40results about How to "Improve convergence accuracy" patented technology

The invention discloses a centralized phased array beam control method and system based on an FPGA, and relates to the technical field of phased array beam control, and the method comprises the steps: carrying out the real-time iterative operation through an incremental adaptive correction kernel in a centralized control FPGA, and obtaining an amplitude-phase correction increment and a corresponding target effective frame number; performing increment differential coding on the amplitude-phase correction increment and the corresponding target effective frame number in a centralized control FPGA (Field Programmable Gate Array) to generate an increment differential beam control message; sending the increment difference beam control message to an array element FPGA (Field Programmable Gate Array) at the front end of an array, executing frame-level atom switching according to a target effective frame number, and fusing the amplitude-phase correction increment and a locally stored old beam control weight in the switching process to obtain a final array element beam control weight; generating a real-time beam control directional diagram based on the final array element beam control weight; the stability and convergence precision of beam control are improved, and the bandwidth requirement of a control link is reduced.

The application provides a UWB and PDR fusion positioning method and system in an indoor shielding environment and a medium, and belongs to the technical field of indoor positioning and multi-source information fusion. The method comprises the following steps: calculating a UWB positioning track and a PDR positioning track. Taking translation, rotation, step scaling and scale change as combined parameters, a track alignment error function is constructed by using the Euclidean distance of the two tracks at each timestamp. An improved vole optimization algorithm is used to solve the optimal combined parameters. An initial population with uniform distribution is generated by mapping a Sobol difference sequence. In the iteration, the UWB track is used as the reference, the PDR track is temporarily transformed by using the combined parameters, the Euclidean distance is calculated, and the vole with the smallest distance is selected to breed the next generation. After reaching the maximum number of iterations, the PDR track is corrected by using the optimal combined parameters at one time, and the UWB track is fused and output. The application does not require a priori model, can simultaneously suppress UWB non-line-of-sight drift and PDR cumulative error, and has high positioning accuracy.

This invention discloses an energy-saving method and system based on data sensing, belonging to the field of energy-saving optimization technology. The method includes: using a sensor array to collect raw multi-source sensor data, performing data preprocessing and feature extraction to generate room occupancy status and a state vector; based on the room occupancy status, inputting the state vector into a macroscopic decision-making agent constructed based on a reinforcement learning algorithm to generate an optimal dynamic weight coefficient vector; based on the optimal dynamic weight coefficient vector, using an improved Hippo optimization algorithm to solve a multi-objective optimization function to generate an optimal equipment control parameter vector; executing the optimal equipment control parameter vector, returning to the sensor data acquisition step, calculating the immediate reward value after the decision, and updating the macroscopic decision-making agent using the corresponding state transition data. This invention solves the problems of poor flexibility, low optimization accuracy, and difficulty in balancing multiple objectives in existing energy-saving control strategies.

This invention relates to the field of multi-agent cooperative control technology, and in particular to a switching multi-agent cooperative tracking control method based on deep reinforcement learning. The method includes building a two-layer independent architecture for each independent agent, constructing a multi-level grouping architecture for the overall multi-agent cluster, proactively adjusting the internal structure of the multi-level grouping architecture based on prediction results, and using causal analysis technology to trace and determine the root causes of sudden changes and select an execution strategy. This invention adopts a two-layer independent architecture scheme, configuring a leader layer and an executor layer for each agent. Through an asynchronous decision-making mechanism with unidirectional instruction transmission, it achieves complete decoupling between global planning and local execution, ensuring the global optimality of the cooperative tracking task, meeting the real-time control requirements in dynamic environments, eliminating the single-point failure risk of centralized decision-making, and improving the cluster decision-making efficiency and operational stability.

The invention belongs to the field of micro-grid cluster system coordination control, and particularly relates to a wind and light storage multi-source micro-grid cluster coordination control method based on a sparrow search algorithm. Aiming at the defect that the convergence speed of the existing sparrow search algorithm is difficult to meet the real-time requirement of micro-grid coordination control, the invention adopts the following technical scheme: the wind-light storage multi-source micro-grid cluster coordination control method based on the sparrow search algorithm comprises the following steps of: formulating respective control models according to different controllers in a wind-light storage multi-source micro-grid; setting PI parameters of each controller according to the control model and the system response; an improved sparrow search algorithm is adopted to update PI parameters of each controller of the micro-grid, including mapping a vector to a symmetric positive definite manifold through a function, and calculating an optimized update step length through Riemannian metric; updating the position according to the optimized updating step length; and judging whether a convergence condition is met or the maximum number of iterations is obtained. The method has the beneficial effects that the convergence speed and precision are improved.

The application relates to a video instance segmentation method based on a dynamic convolution decomposition of a lightweight attention mechanism; the method first inputs a video frame, and a backbone network independently extracts a feature map of each frame in the video; then the feature map extracted by the backbone network enters an HQT encoding and decoding module, the position change of an instance in each frame is accurately positioned by combining an output head, and three prediction branches are used to supervise model training; in the application, a convolution layer of an encoding network adopts dynamic convolution decomposition, a more compact model is obtained, model training is easier, the required parameter quantity is greatly reduced, the training speed is improved, and the training time is shortened; a lightweight HQT encoding and decoding module is provided, which helps to reduce model parameters and improve efficiency; the loss function is improved, model training is more stable, the convergence speed and convergence precision are improved, and the problems of imbalance between foreground and background in samples and imbalance between foreground categories under a long tail condition are relieved.

The invention discloses a mining subsidence prediction parameter solving method based on an improved center collision optimization algorithm, and belongs to the field of mine deformation monitoring data processing. Aiming at the defects that a basic center collision optimization algorithm is prone to premature convergence and insufficient in optimization precision when solving a mining subsidence parameter inversion problem, a Huber loss function is adopted to construct a fitness evaluation model to enhance the robust capability of the algorithm, and Cubic mapping is utilized to generate a random number to improve the stability of a search process, so that the method is suitable for the mining subsidence parameter inversion problem. And a self-adaptive elite-guided Cauchy variation mechanism is introduced to enhance the global exploration capability. According to the method, firstly, a probability integral method parameter inversion problem is constructed into an optimization model, and a population is initialized; performing double-space collaborative search in an original space and a decorrelation space constructed based on principal component analysis, and iteratively updating a population in combination with a dynamic space allocation strategy; and finally outputting an optimal parameter solution. According to the method, the convergence precision, stability and robustness of the algorithm in complex nonlinear parameter inversion are effectively improved.

This invention discloses a hierarchical federated learning method and system based on split meta-learning. The method specifically includes: a cloud server initializing a global model and distributing it to various edge servers; the edge server layer distributing the global model to its respective clients; each client updating the feature extraction module of the global model locally based on local data, while simultaneously freezing the classifier module of the global model; the edge server layer aggregating the updated parameters of the feature extraction modules from multiple clients and performing meta-optimization on the classifier module using a local validation dataset to generate an edge local model; and the cloud server layer periodically collecting model parameters from multiple edge servers and updating the global model using a gradient-sensitive momentum aggregation strategy. This invention, through a three-layer "vehicle-edge-cloud" architecture and a two-level collaborative optimization mechanism, achieves coordinated optimization between privacy protection, communication efficiency, and model adaptability in a vehicle-to-everything (V2X) environment.

The application discloses a kind of optimal arrangement method of non-uniform linear array for direction finding, comprising: based on preset array optimal arrangement model, the position, speed and initial local optimal position of initial quantum particle are generated;Based on preset array optimal arrangement model, the first fitness function based on minimum interval criterion and minimum maximum relative sidelobe level is constructed;Based on the first fitness function, the initial global optimal position of quantum particle is obtained;Based on initial local optimal position and initial global optimal position, the speed and position of quantum particle are updated, and the global optimal position is obtained;Based on global optimal position, the optimal array arrangement result is obtained.The application designs an optimal special array arrangement method based on minimum interval criterion and minimum maximum relative sidelobe level, finds optimal array arrangement mode using discrete quantum particle swarm, and realizes high-precision direction finding of optimal array arrangement under specific conditions and requirements.

This invention relates to the field of mobile communication radio frequency signal processing technology, and particularly to a self-oscillation cancellation method and device based on LSTM. The method involves acquiring a mixed signal from the repeater receiver and a self-oscillation interference reference signal from the repeater coupling end, preprocessing both signals to obtain a preprocessed mixed signal and a self-oscillation interference reference signal. These two signals are then input into a pre-set LSTM model, enabling the model to simultaneously learn the correlation between the interference characteristics in the mixed signal and the pure reference signal, thereby generating a cancellation signal whose amplitude, phase, and timing match the actual self-oscillation interference. After amplitude and phase calibration, this cancellation signal is subtracted from the preprocessed mixed signal in the time domain. The residual interference in the resulting clean, useful communication signal is monitored in real time, and the LSTM model is updated online accordingly.

The invention belongs to the technical field of electric power systems, and particularly relates to a quick-closing valve control method for transient stability of an electric power system based on deep reinforcement learning. Comprising the following steps: establishing a Markov decision model based on power system transient stability emergency control; constructing a quick-closing valve decision framework based on a near-end strategy optimization algorithm; constructing a deep reinforcement learning model for emergency quick-closing valve control; by introducing the deep reinforcement learning method, a more accurate and efficient quick-closing valve control strategy can be formulated, so that when the power system breaks down, the stability of the system is recovered more quickly, and the loss caused by instability is reduced.

The application belongs to the technical field of computer vision, and discloses a three-dimensional point cloud registration method based on point-line feature fusion. The three-dimensional point cloud registration method firstly uses SAC-IA to complete static registration of three-dimensional point clouds on the surface of a target object, and establishes an initial motion transformation relationship from the three-dimensional point clouds of the 3D model of the target object to the three-dimensional point clouds on the surface of the target object. Then, in the dynamic tracking stage, point features extracted by ORB are matched with line segment features extracted by LSD-LBD, coarse registration is realized by combining an EPnPL algorithm, and the robustness of pose estimation is improved by eliminating false matching through RANSAC. Finally, in the fine registration stage, point-to-plane ICP is used to further optimize pose estimation. The method does not need to rely on artificial marking, and can complete stable three-dimensional point cloud registration based on natural features, and can be widely applied to complex scenes such as mechanical product maintenance and industrial assembly.

The present application relates to a distributed photovoltaic MPPT method, system, device and medium, and relates to the technical field of distributed photovoltaic power generation, which comprises the following steps: preprocessing photovoltaic array data, obtaining irradiance, temperature and MPP voltage dataset; determining the optimal ElmanNN hidden layer node number by K-fold cross-validation and RMSE evaluation, and dividing the dataset into a training set, a validation set and a test set; inputting the training set into global search of the firefly algorithm and local optimization of the ant colony optimization algorithm, defining the stable solution as the FA-EAS optimal parameters after convergence; updating the ElmanNN weight threshold based on the optimal parameters, deploying the training model to perform MPP voltage online prediction; inputting the difference between the predicted voltage and the reference voltage into a PID controller, calculating the control amount through the incremental PID, generating a PWM signal to control the duty cycle of the Boost converter, and realizing maximum power point tracking. The present application can improve the tracking speed, stability and success rate of MPPT.

An LMS-FNNCMA algorithm for MIMO equalization belongs to the field of optical fiber communication technology. Switches A1 and A2 are turned on, and the training input signal is passed through a transverse filter. The weight coefficients are then adjusted using the LMS algorithm to continuously approximate the known desired response d(n). After the learning process is complete, the transverse filter reaches its optimal design. Its weight coefficients are then fixed, and switches B1 and B2 are turned on to filter the working input signal. Then, based on the cost function method, a suitable nonlinear activation function is selected through the neural network coefficients continuously adjusted by the blind equalization algorithm to perform a second filtering operation on the working input signal y(n) after the LMS algorithm filtering, finally obtaining the decision output after passing through the decision unit.

This invention discloses an improved near-infrared spectral wavelength selection and modeling method for artificial bee colony algorithms, comprising the following steps: data preparation, chaotic initialization, fitness evaluation, establishing a PLS model based on wavelength subsets, using the regularized fitness function of RMSECV as the fitness Fitnessxi of nectar source xi; searching for foraging bees; searching for observation bees; searching for scout bees; outputting the optimal wavelength subset Sbest and establishing the final PLS model. The beneficial effects of this invention are: the method effectively improves convergence speed and accuracy; chaotic initialization improves the quality of the initial population, avoiding the algorithm from prematurely falling into local extrema; the adaptive perturbation factor balances the algorithm's global search and local exploitation, improving the ability to find the optimal wavelength combination in high-dimensional discrete space, and the selected wavelength subset has stronger representativeness, resulting in a final PLS model with lower RMSEP (root mean square error of prediction) and better robustness.

This invention provides a digital twin method for coal mine mining based on multi-level dynamic error correction and AI optimization. The method includes: monitoring data acquisition and dynamic preprocessing, denoising and spatial consistency detection of the data; construction and preprocessing of the coal mine digital twin; generation of a nonlinear deviation field using highly nonlinear variable interpolation and model mapping methods; Kalman filtering error analysis and correction in continuous analysis steps, combined with spatial mutation step marking to optimize model states; staged sliding window long-term fluctuation analysis, classifying and marking regions based on the absolute value of the error and fluctuations; multi-level weighted error optimization based on genetic algorithms to adjust material parameters and boundary conditions; material constitutive error correction based on dynamic Bayesian methods to gradually enhance the model's adaptability and accuracy; and construction of a deep data-driven hybrid network structure AI optimization and prediction model to capture the dynamic changes in material constitutive features and differences in model constitutive characteristics, achieving accurate digital twin simulation of the entire coal mine mining process.

This application relates to the field of carpooling demand prediction technology, and particularly to a method, device, electronic device, and storage medium for predicting carpooling demand. The method includes: constructing a feature tensor set based on historical order data, meteorological parameters, and regional static feature data; transforming the target low-dimensional statistical features into target high-dimensional semantic features; and mapping the target high-dimensional semantic features to a target orthogonal subspace to generate a feature set that eliminates redundant temporal correlations. The feature set is then used to optimize the hyperparameters of a pre-constructed ensemble learning gradient boosting tree model until an iteration stopping condition is met, thus constructing a carpooling demand prediction model. This model outputs carpooling demand. This solves the problem that related technologies fail to couple the temporal and spatial dependencies of passenger travel demand, and that the prediction models are sensitive to hyperparameters, making it difficult to adapt to unconventional scenarios and output accurate carpooling demand.

This invention relates to the field of array antenna synthesis, and more particularly to a method for synthesizing sparsely distributed planar arrays. To balance the development and exploration processes, this invention employs nonlinear function reconstruction for the arithmetic optimization accelerator in the arithmetic optimization algorithm; it uses the top three best individuals to replace the current best individual for exploration and development and introduces an elite mutation strategy to enhance the algorithm's ability to escape local optima and improve its convergence accuracy; it proposes an adaptive matrix mapping rule to judge the current array element distribution, and if it does not meet the minimum element spacing constraint, it adjusts it through an adjustment strategy, avoiding infeasible solutions while ensuring the degrees of freedom of the array elements.

The application relates to the technical field of big data processing and data quality control, in particular to a data cleaning system based on big data, which comprises a data preprocessing center used for collecting multi-source heterogeneous data, generating dimensionless statistical distribution values, and constructing a weighted directed graph; a state mapping unit used for generating a state space vector; an energy verification unit used for comparing a global consistent energy with a preset safety threshold; if the global consistent energy is greater than the preset safety threshold, a logic conflict signal is generated; if the global consistent energy is less than or equal to the preset safety threshold, a logic consistency signal is generated; a gradient repair unit used for generating a dimensionless repair value; and a feedback control unit used for generating a final cleaning result; the multi-source heterogeneous data is set as the final cleaning result; the application solves the problem that the prior art lacks an automatic correction means for system constraints, and improves the convergence speed and precision of repair.

The application belongs to the technical field of nonlinear system control, and discloses a method and system for introducing composite learning into nonlinear system output feedback adaptive control. The application directly reuses the existing K-filter in standard output feedback backstepping control to construct an extended prediction error, and does not need to establish an additional observer or a state estimation model (such as a series-parallel estimation model, a fuzzy observer, etc.) in parallel with the K-filter. This functional reuse design eliminates the calculation overhead of the additional dynamic system, reduces the storage requirement and real-time calculation burden of the controller, and significantly reduces the hardware implementation cost and system debugging complexity. Meanwhile, through the composite driving of the tracking error and the extended prediction error, the cumulative information is continuously introduced by using the historical memory of the regression quantity, so that the adaptive and self-learning ability of the control system is significantly enhanced, and the parameter estimation convergence and the system response performance are improved.

The application relates to the technical field of multi-robot system, in particular to a multi-robot task allocation optimization method based on a DS-GRCA algorithm, which comprises the following steps: obtaining task data to be allocated and state data of a plurality of robots; based on the task data and the state data, applying an upper-layer global optimization algorithm to group the plurality of robots, to generate a robot grouping scheme; for each robot subgroup in the robot grouping scheme, applying a lower-layer local game algorithm to perform intra-group fine task allocation, and allocating specific tasks to robots in each robot subgroup; integrating allocation results of all robot subgroups, and outputting a final global task allocation scheme. By adopting a hierarchical architecture of upper-layer global optimization + lower-layer local game, the application realizes effective balance between globality and calculation efficiency, so that the application can obtain a high-quality global solution while significantly improving the solution speed.

This invention belongs to the technical field of bearing fault diagnosis. To address the inaccuracy of existing bearing fault diagnosis methods, this invention proposes a fault diagnosis method and system for press bearings based on conditional generative adversarial networks (GANs). A generator is trained using operating condition vectors and Gaussian noise. A discriminator is trained using the health state features generated by the generator and the actual health state features. The generator and discriminator are then subjected to adversarial training to prevent the discriminator from distinguishing between the health state features generated by the generator and the actual health state features, as well as from distinguishing the matching between the health state features generated by the generator and the corresponding operating condition vectors. This results in a well-trained health state model. The health state model is then used to generate a dynamic health baseline for the press shaft to be diagnosed, thereby obtaining the fault diagnosis result for the press shaft and achieving high-precision fault diagnosis and early warning.

This invention discloses a joint spatiotemporal decorrelation suppression method for TomoSAR. First, multi-temporal SLC images are registered and de-skewed to generate an initial complex covariance matrix. Then, homogeneous pixels are identified based on likelihood ratio, and a high-precision complex covariance matrix is ​​obtained through weighted summation. A joint model is constructed by fusing interferometric observations and phase priors, linearized using Taylor expansion, and the phase of the SLC complex image is inverted using iterative weighted least squares. Finally, the decorrelation-suppressed SLC complex image is output, and 3D tomographic reconstruction is completed by combining compressed sensing. This method effectively suppresses spatiotemporal decorrelation noise, accurately recovers complex images, and improves the point cloud density and elevation accuracy of TomoSAR 3D reconstruction, providing a technical reference for application in multi-temporal, multi-baseline TomoSAR.

The application discloses a kind of unmanned aerial vehicle path planning method and system based on hybrid swarm intelligence algorithm, and is related to unmanned aerial vehicle planning technical field.The method comprises: obtaining target task information;The target task information includes including departure location and end position;Utilize hybrid swarm intelligence algorithm, path planning is carried out to the target task information, and the optimal path of unmanned aerial vehicle is determined;The hybrid swarm intelligence algorithm includes improved particle swarm algorithm and sparrow search algorithm;The improved particle swarm algorithm is constructed according to particle swarm algorithm and adaptive t distribution variation operator.The application can improve the speed and accuracy of unmanned aerial vehicle path planning.

The invention relates to a space object ISAR image key point positioning and attitude estimation method. The method comprises the following steps: acquiring an ISAR image of a to-be-processed space object; and inputting the ISAR image into a pre-trained attitude estimation model to obtain an attitude evaluation result of the to-be-processed space object, the attitude evaluation result including an object category, a detection frame, a key point position and a main body attitude angle, and performing parallel processing through the attitude estimation model to obtain four evaluation results. By means of the method, the technical problems that the universality is poor and the real-time requirement cannot be met due to the fact that an existing method only depends on a single feature are effectively solved.

The present application relates to an X-ray security image contraband detection method based on improved YOLOv7, belonging to the field of target detection, comprising the following steps: S1: preprocessing the security data set and randomly dividing it into a training set and a validation set; S2: combining a multi-dimensional efficient channel attention module with a backbone network to construct an efficient backbone network; S3: constructing a transition network between the backbone network and the neck network through a multi-scale feature aggregation module; S4: designing a precise bounding box regression loss EIoUer Loss as the positioning loss of the model; S5: constructing an MME-YOLO security image contraband detection network model through an image preprocessing module, an efficient backbone network, a transition network, a neck network and a detection head; S6: training the MME-YOLO model; S7: verifying the MME-YOLO model.

The invention discloses a lithium battery defect signal extraction method based on improved WOA-VMD, and belongs to the technical field of lithium battery nondestructive testing and signal processing. According to the method, aiming at the problems of low signal-to-noise ratio after ultrasonic signal penetration and difficulty in defect feature extraction caused by an internal multi-layer structure during air coupling ultrasonic detection of a lithium battery, a cosine convergence factor and a nonlinear inertia weight strategy are introduced to improve a traditional whale optimization algorithm; key parameters of variational mode decomposition are optimized in a self-adaptive mode; variational mode decomposition is carried out on the noisy ultrasonic signal by using the optimized parameters, and dominant intrinsic mode function components are screened out according to a correlation coefficient threshold value; and carrying out wavelet threshold denoising on the screened components, and finally reconstructing to obtain a defect characteristic signal with a high signal-to-noise ratio. The method has high adaptability and robustness, is suitable for defect detection of lithium batteries of different specifications, and provides an effective technical means for safety and reliability evaluation of the lithium batteries.

The invention relates to the technical field of analog-to-digital converters, in particular to a sampling time error background calibration method for a time-interleaved analog-to-digital converter, which comprises a feed-forward compensation unit, a feedback compensation unit and a judgment unit, and is characterized in that during initialization, the feed-forward compensation unit is directly used for compensating sampling data of each channel; after each time of compensation, a judgment unit is adopted to judge whether the time mismatch error estimation value is larger than a set threshold value or not, if yes, a feed-forward compensation unit is adopted for next compensation, and otherwise, a feedback compensation unit is adopted for compensation. According to the method, when the cross-correlation function is calculated, the operation complexity is effectively reduced through the moving average filter, and the resource consumption during hardware implementation is equivalently reduced.

This invention provides a population optimization method, system, equipment, and medium based on watershed hydropower complementarity, belonging to the fields of intelligent optimization and computer applications. It includes: obtaining adjustment parameters for the current iteration step; obtaining the original position and fitness of each individual in the current population; calculating rainfall intensity and infiltration capacity; calculating the actual infiltration and surface runoff of each individual; determining the runoff generation pattern of the current individual based on the comparison between the actual infiltration and rainfall intensity, and updating the position of the current individual; performing boundary processing on the updated position to obtain a corrected position; updating the positions of individuals in the current population to obtain an updated population; sorting by fitness; and determining whether a preset termination condition is met; if met, outputting the optimal solution based on the sorting result. This invention introduces the watershed hydropower complementarity mechanism into the population evolution process, which helps coordinate global exploration and local development, improving the accuracy and stability of solving complex optimization problems.

The invention relates to the field of mobile edge computing, and discloses an unmanned aerial vehicle assisted asynchronous federated learning online scheduling method and system based on convergence perception. The method comprises the following steps: establishing a system model and defining model obsolessness; constructing a convergence cost penalty function containing data isomerism and old degree influence on the basis of theoretical derivation; under the condition that energy budget and bandwidth constraints are met, a joint optimization problem with the goal of minimizing long-term convergence cost is established; constructing a virtual queue by using a Lyapunov optimization technology, and converting a long-term problem into a single-slot deterministic optimization sub-problem; and finally, a deep reinforcement learning algorithm based on an encoder-decoder structure is adopted to solve a flight position and a client selection strategy online. The system comprises various modules and computing equipment which are configured to execute the steps of the method. According to the method, model deviation caused by asynchronous updating can be effectively relieved, and the convergence speed and precision of the model are remarkably improved in an energy-limited environment.