Search

Obstacle avoidance is essential for the effective operation of autonomous mobile robots, enabling them to detect and navigate around obstacles in their environment. While deep learning provides significant benefits for autonomous navigation, it typically requires large, accurately labeled datasets, making the data’s preparation and processing time-consuming and labor-intensive. To address this challenge, this study introduces a transfer learning (TL)-based automatic labeling segmentation (ALS) framework. This framework utilizes a pretrained attention-based network, DifferNet, to efficiently perform semantic segmentation tasks on new, unlabeled datasets. DifferNet leverages prior knowledge from the Cityscapes dataset to identify high-entropy areas as road obstacles by analyzing differences between the input and resynthesized images. The resulting road anomaly map was refined using depth information to produce a robust drivable area and map of road anomalies. Several off-the-shelf RGB-D semantic segmentation neural networks were trained using pseudo-labels generated by the ALS framework, with validation conducted on the GMRPD dataset. Experimental results demonstrated that the proposed ALS framework achieved mean precision, mean recall, and mean intersection over union (IoU) rates of 80.31%, 84.42%, and 71.99%, respectively. The ALS framework, through the use of transfer learning and the DifferNet network, offers an efficient solution for semantic segmentation of new, unlabeled datasets, underscoring its potential for improving obstacle avoidance in autonomous mobile robots.

This paper presents a microcontroller-based solution for generating real-time normal walking knee angle of a powered transfemoral prosthetic leg prototype. The proposed control algorithm was used to determine the prosthetic knee angle by utilizing seven hip angle movement features generated from only the inertia measurement unit (IMU) deployed on the prosthetic socket on the thigh of the same side. Then, a proportional–integral–derivative (PID) controller is developed to control the motor to reach the desired knee angle in real time. Furthermore, a novel parallel four-bar linkage-based master–slave validation framework combining a motion capture system was introduced to evaluate the performance of the knee angle generation on a speed-adjustable treadmill with able-bodied subjects. In the framework evaluation, 3 different walking speeds were applied to the treadmill to validate different speed adaptation capabilities of the prosthetic leg control system, precisely 50 cm/s, 60 cm/s, and 70 cm/s. Through the proposed 4-bar linkage framework, the prosthesis’s movement can simulate able-bodied subjects well with maximum RMSE never exceeding 0.27° in the swing flexion phase, 4.4° to 5.8° in the stance phase, and 1.953° to 13.466° in the swing extension phase. The treadmill results showed that the prosthetic leg is able to perform a normal walking gait following different walking speeds of the subject. Finally, a corridor walking experiment with a bypass adapter was successfully performed to examine the feasibility of real prosthetic walking situations.

This paper presents the design and implementation of a flexible manipulator formed of connected continuum kinematic modules (CKMs) to ease the fabrication of a continuum robot with multiple degrees of freedom. The CKM consists of five sequentially arranged circular plates, four universal joints intermediately connecting five circular plates, three individual actuated tension cables, and compression springs surrounding the tension cables. The base and movable circular plates are used to connect the robot platform or the neighboring CKM. All tension cables are controlled via linear actuators at a distal site. To demonstrate the function and feasibility of the proposed CKM, the kinematics of the continuum manipulator were verified through a kinematic simulation at different end velocities. The correctness of the manipulator posture was confirmed through the kinematic simulation. Then, a continuum robot formed with three CKMs is fabricated to perform Jacobian-based image servo tracking tasks. For the eye-to-hand (ETH) experiment, a heart shape trajectory was tracked to verify the precision of the kinematics, which achieved an endpoint error of 4.03 in Root Mean Square Error (RMSE). For the eye-in-hand (EIH) plugging-in/unplugging experiment, the accuracy of the image servo tracking system was demonstrated in extensive tolerance conditions, with processing times as low as 58±2.12 s and 83±6.87 s at the 90% confidence level in unplugging and plugging-in tasks, respectively. Finally, quantitative tracking error analyses are provided to evaluate the overall performance.

Failure mode detection is essential for bearing life prediction to protect the shafts on the machinery. This work demonstrates the rolling bearing vibration measurement, signals converting and analysis, feature extraction, and machine learning with neural networks to achieve failure mode detection for a shaft bearing. Two self-designed bearing test platforms with two types of sensors conduct the bearing vibration collection in normal and abnormal states. The time-domain signals convert to the frequency domain for analysis to observe the dominant frequency between these two types of sensors. In feature extraction, principal components analysis (PCA) combines with wavelet packet decomposition (WPD) to form the two feature extraction methods: PCA-WPD and WPD-PCA for optimization. The features extracted by these two methods serve as input to the long short-term memory (LSTM) networks for classification and training to distinguish bearing states in normal, misaligned, unbalanced, and impact loads. The evaluation arguments include sensor types, vibration directions, failure modes, feature extraction methods, and neural networks. In conclusion, the developed methods with the typical lower-cost sensor can achieve 97% accuracy in bearing failure mode detection.

Accurate segmentation of drivable areas and road obstacles is critical for autonomous mobile robots to navigate safely in indoor and outdoor environments. With the fast advancement of deep learning, mobile robots may now perform autonomous navigation based on what they learned in the learning phase. On the other hand, existing techniques often have low performance when confronted with complex situations since unfamiliar objects are not included in the training dataset. Additionally, the use of a large amount of labeled data is generally essential for training deep neural networks to achieve good performance, which is time-consuming and labor-intensive. Thus, this paper presents a solution to these issues by proposing a self-supervised learning method for the drivable areas and road anomaly segmentation. First, we propose the Automatic Generating Segmentation Label (AGSL) framework, which is an efficient system automatically generating segmentation labels for drivable areas and road anomalies by finding dissimilarities between the input and resynthesized image and localizing obstacles in the disparity map. Then, we train RGB-D datasets with a semantic segmentation network using self-generated ground truth labels derived from our method (AGSL labels) to get the pre-trained model. The results showed that our AGSL achieved high performance in labeling evaluation, and the pre-trained model also obtains certain confidence in real-time segmentation application on mobile robots.

There is evidence of strong association between influenza infections and stroke; however, the influenza vaccination and its effect on strokes is currently unclear. In the present study, Taiwan’s National Health Insurance Database was used in obtaining data for study subjects 55 years and older diagnosed with hypertension (n = 59,251; 25,266 vaccinated and 33,985 unvaccinated subjects) from 2001–2012. Propensity scores were calculated using a logistic regression model to determine the effects of vaccination by accounting for covariates that predict receiving the intervention (vaccine). A time-dependent Cox proportional hazard model was used to calculate the hazard ratios (HRs) for stroke in vaccinated and unvaccinated patients. Influenza vaccination was associated with a 42%, 40% and 44% stroke risk reduction in the entire cohort for all seasons, the influenza season and the non-influenza season, respectively (Adjust hazard ratio [aHR]: 0.58, 95% confidence interval [CI]: 0.56–0.61; aHR: 0.60, 95% CI: 0.56–0.63; aHR: 0.56, 95% CI: 0.52–0.60, for all seasons, the influenza season and the non-influenza season, respectively). The effect of risk reduction by vaccination also revealed a trend of dose dependency. Among subjects between 55 to 64 years old with four or more vaccinations during the study period, there is a 73% risk reduction for stroke during the non-influenza season (aHR: 0.27, 95% CI: 0.20–0.34). In conclusion, the influenza vaccination exerts dose-dependent and synergistic protective effects against stroke in individuals 55 years and older with hypertension.

In the industrial environment, Autonomous Guided Vehicles (AGVs) generally run on a planned route. Among trajectory-tracking algorithms for unmanned vehicles, the Pure Pursuit (PP) algorithm is prevalent in many real-world applications because of its simple and easy implementation. However, it is challenging to decelerate the AGV’s moving speed when turning on a large curve path. Moreover, this paper addresses the kidnapped-robot problem occurring in spare LiDAR environments. This paper proposes an improved Pure Pursuit algorithm so that the AGV can predict the trajectory and decelerate for turning, thus increasing the accuracy of the path tracking. To solve the kidnapped-robot problem, we use a learning-based classifier to detect the repetitive pattern scenario (e.g., long corridor) regarding 2D LiDAR features for switching the localization system between Simultaneous Localization And Mapping (SLAM) method and Odometer method. As experimental results in practice, the improved Pure Pursuit algorithm can reduce the tracking error while performing more efficiently. Moreover, the learning-based localization selection strategy helps the robot navigation task achieve stable performance, with 36.25% in completion rate more than only using SLAM. The results demonstrate that the proposed method is feasible and reliable in actual conditions.

This paper addresses the end-effector position and orientation tracking problem of a 6-degrees of freedom (DOF) rigid link electrically driven (RLED) revolute joint serial robot manipulator with uncertain parameters and external disturbances. System uncertainties and external disturbances are bounded but their upper limits are unknown. The input matrix, which is always invertible and not necessarily a diagonal matrix, is also assumed to be uncertain with a certain bound. An adaptive sliding mode control (ASMC) is designed to solve the tracking problem given these uncertainties. Note that the designed control method can be applied to any RLED serial robot manipulator. No regression matrices are required and the uncertainty upper limits are not necessarily known. Furthermore, the controller computation time can be reduced by considering some parts of the system dynamics and the nth time-derivative reference signals of nth order system, in this case desired joint jerk signals, as unknown bounded uncertainties. The desired trajectory is defined in operational space while the designed ASMC works in joint space. The methods to convert the end-effector pose, velocity, acceleration, and jerk from operational space to joint space are also explained. The Lyapunov stability criterion is applied to prove the convergence of the adaptive gain of ASMC, the sliding surface, and system stability. Finally, simulation results verified ASMC performance.

ZigBee wireless sensor devices possess characteristics of small size, light weight, low power consumption, having up to 65535 nodes in a sensor network, in theory. Therefore, the ZigBee wireless sensor network (WSN) is very suitable for use in developing monitoring and control (MC) applications, such as remote healthcare, industrial control, fire detection, environmental monitoring, and so on. This dissertation is directed towards the research on the issues of communication extension and backup, encountered in creating ZigBee-based MC systems for military storerooms, together with providing associated solutions. We design an embedded gateway that possesses wired network (Ethernet) and wireless communication (GSM) backup capability. The gateway can not only easily extend the monitoring distance of the ZigBee-based MCS, but can also solve the problem that some military zones do not have wire networks or possess communication blind spots. The results of this dissertation have been practically applied in constructing a paradigm monitoring system of a military storeroom. It is believed that the research results could be a useful reference for developing ZigBee-based MCSs in the future.

Design of a new differential-velocity–type compliant joint with two motors is proposed for robotic manipulators. The speed/output torque of the differential-velocity–type joint is synthesized by controlling the differential speeds of a pair of actuators. Descriptions of the design concept, theoretical analysis, and the experiments are presented in this article. The experimental results showed that (1) the stiffness of the differential-velocity–type joint can be regulated by changing the speeds of both motors, (2) the differential-velocity–type joint actually reduced 7.9%−15.7% time compared with the single-drive joint when performing five cycles of reciprocation, and (3) a new design of delta-type robot with differential-velocity–type driving modules was developed for verification. The experimental results showed that the differential-velocity–type delta robot can perform flexible assembly operations within a 0.028-mm tolerance.

The purpose of this research is to develop multi-talented humanoid robots, based on technologies featuring high-computing and control abilities, to perform onstage. It has been a worldwide trend in the last decade to apply robot technologies in theatrical performance. The more robot performers resemble human beings, the easier it becomes for the emotions of audiences to be bonded with robotic performances. Although all kinds of robots can be theatrical performers based on programs, humanoid robots are more advantageous for playing a wider range of characters because of their resemblance to human beings. Thus, developing theatrical humanoid robots is becoming very important in the field of the robot theatre. However, theatrical humanoid robots need to possess the same versatile abilities as their human counterparts, instead of merely posing or performing motion demonstrations onstage, otherwise audiences will easily become bored. The four theatrical robots developed for this research have successfully performed in a public performance and participated in five programs. All of them were approved by most audiences.