In this study, we investigate model structures for neural ODEs to improve the data efficiency in learning the dynamics of control systems. We introduce two model structures and compare them with a typical baseline structure. The first structure considers the relationship between the coordinates and velocities of the control system, while the second structure adds linearity with respect to the control term to the first structure. Both of these structures can be easily implemented without requiring additional computation. In numerical experiments, we evaluate these structure on simulated simple pendulum and CartPole systems and show that incorporating these characteristics into the model structure leads to accurate learning with a smaller amount of training data compared to the baseline structure.

This paper presents a method of pose estimation with camera localization based on a genetic algorithm to derive the joint angle using an inverse kinematics model. In recent years, some open-source libraries for human skeleton tracking that can be implemented on smartphones have been released. Such software can detect joint positions of a skeleton from a 2D camera image; however, the 3D posture cannot be obtained. We therefore propose a method to estimate joint angles by using skeleton data. In the proposed approach, we use a multi-island genetic algorithm to maintain the diversity of the population and design a multi-objective function to improve the robustness. Moreover, structured learning based on topological mapping is implemented in the proposed method to enhance searching efficiency. In an experiment, the proposed method reduced the effect of outliers caused by misdetection. In addition, the structured learning was effective in decreasing the difference between skeleton data and the estimated poses.

Let G(V,E) be a graph with a set of vertices V and a set of edges E. Then, a minimum subset W_l of V is said to be a local metric basis of G if for any two adjacent vertices u,v∈V∖W_l there exists a vertex w∈W_l such that d(u,w) ≠ d(v,w). The cardinality of a local metric basis is referred to as the local metric dimension of the graph G denoted by β_l(G). In this paper, we investigate the local metric dimensions of certain circulant-related architectures such as Harary graphs H_k,n with even k or n, Toeplitz networks, and ILLIAC networks.

Human parsing is one of the basic tasks in the field of computer vision. It aims at assigning pixel-level semantic labels to each human body part. Single human parsing requires further associating semantic parts with each instance. Aiming at the problem that it is difficult to distinguish the body parts with similar local features, this paper proposes a single human parsing method based on the visual attention mechanism. The proposed algorithm integrates advanced semantic features, global context information, and edge information to obtain accurate results of single human parsing resolution. The proposed algorithm is validated on standard look into part (LIP) dataset, and the results prove the effectiveness of the proposed algorithm.

Given the presence of handwritten documents in human transactions, including email sorting, bank checks, and automating procedures, handwritten characters recognition (HCR) of documents has been invaluable to society. Handwritten Chinese characters (HCC) can be divided into offline and online categories. Online HCC recognition (HCCR) involves the trajectory movement of the pen tip for expressing linguistic content. In contrast, offline HCCR involves analyzing and categorizing the sample binary or grayscale images of characters. As recognition technology develops, academics’ interest in Chinese character recognition has continuously increased, as it significantly affects social and economic development. Recent development in this area is promising. However, the recognition accuracy of offline HCCR is still a sophisticated challenge owing to their complexity and variety of writing styles. With the advancement of deep learning, convolutional neural network (CNN)-based algorithms have demonstrated distinct benefits in offline HCCR and have achieved outstanding results. In this review, we aim to show the different HCCR methods for tackling the complexity and variability of offline HCC writing styles. This paper also reviews different activation functions used in offline HCCR and provides valuable assistance to new researchers in offline Chinese handwriting recognition by providing a succinct study of various methods for recognizing offline HCC.

The activation function introduces nonlinearity into convolutional neural network, which greatly promotes the development of computer vision tasks. This paper proposes elastic adaptively parametric compounded units to improve the performance of convolutional neural networks for image recognition. The activation function takes the structural advantages of two mainstream functions as the function’s fundamental architecture. The SENet model is embedded in the proposed activation function to adaptively recalibrate the feature mapping weight in each channel, thereby enhancing the fitting capability of the activation function. In addition, the function has an elastic slope in the positive input region by simulating random noise to improve the generalization capability of neural networks. To prevent the generated noise from producing overly large variations during training, a special protection mechanism is adopted. In order to verify the effectiveness of the activation function, this paper uses CIFAR-10 and CIFAR-100 image datasets to conduct comparative experiments of the activation function under the exact same model. Experimental results show that the proposed activation function showed superior performance beyond other functions.

The operating conditions influence the stability and consumption of a blast furnace. Recognizing these conditions makes changing the burden distribution parameters more efficient. The cooling stave temperature (CST) is a crucial state parameter that indicates the conditions of the process. Owing to the high data volume of the CST and the lack of methods for recognizing the stability of the slag crust, it is difficult for operators to recognize the conditions accurately according to the CST during the ironmaking process. Thus, in this study, a condition recognition method with information granulation for burden distribution in a blast furnace was presented. First, information granulation was employed to reduce the volume of the CST data and present it in a granular form. Then, considering the lack of a method for calculating the similarity of CST information granules, a novel fuzzy similarity calculation method was devised to calculate the membership grades of information granules belonging to different standard granules. Finally, the conditions were recognized according to the membership values. Experimental results based on industrial data demonstrated that the proposed method can be used to recognizes the conditions in the blast furnace.

In this study, a drilling fluid prediction method based on an online sequential extreme learning machine (OS-ELM) is proposed, which is prepared for water well drilling on the muddy clay formation of Tarim Basin, Qinghai Province. First, we investigated the mechanism linking mix ratio to fluid performance, allowing us to employ an OS-ELM algorithm derived from the extreme learning machine. Particularly, the proposed prediction method is bidirectional to identify an appropriate slurry formulation. The forward prediction model is established to predict the fluid performance, where the mud additive contents are inputs, and the drilling fluid properties parameters are outputs. Correspondingly, the backward prediction model is established to modify the slurry formula, where differences in the drilling fluid properties are inputs and percentages of slurry additives amount are output. The simulation results show that the two-direction OS-ELM prediction model can better predict the drilling fluid properties in water well drilling.

In this study, we introduce a system called Matrix Researcher2vec (MResearcher2vec) which generates researcher embedding vectors from their papers and research projects in researchmap and KAKENHI databases. The system includes data on 276,841 researchers, 6,161,592 papers, and research projects. Utilizing natural language processing techniques, the MResearcher2vec model extracts researcher vectors from the papers and research project summaries of KAKENHI grant recipients. The similarity between reseachers is then computed to visualize inter-researcher relationships. The machine learning results have been integrated into a web service, providing a novel approach for academic relationship mining. It can be applied in the matching of research contents and researchers in evaluation of industry-government-academia collaboration and joint research. It contributes in four aspects: (1) exchanges between researchers, (2) creation of opportunities for researchers and companies to connect, (3) further promotion of interdisciplinary research, and (4) reduction of lost opportunities for research institutions to acquire talents.

In the future, considering the expansion of the autonomous driving society, autonomous driving systems that can drive safely and quickly will be required for the purpose of saving lives and transporting goods even on rough road such as snowy, icy, and unpaved roads. In such unknown environments, technologies that combine model-based control and artificial intelligence (AI) are attracting attention for the purpose of ensuring operational stability and reliability. The second author has proposed a vehicle driving model that is robust to road geometry and ever-changing environmental disturbances. This model is based on a two-wheel model, and expresses the error in the position of the center of gravity of the vehicle by the front wheel steering angle deviation, and adaptively estimates this deviation. However, this model has large modeling errors when driving at high velocity on slippery roads. In this study, we extend this model proposed in previous study, and propose a new vehicle driving model that can handle situations such as driving with large acceleration changes and steering on bad roads such as snowy and wet roads. Then, we demonstrate the usefulness of the proposed method in a simulation using vehicle motion analysis software.

Numerical calculations are used in various situations. However, to achieve accurate numerical calculations, accuracy in the calculation method and initial values with high spatial resolution are necessary. Therefore, we propose a new method for estimating spatial values that considers characteristic theory but does not use interpolation. We consider the treatment of the curved characteristic line, which implies that the characteristic speed is altered locally. In the new method named averaging inverse characteristics method (AICM), the locally changing characteristic speed is averaged with the characteristic speed of the previous steps. We calculated the spatial values of the shock tube problem, described by the Euler equation, and examined the accuracy of the AICM by comparing the results of the inverse characteristics method (ICM) proposed in the previous study and the traditional interpolating methods. Compared to other methods, AICM reduced the error to less than 1/10 for all parameters. We determined from these results that the AICM accurately estimates the spatial distribution of problems where characteristic speed has significantly changed.

The illegal dumping of aluminum and plastic into cities and marine areas leads to negative impacts on the ecosystem and contributes to increased environmental pollution. Although volunteer trash pickup activities have increased in recent years, they require significant effort, time, and money. Therefore, we propose automated trash pickup robot, which incorporates autonomous movement and trash pickup arms. Although these functions have been actively developed, relatively little research has focused on trash detection. As such, we have developed a trash detection function by using deep learning models to improve the accuracy. First, we created a new trash dataset that classifies four types of trash with high illegal dumping volumes (cans, plastic bottles, cardboard, and cigarette butts). Next, we developed a new you only look once (YOLO)-based model with low parameters and computations. We trained the model on a created dataset and a dataset consisting of marine trash created during previous research. In consequence, the proposed models achieve the same detection accuracy as the existing models on both datasets, with fewer parameters and computations. Furthermore, the proposed models accelerate the edge device’s frame rate.

Experimental education imparted at universities is essential for students to confirm their theoretical knowledge. Experiments, in general, can deal with real-world data that theory or simulation cannot handle. However, there are experimental subjects where it is difficult to obtain results according to the theory. Among them, heat conduction is a subject wherein it is difficult to obtain theoretical results because it is difficult to establish an experimental environment. Therefore, it is highly important to design experimental content using students’ perspectives, such as theory and practical experiments. Therefore, this study investigates the impact of education on the design and application of experimental apparatus for a heat-transfer experiment, which is a part of an experiment conducted by the Department of Mechanical Engineering, Tokyo University of Technology. Furthermore, we discuss the effectiveness of the proposed experimental system based on students’ behavior, comprehension, satisfaction, and subsequent results.

This study proposes a cloud-edge collaboration framework for temperature regulation in continuous annealing processes. A multiobjective optimization is formulated by ensuring the control accuracy of the temperature to reduce energy consumption and increase efficiency with cloud computing. Based on process analytics, a framework for clustering operating conditions with high real-time requirements is proposed. Further, a recommendation mechanism for furnace temperatures with low real-time requirements is developed in the cloud. Compared with traditional architectures, the cloud-edge collaboration approach improves energy savings and control stability, which demonstrates its effectiveness and practicality.

This study considers the exact wirelength of embedding complete bipartite graphs into wheel related graphs such as wheel graphs, gear graphs, and helm graphs.

Chemical graph theory is a field related to Chemistry which is used to predict the behavior of chemical structures. The current trend in Chemical graph theory is evaluating topological indices. They are the numerical descriptors of the molecular structures derived from their corresponding molecular graph. These indices are studied and used in drugs, pharmaceutical research, and other fields. In this paper, the eccentric connectivity index is computed for nanosheets and nanotubes of SiO₂.

Gripping comfort evaluation was crucial for designing a product with good gripping comfort. In this study, a novel evaluation method using gripping posture image was constructed based on convolutional neural network (CNN). Human subject experiment was conducted to acquire gripping comfort scores and gripping posture images while gripping seven objects with simple shape and eleven manufactured products. The scores and the images were used as training set and validation set for CNN. Classification problem was employed to classify gripping posture images as comfort or discomfort. As a result, accuracies were 91.4% for simple shape objects and 76.2% for manufactured products. Regression problem was utilized to predict gripping comfort scores from gripping posture images while gripping cylindrical object. Gripping posture images of radial and dorsal sides in direction of hand were used to investigate effect of direction of hand on prediction accuracy. Consequently, mean absolute errors (MAE) of gripping comfort scores were 0.132 for radial side and 0.157 for dorsal side in direction of hand. In both problems, the results indicated that these evaluation methods were useful to evaluate gripping comfort. The evaluation methods help designers to evaluate products and enhance gripping comfort.

In the communication-constrained operating environment, a unmanned aerial vehicle (UAV) needs to plan a feasible path from the starting point to the endpoint while planning the node deployment location for multi-hop communication to establish an information pathway. In this study, a new algorithm was designed for joint path and multi-hop communication node location planning in cluttered environments based on rapidly-exploring random trees star (RRT*) algorithm. The maximum communication distance constraint between nodes was obtained based on the signal-free propagation model, whereas the communication node loss and path loss were established as joint optimization objectives. In bidirectional random tree growth, the structure of the trees was optimized according to the value of the loss function, and optimal path and node location planning were finally achieved through continuous growth and iteration. When tested in different complexity-barrier environments and compared to RRT*, Informed-RRT*, and IB-RRT* algorithms, the paths in the planning results of the new algorithm are close to those of the comparison algorithms; however, the number of nodes decreases significantly, which proves the effectiveness of the newly proposed algorithm.

Herein, a dual-branch semantic segmentation model based on depth-separable convolution and attention mechanism is proposed for the real-time and accuracy requirement of semantic segmentation. The proposed approach overcomes the problems of poor segmentation effect and over-simplification of feature fusion arising from the constant downsample operations in semantic segmentation. The network is divided into spatial detail and semantic information paths. The spatial detail path utilizes a smaller downsample multiplier to maintain resolution and efficiently extract spatial information. The semantic information path is constructed by a non-bottleneck residual unit with dilated convolution; it extracts semantic features. For the feature aggregation problem, the feature-guided fusion module is designed to assign different weights to the parts of the two paths and fuse them to obtain the final output. The proposed algorithm achieves a segmentation accuracy of 69.6% and speed of 70 fps on the Cityscapes dataset, with a model parameter count of only 0.76 M, thus indicating some advantages over recent real-time semantic segmentation algorithms. The proposed method with depth separable convolution and attention mechanism can effectively extract features and compensate for the loss of accuracy caused by downsampling. The experiments demonstrate that the proposed fusion module outperforms other methods in fusing different features.

This article examines the stability problem of systems with interval time-varying delays. In the derivation of Lyapunov–Krasovskii functional (LKF), non-convex higher-degree polynomials may arise with respect to interval time-varying delays, making it difficult to determine the negative definiteness of LKF’s derivative. This study was conducted to obtain stability conditions that can be described as linear matrix inequalities (LMIs). By considering the idea of matrix transition and introducing the delay-dependent augmented vector, a novel higher-degree polynomial inequality is proposed under the condition that the lower bound of the polynomial function variable is non-zero, which encompasses the existing lemmas as its special cases. Then, benefiting from this inequality, a stability criterion is derived in terms of LMIs. Finally, several typical examples are presented to verify the availability and strength of the stability condition.

Rehabilitation evaluation is an important part of rehabilitation training. It is closely related to the robot-assisted training effect. Different rehabilitation robots need different methods to evaluate patients. Rehabilitation training is a long process, and the patient’s performance scores will continue to change. A lower-limb rehabilitation robot needs a dynamic performance score to evaluate rehabilitation’s effects. This study used an analytic hierarchy process and fuzzy comprehensive evaluation methods to establish a rehabilitation evaluation system for lower-limb rehabilitation robots. A multi-scale personalized rehabilitation plan is conceived, based on the evaluation system and the combination of objective factors. This method dynamically adjusts the plan according to the rehabilitation situation of patients, which is beneficial to the improvement of the efficiency and initiative of training.

To overcome the previous challenges and to mitigate the retransmission and acknowledgment of LoRa for electric shelf labels, the data parallelism model is used for transmitting the concurrent data from the network server to end devices (EDs) through gateways (GWs). The EDs are designated around the GWs based on machine clustering to minimize data congestion, collision, and overlapping during signal reception. Deployment and redeployment of EDs in the defined clusters depend on arithmetic distribution to reduce the near-far effect and the overall saturation in the network. To further improve the performance and analyze the behavior of the network, constant uplink power for signal-to-noise (SNR) while dynamic for received signal strength (RSS) has been proposed. In contrast to SNR, the RSS indicator estimates the actual position of the ED to prevent the capture effect. In the experimental implementation, downlink power at the connected EDs in the clusters illustrates higher values than the defined threshold.

In the quest for human-robot interaction (HRI), leading to the development of emotion recognition, learning, and analysis capabilities, robotics plays a significant role in human perception, attention, decision-making, and social communication. However, the accurate recognition of emotions in HRI remains a challenge. This is due to the coexistence of multiple sources of information in utilizing multimodal facial expressions and head poses as multiple convolutional neural networks (CNN) and deep learning are combined. This research analyzes and improves the robustness of emotion recognition, and proposes a novel approach that optimizes traditional deep neural networks that fall into poor local optima when optimizing the weightings of the deep neural network using standard methods. The proposed approach adaptively finds the better weightings of the network, resulting in a hybrid genetic algorithm with stochastic gradient descent (HGASGD). This hybrid algorithm combines the inherent, implicit parallelism of the genetic algorithm with the better global optimization of stochastic gradient descent (SGD). An experiment shows the effectiveness of our proposed approach in providing complete emotion recognition through a combination of multimodal data, CNNs, and HGASGD, indicating that it represents a powerful tool in achieving interactions between humans and robotics. To validate and test the effectiveness of our proposed approach through experiments, the performance and reliability of our approach and two variants of HGASGD FER are compared using a large dataset of facial images. Our approach integrates multimodal information from facial expressions and head poses, enabling the system to recognize emotions better. The results show that CNN-HGASGD outperforms CNNs-SGD and other existing state-of-the-art methods in terms of FER.

In this study, for a linear singular system, the dissipativity and disturbance-rejection problems are considered simultaneously. An improved equivalent-input-disturbance (IEID) method has shown good disturbance-rejection performance for linear systems. Therefore, the objective of this study is to obtain a satisfactory disturbance-rejection performance and dissipativity performance level based on the IEID method for singular systems. First, the influence of exogenous disturbances on the system is estimated based on the IEID method. The estimate is added to the control input channel to offset this influence. A necessary and sufficient condition is obtained to ensure that the singular system is admissible and satisfies dissipativity performance level. Subsequently, a state-feedback controller is designed based on the admissibility condition. Finally, a numerical example is used to demonstrate the validity of the proposed method.

A novel method for predicting the thicknesses of plastics based on continuous-wave terahertz (THz) frequency-domain spectroscopy (THz-FDS) is presented in this study. Initially, the target material’s THz refractive index is determined from the phase information provided by the coherent nature of THz-FDS. For thickness prediction, the optimal frequency band with a high signal-to-noise ratio and minor water vapor absorption is chosen first. The optical path along which the THz wave passes through a sample with unknown thickness is extracted from the phase delay information. The physical thickness of the sample is then determined using the calibrated refractive index obtained in the first step. Teflon, a classical plastic material, is utilized to illustrate the proposed process. A remarkable consistency with an overall relative difference of only 0.45% is revealed between the THz-FDS predicted and caliper measured thicknesses. The proposed method is expected to significantly expand the capabilities of THz spectroscopy.