetic data.

The invention relates to a seismic phase pickup method based on time-frequency analysis and deep learning, and belongs to the field of seismology, a deep neural network comprising an empirical mode decomposition module, a residual shrinkage module, an LSTM layer, a transformer layer, an up-sampling layer and a convolutional layer is established, a linear rectification function is adopted as an activation function, an Adam adaptive optimization function is adopted for back propagation, and a seismic phase is obtained. And taking a binary cross entropy loss function as a loss function, and taking a mean absolute error function, an error standard deviation, an accuracy rate, a recall rate and an F1 score as evaluation indexes to obtain a deep neural network model. When the frequencies of the noise, the P wave and the S wave are different, empirical mode decomposition can effectively separate the noise, the P wave and the S wave, so that the network is helped to better identify the P wave and the S wave. Even for broadband noise, the noise can be decomposed into different intrinsic mode functions, so that the noise content in the intrinsic mode functions where P waves and S waves are located is reduced. The time-frequency analysis method can effectively improve the anti-noise capability of the network.

The invention discloses a three-dimensional structure modeling method based on zero sample semantic segmentation and physical attribute estimation, which belongs to the technical field of image data processing, and comprises the following steps: constructing an image set D1; generating an initial three-dimensional point cloud model M0 based on D1; the M0 is divided into S sub-point clouds; generating a physical attribute graph of the s sub-point cloud Ms under each visible view angle, and constructing a multi-view-angle physical attribute consistency error function Ls; adjusting the Ms by minimizing the Ls to obtain an optimized sub-model Ms' and an optimized attribute atlas Cs; and constructing an optimized point cloud set G by using the sum Cs of all the sub-point clouds, and training a three-dimensional conditional diffusion model for three-dimensional reconstruction. According to the method, fine modeling of a target structure and details can be realized, the reconstruction efficiency and authenticity are improved, controllable material and style reconstruction is supported, and the method has the advantages of weakening supervision dependence and enhancing interpretability and adjustability of the model. The method is suitable for the fields of virtual reality, industrial digital modeling, element universe scene generation and the like.

The invention relates to the technical field of deformation detection, in particular to a method for measuring the deformation of a rotary body part, which comprises the following steps of: arranging measuring points on a target rotary body part, and performing dotting measurement by using a three-coordinate measuring instrument to obtain dotting data which are three-dimensional coordinates; based on the dotting data, calculation is carried out, and deformation information of the target rotary body part is obtained, the calculation comprises the steps of determining a data source, normalizing coordinates, converting a coordinate form, sequencing path measuring points, compensating a measuring head, calculating a position-error function, calculating all-directional axial deformation and radial deformation, speculating all-position deformation, and performing any combination calculation in multi-coordinate system data integration. According to the invention, the deformation of the rotary part can be measured conveniently, quickly and efficiently.

The invention discloses a calibration method for sampling time offset mismatch of TI-ADC for SerDes and a storage medium. The calibration method comprises the following steps: performing operation on a current output value obtained by quantization of each sub-ADC and a previous beat of data recovered by judgment of SerDes; taking a certain sub-ADC channel as a reference to obtain an error function of offset of sampling time of other sub-ADCs relative to the reference time; and the sampling time of the sub-ADCs is adjusted by adopting an LMS algorithm to minimize an error function, so that the calibration of the sampling time offset mismatch is completed. The storage medium is used for storing a computer program for executing the method. The method has the advantages of simple principle, simplicity and convenience in operation, wide application range, good calibration effect, high precision and the like.

The invention discloses an intelligent flood forecasting method and system fusing flood process information physical characteristics, and solves the problems of poor robustness, strong dependence on data quality, weak interpretability and the like of deep learning in flood forecasting modeling. A new data processing mode and an error function are designed by analyzing the main body evolution rule and noise of the flood process, and the quality of model input data is improved; the method comprises the following steps: designing a two-dimensional hidden layer according to hydrometeorological element characteristics, constructing an ensemble attention mechanism screening architecture, and proposing a flood recycling network (FRN) forecasting architecture fused with flood process physical characteristics; an intelligent FRN flood forecasting model fusing flood physical characteristics is constructed, and FRN model visualization application software integrating interpretability and a long forecasting period is developed. The method can improve the accuracy and interpretability of forecasting, and promotes the innovative application of an artificial intelligence method in the field of water science.

The invention relates to a method for eliminating a double-planetary-calendar forecast error based on post-event precise ephemeris. The method comprises the following steps: forecasting a satellite position of the day at an ephemeris reference moment based on double planetary calendars and an SDP4 model; calculating the satellite position of the ephemeris reference moment on the day based on the post-event precise ephemeris of the double planetary calendars on the ephemeris reference moment on the day, and taking the satellite position as a reference satellite position; acquiring a satellite position forecast error by comparing the difference between the forecast satellite position and the reference satellite position of the day at the ephemeris reference moment, and fitting the satellite position forecast error of the satellite in one orbit period to obtain an error function of the double-planetary-calendar forecast satellite position; and performing error elimination on the forecast satellite position of the target date according to the error function to obtain the forecast satellite position after error elimination. By adopting the method, the satellite position forecasting error of the double planetary calendars can be eliminated, and the precision of the double planetary calendars for medium and long term forecasting is improved.

The invention discloses a doubly-fed fan parameter identification method based on an improved sparrow algorithm, and the method comprises the steps: analyzing the sensitivity of doubly-fed fan parameters, selecting key control parameters, building a doubly-fed fan low-voltage ride-through transient model in a PSCAD (Power System Computer Aided Design), obtaining a plurality of groups of output data through adjusting the key control parameters, and carrying out the identification of the doubly-fed fan parameters. Constructing a multi-target error function of the original model and the identification model; the improved sparrow algorithm is adopted to optimize and solve the established multi-target error function based on the key control parameters and the output data, the identification result of the to-be-identified control parameters of the doubly-fed fan is obtained, an identification model obtained through identification and an original model are compared and verified, the key control parameters in the transient model of the doubly-fed fan can be accurately and reliably identified, and the identification accuracy of the to-be-identified control parameters of the doubly-fed fan is improved. The problems that a traditional doubly-fed fan parameter identification method is low in identification precision, poor in adaptability to extreme working conditions and low in consistency of multiple identification results are effectively solved.

The invention belongs to the technical field of robot safety control, and discloses a safe motion planning method for a dual-redundancy mechanical arm. And a linear mapping relation between the joint space speed of the redundant mechanical arm and the speed of an end effector is established. And establishing an end effector speed constraint, a safety distance constraint and a joint amplitude limiting constraint, and fusing a differential kinematics model and the three constraints to establish a multi-target combined constraint model. The optimal decision variable meets the Carlo demand-Kuhn-Tuck condition, the original expression form of the Carlo demand-Kuhn-Tuck condition is defined, and a nonlinear complementary function is introduced to reconstruct the Carlo demand-Kuhn-Tuck condition. Based on an improved Carlo demand-Kuhn-Tuck condition, a vector error function is defined, and an adaptive return-to-zero neural network is designed; and performing iterative solution on the vector error function through an adaptive return-to-zero neural network to obtain an optimal decision variable of the multi-target joint constraint model, taking the optimal decision variable as a mechanical arm joint motion control instruction, and outputting the mechanical arm joint motion control instruction to complete mechanical arm motion planning.

The invention discloses an improved UWB error compensation positioning method based on a VEGA + algorithm, and relates to the technical field of indoor accurate positioning, and the method mainly comprises the steps: carrying out the preprocessing of original collection data, obtaining a preliminary sample set, carrying out the secondary sampling through a VEGAS + algorithm, obtaining a complete sample set, and carrying out the calculation of the complete sample set; and constructing a space error function by using a sparse polynomial to obtain an error model, and performing error prediction and compensation on a UWB real-time positioning result to obtain a corrected position signal. By implementing the improved UWB error compensation positioning method based on the Vegas + algorithm provided by the invention, delay and positioning errors can be reduced, and the long-term stability and environmental adaptability of the system can be improved.

The invention provides a low earth orbit satellite network slice resource allocation method and system based on reinforcement learning and a storage medium, and the method comprises the steps: taking a current observation state, a current hypothesis function and a current error function corresponding to each slice in a low earth orbit satellite network as input, and executing a decision operation through a reinforcement learning algorithm, in each round of decision-making operation, the satellite resource allocation amount, the safety margin and the service level prediction value corresponding to each slice are used as output and are sent to the corresponding slice, so that each slice can be based on a pre-stored slice-satellite corresponding relation; and allocating resources to the corresponding low-orbit satellites according to the satellite resource allocation quantity output by each round of decision operation. According to the invention, efficient resource utilization rate and stable service quality guarantee can be satisfied at the same time.

The invention provides an irrigation area gate overflow coefficient and hydrodynamic parameter calibration method and system, and the method comprises the steps: constructing a one-dimensional grid model, embedding the design and mapping parameters of a channel section into the one-dimensional grid model, and constructing a hydrodynamic data matrix; calibrating an overflow coefficient reference value of the gate by using water conservancy process data and operation conditions actually measured at the gate; deducing a water level flow calculation field; uncertain undetermined parameters are used as background values, an error function is constructed based on the relative deviation degree of the background values, and a data assimilation optimization algorithm model is established; a data assimilation optimization algorithm model is used for obtaining hydrodynamic parameters of an irrigation area canal system and an overflow coefficient reference value of a gate; in the next period, the steps are repeated, the calibration result is updated, automatic calibration of the overflow coefficient and the hydrodynamic parameter of the gate of the irrigation area can be achieved, and the method is suitable for water resource allocation and water volume deduction calculation of a digitized and informationized irrigation area canal system of the irrigation area, and assists in analysis and optimization of an irrigation area scheduling scheme.

The invention provides an APU characteristic correction method and device, and belongs to the field of aviation electromechanics. The method provided by the invention comprises the following steps: establishing an APU system simulation model; determining a plurality of correction parameters and correction targets; calculating a calculation value corresponding to the correction target based on the APU system simulation model and the correction parameter; constructing a correction error function based on the calculated value of the correction target and a test value of the correction target obtained through a test; determining an updated component in the APU, and determining an optimization strategy based on the identity of the updated component; based on the optimization strategy, executing a particle swarm algorithm to obtain an optimal correction parameter corresponding to the correction error function; and updating the APU system simulation model based on the optimal correction parameter to obtain a corrected APU characteristic model. According to the APU characteristic correction method and device provided by the invention, the accuracy of the model correction result can be improved after the local parts of the APU are updated.

The invention provides an SLAM (Simultaneous Localization and Mapping) optimization method and equipment based on point cloud registration and adaptive resolution, and a medium, and relates to the technical field of AGV (Automatic Guided Vehicle) positioning and mapping, and the method comprises the following steps: (1) inputting and preprocessing sensor data; (2) front-end scanning and matching: based on the preprocessed data, carrying out point cloud registration through a GICP algorithm, updating the pose of the current frame and constructing a local subgraph; (3) global SLAM: receiving sensor data, poses and subgraph data from the front-end scanning and matching step, and carrying out loopback detection and global optimization; wherein a self-adaptive resolution switching strategy based on environment complexity is adopted in the loopback detection process; according to the GICP algorithm, the surface of the point cloud is modeled as Gaussian distribution, local features are described through a covariance matrix, and a target error function constructed based on the mahalanobis distance is optimized to estimate pose transformation. According to the method, the local matching precision and robustness are improved by replacing ICP with GICP, and the global calculation efficiency is adjusted and optimized in combination with the dynamic resolution.

The invention discloses a multi-joint torque control method of a mechanical arm for cleaning foreign matters on a power distribution line. The multi-joint torque control method comprises the following steps: S1, setting an initial value and an expected track of a joint in the mechanical arm; s2, a mechanical arm multi-joint torque control model is constructed based on a fuzzy neural network; s3, performing initialization processing on the fuzzy rule, and determining an initial value of a fuzzy neural network parameter; s4, constructing an error function to calculate an error, and performing adjustment and iterative correction on each parameter; s5, the error and the error change rate are calculated to serve as input of the fuzzy neural network, and the control torque parameter of the mechanical arm serves as output; s6, performing iterative optimization processing on the output parameters of the fuzzy neural network based on a particle swarm and grey wolf hybrid optimization algorithm to obtain a global optimal solution of the output parameters of the fuzzy neural network, and obtaining the joint torque of the mechanical arm; according to the method, the joint output torque can be accurately controlled, the joint position tracking error is reduced, and the overall robustness and stability of a mechanical arm control system are improved.