Chia-Jen Lin

When autonomous mobile robots (AMRs) travel in formation, mutual occlusion reduces the map features available to adaptive Monte Carlo localization (AMCL) and introduces robot body returns. This study proposes a Virtual-AMR LiDAR fusion method that transforms synchronized 2D scans into a common frame, tracks robot contours, estimates relative pose by KD-tree overlap matching, removes robot returns, and fuses environmental measurements for AMCL. The method was evaluated using two AMRs in a 3.1 m × 10 m indoor area with 4 m straight-line and turning trajectories. Four branch-wise configurations were tested in five repeated trials. Against an internal reference trajectory constructed from the planned path and motor encoder measurements rather than an independent ground truth system, the complete method achieved trial-level 2D RMSE values of 0.0412 ± 0.0083 m for the straight-line trajectory and 0.0511 ± 0.0104 m for the turning trajectory (mean ± sample standard deviation across five trials), while reducing localization drift and preventing the AMCL divergence observed under the tested mutual occlusion conditions.

This study proposes a Digital Twin-integrated multi-objective dynamic operation scheduling (MO-DOS) framework for pipeline planning in industrial oil transportation. The framework addresses the complexity of multi-objective scheduling under practical operational constraints, including transport duration, energy efficiency, resource utilization, and safety risk. Pareto front optimization is systematically integrated with the Dynamic Window Approach (DWA) to enable real-time dynamic routing under these competing objectives. To improve performance in complex field environments, a field-adaptive heuristic extension, DWA-H, is incorporated into DWA. By integrating high-fidelity virtual models with physical pipeline networks, the Digital Twin-based platform supports real-time scheduling, monitoring, and optimization in industrial oil processing facilities. The proposed framework is validated using real transportation work orders and benchmarked against the traditional First-In-First-Out (FIFO) strategy and the standard MO-DOS with DWA method. Experimental results demonstrate that the proposed Digital Twin-enabled MO-DOS framework improves oil transport efficiency by approximately 54.7% and reduces operational costs by about 25%. This study establishes a scalable and deployable multi-objective scheduling framework, demonstrating the integration of Digital Twin technology and dynamic optimization for industrial applications.

In this article, we propose a novel real-time terrain recognition and slip estimation method for quadruped robots using proprioceptive sensors and temporal convolutional networks (TCNs). As quadruped robots are increasingly deployed in complex environments, accurate terrain understanding is crucial. External sensors can be affected by lighting variations, occlusion, reflective surfaces, and others. To overcome these challenges, we propose a proprioceptive sensing-based complementary perception module with a TCN, enabling reliable real-time terrain recognition while reducing dependence on external perception. The TCN model effectively captures temporal dependencies in sensor signals, enabling precise and robust detection. The framework is validated through extensive real-world experiments and deployed on an embedded edge computing platform for real-time operation. Results show that the proposed TCN method achieves 98.8% recognition accuracy, outperforming the baseline models compared in this study. In addition, this study analyzes how locomotion speed and environmental conditions affect slip in quadruped robots. These findings confirm that quadruped robots can not only recognize terrain types but also detect surface states, enabling safer and more adaptive locomotion. Therefore, the proposed system is a cost-effective, robust, and low-latency solution for real-time terrain recognition, providing a strong foundation for future deployment across more diverse terrains.

This study proposes a hybrid control framework that integrates a deep Q-network (DQN), adaptive proportional–integral–derivative (PID) control, and multisensor fusion via an extended Kalman filter (EKF) to enhance the accuracy, stability, and adaptability of autonomous mobile robots (AMRs) during docking tasks in complex indoor environments. A neural network dynamically tunes PID parameters based on the robot’s state, combining the robustness of classical control with the flexibility of learningbased methods. For localization, AprilTag visual markers are fused with multisensor data through EKF, yielding more accurate state estimation. A task-specific reward function incorporates target distance, angular deviation, collision penalties, and docking incentives, guiding the learning process toward smooth and efficient trajectories. Cosine-based angular velocity modulation and a LiDAR-triggered mode selector enable seamless switching between DQN–PID control and a modified DQN policy with smoother motion and faster reward convergence. While conventional DQN suffers from unsmooth motion and slower reward convergence, experimental results in both simulated and real-world environments show that the proposed switching framework achieves nearly 100% docking success, greatly surpassing the DQN-only approach, which gained only 59%. These results demonstrate clear advantages in convergence speed, trajectory smoothness, and robustness, confirming the framework’s suitability for real-world autonomous docking applications.

This study proposes a data re-indexing method that integrates the Fast Fourier Transform (FFT) and a Genetic Algorithm (GA) to reconstruct features of back electromotive force (Back-EMF) from ultra-low sampling rate data (ULSD) obtained via a low-speed serial interface (e.g., RS-485) embedded in motor drives. Initially, a demagnetization model is developed to extract Back-EMF features. Subsequently, the ULSD is re-indexed based on varying numbers of sampling points per electrical period, and corresponding waveforms are reconstructed. FFT is applied to analyze rearranged waveforms, and two adjacent sampling configurations with the highest fundamental frequency amplitudes are identified. Finally, the GA optimizes the index assignment of the ULSD using these two candidate configurations, thereby enhancing the reconstruction of the Back-EMF waveform. Experimental results demonstrate that the proposed ULSD re-indexing method reconstructs Back-EMF waveforms. These waveforms are then used as inputs to a one-dimensional convolutional autoencoder (1-D CAE) combined with a multilayer perceptron (MLP) for demagnetization fault (DMF) diagnosis. The proposed approach achieves a high diagnostic accuracy of 96.7%, while reducing noise and sampling errors, and without incurring additional external signal capture interfaces (ESCIs) hardware costs. These findings validate the feasibility and effectiveness of applying rearranged ULSD for low-cost and reliable fault diagnosis in IPMSM systems.

This study proposes a multi-objective optimization scheduling method for edible oil transportation in smart manufacturing, focusing on centralized control and addressing challenges such as complex pipelines and shared resource constraints. The method employs the A* and Dijkstra pathfinding algorithm to determine the shortest pipeline route for each task, and estimates pipeline resource usage to derive a node cost weight function. Additionally, the transport time is calculated using the Hagen–Poiseuille law by considering the viscosity coefficients of different oil types. To minimize both cost and time, task execution sequences are optimized based on a Pareto front approach. A 3D digital model of the pipeline system was developed using C#, SolidWorks Professional, and the Helix Toolkit V2.24.0 to simulate a realistic production environment. This model is integrated with a 3D visual human–machine interface(HMI) that displays the status of each task before execution and provides real-time scheduling adjustment and decision-making support. Experimental results show that the proposed method improves scheduling efficiency by over 43% across various scenarios, significantly enhancing overall pipeline transport performance. The proposed method is applicable to pipeline scheduling and transportation management in digital factories, contributing to improved operational efficiency and system integration.

In recent years, advancements in artificial intelligence (AI) have become an essential study for machine learning. The use of AI with the Robot Operating System (ROS) enables mobile robots to learn and move autonomously. Mobile robots can now be widely used in industrial and service sectors. Generally, robots have been operated on fixed paths requiring set points to function. This study utilizes Deep Q-Network (DQN) incorporating filtering to train and reward AprilTag images, paths, and obstacle avoidance. Training is conducted in a Gazebo simulation environment, and the collected data is verified on physical mobile robots. The DQN network excels in computing complex functions; AprilTag provides X, Y, Z, Pitch, Yaw, and Roll data. By employing DQN methods, recognition and path accuracy are simultaneously enhanced. The constructed DQN network can endow mobile robots with autonomous learning capabilities.

The challenge in developing an AI deep learning model for motor health diagnosis is hampered by the lack of sufficient and representative datasets, leading to considerable time and resource consumption in research. Therefore, this paper focuses on the analysis of the second harmonic component fault characteristic induced by inter-turn short circuits (ITSCs) in phase voltages. First, it establishes a coil inter-turn short-circuit fault (ITSCF) model of the motor to identify the twice-frequency q-axis voltage error characteristics. Subsequently, it develops simulation programs by integrating control and fault models in MATLAB/Simulink/Simscape to observe and analyze the q-axis voltage and circulating current errors caused by the short circuit. Finally, a discrete wavelet transform method is established to analyze the q-axis synchronous reference frame voltage. By applying the energy-based method to extract the twice-frequency voltage error characteristics, the approach successfully detects the error features and confirms ITSCF in the motor. The contributions of this paper include not only the development of an ITSCF characteristic model for the motor but also the successful application of wavelet transform to effectively analyze the time-frequency characteristics of its signals. This approach can serve as a valuable reference for the design of deep learning models in future AI applications.