The main method to diagnose COVID-19 is a nucleic acid test from a throat swab. Routine manual collection methods expose medical personnel to high-risk environment, which has a high risk of cross-infection. A throat swab sampling robot was developed to take the place of medical staff. The automatic segmentation of M-region in the pharyngeal swab image, which plays a core guiding role when the robot takes a throat swab sample. Aiming at the problem of discontinuous or fuzzy boundary in M -region of oral cavity, the segmentation accuracy is affected. An improved U -Net model is proposed and a new multi-scale feature fusion module with channel attention mechanism is presented. The ability of adaptive learning is enhanced and the segmentation precision of M -region with discontinuous or fuzzy edges is increased. Oral images of 45 volunteers were collected for training and testing. Experimental results showed that the model could accurately segment M-region in pharyngeal swab images, and compared with other segmentation networks, it has better indexes of segmentation precision.
On current research and development of autonomous driving mobile system, urgent stopping control in high speed is the hottest research topic. We proposed a group emergency stop mechanism for multi-mobile robot system, which is based on motion control of multi-lane highway traffic. Lane based sub-group and inter-group interaction control scheme are defined and group emergency stop behavior have been illustrating via numerical simulation. In this paper, the group effect has been analyzed on the performance on coping with uncertain sensing delay. Simulation results indicates that Group Emergency Stopping Behavior contributes to reduce the opportunity of rear-end collision due to the sensing delay.
In this paper, we investigate the framing of a newly developed social robot for the application in industrial work environments, e.g. in logistics. To achieve a bond between human and robot and to raise the acceptance towards the robot, framing can be a useful approach since robots lack abilities like emotional intelligence. Therefore, an online survey has been conducted and is evaluated using the TAM3 and Godspeed questionnaire. The survey is focused on three hypotheses involving the relation of anthropomorphous framing and user acceptance. The results give insights on the perception of the robot as anthropomorphic and the framing effect of the robot.
The balance goal of the existing offline dynamic balance detection method is to minimize the amplitude of each vibration measurement point. After balancing, many high-speed rotors need to be equipped with an online dynamic balance system to maintain the rotor at a higher speed and run smoothly. This shows the existing offline dynamic that has much room for improvement in balance. A detection method that can measure the eccentric mass at each position of the rotor axis is proposed. Now the goal is to fit the curve of the center of inertia closer to the geometric center, hoping to minimize the eccentric mass on the rotor.
This paper presents a perspective on some issues related to safety in the context of autonomous surgical robots. To meet the challenge of safety certification and bring about acceptance of the technology by the public, we propose principles for a design paradigm that goes in the direction of safety by construction: design with certification in mind, clearly distinguish the notion of safety from th...
The permanent magnetic suspension dust-free conveying system is introduced. According to the suspension force model and the prototype structure, a nonlinear model of the control decoupling system is established. Using the coordinate transformation decoupling strategy, four-point independent control is changed to a three-degree-of-freedom control suspension vehicle. Finally, PID control and fuzzy PID control are used to simulate the displacement step of the permanent magnet levitation vehicle with three degrees of freedom. The simulation results show that the PID controller has a faster response speed and can return to a stable position after several oscillations. The adjustment time is 0.3 seconds, but the overshoot is relatively large, about 8%. For the fuzzy PID platform, the rise time is about 0.2s, the steady-state error is about 1%, there is no overshoot, and the system has good rigidity, damping characteristics, and robustness.
This work has been carried out to implement the Rapidly exploring Random Tree (RRT) Algorithm for designing a path in a 2D workspace cluttered with different human activities perceived as obstacles of different densities in a collaborative human-robot search operation-based Industry 4.0 environment. RRT based learning is an autonomous path planning algorithm suitable for perceiving environments with differential and non-holonomic constraints. Paths have been generated between an initial point, START to destination point, GOAL amidst uncertain temporal objects met when they are near and far. The algorithm starts to expand the search tree from the START node by randomly exploring the workspace until the GOAL is reached in the presence of these obstacles. The solution is a locus of points in the workspace between source and destination. This has been done for sparsely distributed human activity as obstacles encapsulating waypoint and lower and higher obstacle densities as observed in a real collaborative search environment. The generated paths were evaluated for closeness and impact of different kinds of dynamic obstacle configurations in a real experimental test bed.
Assisted drainage robot has the function of helping users to complete toilet requirements. However, without knowing the sitting posture and height of the user, it is difficult for the robot to find a suitable position for the user. Some scholars have studied the reinforcement learning control robot according to the sitting posture model to find the user's sitting position, but the moving speed and the process of finding the sitting position are slow. In order to overcome the shortcomings of the previous research, this paper proposes a control method based on the combination of Q learning and evaluation reward. First, the robot obtains the distance between human and robot as the motion state. Then, the best exercise strategy is calculated through Q learning, and the best exercise strategy is moved to the appropriate sitting position of the user. Finally, the seat position is roughly determined through reward evaluation. Experimental simulation shows that the method of finding the user's sitting point based on Q learning control method is better than the reinforcement learning control method based on sitting posture model.
Falling from bed during the process of getting out of bed is a serious accident for disabled and elderly patients. It's crucial to accurately predict the intention to get out of bed. However, simply relying on the change of pressure or posture may cause misprediction. In this paper, we propose a novel posture recognition method based on time series analysis and supervised learning algorithm. Firstly, we depict the relationship of posture, pressure and the intention to get out of bed through the preliminary experiment. Secondly, we analyze the pressure changes in continuous time and identify the process of leaving bed by means of comparing the target time series. To reduce the misjudgment of behavior caused by differences in individual behavior time, dynamic time warping (DTW) algorithm is utilized to improve the traditional Euclidean distance calculation method. Thirdly, we compare the accuracy of four different supervised learning arithmetics on posture recognition. Nonlinear SVM achieved the best recognition of 96.8%. Finally, we combine the results of time series analysis and the classification results of supervised learning algorithms to predict the intention to get out of bed. The experimental result shows that our proposed method can accurately predict the intention to get out of bed.
We present solutions for autonomous vehicles in limited visibility scenarios, such as traversing T-intersections, as well as detail how these scenarios can be handled simultaneously. The approach models each problem separately as a partially observable Markov decision process (POMDP). We propose an approach for integrating limited visibility within a POMDPs and implementing them on a physical robot. In order to address scalability challenges, we use a framework for multiple online decision-components with interacting actions (MODIA). We present the novel necessary architectural details to deploy MODIA on an actual robot. The entire approach is demonstrated on a fully operational autonomous vehicle prototype acting in the real world at two different T-intersections.
The shortage of young labor in construction sites can be addressed effectively if fully autonomous material transportation can be realized. In this work, we present a gate-type robot that can transport carts (loaded with building materials) up to five times of its own weight. Additionally, various types of measures including mechanical, electricidal, and control strategies have been considered for the safety of the workers, robots, and surroundings. The presented robots have been evaluated in real construction sites considering both manipulation capacity and safety performance.
Forward fall is a major risk for the elderly. In the case of a trip and fall, it is known that the length of the recovery step affects the ability to avoid the fall. However, the effects of recovery step length on the severity of injuries are unclear. Therefore, this study was conducted with the objective of investigating the effects of the recovery step length on the contact velocity with the ground by estimating the natural fall motion using a motion simulator. In this study, we developed a simulator for estimating the fall motion which started from the motion obtained in an experiment. In the experiment, the recovery step of the participants was classified as a short-step case or long-step case. The simulation results showed that the contact velocity had a correlation with the fall velocity of the center of mass (r = 0.52) and the simulation duration (r = -0.50) in the short-step case, but no correlation was observed in the long-step case. It was suggested that a long recovery step could result in an increase in the types of fall motion and thus variations in fall injuries.
To investigate the mechanism of internal bleeding resulting from human-robot interactions, an in-vivo microscopy technique is developed. A transparent indenter was loaded onto a subject (an anesthetized transparent glass catfish). The affected blood vessels were investigated though the intender using a microscope equipped with a video camera. The internal bleeding, changes in the arrangement of secondary vessels, and bone contusions caused by loading were observed via the proposed microscopy technique, which is expected to capture information how the vessels fracture and cause the internal bleeding in order to discuss the safety criteria of robots.
In recent years, intelligent elderly care equipment represented by personal care robots has gradually become a focus of academic research in various countries. A reliable navigation system is a foundation for personal care robots to perform complex tasks. Traditional nursing robots are difficult to apply to the real home living environment due to the simple operating environment of the navigation system. This paper proposes a personal care robot navigation system employed in a home environment, which uses a multi-sensor data fusion method to sense the surrounding environment information of the robot accurately. Thus, the robot able to perform complex navigation tasks in the home environment, reliably and stably. The navigation system is based on the seed robot hardware platform and ROS (Robot Operating System) operating system. Experimental results show that the navigation system can achieve reliable operation of personal care robots in complex home environments.
Monitoring human-robot interaction (HRI) process of people with mobility inconvenience when to employ multi-robots is an effective approach to mitigate human error and enhance safety. However, the interaction process of transfer that occurs frequently in daily life has not been studied. In this study, we proposed a comprehensive transfer monitoring method according to the mental states monitoring by functional Near-InfraRed Spectroscopy (fNIRS) and subjective workload questionnaire. 7 subjects were to perform transfer behavior under two contrasted levels (self-rising transfer vs. assisted-rising transfer by welfare-robots). After removing physiological noises, six oxygenated and deoxygenated hemoglobin (HbO and HbR) features-mean, slope, variance, peak, skewness, and kurtosis-were calculated. All possible 2- and 3-feature combinations of the calculated features were then used to classify self-rising transfer vs. assisted-rising transfer by linear discriminant analysis (LDA). Then, we established the Subjective Workload Assessment System (SWAS) for 2 subjects, which was used to make an evaluation for HRI. The experiment results demonstrated that the optimal feature-combination selection by mean and peak values, which provided a sound theoretical basis for distinguishing mental states of transfer tasks. The SWAS effectively quantify the mental states and physical load of users when to implement multiply welfare-robots. It was contributed to guide the motion planning of multiply welfare-robots for the ultimate goal of safe transfer, and also could be applied in similar scenarios.
Deep learning gesture recognition based on surface electromyography (sEMG) is playing an increasingly important role in prosthetic hand control. In order to improve the recognition rate of multi-modal EMG signals, this paper proposes a feature model construction and optimization method based on multi-channel EMG signal amplification unit. And through CNN and LSTM (CNN+LSTM) deep learning model, the recognition rate and acquisition window are trained. Use the established time series surface EMG image to construct a feature model to solve the recognition problem of multi-modal surface EMG signal. The experimental results show that under the same network structure, the EMG signal processed by Fast Fourier Transform (FFT) as the characteristic value has better performance.
In this paper, we investigate the framing of a newly developed social robot for the application in industrial work environments, e.g. in logistics. To achieve a bond between human and robot and to raise the acceptance towards the robot, framing can be a useful approach since robots lack abilities like emotional intelligence. Therefore, an online survey has been conducted and is evaluated using the TAM3 and Godspeed questionnaire. The survey is focused on three hypotheses involving the relation of anthropomorphous framing and user acceptance. The results give insights on the perception of the robot as anthropomorphic and the framing effect of the robot.
This paper proposes two quasi-zero power control methods for permanent magnet levitation systems. The system consists of a radially magnetized permanent magnet, two symmetrically arranged iron cores, a servo motor with an encoder, and a bar-shaped suspended object. In this system, the servo motor drives the permanent magnet to rotate, changing the magnetic flux through the suspended object, and realizes the non-contact suspension of the suspended object. Due to the special structure of the system, the system has quasi-zero power characteristics. In this paper, the constant angle control method and the constant air gap control method are used to study the quasi-zero power characteristics of the system under the action of radial interference force. The results show that both control methods have lower current consumption. Under the constant air gap quasi-zero power state, it can not only reduce energy consumption but also ensure stability.
The continuous emergence of new technologies has contributed to the impending reality of service robots an upcoming reality. When interacting with humans, robots must adapt to changing environments. Hence, service robots at home need learning capabilities to acquire new knowledge and merge it with their own. In this study, we have developed a system for learning the ontologies of new concepts, combining textural knowledge, visual analysis, and user interaction. In this system, the robot is provided with an essential feature to adapt to the home environment. We focus on the learning of new ontological concepts oriented toward service robot applications. We propose combining textural knowledge, visual analysis, and user interaction to determine the correct placement of the new concepts in the ontology structure. We aim to enable the robot to extend its ontological knowledge as needed. We conducted a set of experiments to show the applicability of the presented method and the advantage of conceptualizing objects in ontological knowledge. The experiments consisted of two parts: concept learning experiments and experiments with an integrated robot system. In the former, the robot had to conceptualize a set of new objects in its ontological knowledge, and in the latter, the robot was asked to search and find the new objects learned.
Aiming for the wide use of field robots in the infrastructure inspection and maintenance purposes, Japan Ministry of Economy, Trade and Industry (METI) and New Energy and Industrial Technology Development Organization (NEDO) issued the Performance Evaluation Procedures for each of aerial, aquatic and ground robots in 2018. Upon receiving the issue of these procedures, a three-year training course has started since late JFY2018 for each of these three fields, which includes theoretical and practical trainings. The goal of the course is to train personnel who is able to design and implement a scenario-based robot performance test for the designated purposes. The author has acted as a principal lecturer of the aquatic robot course. In this paper, the achievement of the course and the problems towards standardization of the robot performance evaluation found through the course are discussed.
