Keywords: Robot Mechanism and Control
Abstract: This study presents a design incorporating a built-in spring to enhance the gripping force of a two-finger parallel gripper. Conventional grippers grasp objects through the parallel linear motion of two tips. Various mechanisms can generate this motion, among which the rack-and-pinion gear mechanism features rack gears moving linearly in opposite directions around a central pinion gear. This study proposes a method to enhance gripping force by connecting the gripper’s internal frame and the rack gear within such a mechanism. Additionally, the proposed design maintains a certain level of gripping force by the restoring force of the spring even in the event of power loss. The gripping force and the required motor torque in the proposed method are analyzed. The experiments were conducted to compare the gripping force with and without the spring, and to evaluate the effectiveness of the proposed mechanism by measuring the gripping force generated solely by the spring when the motor power was turned off.

Keywords: Control Theory and Applications
Abstract: This paper investigates the oscillation and asymptotic behavior of a class of third-order nonlinear neutral dynamic equations on time scales with distributed deviating arguments. New criteria are established by utilizing Riccati transformations and integral averaging techniques. The obtained results extend and improve some known results in the literature. Two examples are given to illustrate the main results.

Keywords: Control Theory and Applications
Abstract: This paper investigates the solvability of the initial value problem for a class of variable-order fractional differential equations, which is a prerequisite for discussing nonlinear control theory for such problems. The core of this study is to employ the Banach's fixed-point theorem and the solutions of constant-order fractional differential equations established in this article to examine the solvability of variable-order fractional differential equations.

Keywords: Control Theory and Applications
Abstract: In this paper, with the help of generalized intervals and piecewise constant functions, by transforming it into some equivalent integral equations, the existence of solutions for initial value problem of variable-order fractional hybrid differential equations is investigated, which is the key point for control theory for such problems. All results of this paper are grounded in a fixed point theorem in the Banach algebra due to Dhage. Ultimately, an example is constructed to illustrate the validity of our primary results.

Keywords: Biomedical Instruments and Systems, Sensors and Signal Processing, Artificial Intelligence Systems
Abstract: Wireless Capsule Endoscope (WCE) has improved quality of Gastrointestinal (GI) screening. However, current WCE relies on passive locomotion, which may lead to misdiagnosis. To address this, active locomotion has been developed, though effective feedback algorithm remains limited. By employing data-driven ANN model, we could map the complex relationship between the pose of a magnet embedded inside the capsule and the magnetic field sensed by magnetic sensor. Additionally, the workspace is extended by integrating a robotic manipulator. Using a 3D printed model to mimic small intestine trajectory, the system was successfully localizes and tracks the capsule pose with RMSE of 1.69±0.6mm for position and 1.62±0.4° for orientation. The proposed approach in this study demonstrates potential for real-world application in the future.

Keywords: Robot Vision, Autonomous Vehicle Systems, Artificial Intelligence Systems
Abstract: To enable outdoor mobile robots to understand their surroundings, RGB-D semantic segmentation (SS)—which leverages both color (RGB) and depth (D) information—is an effective approach. Real-time performance is essential in this context, yet most existing RGB-D SS methods struggle with high-resolution inputs due to complex architectures and costly depth preprocessing. We propose a real-time RGB-D SS network by extending an existing real-time RGB SS model with depth integration, achieving both speed and accuracy. Our method introduces three key modules: (1) a Scale-Invariant Depth Encoder (SIDE) for efficient depth feature extraction, (2) an Attentive Feature Fusion Module (AFFM) for attention-based RGB-D fusion, and (3) a Noise Robust Guiding Module (NRGM) to handle depth noise without extra inference cost. By applying these modules to RGB networks, we create RGB-D models that match state-of-the-art accuracy while running at nearly one-third the computation time.

Keywords: Exoskeleton Robot, Rehabilitation Robot, Human-Robot Interaction
Abstract: In this study, we developed a wearable device that can support the extension movement of the user's trunk and lower extremities, and analyzed its assistive effects. The developed wearable device was configured to support the user's walking and stair climbing by linking with an external 4DoF actuator using a foot pressure sensor as input. Through the EMG sensor, the muscle activity of the left and right gluteus maximus and quadriceps muscles decreased by 25.5% for walking and by 8.73% for stair climbing.

Keywords: Artificial Intelligence Systems, Control Theory and Applications, Autonomous Vehicle Systems
Abstract: With the development of various air-to-air missiles, missile-based engagement has become the predominant mode of air combat. This evolution introduces discrete control variables (missile launch decisions) into the decision making process, thereby altering the granularity of traditional control spaces. The resulting hybrid action space problem, arising from the combination of discrete missile launch decisions and continuous maneuvering decisions, imposes more stringent requirements on air combat decision making methods. To address the problem of hybrid action spaces arising from the discrete missile launch decisions and continuous maneuvering decisions during the training process of reinforcement learning (RL) algorithms. We propose HPPO-RS, a novel air combat decision making training algorithm that integrates a multi-agent framework with hybrid proximal policy optimization (HPPO). Our algorithm learns both missile launch and maneuvering strategies in a competitive multi-agent environment, utilizing a potential function based reward shaping to overcome sparse reward challenges. Through extensive simulation experiments, we validate the effectiveness of our approach. The results demonstrate that HPPO-RS can generate effective maneuvering strategies and missile launch strategies, leading to tactically sound air combat behaviors.

Keywords: Control Theory and Applications
Abstract: In this paper, we deal with the solvable conditions for system of fractional differential inclusions with integral boundary value conditions depending a parameter. We establish a set of sufficient conditions for the existence of solutions of fractional differential inclusions. We use nonlinear alternative of Leray-Schauder type and corresponding corollary, Covitz and Nadler multivalued contraction principle, which is based on Bressan-Colombo selection theorem for lower semicontinuous multivalued maps with decomposable values to prove our main results. Some examples are also given to illustrate our main results.

Keywords: Sensors and Signal Processing, Multimedia Systems, Industrial Applications of Control
Abstract: This paper presents a hybrid sensor fusion system integrating Ultra-Wideband (UWB) and Radio Frequency (RF) communication enhanced by an Adaptive Kalman Filter (AKF) for robust victim localization in fire scenarios. The system utilizes LoRa-based RF for long-range detection and UWB for short-range precise localization. Wearable IMU sensors track victim motion, while AKF dynamically compensates RF noise fluctuations caused by smoke and structural interferences. Experimental results from simulations and real-world fire conditions confirm improved localization accuracy, reduced rescue time, and enhanced firefighter safety.

Keywords: Navigation, Guidance and Control, Artificial Intelligence Systems, Control Theory and Applications
Abstract: This paper addresses the multi-party optimal control problem of real-time nonlinear systems in the active defense process of the aircraft and defending missile, proposing a Dual-Action-Dependent Heuristic Dynamic Programming Algorithm based on Three-Dimensional Zero-Control Miss Distance (DADHDP). The algorithm combines approximate dynamic programming with the Actor-Critic framework in reinforcement learning, effectively approximating the infinite-horizon optimal control problem of complex nonlinear systems by iterating dynamic adversarial information in real-time. Compared with traditional methods, DADHDP introduces three-dimensional zero-control miss distance, an Actor network structure with embedded saturation constraints, and a non-quadratic cost functional in the algorithm design, which maintains system stability under control input limitations (such as saturation constraints), enhances solution efficiency, and effectively avoids control failure due to constraint violations in traditional methods.

Keywords: Navigation, Guidance and Control, Sensors and Signal Processing
Abstract: High-precision positioning (HPP) is crucial in modern control and automation systems, since it has a direct im- pact on the performance, safety, and reliability of autonomous vehicles and robotic systems. Direction of Arrival (DoA) estimation plays a vital role in HPP, and Singular Value Decomposition (SVD) serves as a key matrix decomposition method in various DoA algorithms by separating signal and noise subspaces. However, SVD operations are computationally intensive, posing challenges for deployment on general-purpose processors, especially in systems with strict on latency and power constraints. In this paper, we design a stream-based input/output interface for a Xilinx-provided SVD function using the Advanced eXtensible Interface (AXI) protocol through high-level synthesis (HLS), and implement an accelerated SVD Intellectual Property (IP) core using register-transfer level (RTL) synthesis. To improve throughput, the AXI stream interface is parallelized, enabling simultaneous processing of input and output data streams and reducing the data transfer overhead during SVD computation. The synthesized SVD IP core for the Zynq UltraScale+ ZCU102 evaluation board operates at a maximum clock frequency of 90.09 MHz. It uses 36,402 Look-Up Tables (LUTs) and 49,607 flip-flops (FFs), with an initiation interval (II) of 12,979 cycles and a latency of 13,012 cycles for a 4×4 input matrix.

Keywords: Sensors and Signal Processing, Autonomous Vehicle Systems
Abstract: In this paper, we propose image enhancement under photon-starved conditions utilizing an improved photon counting method. Conventional photon counting method has some problems that are required to determine how many photons apply to the images, and to find the optimal number of photons at the images. Additionally, the image applied to photon counting method presents some problems to solve, such as a lot of photons shot noise and the vanished information of bright region in the images. To solve these problems, we propose photon counting method using the optimal number of photons and overlapping matrix. To validate our proposed method, we compare with conventional photon counting method and histogram matching algorithm. Furthermore, we also evaluate the efficiency of our proposed method through the no-reference image quality assessments (NR-IQA), such as Naturalness Image Quality Evaluator, Perception-based Image Quality Evaluator, Blind/Referenceless Image Spatial Quality Evaluator, and Blind Image Integrity Notator using DCT Statistics.

Keywords: Sensors and Signal Processing, Autonomous Vehicle Systems
Abstract: In recent years, imaging techniques are close to our daily lives, such as drive recorders or around-view monitors. However, it is anticipated that scattering media, such as fog or smoke, decreases the accuracy of these techniques. Therefore, a lot of researches have been conducted on scattering media removal, one of them is Peplography. Peplography is an algorithm that visualizes objects by estimating and removing scattering media and emphasizing object information by photon-counting algorithm. However, conventional Peplography has low accuracy when visualizing objects. To improve this accuracy, we focus on a parameter in estimating the scattering media called windowing size. It is the size of the local region used to estimate the turbidity of the scattering media. In this research, we combine the object information acquired from two windowing sizes, based on that the obtained object information varies with the windowing size. In addition, we propose a new photon-counting algorithm that estimates photons in a region-by-region method to visualize objects with high accuracy. To confirm the accuracy of the proposed method, we show the visual and numerical experimental results in this paper.

Keywords: Biomedical Instruments and Systems, Artificial Intelligence Systems, Sensors and Signal Processing
Abstract: Digital holographic microscopy (DHM) is a technique that uses the interferometric nature of light to obtain three-dimensional (3D) information of the object. In recent years, research has been conducted to analyze a hologram image of red blood cell (RBC) taken with DHM to identify whether they are normal or abnormal, and to diagnose diseases such as malaria and COVID-19. Conventional methods used in many of these studies have analyzed and discriminated shapes by creating 3D profiles from individual RBCs and inputting the computable shape features into machine learning models. However, these methods identify RBCs on a per RBC basis and have the problems of high time cost for correct labeling and the inability to comprehensively analyze RBCs in an image. Therefore, in this study, we propose a new method that comprehensively analyzes and identifies RBCs in an image using deep learning. In the proposed method, the information of shape is acquired for RBCs in an image as two-dimensional waveform data information and synthesized into a single time series data. These temporal series data are input to a deep learning model to analyze and identify each image unit.

Keywords: Biomedical Instruments and Systems, Artificial Intelligence Systems, Sensors and Signal Processing
Abstract: Digital holographic microscopy (DHM) can obtain three-dimensional (3D) profiles of objects based on phase information obtained through the interference of light. In automatic disease diagnosis based on machine learning using DHM, accurate cell detection for the construction of training datasets is essential.　In this paper, we propose a highly accurate cell segmentation method for automated diagnostic support of DHM.　The proposed method combines the segmentation algorithm SAM and the K-means clustering to construct an automatic cell detection algorithm that does not require new training.　We use the proposed method to perform segmentation on hologram images of red blood cells taken using microscope lenses with magnifications of 40× and 20×, and calculate the detection rate and false positive rate.　The proposed method enables highly accurate detection of cells in images with different magnifications in DHM.　Furthermore, it also contributes to improve the quality of training data when applying machine learning to DHM.

Keywords: Autonomous Vehicle Systems, Robot Vision, Control Theory and Applications
Abstract: This paper presents a real-time algorithm for detecting and localizing contact bars in pantograph-based automatic charging systems used in electric bus depots. The algorithm leverages deep learning-based image processing to identify the contact bar, track the motion of the movable connector, and compute calibration parameters for accurate alignment. It accommodates five distinct charging scenarios and has been experimentally validated. Performance evaluation using Multiple Object Tracking Accuracy (MOTA) confirms its high accuracy in real-time tracking and position estimation. This method is expected to significantly improve the precision and reliability of high-capacity automatic charging for electric buses, thereby contributing to the development of smart transportation infrastructure.

Keywords: Robot Mechanism and Control, Artificial Intelligence Systems, Industrial Applications of Control
Abstract: This paper proposes a MAPPO-based motion planning framework for multi-robot manipulators operating in shared workspace environments. Each robot models its physical structure using multiple line segments representing key links and incorporates neighboring robots’ link positions into its observation space. This line-based state representation is both computationally efficient and suitable for accurately detecting potential collisions. A reward function penalizes unsafe proximity based on inter-line distances and encourages smooth, goal-directed trajectories. The framework follows the centralized training and decentralized execution paradigm, enabling agents to coordinate effectively while acting independently. We validate our method using Doosan A0912 robotic arms in NVIDIA Isaac Sim under two experimental scenarios involving overlapping workspaces. The proposed policy achieves high success rates and low task completion times across both environments, demonstrating its robustness and adaptability. Our results confirm that line-based MAPPO provides a practical and scalable solution for safe and efficient motion planning in collaborative robotic systems. Moreover, the method shows faster convergence and reduced computational overhead compared to key point-based representations, though its reliance on simplified link models may limit applicability to tasks requiring precise end-effector orientation. Future work will address real-world deployment and extend the approach to multi-robot and heterogeneous teams to ensure scalability and robustness.

Keywords: Robot Mechanism and Control, Robotic Applications, Artificial Intelligence Systems
Abstract: The research of multimodal learning in the field of in-hand manipulation has made great progress in recent years. However, existing algorithms often face challenges in finding an optimal balance between performance and complexity, which hinders their practical application. In response, we presented a novel visual-tactile fusion network for in-hand manipulation. We choose Masked MultiModal Learning (M3L) framework as the baseline model and use adaptive Nesterov momentum algorithm as a novel network parameter update strategy. In this way, our novel network enjoys the combined merit of M3L and adaptive Nesterov momentum algorithm, namely stronger generalization capabilities and faster convergence speed. The results of experiment show that our method achieves outstanding performance across a range of manipulation tasks compared with existing methods. In conclusion, this research introduces an advanced model for in-hand manipulation, especially in areas requiring precise manipulation and real-time adaptation, ultimately leading to more practical solutions for real-world robotic applications.

Keywords: Artificial Intelligence Systems, Robotic Applications
Abstract: Automated planning enables intelligent agents to generate action plans for user-defined goals without human intervention. Heuristic search is a leading approach in automated planning, and a common technique for deriving heuristics is to solve a simplified, relaxed version of the original problem, where the solution provides an estimate to guide the search. However, conventional abstraction techniques, such as pattern database (PDB) heuristics, typically ignore entire propositional variables and fail to account for the syntactic or causal relevance of action components. This paper presents a novel abstraction method that selectively removes preconditions and effects from action schemas based on their causal relevance to the goal. Using a Causal Operator Network (CON) that encodes hierarchical dependencies among actions, the method identifies and prunes irrelevant literals through a two-phase process, resulting in a relaxed planning problem with a reduced state space. Our approach provides finer-grained, goal-sensitive abstraction at the operator level. Experimental results across both benchmark and custom domains demonstrate notable improvements in planning efficiency, including significant reductions in search tree size and computation time.

Keywords: Robot Mechanism and Control
Abstract: Humanoid robots capable of performing music particularly drum-playing robots have great potential in entertainment and human–robot interaction. While our previous studies have focused on expressing human-like movements using a large number of degrees of freedom (DOFs), this study proposes a minimally designed and low degree-of-freedom humanoid robot capable of performing drum sequences. We developed an event sequence generator that converts MIDI music data into the event sequence format executable for the developed humanoid robot and automatically assign the hands and instruments to be played. We also created a trajectory generator consisting of several modules. The joint angles are generated in the task space and joint space, and then synthesized accordingly. To address potential self-collisions during fast and complex motions, a real-time collision prediction algorithm based on the Separating Axis Theorem (SAT) is applied. If detected, the collision avoidance module changes the event sequence or task space trajectory following several steps. Experimental results show that even with a simple hardware configuration, it can stably and accurately perform complex drum patterns.

Keywords: Control Devices and Instruments, Industrial Applications of Control, Biomedical Instruments and Systems
Abstract: Human technology has reached a level where it can observe the nanoscale world at the atomic level. This achievement is primarily due to the invention of a groundbreaking technology: the atomic force microscope (AFM). Through the AFM, humanity has gained a deeper understanding of the world of atoms and molecules, leading to revolutionary advancements in various science and engineering fields. This paper reviews the basic concepts, history, operating principles, and usage procedures of the AFM, and provides an overview of its application areas.

Keywords: Robotic Applications, Artificial Intelligence Systems, Robot Vision
Abstract: Learning manipulation skills from human demonstrations traditionally relies on structured datasets, teleoperations, or environment-specific supervision, limiting scalability and generalization. In this paper, we introduce REPLAY, a novel imitation learning framework that enables robots to acquire intent-aware behaviors directly from raw monocular videos, including unstructured sources such as YouTube. REPLAY decomposes demonstration videos into a sequence of semantically meaningful sub-tasks through action segmentation, scene and task reasoning via vision-language models, and fine-grained action understanding using 3D human pose estimation. These extracted trajectories are then retargeted to robot embodiments through intent-aware motion adaptation that accounts for embodiment differences and environmental constraints. To support scalable evaluation, we also present Video2Sim, a method that reconstructs realistic 3D simulation environments directly from demonstration videos, enabling repeatable testing and training. We demonstrated that REPLAY outperformed strong baselines in both motion fidelity and task success on complex object manipulation tasks, illustrating its potential as a scalable, generalizable approach for real-world imitation learning from in-the-wild video data.

Keywords: Artificial Intelligence Systems, Industrial Applications of Control, Robotic Applications
Abstract: Manipulators are widely utilized in industrial automation, and their efficient control remains a critical area of research. Controlling multiple manipulators in a shared workspace for industrial automation is challenging due to the need for real-time collision avoidance. This study tackles dynamic obstacle avoidance for collaborative pick-and-place tasks using Multi-Agent Proximal Policy Optimization (MAPPO). We designed a specialized collision reward function based on the spatial relationships of spherical representations on manipulator links. This encourages RL agents to autonomously discover collision-free paths and maintain safe distances. Our method's feasibility, effectiveness, and robustness were validated through comprehensive experiments in the high-fidelity physics engine, Isaac Sim. This simulation-based approach facilitates the smooth transition of learned policies to real robotic systems.

Keywords: Robotic Applications, Artificial Intelligence Systems
Abstract: Many service robot tasks, particularly cooking, require long-horizon planning in complex household environments. Classical task and motion planning (TAMP) methods have been widely adopted, but they suffer from limited flexibility in search-based approaches or fail to guarantee feasibility in learning-based approaches. We propose a modular robotic intelligence system for cooking tasks that integrates an object-oriented task planner (OO-TP), a robot task planner, and a motion planner. The OO-TP, trained on cooking recipe data, generates object-centric subgoals that capture sequential dependencies in long-horizon cooking processes. These subgoals are refined into feasible robot actions using a symbolic planner and an object knowledge base, ensuring both generalization and executability. Our proposed system achieves a 97.35% success rate across diverse recipes and demonstrates portability on heterogeneous robot arms without manual fine-tuning, highlighting its adaptability to long-horizon cooking tasks within the TAMP framework.

Keywords: Control Theory and Applications, Robotic Applications, Process Control Systems
Abstract: This study introduces a learning-based approach for optimal lane-changing control using real-time sensor data from a remote-controlled (RC) car. By employing an Adaptive Dynamic Programming (ADP) algorithm and integrating enhanced perception from GPS, IMU, and cameras interfaced with an Nvidia Jetson AGX Xavier board, the RC car can perceive its surroundings and make informed lane-changing decisions in the face of obstacles. The significance of this research lies in its rigorous experimental validation, emphasizing the practical applicability and robustness of our learning-based control algorithm. Our results confirm the algorithm's adaptability to parameter variations and its ability to execute lane-changing maneuvers in realistic scenarios, thereby bridging the gap between theoretical innovation and real-world application.

Keywords: Autonomous Vehicle Systems, Artificial Intelligence Systems, Robotic Applications
Abstract: Autonomous driving (AD) technology is shifting from traditional rule-based algorithms to data-driven end-to-end (E2E) algorithms to overcome their limitations in complex urban environments. However, E2E algorithms are highly sensitive to the training dataset and often suffer from causal confusion, which is learning spurious correlations, as well as data imbalance, both of which degrade generalization and safety. Moreover, since E2E algorithms typically rely on synchronized sequential inputs to process temporally continuous observations, they often require a predefined inference interval. This constraint limits control frequency and makes the algorithms vulnerable to asynchronous sensor inputs in real-time applications. To mitigate these issues, we propose a purpose-driven dataset construction strategy to alleviate causal confusion and data imbalance. We further develop an adaptive sequential inference framework that removes fixed timing constraints of sequential networks and enables real-time operation with asynchronous sensor inputs. Experiments across diverse scenarios, including downtown traffic and school zones, show robust edge-case performance, over 70% improvement in driving score compared to the baseline, and a second-place finish in the 2025 Hyundai Motor Group (HMG) AD Challenge conducted in real time on the MORAI digital-twin simulator. These results demonstrate that data-centric design combined with simulation based evaluation can substantially enhance the real-world deployability of E2E AD systems.

Keywords: Navigation, Guidance and Control, Artificial Intelligence Systems, Autonomous Vehicle Systems
Abstract: Recent years have witnessed a significant increase in demand for robots capable of autonomously performing object navigation tasks in cluttered indoor environments. Existing approaches to object navigation either focus solely on navigating to a single target object or adhere to a predefined sequence when multiple targets are involved. Both approaches often result in inefficient paths and low success rates. To overcome these limitations, we propose a modular framework for efficient Multi-object navigation that (1) dynamically optimizes the target visitation order and (2) adaptively adjusts path- planning costs based on the current exploration context. First, a pretrained semantic map-based inference network predicts spatial distributions of potential target locations. These predicted distributions are then combined with geodesic distances to calculate target values, enabling the agent to dynamically select the most beneficial next target. Navigation toward the chosen goal is executed by a Fast Marching Method planner operating on a cost map, whose costs are adaptively updated based on predicted probabilities and visitation history, thereby balancing exploration and efficient navigation. Experiments conducted on the HM3D datasets demonstrate that the proposed method significantly improves success rates and path efficiency compared to fixed-order baselines, showcasing its practical applicability to real-world robotic missions.

Keywords: Robotic Applications, Human-Robot Interaction, Exoskeleton Robot
Abstract: Assessing safety of robot systems is essential for their employment in real-world environments. Especially applications that involve humans must be certified with respect to strict safety requirements. While existing works in robotics literature presents various methods to circumvent dangerous events (e.g., collisions, singularities,...), it remains unclear to date textit{how} exactly a robot system must be evaluated to be considered as textit{safe}. As a logical consequence, the safety critical variables are not clearly identified, causing difficulties in deriving measures for safety assurance on technical level apart from collision and singularity avoidance. One possibility to address this issue lies in deriving a mapping between the robot system parameters and the safety requirements in the official standards. This paper specifically refers to the safety standards provided by the International Organization for Standardization (ISO) to derive an approach for safety evaluation of robot systems. Based on methods from prior work, this contribution provides one step towards the development of a generic safety evaluation technique of robot systems.

Keywords: Navigation, Guidance and Control, Artificial Intelligence Systems, Autonomous Vehicle Systems
Abstract: Object Navigation (ObjectNav) remains a central challenge in embodied AI—agents must interpret visual cues and natural-language goals to locate objects in previously unseen environments, balancing high-level reasoning with low- level control. While supervised and pretrained model-based methods have demonstrated great performance, they often struggle to prioritize candidate goals effectively at each decision step. To address this, we introduce two key contribu- tions. First, a confidence module that joins channel- and spatial-attention over semantic maps with CLIP-based goal-text embeddings and pose embeddings, and refines these joined features via cross-attention to score frontier candidates. Sec- ond, an exponential coverage reward function that promotes efficient exploration and stabilizes reinforcement-learning training. In our approach, channel attention aggregates per-channel statistics to recalibrate feature importance, spatial attention highlights critical regions, and pose and textual embeddings guide goal selection. A subsequent cross-attention layer produces confidence scores for frontier edges. Empirical evaluations on standard ObjectNav benchmarks demon- strate significant gains in Success Rate and Success weighted by Path Length over competitive baselines. Ablation studies confirm that both the attention-based fusion and exponential reward are pivotal to these improvements.

Keywords: Autonomous Vehicle Systems, Artificial Intelligence Systems, Robot Vision
Abstract: Traffic controllers play an important role in managing traffic flow at construction and accident sites. However, they are rarely encountered in typical driving environments. As a result, most existing road scene datasets contain very few traffic controllers. These individuals are usually labeled as ordinary pedestrians, rather than as a separate class. Consequently, autonomous driving systems trained on such datasets cannot distinguish traffic controllers from regular pedestrians, and are unable to respond appropriately in these special situations. To address traffic controller data scarcity, we propose a two-stage data augmentation framework. In the first stage, we introduce a Baton-aware Pose-guided Diffusion Model (B-PDM), which generates diverse images of traffic controllers by controlling both body pose and the pose of the baton. In the second stage, we propose a Human Height Estimation Module (HHEM) that predicts the appropriate height for each traffic controller based on the scene context. Using these predictions, we realistically paste the generated controllers onto various road background images. Our framework enables efficient creation of large-scale, realistic training data that includes traffic controllers. We expect that this approach will facilitate robust learning of traffic controller recognition in autonomous driving systems. Ultimately, our method has the potential to help autonomous vehicles better handle rare but critical traffic scenarios, and may contribute to progress toward fully autonomous driving.

Keywords: Control Devices and Instruments, Control Theory and Applications, Sensors and Signal Processing
Abstract: This paper proposes the Line of Sight (LOS) orientation cascade-type stabilization controller of a two-axis gimbal with a novel integrated coordinate approach, the Hybrid Coordinate System (HCS), which combines joint and inertial coordinate frames. A robust controller is designed using this system, employing feed-forward (FF) stabilization and a disturbance observer (DOB) for each joint within the new coordinate system to eliminate external disturbances, internal joint friction, coupling torque, and model uncertainty. Controller parameters and controllable frame designs are determined through HCS model with multi-sensor configuration. Additionally, a filter is designed to address the drift issue that may occur in the given system. To verify the proposed control algorithm, a controller analysis and comparison with other controllers are conducted through actual external environment experiments. The HCS approach showed improved LOS stabilization performance compared with other methods.

Keywords: Control Theory and Applications, Navigation, Guidance and Control, Artificial Intelligence Systems
Abstract: Generalized planning using deep reinforcement learning (RL) combined with graph neural networks (GNNs) has shown promising results in various symbolic planning domains described by PDDL. However, existing approaches typically represent planning states as fully connected graphs, leading to a combinatorial explosion in edge information and substantial sparsity as problem scales grow, especially evident in large grid-based environments. This dense representation results in diluted node-level information, exponentially increases memory requirements, and ultimately makes learning infeasible for larger-scale problems. To address these challenges, we propose a sparse, goal-aware GNN representation that selectively encodes relevant local relationships and explicitly integrates spatial features related to the goal. We validate our approach by designing novel drone mission scenarios based on PDDL within a grid world, effectively simulating realistic mission execution environments. Our experimental results demonstrate that our method scales effectively to larger grid sizes previously infeasible with dense graph representations and substantially improves policy generalization and success rates. Our findings provide a practical foundation for addressing realistic, large-scale generalized planning tasks.

Keywords: Exoskeleton Robot, Human-Robot Interaction, Robot Mechanism and Control
Abstract: This paper proposes a control method for power augmentation of an antagonistic pneumatic artificial muscle(PAM) system. While PAMs are a suitable actuator for exoskeletons due to their high power-to-weight ratio, they exhibit nonlinearity and model uncertainties. In order to achieve power augmentation effects using antagonistic PAM system, the system must be controlled in a way that minimizes contact force with the wearer and enhances robustness against disturbances and uncertainties. For this purpose, robust controller that can minimize contact force was developed. the contact force is measured through force/torque sensors for feedback control, and an observer is used to estimate the disturbance and uncertainties on the system. To apply this to PAMs, a simple force model of the pneumatic actuator is presented, and a torque model is derived for the joint of an antagonistic PAM system in which two PAMs are connected in parallel to enable bidirectional rotation. The proposed control algorithm was verified through one-degree-of-freedom test platform. Experiments confirmed that the power augmentation effect of antagonistic PAM system was achieved.

Keywords: Robotic Applications, Navigation, Guidance and Control
Abstract: Autonomous Mobile Robots (AMRs) are increasingly deployed in industrial settings, where they perform various tasks that replace human labor, thereby boosting operational efficiency. In such smart manufacturing environments where multiple types of robots and vehicles coexist, effective management of these dynamic obstacles during AMR navigation is crucial. However, many existing studies are limited to local path planning, focusing primarily on real-time recognition and evasion based on sensor data. In this study, we propose a time-aware costmap framework that leverages prior temporal information about areas frequently traversed by dynamic obstacles. By integrating a time-bound dynamic obstacles layer into the global costmap as a plugin, the proposed approach fits seamlessly into the existing ROS 2 navigation stack and achieves smoother and less disruptive navigation at the global path planning level. We validate the performance of the proposed framework in a Gazebo simulation environment modeled after a milk-run distribution system.

Keywords: Control Theory and Applications
Abstract: Bridge cranes are critical for industrial automation but suffer from vibrations that compromise safety and efficiency. Suppressing these vibrations remains a key challenge in crane control systems. This paper investigates the vibration suppression problem of the primary girder of a bridge crane. To address nonlinear time-varying actuator failures, we develop a novel adaptive fault-tolerant in-domain control scheme using the lower bound method, derived from the primary girder's partial differential equation (PDE) model. The proposed controller guarantees effective vibration suppression. Furthermore, Lyapunov stability analysis rigorously demonstrates the boundedness of all closed-loop signals. Numerical simulations validate the effectiveness of the proposed control scheme in suppressing vibrations under actuator failures. In addition, the method exhibits broad applicability to flexible mechanical systems with structural dynamics analogous to bridge crane girders.

Keywords: Control Theory and Applications
Abstract: This paper investigates global adaptive output-feedback stabilization for a class of nonlinear time-delay systems with function control coefficients. In contrast to previous studies, this paper allows the control coefficients to be output-dependent functions. Additionally, the derivative of the unknown time-varying delay is not required to have a known upper bound. To overcome this challenge, an adaptive output-feedback control strategy is proposed. In particular, a dynamic gain is designed to compensate for uncertainties from function control coefficients and unknowns in nonlinearities. An observer is then constructed to rebuild the unmeasured states, followed by the design of an adaptive output-feedback controller to ensure global stabilization of the closed-loop system. A simulation example is provided to demonstrate the effectiveness of the proposed control strategy.

Keywords: Control Theory and Applications
Abstract: This paper investigates global practical tracking via adaptive output-feedback for a class of uncertain nonlinear time-delay systems with function control coefficients. To address this issue, a novel high gain is introduced to overcome additional system nonlinearities and serious unknowns, and then an adaptive high-gain observer with appropriate design parameters is constructed to rebuild the system unmeasured states. Subsequently, an adaptive output feedback controller is designed to ensure that the tracking error can be adjusted to an arbitrarily small interval after a finite time, while keeping all closed-loop signals bounded. By utilizing the Lyapunov-Krasovskii functional, the time-varying delay phenomenon is successfully handled. Finally, a numerical example is provided to demonstrate the effectiveness of the results.

Keywords: Autonomous Vehicle Systems, Artificial Intelligence Systems, Civil and Urban Control Systems
Abstract: We introduce the multi-vehicle road graph delivery problem, aimed at solving real-time delivery tasks using autonomous vehicle fleets. Our approach utilizes road graph representations to accurately capture the characteristics of urban road networks. To address this problem efficiently, we propose a multi-agent reinforcement learning (MARL) framework incorporating an attention-based state encoder, which effectively encodes the road network structure and package information. Our modified implementation of a multi-agent Tsallis actor-critic (MATAC) algorithm, combined with the state encoder, is trained to collaboratively minimize delivery makespan using individualized rewards that encourage cooperative vehicle routing behaviors. Experimental results on multiple benchmark maps demonstrate that our algorithm significantly reduces the makespan more than heuristic approaches, while achieving inference times much faster than an exact optimization method with competitive solution quality. These results highlight the applicability of our method for large-scale real-time delivery scenarios involving autonomous vehicles.

Keywords: Information and Networking, Control Theory and Applications
Abstract: This extended abstract summarizes a novel CPS attack named disturbance-induced denial-of-service (DIDoS). Unlike traditional DoS, which overwhelms network resources via external flooding, DIDoS exploits event-triggered control (ETC) in physical systems. A carefully crafted disturbance causes excessive event-triggered transmissions, saturating the shared network and destabilizing all plants connected to it. We describe ETC models, derive stability conditions under time-varying network delays, and validate the destabilizing effect of DIDoS with IEEE 802.11-based simulations. Results show that even a small number of compromised physical systems can saturate the network and cause instability for all.

Keywords: Control Theory and Applications
Abstract: This paper presents the development and validation of a sampled-data control framework for precise speed regulation of a brushed DC (BDC) motor coupled with a generator load. A detailed mathematical model of the coupled motor-generator system is established, incorporating PWM switching noise and parameter uncertainties to reflect real-world physical dynamics and provide a practical foundation for controller design. To achieve accurate control performance on a digital control platform, a sampled-data controller is designed based on a lightweight looped functional that nonetheless retains a new cross term. The effectiveness of the proposed controller is validated through simulation results using Simscape, a physical modeling tool, which demonstrates its feasibility in sampled-data control environments for coupled electromechanical systems.

Keywords: Control Theory and Applications
Abstract: This paper proposes a robust control framework using a disturbance observer (DOB) based on a linear dynamic model for a hybrid gantry stage exhibiting Linear Parameter Varying (LPV) characteristics due to its asymmetric structure and varying mass distribution. The system includes rigid and compliant actuators in parallel along the Y-axis, forming an asymmetric Multi-Input Multi-Ouput (MIMO) configuration, and its dynamics change with the position of the payload on the crossbeam. Frequency response functions are measured under various payload conditions, and Canonical Polyadic Decomposition (CPD) is applied to extract a nominal model commonly applicable across conditions. The DOB designed based on this model satisfies robust stability under different loads, verified through theoretical and simulation analyses. This study provides a practical methodology for disturbance rejection in LPV MIMO systems with complex structures.

Keywords: Artificial Intelligence Systems
Abstract: Existing autonomous robots rely on LiDAR and cameras to perceive their surroundings, but they remain vulnerable to occluded threats such as obstacles behind corners. Therefore, auditory perception is essential for safe operation in complex environments. Prior sound event localization and detection (SELD) approaches have employed First-Order Ambisonics with four microphones (FOA) with Transformer or RNN-based models. However, these architectures require large parameter counts and incur high latency, making them unsuitable for real-time operation. In this paper, we propose MMT-Net, a lightweight SELD model that uses only two microphones and integrates a Transformer-based front-end with a Multi-scale Modern Temporal Convolutional Network (MM-TCN) back-end. Compared to Conformer-based models, MMT-Net reduces the parameters count by ≈ 7× (to ∼ 2M) and achieves ≈ 3× faster inference speed with only a 1% drop in F1-score. Furthermore, on NVIDIA Jetson Xavier NX/Orin AGX, MMT-Net runs at over 20 fps, demonstrating real-time capability on embedded platforms.

Keywords: Autonomous Vehicle Systems
Abstract: Precise longitudinal control in autonomous driving has been challenged by abrupt actuator dynamics changes. While Cooperative Adaptive Cruise Control (CACC) leverages vehicle-to-vehicle communication to improve upon Adaptive Cruise Control (ACC), it struggles with varying intrinsic parameter due to vehicle and environmental variation. Existing methods treat actuator behavior as static, relying on disturbance estimation, leading to jerky motion. This paper proposes a novel endogenous uncertainty compensator approach using self-adaptation feedforward controller with nonlinearly estimated response parameter in real-time. CACC Simulation results showed up to 96 [%] spacing error reduction, and accurate actuator parameter estimation, demonstrating strong potential for improving stability and safety in CACC.

Keywords: Robot Vision, Sensors and Signal Processing, Robotic Applications
Abstract: A major obstacle to the development of effective monocular depth estimation algorithms is the difficulty in obtaining high-quality metric depth data that corresponds to real-world RGB images. Collecting this data is time-consuming and costly, and even data collected by modern sensors has limited range or resolution, and is subject to inconsistencies and noise. Data generated in simulation avoids these problems with accurate depth information, but models trained on synthetic data often do not transfer well to real world applications. To combat this, we propose a method of data generation in simulation using 3D synthetic environments and CycleGAN domain transfer to increase the realism of simulated images. We analyze this data generation method by training multiple depth estimation models on different datasets, including synthetic and domain-transferred data. We evaluate the performance of the models on the NYU-Depth V2 dataset to verify the generalizability of the approach and show that GAN-transformed data effectively helps to bridge the gap between simulated and real-world data in depth estimation.

Keywords: Autonomous Vehicle Systems, Information and Networking
Abstract: Digital twin technology enables realistic, repeatable testing with minimal cost and enhanced safety. It supports cost-effective scenario exploration and systematic validation of autonomous systems. It bridges the reality gap by synchronizing physical platform data with its simulation counterpart in real time. This paper presents a low-cost digital twin framework integrating the open-source CARLA simulator with a vehicle-like US-platform. Two interconnected PCs exchange motion, actuator, and sensor data to achieve bidirectional communication. US-platform captures GPS-IMU fused odometry, steering angle, speed, brake pressure, and gear status. Digital twin framework could support waypoint-based autonomous guidance and remote control to handle edge-case failures.

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Sensors and Signal Processing
Abstract: This paper presents a practical approach for deploying an autonomous valet parking (AVP) system. The system uses low-cost hardware including a solid-state LiDAR and a GNSS RTK receiver. Point cloud mapping is conducted using a SLAM framework with scan matching and pose optimization. A vector map with semantic labels is built on top of the aligned 3D point cloud map. Coordinate registration is performed using GNSS-based MGRS transformation and ICP alignment. The mapping result shows sufficient accuracy for AVP localization requirements in real-world conditions. Autoware Universe is utilized to integrate mapping, localization, and path planning components. A parking trajectory is generated using the Hybrid A* algorithm in a simulated parking lot. Simulation results demonstrate stable planning performance without collision in narrow spaces. The proposed method reduces system cost while maintaining practical feasibility for deployment. Future work includes integration with object detection and dynamic behavior planning modules.

Keywords: Robot Mechanism and Control, Robotic Applications
Abstract: This paper presents the design and implementation of a multi-functional logistic robot. The robot has two wheels to balance and move with two arms to handle an object. The robot can transform Segway mode into bicycle mode, or vice versa. Two wheels can control balancing motion in the Segway mode and reaction wheels can control balance in the bicycle mode. Multi-functional design has been demonstrated by implementing the logistics robot.

Keywords: Navigation, Guidance and Control, Control Theory and Applications, Industrial Applications of Control
Abstract: This study proposes a path tracking control method to enhance the docking precision of autonomous mobile robots (AMRs) operating in constrained environments. The proposed controller builds upon a previously developed smooth law control scheme, and its performance is improved by automatically adjusting the directional and lateral error gains through Bayesian Optimization, according to the curvature characteristics of the trajectory. Through this approach, the controller is designed to maintain high tracking accuracy even under curvature constraints during curved-path docking. Furthermore, the proposed controller is enhanced via integration with PID control, enabling efficient convergence of heading and lateral errors in segments with high curvature variation, while maintaining overall system stability. Experimental validation conducted in a narrow, factory-like environment demonstrates that the proposed method outperforms conventional PID, smooth law-based control approaches, and Model Predictive Path Integral (MPPI) control approaches in terms of docking accuracy, robustness to curvature variation, and control reliability. The proposed controller achieved an average final position error of 3.209 mm and an average final yaw error of 0.0072 rad. These results indicate that the controller effectively extends the straight-line segment before docking even in high-curvature scenarios, thereby significantly improving final alignment precision and achieving stable path convergence without collisions in confined spaces. These findings suggest that the proposed control method offers a practical and reliable solution for high-precision docking in industrial environments.

Keywords: Artificial Intelligence Systems
Abstract: To find more suitable algorithms and models for heart disease prediction, this paper uses seven machine learning algorithms, including Naive Bayes, logistic regression, support vector machine, random forest, k-nearest neighbors, neural network, and XGBoost, to establish training models, and sets corresponding parameters for each algorithm. These models perform binary classification on test samples to predict whether the samples have heart disease. Based on various evaluation indicators, the prediction effects of these models are compared. The results show that models trained by support vector machine and k-nearest neighbors algorithms have the best performance, and the evaluation indicators of the remaining models are also favorable, demonstrating that machine learning algorithms are effective for heart disease prediction. The Bayesian optimization method is used to search for optimal parameters to optimize the XGBoost model, and a comparative analysis is conducted with the preoptimized model. The results show that the prediction performance of the model is significantly improved after optimization.

Keywords: Autonomous Vehicle Systems, Artificial Intelligence Systems
Abstract: As autonomous driving systems become more prevalent, ensuring the safety of pedestrians is a critical challenge that must be addressed. One of the most crucial components of pedestrian safety systems is the ability to estimate pedestrian trajectories accurately in the vehicle coordinates. However, existing approaches primarily concentrate on 2D image-plane bounding-box forecasts, which are often inadequate for real-world applications. Other methods rely on expensive 3D sensors such as LiDAR, which hinders their commercialization. Some monocular camera-based approaches have explored 3D localization, but most remain limited to frame-level estimates, lacking the temporal continuity needed for reliable trajectory analysis. To overcome these limitations, this paper introduces EPIC: Ego-centric Pedestrian Trajectory Estimation via Instance-aware Center-focused Depth Processing, a cost-efficient system for estimating 3D pedestrian trajectories using only a monocular camera. This approach integrates instance segmentation, object tracking, monocular metric depth estimation, and a novel adaptive center-focused depth extraction with Gaussian-weighted smoothing, enabling trajectory estimation in the vehicle coordinate frame. This paper evaluates the proposed system on the nuScenes dataset, demonstrating the feasibility of monocular trajectory estimation for pedestrian safety applications.

Keywords: Artificial Intelligence Systems, Navigation, Guidance and Control, Control Theory and Applications
Abstract: Proximal Policy Optimization (PPO) trains a stochastic policy, yet evaluation is often performed deterministically by taking the argmax of the policy's output. This paper empirically analyzes the performance discrepancies between stochastic and deterministic evaluation across four Atari games. Our findings reveal that the choice of evaluation protocol significantly impacts outcomes and is environment-dependent. Stochastic evaluation can yield higher mean rewards in complex environments like Breakout, where its inherent exploration helps avoid suboptimal greedy paths. Conversely, deterministic evaluation achieves superior and perfectly consistent performance in environments like Pong where a stable optimal policy is learned. We show that stochastic evaluation consistently produces higher reward variance, driven by its sampling of lower-probability actions. These results underscore that deterministic evaluation may not fully capture the robustness of a trained policy, and a combined evaluation approach offers a more comprehensive assessment.

Keywords: Robot Mechanism and Control, Robotic Applications, Autonomous Vehicle Systems
Abstract: Recent advances in mobile robotics have underscored the need for adaptable locomotion systems capable of traversing complex and unstructured terrains. While wheeled robots provide high speed and efficiency on flat surfaces, their mobility is limited in irregular environments. Conversely, legged robots offer superior terrain adaptability but involve greater mechanical and control complexity. To address these trade-offs, we present a novel quadruped robot equipped with a transformable wheel-leg mechanism featuring a continuous, curvature-optimized structure. This design preserves wheel integrity during transformation, enabling both stable rolling and robust quadrupedal locomotion. Each limb module operates compactly with only two actuators. A depth-based perception system and lightweight terrain classification algorithm allow autonomous mode switching between wheel and leg configurations in real time. Experimental results demonstrate the robot’s ability to navigate uneven terrain with improved stability and adaptability, highlighting its potential for future applications in multi-modal locomotion and manipulation.

Keywords: Control Theory and Applications, Industrial Applications of Control
Abstract: Regulating the liquid level within tanks is a common necessity across various industrial sectors, including chemical processing and oil refining. Maintaining the liquid level at a desired value is crucial for efficient operation. Therefore, it is necessary to design controllers that can reduce error by tuning their parameters. This study presents a tuned PID controller based on a hybrid Particle Swarm Optimization and Gravitational Search Algorithm (PSOGSA) to control liquid level regulation in a tank. A mathematical model of the water tank level is developed using mass balance principle. The controller performance is evaluated using Integral Square Error (ISE), Integral of Absolute Error (IAE), Integral Time Absolute Error (ITAE) and Mean Squared Error (MSE) and are considered as objective functions for the PSOGSA. The control system is simulated using MATLAB, and the simulation results are compared. The results shows that the Hybrid Particle Swarm Optimization and Gravitational Search Algorithm (PSOGSA) effectively controls the liquid level system. Notably, the PID controller parameters tuned technique yields superior performance.

Keywords: Robotic Applications, Robot Vision, Artificial Intelligence Systems
Abstract: This paper presents Level-LOC, a Loop-Optimized GNSS–Visual–Inertial SLAM Using Multi-Height Se- mantic Building Contour features to improve loop closure robustness in large-scale urban environments. Building and ground regions are segmented using YOLOv11, and stereo depth is fused to generate a pseudo-LiDAR point cloud. Static contour points are sampled at fixed height intervals from 1 m to 5 m above the estimated ground plane and projected into birds-eye-view contour layers. During loop closure, a contour-based geometric verification step filters DBoW3 visual loop candidates, rejecting false positives caused by dynamic objects and illumination changes. On a 1 km handheld urban dataset, the proposed method reduces the false positive rate from 6.41 % to 2.56 % (a 60 % reduction) and increases precision from 93.59 % to 97.44 %. These results demonstrate that integrating structured semantic contours significantly enhances loop detection accuracy in outdoor SLAM.

Keywords: Control Theory and Applications, Navigation, Guidance and Control
Abstract: This paper investigates the distributed leader-following consensus problem of multi-agent systems with Lipschitz nonlinear dynamics via an event-triggered control approach. To address this problem, a fully distributed adaptive event-triggered control strategy is proposed, which consists of a distributed adaptive control law and a novel event-triggering mechanism. The control law incorporates adaptive coupling gains for both the leader and the followers, and the event-triggering mechanism ensures strictly positive minimum inter-event times to exclude Zeno behavior. It is shown that under the proposed control strategy, the states of all agents asymptotically reach consensus with the leader, and the minimum inter-event times are guaranteed to be strictly positive. The effectiveness of the proposed strategy is demonstrated through simulations.

Keywords: Robot Mechanism and Control, Artificial Intelligence Systems, Robotic Applications
Abstract: Stable locomotion of humanoid robots in complex and unstructured terrains remains a significant challenge, primarily due to their intricate and highly coupled dynamics, as well as the inherent sim-to-real gap that complicates direct transfer of simulated policies to physical robots. This paper introduces a novel Motor-Guided Randomization Reinforcement Learning (MGRL) framework, specifically designed to substantially improve both the robustness and environmental adaptability of bipedal locomotion. At its core, MGRL incorporates an adaptive noise perturbation during reinforcement learning training. This perturbation, dynamically adjusted based on the robot's motor output torque, effectively simulates the inevitable uncertainties and non-idealities present in real-world motor performance, such as friction, backlash, and varying efficiency. By systematically exposing the policy to these physics-level non-idealities during training, MGRL significantly enhances the robot's walking stability, disturbance rejection capabilities, and overall task success rate across a wide array of challenging terrains, including steep slopes, irregular steps, and highly uneven surfaces. Furthermore, this principled approach not only mitigates issues arising from discrepancies between simulated and real-world physics but also prevents the policy from over-adapting to overly idealized or unrealistic disturbances, leading to more transferable and reliable behaviors.

Keywords: Control Theory and Applications
Abstract: This paper employs an integral-based event-based control strategy to address the consensus problem of Lipschitz nonlinear multi-agent systems. Assuming that the communication topology contains a directed spanning tree with the leader as the root, the controller for each agent is updated only when its event conditions are satisfied. Under the premise that the assumptions hold, we first use the integral inequality method to solve the consensus problem of leader-following nonlinear multi-agent systems. Then, introducing the combined states of agents, where each agent samples the relative information with its neighbors, the consensus problem of such systems is solved by the Lyapunov function method. Theoretical analysis shows that under this integral-based event-based control strategy, the leader-following multi-agent system can be guaranteed to achieve consensus, and the event interval has a lower bound. Simulation results verify the effectiveness of the obtained conclusions.

Keywords: Control Devices and Instruments
Abstract: This paper investigates the impact of temperature variations on the failure rates and lifetime prediction of FPGA and CPU components in real-time Ethernet-based semiconductor control systems. Temperature-induced stress can degrade the performance and reliability of these critical components, leading to system failures and increased maintenance costs. Using data collected from FDC (Fault Detection & Classification) systems, this study applies predictive maintenance models, including Arrhenius and Weibull distributions, to analyze failure patterns. Furthermore, AI-based machine learning approaches such as LSTM, Random Forest, and XGBoost are utilized to enhance lifetime prediction accuracy. Experimental validation is performed through thermal stress simulations, real-time monitoring of industrial semiconductor equipment, and finite element analysis (FEA) for thermal behavior evaluation. The findings contribute to optimizing semiconductor manufacturing control systems by providing a robust predictive maintenance framework.

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Robotic Applications
Abstract: This paper introduces safe segments forwarding-based exploration (SSF-Exploration), a novel exploration approach that enhances efficiency by significantly reducing idle time and unnecessary movements for unmanned aerial vehicles (UAVs). Unlike traditional naive frontier exploration approaches, which frequently stop and hover to replan paths upon detecting collisions from new sensor scan data, SSF-Exploration tracks the safe segments of the previously computed path up to the last collision-free point and seamlessly replans a new path from that point to the goal position. Through comprehensive simulations conducted in a complex maze environment, SSF-Exploration demonstrated enhanced exploration performance by substantially reducing idle time by approximately 44.15%, shortening traveled distances, and increasing average exploration speeds compared with naive frontier exploration.

Keywords: Industrial Applications of Control, Control Theory and Applications
Abstract: This paper addresses the control problem of a spinning flexible cantilever beam attached to a translating base. When the beam spins, both transverse and lateral vibrations occur simultaneously, even without external forces acting along the lateral axis. A dynamic model for the system is established based on Hamilton’s principle. The developed model analysis indicates that the dynamic behavior of lateral and transverse vibrations in a spinning beam is coupled. Consequently, it is feasible to apply an input shaping control technique to suppress transverse vibrations, which also suppresses lateral vibrations. A control strategy that integrates a PD controller with input shaping techniques is developed to ensure precise spindle positioning and vibration suppression. The effectiveness of this control approach is verified through simulations performed in MATLAB.

Keywords: Robot Vision, Sensors and Signal Processing, Autonomous Vehicle Systems
Abstract: Stereo‐matching networks trained exclusively on synthetic imagery suffer severe accuracy drops in real-world scenes because of the domain shift. This paper introduces a three-stage, label-free adaptation pipeline that closes this gap without any target-domain depth ground truth. (i) Entropy-filtered pseudo labels supervise only the most reliable pixels, (ii) entropy minimization sharpens the entire cost volume, and (iii) an epipolar-aware contrastive learning suppresses sensitivity to color while reinforcing geometry-aware distinctiveness. Applied to FastACVNet[1], this method reduces KITTI-2015[18] D1-all and EPE, yet still runs in real time at 15 FPS on a single RTX 3090. The approach requires no image translation, extra labels, or runtime overhead, providing a plug-and-play upgrade for stereo perception in autonomous-driving and robotics applications.

Keywords: Sensors and Signal Processing, Biomedical Instruments and Systems
Abstract: This study investigates whether biomimetic temporal encoding of electrotactile stimuli can reduce tingling sensations and enhance artificial pressure perception. Providing intuitive artificial tactile sensations is critical in mixed reality (MR) applications, where electrotactile stimulations have emerged as a promising means. Yet, electrotactile pressure sensations often accompany unnatural tingling, particularly at high stimulation frequencies. We hypothesized that mimicking the spike patterns of cutaneous afferents may mitigate tingling by reducing temporal incongruency in peripheral neuronal responses. To test this, we developed two biomimetic stimulation patterns based on physiological models and compared them with a conventional control stimulus. Six participants received electrotactile stimuli on the fingertip via surface electrodes and reported their perceived tingling-to-pressure ratio and dominant sensation. Results showed that the second biomimetic pattern, which incorporated adaptation observed in slowly adapting type I (SA-I) afferent, elicited more perceptual responses of pressure as the dominant sensation, while reducing tingling. These findings suggest that biomimetically patterned electrical stimulation can improve tactile perceptual quality.

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Robotic Applications
Abstract: This paper presents a cooperative localization method for a system composed of an autonomous surface vehicle (ASV) and an autonomous underwater vehicle (AUV). While the ASV can perform accurate localization using GNSS and inertial sensors, the AUV relies on dead-reckoning, which leads to unbounded error accumulation over time. To overcome this limitation, a cooperative approach is proposed where the AUV receives relative range and bearing measurements from the ASV via acoustic communication. These measurements are integrated into the AUV’s localization framework using an extended Kalman filter (EKF), enabling correction of its position estimates. The proposed method is implemented using an inverted-USBL configuration for acoustic sensing and was validated through an inland water experiment with real sensor-equipped platforms. Experimental results confirm that the cooperative localization system effectively constrains position error growth and enhances the reliability of underwater navigation.

Keywords: Autonomous Vehicle Systems, Human-Robot Interaction, Control Theory and Applications
Abstract: This paper proposes a robot navigation algorithm for human-following based on trajectory forecasting. In home environments, the target-following task is becoming increasingly important for achieving effective human-robot interaction. However, vision-based human-following methods often suffer from the presence of obstacles. Therefore, we employ EKF-based trajectory forecasting to enable the robot to follow the person when they disappear from the field of view. Based on the occupancy grid map in SLAM, the mobile robots determine the goal location for human-following. When the target remains in vision, the home robots move a location 1 meter in front of the target's position as its goal point. On the other hand, one of the reasonable point in trajectory forecasting is selected when the target disappears from the field of view.

Keywords: Human-Robot Interaction, Sensors and Signal Processing, Biomedical Instruments and Systems
Abstract: Gait analysis is vital because deviations from normal walking patterns can indicate potential abnormalities or diseases. In the past, gait evaluation relied on visual observation; however, due to low inter-rater reliability and a heavy dependence on clinical expertise, instrumented gait analysis has been proposed as an alternative. Various systems have been developed and studied to measure and analyze gait parameters quantitatively. However, well-known commercial gait analysis systems with high accuracy are often expensive or require the attachment of sensors to the body. To address these limitations, vision-based systems have been introduced; however, they tend to show reduced accuracy in certain parameters such as step length and step width. In this study, we propose a tactile sensor-based treadmill system that was specifically developed to enable low-cost measurement of spatial gait parameters without the need for any wearable sensors. To validate the accuracy of the proposed system, a pilot test was conducted using an infrared marker-based motion capture system for comparison. Experimental results demonstrate that our system achieved improved performance in a step width measurement compared to previous studies. In future work, we aim to validate the system with a larger participant pool and enhance its precision by integrating methods such as artificial intelligence or multi-sensor fusion.

Keywords: Human-Robot Interaction, Robot Vision, Robotic Applications
Abstract: Robot manipulation has attracted considerable attention in the development of robotic intelligence due to its high applicability in real-world industries. In particular, the importance of visual feedback tele-operation is increasingly emphasized in robot manipulation because of the potential hazards in industrial environments. Moreover, it can be employed for data collection to facilitate the development of imitation learning. In this paper, we propose a visual feedback tele-operation approach for robot manipulation using 3D printing technology. By sharing the robot’s view through visual SLAM and employing ROS2, the actual robot manipulation can achieve in a manner that replicates the movements of the 3D-printed object.

Keywords: Robotic Applications, Industrial Applications of Control
Abstract: This paper introduces a framework for projecting predefined 2D path patterns from a unit square onto 3D single-patch surfaces. By leveraging As-Rigid-As-Possible parameterization, the method ensures low geometric distortion flattening of 3D meshes onto 2D domains. A bilinear Coons patch interpolation establishes a smooth path mapping between square and the parameterized boundary. The 2D path pattern is then accurately mapped back to the 3D surface using barycentric coordinate conversion. The proposed approach enables geometry-aware trajectory generation by mapping 2D path patterns onto 3D surfaces in a way that conforms to the underlying geometry and distributes the pattern symmetrically with respect to four user-defined control points. We validated the effectiveness of the generated trajectory using robotic simulation.

Keywords: Navigation, Guidance and Control, Control Theory and Applications
Abstract: Heterogeneous autonomous small aerial vehicles are gradually becoming a major component of the low-altitude economy. This study addresses the problem of cooperative trajectory planning and optimization for heterogeneous aerial vehicle systems operating under cooperative control frameworks. Focusing on cooperation between fixed-wing and quadrotor platforms, a trajectory generation method is developed based on quaternion-represented Pythagorean hodograph curves, which explicitly accounts for the distinct dynamic constraints inherent to each vehicle type. To enhance robustness under uncertainty, a risk-averse optimization framework is introduced to model constraints on the tracking state and control variables of aerial vehicles. A dynamic risk measure model is proposed to quantitatively capture the uncertainty and constraint violations associated with the dynamics of heterogeneous systems, serving as a foundation for risk-aware trajectory optimization and control. These constraints are integrated into a model predictive control architecture to achieve optimal and reliable cooperation. The effectiveness and practical applicability of the proposed approach are validated through a series of experiments involving a five-agent heterogeneous aerial vehicle formation.

Keywords: Artificial Intelligence Systems
Abstract: Accurate and real-time lane detection is essential for autonomous driving systems. While anchor-based methods such as LaneATT have demonstrated strong performance in both speed and accuracy, they often lack spatial understanding that could be provided by dense semantic segmentation. Conversely, segmentation-based approaches struggle to capture instance-level lane structures. Notably, enhancing vectorized lane detection with auxiliary segmentation supervision has not been widely addressed. Therefore, in this work, we propose a dual-task extension to LaneATT by integrating an auxiliary segmentation head. Our architecture jointly learns vectorized lane representations and pixel-wise lane masks using a shared backbone. It is a hybrid vectorized lane detection model that incorporates auxiliary semantic segmentation to guide spatial reasoning in challenging road scenarios. Experiments on the TuSimple dataset demonstrate that our method improves lane detection accuracy and robustness while preserving real-time performance. Ablation studies further validate the effectiveness of different backbones, segmentation heads, and loss weights. Our proposed approach maintains real-time performance while offering improved spatial consistency.

Keywords: Control Theory and Applications, Information and Networking, Navigation, Guidance and Control
Abstract: This paper addresses the problem of locating and suppressing a passive eavesdropper in a multi-agent system (MAS) with unknown adversary position. Unlike existing works that assume the eavesdropper's location is known or detectable, we consider a more realistic scenario where the agents must infer the eavesdropper's position based solely on the intensity of intercepted signals. To exploit the distance-dependent nature of signal strength, we develop a distributed unscented Kalman filter (DUKF) that enables each agent to estimate the eavesdropper's position using noisy, partial observations and local communication. Once localization is achieved, a control strategy based on the separation principle is applied to guide agents in encircling the estimated position cooperatively. The entire framework is fully decentralized, scalable, and suitable for operation in adversarial environments. Simulation results using a satellite formation system verify the effectiveness of the proposed approach in achieving accurate localization and coordinated encirclement under measurement uncertainty.

Keywords: Autonomous Vehicle Systems
Abstract: Occupancy Network (OCC) has been widely applied to 3D scene perception in autonomous driving, owing to their capability of jointly modeling geometric and semantic information. However, constructing large-scale OCC maps is often time-consuming. To improve mapping efficiency, the collaboration of multiple agents becomes crucial. Nevertheless, limited inter-agent communication bandwidth and the large data volume of dense 3D scene representations have become major bottlenecks for multi-agent collaborative OCC mapping. To address these challenges, this study investigates the design of a communication-efficient multi-agent mapping framework. Specifically, we 1) propose a novel multi-agent 3D OCC mapping framework with high communication efficiency; 2) introduce a semantic topological map method to assist overlap scene detection, improving the detecting success rate while requiring only minimal extra communication; 3) present an efficient transmission strategy that significantly reduces communication cost through the transmission of low-resolution images in place of voxelized OCC representations. Experiments on autonomous driving datasets validate the feasibility of the proposed framework and demonstrate its potential for multi-agent collaborative 3D dense scene mapping.

Keywords: Control Theory and Applications, Robot Mechanism and Control
Abstract: This paper introduces a tracking algorithm specifically designed for constrained Euler-Lagrange systems (ELS). To address the constraint issues associated with position tracking errors and control torque, we incorporate state-dependent nonlinear functions and hyperbolic tangent functions. The sliding-mode control and dynamic surface control methods are employed to rigorously ensure that both position error and control torque remain within predefined, time-varying constraints. As a result, the closed-loop ELS achieves semi-global practical stability. Additionally, by properly selecting control parameters and constraints, we can guarantee the steady-state and transient behavior of the position tracking error. Theoretical findings are corroborated through both numerical simulations and experimental validations.

Keywords: Biomedical Instruments and Systems, Artificial Intelligence Systems
Abstract: Cervical cancer is the fourth leading cause of cancer-related deaths among women worldwide, with a particularly high burden in low-income countries. To address the need for cost-effective and accessible screening solutions, we propose a novel Vision Transformer (ViT)-based model for classifying cervical cancer from tele-cervicography images. To enhance classification performance, the model automatically extracts clinically meaningful anatomical features—such as the cervical region, external os location, and acetowhite lesion distribution—and integrates them into a dual attention bias mechanism that directs the model’s focus to key diagnostic areas. This mechanism emphasizes both the global cervical area and the spatial relationship between acetowhite lesions and the external os, enabling the model to attend to clinically significant regions more effectively. Trained on a dataset of over 17,000 tele-cervicography images, the proposed approach achieved a 2.04% improvement in accuracy, a 4.89% increase in specificity, and a 4.07% improvement in precision compared to the baseline ViT, while maintaining comparable sensitivity. Furthermore, the automated preprocessing pipeline enables fully automated clinical workflows, offering a scalable diagnostic mechanism suitable for real-world medical applications.

Keywords: Navigation, Guidance and Control
Abstract: Target localization is a critical task in various applications, such as search and rescue, surveillance, and wireless sensor networks. When a target emits a radio frequency (RF) signal, spatially distributed sensors can collect signal measurements to estimate the target’s location. Among various measurement modalities, received signal strength (RSS) is particularly attractive due to its low cost, low power consumption, and ease of deployment. While particle filtering has previously been applied to RSS-based target localization, few studies have systematically analyzed its performance under varying sensor geometries and RSS noise levels. This paper addresses this gap by designing and evaluating a particle filtering algorithm for localizing a stationary target. The proposed method is compared with a conventional RSS-based trilateration approach across different sensor configurations and noise conditions. Simulation results indicate that particle filtering provides more accurate target localization than trilateration, particularly in scenarios with unfavorable sensor geometries and high RSS noise.

Keywords: Control Theory and Applications, Navigation, Guidance and Control, Robot Mechanism and Control
Abstract: This paper presents a trajectory-tracking control framework that integrates a data-driven error dynamics model—learned from demonstration data via a Neural Ordinary Differential Equation–Recurrent Neural Network (NODE–RNN) architecture—with a nonlinear model predictive control (NMPC) scheme. Conventional NMPC relies on fixed analytical error models that are inflexible to environmental variations, vulnerable to disturbances, and sensitive to model uncertainty. To address these limitations, we integrate a recurrent neural network (RNN) that preserves temporal information in its hidden states and captures nonlinear dynamics with a Neural ODE, which models the continuous‐time error derivatives from those hidden states via Euler integration. The resulting learned model replaces the conventional analytical error dynamics within the NMPC predictive function. Through simulation studies on demonstration datasets with diverse distributions, the proposed NODE–RNN–NMPC framework is shown to converge to the desired trajectory more rapidly and precisely than conventional baseline controllers

Keywords: Autonomous Vehicle Systems, Navigation, Guidance and Control, Sensors and Signal Processing
Abstract: With continuous advances in ocean robotics and sensing technologies, the use of multiple underwater vehicles has attracted increasing attention in various scientific and engineering applications. Since GPS signals are not available under the water surface, navigation for multiple underwater vehicles typically relies on relative measurements between each vehicle and a GPS-equipped surface vehicle acting as a reference beacon. This study presents the design of a tracking filter for estimating the trajectories of multiple underwater vehicles in 3D space using such relative information. In particular, a joint probabilistic data association filter (JPDAF) is employed to achieve robust trajectory estimation in the presence of cluttered measurements caused by sensor noise, reflections from physical objects, or environmental disturbances. The performance of the JPDAF is evaluated and compared with that of a tracking filter based on the extended Kalman filter with using a nearest-neighbor data association approach through numerical simulations.

Keywords: Robot Mechanism and Control, Control Theory and Applications
Abstract: Cable-suspended aerial systems enable drones to carry heavier payloads by offloading the weight onto a supporting cable, but they introduce complex cable dynamics and the risk of excessive cable tension. This paper proposes a real-time tension estimator that leverages a nonlinear disturbance observer (NDOB) framework for modeling and tracking cable tension as an external disturbance. We then integrate the estimated tension into a tension-estimation-based control (TEC) scheme. The TEC compensates for unwanted lateral tension disturbances, preserves beneficial vertical support, and ensures that cable tension remains within safe limits. Simulation results demonstrate that the estimator accurately tracks true cable tension, and that the TEC eliminates steady-state position and attitude errors while keeping tension below the safety threshold. The proposed approach accurately estimates cable tension, thereby simplifying the control of cable-suspended aerial systems and enhancing both regulation accuracy and system safety, all without additional sensors.

Keywords: Control Theory and Applications, Artificial Intelligence Systems, Autonomous Vehicle Systems
Abstract: This paper presents a hierarchical control architecture that resolves the critical trade-off between landing accuracy and fuel consumption for autonomous powered descent. We compare a proposed controller, which integrates a Reinforcement Learning (RL) based guidance policy with a Sliding Mode Controller (SMC) tracker, against a conventional SMC baseline. The objective is to leverage the high robustness of SMC while using a learned policy to improve both accuracy and fuel efficiency. Through over 500 Monte Carlo simulations with system uncertainties, we demonstrate that the hierarchical controller improves landing accuracy by over 98% while reducing fuel consumption by approximately 13%. These findings demonstrate that integrating learning-based guidance with classical robust controllers offers a promising path toward developing highly efficient and reliable autonomous systems.

Keywords: Navigation, Guidance and Control, Sensors and Signal Processing, Autonomous Vehicle Systems
Abstract: As the vulnerabilities of global navigation satellite systems (GNSS) have become more widely recognized, the need for complementary navigation systems has grown. Medium frequency ranging mode (MF R-Mode) has gained attention as an effective backup system during GNSS outages, owing to its strong signal strength and cost-effective scalability. However, to achieve accurate positioning, MF R-Mode requires correction for the additional secondary factor (ASF), a propagation delay affected by terrain. The temporal variation of ASF, known as temporal ASF, is typically corrected using reference stations; however, the effectiveness of this method decreases with distance from the reference station. In this study, we analyzed the correlation between temporal ASF and meteorological factors to evaluate the feasibility of predicting temporal ASF based on meteorological factors. Among these factors, temperature and humidity showed significant correlations with temporal ASF, suggesting their potential utility in ASF correction.

Keywords: Control Theory and Applications, Robot Vision, Robotic Applications
Abstract: Venous insufficiency in the lower limbs is a disease in which the veins have difficulty returning blood to the heart, and as the condition worsens, it can significantly adversely affect daily life. Ultrasound examination is a widely used approach due to its non-invasive nature, but conventional manual scanning is highly dependent on operator, resulting in inconsistent image quality and poor reproducibility. To address this issue, we first analyze the image similarity between a reference image and an image acquired during re-examination at the same contact point and evaluate image reproducibility using the Structural Similarity Index (SSIM). Subsequently, we propose an algorithm that optimizes the probe orientation in real-time to enhance the SSIM between reference image and current one. Discrepancies between reference and follow-up images, caused by factors such as contact variation or patient movement, are compensated by adjusting the roll and pitch angles of the probe using a gradient ascent method based on SSIM. The proposed method is expected to obtain more consistent ultrasound image, leading to reliable follow-up diagnosis by monitoring the target vessels or tissues within the image.

Keywords: Sensors and Signal Processing, Rehabilitation Robot, Exoskeleton Robot
Abstract: This study proposes a geometry-aware gait phase classification framework designed for real-time application using a single inertial measurement unit (IMU) placed at the center of the pelvis. Our approach employs a relevant classification scheme consisting of three gait combinations: (1) left swing with right stance, (2) right swing with left stance, and (3) double stance. To extract meaningful temporal dynamics from a single IMU, raw signals are encoded into 6 times 6 covariance matrices, which are then embedded into a nonlinear feature space using a Gaussian kernel. This embedding captures underlying gait structure while respecting the geometry of the SPD manifold. A lightweight Extreme Learning Machine (ELM) classifier processes the resulting kernel vectors for efficient phase recognition. The proposed system was validated on data from eight healthy participants using four different geometry-aware distance metrics. Among them, the Log-Euclidean distance yielded the highest average classification accuracy of 89.18% ± 1.51% across five trials. These results demonstrate that even with a single pelvic IMU, our method can reliably distinguish bilateral gait phase dynamics, highlighting its potential for future application in hemiplegic gait assessment and wearable assistive robotics.

Keywords: Autonomous Vehicle Systems
Abstract: This study presents an advanced hybrid powertrain with dual energy regeneration for hydraulic excavators to address fuel consumption and emission challenges. The system integrates a hydrostatic transmission (HST) for torque optimization and a hydraulic pump/motor (HPM) capable of pump–motor switching. Two regeneration modes are considered: a fixed hydraulic motor and the HPM motor mode. To coordinate energy flows, an Equivalent Consumption Minimization Strategy (ECMS) is applied, balancing ICE–EMG power and battery state-of-charge (SOC). Simulation and laboratory experiments confirm up to 61.64% energy savings compared to conventional excavators, with regeneration efficiency reaching 73% in HPM mode. Furthermore, battery life is extended by 16.09% compared with EHCVP II systems. These results highlight the potential of combining HST, dual-mode HPM, and online EMS for efficient and sustainable excavator operation.

Keywords: Robotic Applications, Autonomous Vehicle Systems, Robot Vision
Abstract: An orchard monitoring system can contribute to improving the efficiency of agricultural operations by acquiring various types of information, including images of fruit trees. However, outdoor photography often suffers from reduced　visibility due to the effects of backlighting. Although　previous studies have addressed this problem using image　processing techniques, these methods are often ineffective under severe backlighting conditions. Therefore, in this　study, we propose a monitoring system that mitigates backlighting effects by capturing images from the forward-lit direction. This system consists of 3D point cloud map　generation, automatic fruit tree position estimation, calculation of shooting positions, autonomous navigation and automatic photography, and a GUI application. We confirmed that the system successfully performed stable navigation and automatic acquisition of obvious fruit tree images in a test field.