Abstract:On-line chatter detection can avoid unstable cutting through monitoring the machining process. In order to identify chatter in a timely manner, an improved Support Vector Machine (SVM) is developed in this paper, based on extracted features. In the SVM model, the penalty factor (c) and the core parameter (g) have important influence on the classification, more than from Kernel Functions (KFs). Hence, first the classification results are conducted using different KFs. Then two methods are presented for exploring the best parameters. The chatter identification results show that the Genetic Algorithm (GA) approach is more suitable for deciding the parameters than the Grid Explore (GE) approach.

Abstract:In constrained motion control of a robot, the interaction force is an important variable, which directly describes the state of interaction. It is required in a number of algorithms for interaction control. Desirably, the interaction force has to be measured by force sensors. However, there are inherent limitations with force sensors, such as the cost, sensing noise, limited bandwidth, and the difficulty of physical location at the required place, which is dynamic. In the present paper, the interaction force is estimated by using high order sliding mode observers. An adaptive version of a high order sliding mode observer is developed to robustly reconstruct the interaction force. Experimental results are given to show the effectiveness of the developed algorithms.

Abstract:Detection and recognition of a stairway as upstairs, downstairs and negative (e.g., ladder, level ground) are the fundamentals of assisting the visually impaired to travel independently in unfamiliar environments. Previous studies have focused on using massive amounts of RGB-D scene data to train traditional machine learning (ML)-based models to detect and recognize stationary stairway and escalator stairway separately. Nevertheless, none of them consider jointly training these two similar but different datasets to achieve better performance. This paper applies an adversarial learning algorithm on the indicated unsupervised domain adaptation scenario to transfer knowledge learned from the labeled RGB-D escalator stairway dataset to the unlabeled RGB-D stationary dataset. By utilizing the developed method, a feedforward convolutional neural network (CNN)-based feature extractor with five convolution layers can achieve 100% classification accuracy on testing the labeled escalator stairway data distributions and 80.6% classification accuracy on testing the unlabeled stationary data distributions. The success of the developed approach is demonstrated for classifying stairway on these two domains with a limited amount of data. To further demonstrate the effectiveness of the proposed method, the same CNN model is evaluated without domain adaptation and the results are compared with those of the presented architecture.

Abstract:Sleep apnea (SA) is a common sleep disorder. Identifying patients at risk by means of comprehensive monitoring that requires overnight stay at professional sleep clinics are costly and inconvenient and can lead to unreliable results in view of the unfamiliar sleep environment. Existing wearable devices for sleep monitoring, which can be used in a familiar home environment, do not provide the same comprehensive monitoring as through clinical monitoring. The larger objective of the present work is to develop a sleep monitoring system for home use, which can provide comprehensive monitoring. In the development in this paper, machine learning (ML) models are explored for the classification of SA and sleep stages using multisensory data, without neglecting any of the required signals. The data acquired through the sensors are normalized, their features are extracted using Composite Multiscale Sample Entropy (CMSE) and are standardized using a robust scaling algorithm. Processed features are classified using a Neural Network (NN) and the obtained results for the SA classification are compared with those obtained by using a Support Vector Machine (SVM) approach. The impact of neglecting signals when classifying sleep stages is analyzed as well. The results are presented in the paper and observations are made. The NN model trained with the Bayesian regularization algorithm has provided an overall average accuracy of 94.5% and performed slightly better than when trained using the scaled conjugate gradient backpropagation algorithm (93.2%). The SVMs have yielded lower accuracy levels compared to the NNs (<92%). It is observed that the use of all 14 signals for SS classification yields an overall test accuracy of 72.3%, which is higher than that when one or few signals are used. It is concluded that ML models are effective in classifying sleep data from multiple sensors. Accuracy levels are higher when fused multisensory data are used as inputs. Furthermore, NN models are found to be better suitable in practical application and can be incorporated into an inexpensive and convenient wearable device that can carry out comprehensive monitoring.

Abstract:As unmanned electric wheeled mobile robots have been increasingly applied to high-speed operations in unknown environments, the wheel slip becomes a problem when the robot is either accelerating, decelerating, or turning at high speed. Ignoring the effect of wheel slip may cause the mobile robot to deviate from the desired path. In this paper a recently proposed method is implemented to estimate the surface conditions encountered by an unmanned wheeled mobile robot, without using extra sensors. The method is simple, economical and needs less processing power than for other methods. A reaction torque observer is used to obtain the rolling resistance torque and it is applied to a wheeled mobile robot to obtain the surface condition in real-time for each wheel. The slip information is observed by comparing the reaction torque of each wheel. The obtained slip information is then used to control the torque of both wheels using a torque controller. Wheel slip is minimized by controlling the torque of each wheel. Minimizing the slip improves the ability of the unmanned electric wheeled mobile robot to navigate in the desired path in an unknown environment, regardless of the nature of the surface.

Abstract:Automobile accidents cost over a trillion-dollar every year and this figure will continue increasing without employing new technological solutions. Among these solutions, the automated lane-keeping system is one of the promising ones and such a system consists of two essential technologies: road detection and steering control. In this paper, novel lane keeping algorithms are proposed and are implemented using only a single off-the-shelf wide-angle camera as input. The implemented system is verified, through both simulation and experiments, and is found providing satisfactory performance for an automated lane-keeping system. When compared to the state-of-the-art lane-keeping systems, the implemented system can perform consistently across various ambient light conditions including the most challenging ones.

Abstract:Online sensing can provide useful information in monitoring applications, for example, machine health monitoring, structural condition monitoring, environmental monitoring, and many more. Missing data is generally a significant issue in the sensory data that is collected online by sensing systems, which may affect the goals of monitoring programs. In this paper, a sequence-to-sequence learning model based on a recurrent neural network (RNN) architecture is presented. In the proposed method, multivariate time series of the monitored parameters is embedded into the neural network through layer-by-layer encoders where the hidden features of the inputs are adaptively extracted. Afterwards, predictions of the missing data are generated by network decoders, which are one-step-ahead predictive data sequences of the monitored parameters. The prediction performance of the proposed model is validated based on a real-world sensory dataset. The experimental results demonstrate the performance of the proposed RNN-encoder-decoder model with its capability in sequence-to-sequence learning for online imputation of sensory data.

Abstract:Needle insertion procedures have been gaining in popularity with medical communities and patients over the recent years. Currently, their applications span a wide range of diagnostic and therapeutic procedures and there is still a growing tendency toward integrating needle insertion into other procedures and surgeries. Its less invasive nature, which is performed locally on the body, largely explains this growing trend. This results in less intraoperative tissue damage and shorter post-operative recovery time. Many procedures like biopsy, deep brain stimulation, and cancer treatments are done using needles/catheters. However, despite all the advantages of needle insertion procedures, the inherent complications resulting from them, such as tissue deformation, needle deflection, tissue inhomogeneity, patient variability, and associated uncertainty, can hardly be missed. Therefore, a needle insertion procedure requires that we address promising aspects and associated concerns. Against this backdrop, this paper provides a review of some of the main issues associated with a generic needle insertion procedure.

Abstract:There are many challenges for robot navigation in densely populated dynamic environments. This paper presents a survey of the path planning methods for robot navigation in dense environments. Particularly, the path planning in the navigation framework of mobile robots is composed of global path planning and local path planning, with regard to the planning scope and the executability. Within this framework, the recent progress of the path planning methods is presented in the paper, while examining their strengths and weaknesses. Notably, the recent developed Velocity Obstacle method and its variants that serve as the local planner are analyzed comprehensively. Moreover, as a model-free method that is widely used in current robot applications, the reinforcement learning-based path planning algorithms are detailed in this paper.

Abstract:Gear transmissions are widely used in industrial drive systems. Fault diagnosis of gear transmissions is important for maintaining the system performance, reducing the maintenance cost, and providing a safe working environment. This paper presents a novel fault diagnosis approach for gear transmissions based on convolutional neural networks (CNNs) and decision-level sensor fusion. In the proposed approach, a CNN is first utilized to classify the faults of a gear transmission based on the acquired signals from each of the sensors. Raw sensory data is sent directly into the CNN models without manual feature extraction. Then, classifier level sensor fusion is carried out to achieve improved classification accuracy by fusing the classification results from the CNN models. Experimental study is conducted, which shows the superior performance of the developed method in the classification of different gear transmission conditions in an automated industrial machine. The presented approach also achieves end-to-end learning that can be applied to the fault classification of a gear transmission under various operating conditions and with signals from different types of sensors.