Manual Tuning Research Articles

Virtually constrained generalized relative motion modeling and a control parameter optimizer for automatic carrier landing

PurposeThis paper aims to propose a novel control scheme and offer a control parameter optimizer to achieve better automatic carrier landing. Carrier landing is a challenging work because of the severe sea conditions, high demand for accuracy and non-linearity and maneuvering coupling of the aircraft. Consequently, the automatic carrier landing system raises the need for a control scheme that combines high robustness, rapidity and accuracy. In addition, to exploit the capability of the proposed control scheme and alleviate the difficulty of manual parameter tuning, a control parameter optimizer is constructed.Design/methodology/approachA novel reference model is constructed by considering the desired state and the actual state as constrained generalized relative motion, which works as a virtual terminal spring-damper system. An improved particle swarm optimization algorithm with dynamic boundary adjustment and Pareto set analysis is introduced to optimize the control parameters.FindingsThe control parameter optimizer makes it efficient and effective to obtain well-tuned control parameters. Furthermore, the proposed control scheme with the optimized parameters can achieve safe carrier landings under various severe sea conditions.Originality/valueThe proposed control scheme shows stronger robustness, accuracy and rapidity than sliding-mode control and Proportion-integration-differentiation (PID). Also, the small number and efficiency of control parameters make this paper realize the first simultaneous optimization of all control parameters in the field of flight control.

Inverse design of twisted bilayer graphene metasurface for terahertz absorption broadening based on artificial neural network

With the development of terahertz (THz) technology, achieving ultra-wideband absorption of THz waves has become a significant challenge that researchers are striving to address. Different from the complex stacking of multiple graphene layers in the traditional design, inspired by the twisted magic angle of graphene structures, a twisted bilayer graphene metasurface absorber (TBGMA) is innovatively investigated for THz wave absorption broadening inverse design based on artificial neural network (ANN). Compared with the traditional manual parameter tuning, the ANN-based inverse design method can quickly accomplish the selection of the structural parameters and twist angles of the TBGMA, realizing the ultra-wideband absorption of THz waves (6.046 THz). The relationship between twisted SGLs and broadened effective absorption bandwidths is thoroughly analyzed through a combination of the electric field distribution and effective medium theory (EMT). Additionally, the tunability and absorption spectra of TBGMA under different incidence conditions have been discussed. The study of the twisted angle of SGLs can provide a unique idea for expanding the absorption bandwidth, and the inverse design of structural and angle parameters in conjunction with ANN is expected to be extended to the design of other precision structures in nanophotonics.

Enhancing Rice Production Prediction: A Comparative Machine Learning Analysis of Climate Variables

This study aims to enhance rice production prediction through a comparative analysis of machine learning models utilizing climate variables. Eight models were assessed on a predetermined dataset, with Support Vector Regression (SVR) emerging as the top performer. Following the identification of significant climate variables influencing rice production, the models underwent evaluation using two hyperparameter approaches: random search and manual tuning. SVR outperformed other models, achieving impressive metrics with MAE 0.180, MSE 0.186, RMSE 0.431, and an exceptionally low MAPE of 0.020. Key factors influencing rice production included productivity and area, along with humidity, rainfall, temperature, wind velocity, and sunshine duration. Favorable conditions for rice output encompassed low humidity, moderate rainfall, increased wind speed, and prolonged sunshine, while rainfall and temperature exhibited minimal impact. The success of random search emphasizes the importance of effective hyperparameter tuning. This research provides valuable insights for enhancing rice production prediction.

A Bayesian optimized machine learning approach for accurate state of charge estimation of lithium ion batteries used for electric vehicle application

In recent times, battery technology used in electric vehicles has drawn numerous researchers' attention. Monitoring the battery condition, particularly the State of charge, is required to ensure the battery's safe and reliable performance. Despite the fact that many SOC estimation methods have been proposed, further research is necessary to identify an approach that adapts versatile lithium-ion battery chemistries. In the recent study it has been demonstrated that Machine Learning approaches have better prediction accuracy compared to conventional methods. To maximize the performance of Machine learning models, it is imperative to select the optimal hyperparameters and employ appropriate input parameters. At present, researchers employ, established heuristics methods to select hyperparameters, which may involve manual tuning or exhaustive search techniques such as random search and grid search. These techniques make the models less accurate and inefficient. In this paper, a systematic, automated process for selecting hyperparameters with a Bayesian optimization algorithm is proposed. In addition, along with the battery parameters (voltage, current and temperature), vehicle velocity, road condition, motor characteristics and environmental conditions are used as the input parameters for accurate SOC prediction. The highly correlated input features are selected through the MRMR algorithm. The performance of six ML algorithms, namely SVM, ANN, GPR, Ensembler, linear regression and Decision Tree, is tested and validated with and without hyperparameter tuning for different data sets. The experimental results demonstrate that the hyperparameter-tuned model outperforms the standard model in estimating the State of charge (SOC). In addition, the model's estimation error is <1 % across a range of vehicle velocity, road condition, motor torque, battery voltage, current, and temperature.

Application research on improved genetic algorithm and active disturbance rejection control on quadcopters

For nonlinear, strongly coupled, underdriven quadcopters in the context of modeling complexity and demanding performance requirements for the controller, this paper proposes a strategy based on an improved genetic algorithm to optimize the active disturbance rejection control (ADRC) controller. To make the quadcopter continue to fly stably in a complex environment, the dynamics model of the quadcopter was firstly established, the mathematical model was simplified according to the real world, and the ADRC controller of the quadcopter was designed. Given a large number of ADRC controller parameters, the difficulty of manual tuning and obtaining the optimal control effect, and the shortcomings of the genetic algorithm in solving the problem of local optimal and precocious convergence, a control strategy based on improved genetic algorithm to optimize ADRC’s parameters is proposed to improve the genetic diversity in the population and enhance the adaptability of individuals to the environment, ITAE (Integral-of-Time-multiple Absolute Error) evaluation index is selected as the fitness value. Finally, the model of the control system is built according to the real aircraft. The application results prove that the altitude, attitude of the quadcopter are controlled stably, and it is verified that the control strategy based on the improved genetic algorithm optimizing ADRC has faster rapidity, stronger tracking performance, and robustness in altitude, attitude control of the quadcopter, which has greater practical application value.

Performance analysis of genetically optimized 1D-convolutional neural network architecture for rotor system fault detection and diagnosis

Rotor mass imbalance is a critical problem in industries, which can lead to a breakdown or a catastrophic failure if unattended. Hence, Industry 4.0 uses deep learning approaches like convolutional neural network (CNN) for rotor fault detection and diagnosis of Industrial machines and turbomachinery where multifunctional structures are used. In machinery fault diagnosis, manual tuning of CNN hyperparameters is widely used, which is a tedious, time-consuming, and challenging task to achieve effective feature extraction. Additionally, overfitting of CNN during its training is an obstacle to achieving higher prediction accuracy. Hence, to address these issues, this research focuses on multi-class mass imbalance fault diagnosis using genetically optimized deep learning architecture of 1D-CNN to achieve higher diagnosis capability. The performance of the proposed methodology is evaluated through various case studies using experimental data from the in-house developed test rig. The test results are compared against various 1D-CNN architectures in which the hyperparameters and effective dropout layer positioning are genetically optimized. Also, the performance is benchmarked against standard machine learning algorithms. The results show that genetic algorithm (GA)-optimized 1D-CNN with dropout achieves highest fault prediction accuracy of 97.47% with reduced depth of CNN (with three convolution layers) compared to 84.16% of manually tunned CNN architecture (with seven convolutional layers) and outperformed standard machine learning algorithms. Best drop positioning (end of feature extraction part) reduced the learnable parameters to 95.9% with the highest prediction accuracy of 97.47% compared to adding dropout at all the CNN layers. The proposed GA-optimized 1D-CNN with effective dropout positioning eliminates human intervention and reduces the depth of CNN architecture; this, in turn, reduces computational load and time by reducing learnable parameters of CNN (network weights) with the highest prediction accuracy. Hence, the proposed approach shows promise in enhancing the performance and contributing to the advancement of 1D-CNN for rotor system fault detection and diagnosis.

Prediction of chlorine residual in effluent based on SSA-SFA-CFBLS for water treatment process parameters

Accurate parameter prediction of chlorine residual in effluent is necessary for nonlinear, long-delayed water treatment process to raise water quality. To improve the prediction model precision and computational efficiency, a cascade broad learning system based on the sparrow search algorithm (SSA) and slow feature analysis (SFA) is proposed in this paper. First, the SFA method is introduced to extract the essential characteristics of water monitoring data as the input of the prediction model. Then, a cascaded broad learning system is adopted to establish a prediction model for residual chlorine in water works effluent. The cascade broad learning can deal well with online prediction. Furthermore, the SSA is utilized to obtain the optimal hyperparameters of the established model, which can avoid the complex and time-consuming manual parameter tuning process. Finally, the comparison experiment with several methods is carried out. The experimental results show that the proposed method saves more computational resources, and its time consumption is only 16.8% of that of the comprehensive deep learning model with basically the same accuracy. The prediction accuracy is improved, and its prediction accuracy is improved by 7.6% on average compared with most traditional deep learning models such as long short term memory (LSTM), recurrent neural network (RNN) and fated recurrent unit (GRU).

Magnetorheological Damper Force Prediction using Particle Swarm Optimization and Long Short-Term Memory Model

Smart materials, like magnetorheological (MR) fluid, are gaining attention for their ability to rapidly change properties under magnetic influence, making them useful in vibration control systems for vehicles, medical devices, and civil engineering structures. Common parametric models, such as Bouc-Wen and Bingham, are traditionally employed to model MR damper dynamics behavior. However, the manual tuning of numerous parameters in these models increases complexity and hinders the identification of inverse models, potentially leading to unpredictable optimum target forces. In response to these challenges, this study suggested a non-parametric approach using Long Short-Term Memory (LSTM) models to predict the optimum target force of MR dampers. Unlike parametric models, LSTM models capture dynamic behavior without the need for extensive manual tuning. To optimize the LSTM model, Particle Swarm Optimization (PSO) is employed to fine-tune hyperparameter values, ensuring robust performance. The proposed non-parametric method, specifically the PSO-LSTM model, demonstrates faster processing times compared to traditional parametric approaches. The proposed model produced an accurate damping force prediction with a root mean square error of less than 5%, This novel approach simplifies the modeling process and offers an efficient and precise alternative to traditional parametric methods.

Grid-less wideband direction of arrival estimation based on variational Bayesian inference.

Many recent works have addressed the problem of wideband direction of arrival (DOA) estimation using grid-less sparse techniques, and these methods have been shown to outperform the traditional wideband DOA estimation methods. However, these methods often suffer from the problem of requiring manual parameter tuning or high computational complexity, which reduces their practicality. To alleviate this problem, a grid-less wideband DOA estimation method based on variational Bayesian inference is proposed in this paper. The method approximates the posterior probability density function of DOA with the help of variational Bayesian inference, which does not require manual adjustment of parameters and can obtain accurate DOA estimation results with low computational complexity. Numerical simulations and real measurement data processing show that the proposed method has a higher DOA estimation accuracy than other grid-less wideband methods while providing higher computational speed.

Towards an optimized heat transfer process in vaporizing liquid microthrusters using pulsed heating control

The Institute for Microelectronics and Microsystems of the Italian National Research Council (Italy) has designed and fabricated a water-fed MEMS vaporizing liquid microthruster (VLM) in collaboration with the University of Salento (Italy) and KU Leuven (Belgium). Previous studies have demonstrated the device’s functioning and highlighted the presence of a strong thermoelectric-hydraulic coupling that severely affects the heating and propulsive efficiencies and reliability of the device. In this regard, the current work provides a preliminary experimental investigation of a pulsed heating system actively controlled by temperature using PI-D logic (proportional and integrative actions applied to the error; derivative action applied to the output). The overall performance assessment focuses on evaluating power consumption during steady-state operation and analyzing the dynamic response of the VLM undergoing such an actively controlled pulsed heating, which has been built using both manual and data-driven offline adaptive tuning of the PID parameters. Concerning the steady-state operation, results highlight a promising enhancement of the heating efficiency to values above 0.9. The dynamic response analysis shows that operating with a single set of fixed PID parameters is not feasible and that the manual tuning is a trial-and-error approach highly dependent on the operator’s experience, operating conditions, and reference temperature profile. Furthermore, shorter rise and response times require a higher proportional gain, and the overshoot experienced when crossing the saturation temperature cannot be avoided. In this paper, we show that the response of the data-driven adaptive controller solves these issues, exhibiting a reduced maximum settling time from around 59 s to less than 10 s, while the overshoot is avoided thanks to the temporal adjustment of the PID parameters.

Research on emotion recognition method based on IWOA-ELM algorithm for electroencephalogram

Emotion is a crucial physiological attribute in humans, and emotion recognition technology can significantly assist individuals in self-awareness. Addressing the challenge of significant differences in electroencephalogram (EEG) signals among different subjects, we introduce a novel mechanism in the traditional whale optimization algorithm (WOA) to expedite the optimization and convergence of the algorithm. Furthermore, the improved whale optimization algorithm (IWOA) was applied to search for the optimal training solution in the extreme learning machine (ELM) model, encompassing the best feature set, training parameters, and EEG channels. By testing 24 common EEG emotion features, we concluded that optimal EEG emotion features exhibited a certain level of specificity while also demonstrating some commonality among subjects. The proposed method achieved an average recognition accuracy of 92.19% in EEG emotion recognition, significantly reducing the manual tuning workload and offering higher accuracy with shorter training times compared to the control method. It outperformed existing methods, providing a superior performance and introducing a novel perspective for decoding EEG signals, thereby contributing to the field of emotion research from EEG signal.

NnU-Net versus mesh growing algorithm as a tool for the robust and timely segmentation of neurosurgical 3D images in contrast-enhanced T1 MRI scans

PurposeThis study evaluates the nnU-Net for segmenting brain, skin, tumors, and ventricles in contrast-enhanced T1 (T1CE) images, benchmarking it against an established mesh growing algorithm (MGA).MethodsWe used 67 retrospectively collected annotated single-center T1CE brain scans for training models for brain, skin, tumor, and ventricle segmentation. An additional 32 scans from two centers were used test performance compared to that of the MGA. The performance was measured using the Dice-Sørensen coefficient (DSC), intersection over union (IoU), 95th percentile Hausdorff distance (HD95), and average symmetric surface distance (ASSD) metrics, with time to segment also compared.ResultsThe nnU-Net models significantly outperformed the MGA (p < 0.0125) with a median brain segmentation DSC of 0.971 [95CI: 0.945–0.979], skin: 0.997 [95CI: 0.984–0.999], tumor: 0.926 [95CI: 0.508–0.968], and ventricles: 0.910 [95CI: 0.812–0.968]. Compared to the MGA’s median DSC for brain: 0.936 [95CI: 0.890, 0.958], skin: 0.991 [95CI: 0.964, 0.996], tumor: 0.723 [95CI: 0.000–0.926], and ventricles: 0.856 [95CI: 0.216–0.916]. NnU-Net performance between centers did not significantly differ except for the skin segmentations Additionally, the nnU-Net models were faster (mean: 1139 s [95CI: 685.0–1616]) than the MGA (mean: 2851 s [95CI: 1482–6246]).ConclusionsThe nnU-Net is a fast, reliable tool for creating automatic deep learning-based segmentation pipelines, reducing the need for extensive manual tuning and iteration. The models are able to achieve this performance despite a modestly sized training set. The ability to create high-quality segmentations in a short timespan can prove invaluable in neurosurgical settings.

A generative adaptive convolutional neural network with attention mechanism for driver fatigue detection with class-imbalanced and insufficient data

Over the past few years, fatigue driving has emerged as one of the main causes of traffic accidents, necessitating the development of driver fatigue detection systems. However, many existing methods involves tedious manual parameter tunings, a process that is both time-consuming and results in task-specific models. On the other hand, most of the researches on fatigue recognition are based on class-balanced and sufficient data, and effectively “mine” meaningful information from class-imbalanced and insufficient data for fatigue recognition is still a challenge. In this paper, we proposed two novel models, the attention-based residual adaptive multiscale fully convolutional network-long short term memory network (ARMFCN-LSTM), and the Generative ARMFCN-LSTM (GARMFCN-LSTM) aiming to address this issue. ARMFCN-LSTM excels at automatically extracting multiscale representations through adaptive multiscale temporal convolutions, while capturing temporal dependency features through LSTM. GARMFCN-LSTM integrates Wasserstein GAN with gradient penalty (WGAN-GP) into ARMFCN-LSTM to improve driver fatigue detection performance by alleviating data scarcity and addressing class imbalances. Experimental results show that ARMFCN-LSTM achieves the highest classification accuracy of 95.84% in driver fatigue detection on the class-balanced EEG dataset (binary classification), and GARMFCN-LSTM attained an improved classification accuracy of 84.70% on the class-imbalanced EOG dataset (triple classification), surpassing the competing methods. Therefore, the proposed models are promising for further implementations in online driver fatigue detection systems.

Pseudo-L0-Norm Fast Iterative Shrinkage Algorithm Network: Agile Synthetic Aperture Radar Imaging via Deep Unfolding Network

A novel compressive sensing (CS) synthetic-aperture radar (SAR) called AgileSAR has been proposed to increase swath width for sparse scenes while preserving azimuthal resolution. AgileSAR overcomes the limitation of the Nyquist sampling theorem so that it has a small amount of data and low system complexity. However, traditional CS optimization-based algorithms suffer from manual tuning and pre-definition of optimization parameters, and they generally involve high time and computational complexity for AgileSAR imaging. To address these issues, a pseudo-L0-norm fast iterative shrinkage algorithm network (pseudo-L0-norm FISTA-net) is proposed for AgileSAR imaging via the deep unfolding network in this paper. Firstly, a pseudo-L0-norm regularization model is built by taking an approximately fair penalization rule based on Bayesian estimation. Then, we unfold the operation process of FISTA into a data-driven deep network to solve the pseudo-L0-norm regularization model. The network’s parameters are automatically learned, and the learned network significantly increases imaging speed, so that it can improve the accuracy and efficiency of AgileSAR imaging. In addition, the nonlinearly sparsifying transform can learn more target details than the traditional sparsifying transform. Finally, the simulated and data experiments demonstrate the superiority and efficiency of the pseudo-L0-norm FISTA-net for AgileSAR imaging.

MetaLo: metabolic analysis of Logical models extracted from molecular interaction maps.

Molecular interaction maps (MIMs) are static graphical representations depicting complex biochemical networks that can be formalized using one of the Systems Biology Graphical Notation languages. Regardless of their extensive coverage of various biological processes, they are limited in terms of dynamic insights. However, MIMs can serve as templates for developing dynamic computational models. We present MetaLo, an open-source Python package that enables the coupling of Boolean models inferred from process description MIMs with generic core metabolic networks. MetaLo provides a framework to study the impact of signaling cascades, gene regulation processes, and metabolic flux distribution of central energy production pathways. MetaLo computes the Boolean model's asynchronous asymptotic behavior, through the identification of trap-spaces, and extracts metabolic constraints to contextualize the generic metabolic network. MetaLo is able to handle large-scale Boolean models and genome-scale metabolic models without requiring kinetic information or manual tuning. The framework behind MetaLo enables in depth analysis of the regulatory model, and may allow tackling a lack of omics data in poorly addressed biological fields to contextualize generic metabolic networks along with improper automatic reconstructions of cell- and/or disease-specific metabolic networks. MetaLo is available at https://pypi.org/project/metalo/ under the terms of the GNU General Public License v3.

Two-layer Ensemble of Deep Learning Models for Medical Image Segmentation

One of the most important areas in medical image analysis is segmentation, in which raw image data is partitioned into structured and meaningful regions to gain further insights. By using Deep Neural Networks (DNN), AI-based automated segmentation algorithms can potentially assist physicians with more effective imaging-based diagnoses. However, since it is difficult to acquire high-quality ground truths for medical images and DNN hyperparameters require significant manual tuning, the results by DNN-based medical models might be limited. A potential solution is to combine multiple DNN models using ensemble learning. We propose a two-layer ensemble of deep learning models in which the prediction of each training image pixel made by each model in the first layer is used as the augmented data of the training image for the second layer of the ensemble. The prediction of the second layer is then combined by using a weight-based scheme which is found by solving linear regression problems. To the best of our knowledge, our paper is the first work which proposes a two-layer ensemble of deep learning models with an augmented data technique in medical image segmentation. Experiments conducted on five different medical image datasets for diverse segmentation tasks show that proposed method achieves better results in terms of several performance metrics compared to some well-known benchmark algorithms. Our proposed two-layer ensemble of deep learning models for segmentation of medical images shows effectiveness compared to several benchmark algorithms. The research can be expanded in several directions like image classification.

Advancing Cognitive Radio in 5G Networks: A Survey of Flexible and Reconfigurable Antenna Technologies

Abstract: Nowadays, the communication industry is looking for a cognitive radio platform that can be used to handle multiple frequency bands, utilize various transmission protocols and be reconfigurable. For spectrum allocation, two methodologies have been adopted for CR between users who are licensed (PU) and those who are not (SU). They are spectrum overlay and spectrum underlay. Secondary users need to operate below the principal users' noise level therefore, significant restrictions resisted on the strength of their transmission in spectrum underlay methodology. Spectrum underlay is a type of underlay that is used in the construction of the unoccupied section of the spectrum searched by CR in the spectrum overlay process. A reconfigurable antenna is needed on RF front-end side of CR, which will a range of completely to span a vast array of frequencies specified sub-bands of frequency, and this is often a major design challenge. An antenna with a flexible monopole for Ultra-wideband equipment competent to function between the frequency range of 3.1 GHz to 10.6 GHz. Designs of reconfigurable antennas reported in the literature are mostly observed to be capable of manual tuning or switching between two or a few frequency bands. Hence, the requirement of new designs of integrated UWB and reconfigurable narrowband antennas that operate from 2 GHz and up to 12 GHz frequency range with a high degree of automated tunability or switching facility into different narrow bands would have a strong application potential for CRs.

Automatically discovering ordinary differential equations from data with sparse regression

Discovering nonlinear differential equations that describe system dynamics from empirical data is a fundamental challenge in contemporary science. While current methods can identify such equations, they often require extensive manual hyperparameter tuning, limiting their applicability. Here, we propose a methodology to identify dynamical laws by integrating denoising techniques to smooth the signal, sparse regression to identify the relevant parameters, and bootstrap confidence intervals to quantify the uncertainty of the estimates. We evaluate our method on well-known ordinary differential equations with an ensemble of random initial conditions, time series of increasing length, and varying signal-to-noise ratios. Our algorithm consistently identifies three-dimensional systems, given moderately-sized time series and high levels of signal quality relative to background noise. By accurately discovering dynamical systems automatically, our methodology has the potential to impact the understanding of complex systems, especially in fields where data are abundant, but developing mathematical models demands considerable effort.

ChatGPT: A Breakthrough in Developing Human-Like Robots with Natural Language Capabilities

Robotic systems often require engineers to write code to specify the desired behaviour of the robots. This process is slow, costly, and inefficient, as it involves multiple iterations and manual tuning. ChatGPT is a tool that leverages a large language model (LLM) to enable natural language interaction, code generation, and learning from feedback for robotic applications. ChatGPT allows users, who may not have technical expertise, to provide high-level instructions and feedback to the LLM, while observing the robot's performance. ChatGPT can produce code for various scenarios of robots, using the LLM's knowledge to control different robotic factors. ChatGPT can also be integrated with other platforms, such as Snapchat and Duolingo, to enhance the user experience and management. ChatGPT is a novel tool that facilitates a new paradigm in robotics, where users can communicate with and teach robots using natural language. Keywords: ChatGPT, Large Language Model, Natural Language Processing, Human Robot Interaction

Positioning by floors based on WiFi fingerprint

WiFi-based indoor positioning technology has gradually become a hot research topic in the field of indoor positioning, but the development of this technology has been facing the challenge of susceptibility to environmental interference. Therefore, in this paper, the kernel function method (KFM) with stronger interference resistance is used for positioning, and the adaptive algorithm is proposed for the time-consuming and laborious problem of manual parameter tuning, which incorporates the ideas of cross-validation and iteration. In addition, too many wireless access points (APs) mean higher computational cost and longer positioning time, so it is necessary to choose reasonable APs for positioning. In this paper, we use the random forest (RF) algorithm to assess the importance of APs and filter out a small number of APs with high importance. Considering the obvious differences in the WiFi signals received on different floors, a system framework for positioning by floors based on WiFi fingerprints is proposed. In the offline phase, the fingerprint library is first divided according to floors, and then perform separately AP selection and parameter tuning for each sub-fingerprint library. In the online phase, support vector machine is used to discriminate the floors first, and then KFM is used for planar positioning. Experiments are conducted on the public dataset, and the results show that the proposed algorithm has higher positioning accuracy, more robustness, and less time-consuming compared to several common algorithms.