Movement Accuracy Research Articles

Detailed characterization of neural selectivity in free viewing primates

Fixation constraints in visual tasks are ubiquitous in visual and cognitive neuroscience. Despite its widespread use, fixation requires trained subjects, is limited by the accuracy of fixational eye movements, and ignores the role of eye movements in shaping visual input. To overcome these limitations, we developed a suite of hardware and software tools to study vision during natural behavior in untrained subjects. We measured visual receptive fields and tuning properties from multiple cortical areas of marmoset monkeys who freely viewed full-field noise stimuli. The resulting receptive fields and tuning curves from primary visual cortex (V1) and area MT match reported selectivity from the literature which was measured using conventional approaches. We then combined free viewing with high-resolution eye tracking to make the first detailed 2D spatiotemporal measurements of foveal receptive fields in V1. These findings demonstrate the power of free viewing to characterize neural responses in untrained animals while simultaneously studying the dynamics of natural behavior.

Basketball Games and Early Childhood Motoric Development: Systematic Literature Review

This study aims to examine literature that focuses on research on early childhood basketball learning and motor development. Using the search term "basketball game and children's motor development", two databases (Scopus and Google Scholar) were selected to select journal articles. To achieve the measurement objectives, the collected literature is examined using the Critical Appraisal table, then a comparison of the measurement results is carried out. Basketball games and children's motoric development are discussed in 17 literatures. The results of several journals revealed several significant things, including the fact that playing basketball improves students' motor skills by increasing movement speed, accuracy, and strength in static and dynamic spaces, as well as game performance. Basketball players compete at various levels of competition. In addition, children who previously experienced delays in these movements at the level of fitness, behavior, knowledge and basic skills, which include agility, balance and coordination, were greatly influenced by the basic movements and basic skills of mini basketball. Basketball also has many benefits for your health and happiness. It is considered an easy game to play because of the satisfying psychosocial interactions. generate profits that last into adulthood.

Effective Modeling on Learning Ballet Online

After COVID-19, face-to-face learning was changed to online learning. However, very few effective online learning methods were available regarding physical education. Therefore, this study aimed to examine the modeling effects on learning ballet movement in the online system. We aimed to find effective modeling presentations based on objective information, expert assessments, and a kinematic approach. The study included 36 individuals who were divided into an expert modeling group, a self-modeling group, and controls. Participants performed 60 trials of Pas de basque in the acquisition phase and 10 trials without a demonstration video after 24 h. 10 min later, the reversed Pas de basque was conducted for the retention test. All groups showed improved performance after the acquisition phase, which indicated that the modeling presentation was effective despite adopting an online learning system. However, higher expert scores and more accurate joint movements were shown in the expert modeling group compared to the other groups. Therefore, expert modeling seems to be the most effective method for learning high-difficulty tasks with jumps and turns.

Accurate Movement Detection of Artificially Intelligent Security Objects

A number of intelligent applications demand multiple item detection and tracking as vital elements. Although object search links the identified object through a series of frames, object detection pinpoints the thing's location in a scene. Over the last few decades, a wide range of approaches has been created, which may be divided into 2D and stereo-based 3D techniques. When used in limited circumstances, most of these strategies yield trustworthy findings. These limiting presumptions are used to determine the number of complex elements that object detection and tracking naturally entail. Environmental factors, object appearance, flow density, backdrop colour intensity information, the amount of time an object is present in the scene, object occlusion, a scene's maximum number of things, etc., are among the most often held presumptions. In real-time applications, these approaches' dependability is not assured. A modern surveillance system needs reliable object identification and tracking in an open area.

HUMAN GAIT ANALYSIS USING HYBRID CONVOLUTIONAL NEURAL NETWORKS

Human gait analysis is a promising method of researching on human activities like walking or sitting. It reflects the habits of one person and can be observed in any activity that person performs. The patterns in human movements are influenced by many factors, including physiology, social, psychological, and health factors. Differences in limb movements help identify gait patterns, which are often measured using inertial measurement unit sensors (IMU) like gyroscopes and accelerometers placed in various locations throughout the body. This paper analyses the combination of IMU sensors and electromyography sensors (EMG) to improve the identification accuracy of human movements. We propose the hybrid convolutional neural network (CNN) and long short-term memory neuron network (LSTM) for the human gait analysis problem to achieve an accuracy of 0.9418, better than other models including pure CNN models. By using CNN's image classification advancements, we analyse multivariate time series sensor signals by using a sliding window to transform sensor data into image representation and principal component analysis (PCA) to reduce the data dimensionality. To tackle the dataset imbalance issue, we re-weight our model loss by the inverse effective number of samples in each class. We use the human gait HuGaDB dataset with unique characteristics, for gait analysis.

Utility and Usability of Two Forms of Supplemental Vibrotactile Kinesthetic Feedback for Enhancing Movement Accuracy and Efficiency in Goal-Directed Reaching.

Recent advances in wearable sensors and computing have made possible the development of novel sensory augmentation technologies that promise to enhance human motor performance and quality of life in a wide range of applications. We compared the objective utility and subjective user experience for two biologically inspired ways to encode movement-related information into supplemental feedback for the real-time control of goal-directed reaching in healthy, neurologically intact adults. One encoding scheme mimicked visual feedback encoding by converting real-time hand position in a Cartesian frame of reference into supplemental kinesthetic feedback provided by a vibrotactile display attached to the non-moving arm and hand. The other approach mimicked proprioceptive encoding by providing real-time arm joint angle information via the vibrotactile display. We found that both encoding schemes had objective utility in that after a brief training period, both forms of supplemental feedback promoted improved reach accuracy in the absence of concurrent visual feedback over performance levels achieved using proprioception alone. Cartesian encoding promoted greater reductions in target capture errors in the absence of visual feedback (Cartesian: 59% improvement; Joint Angle: 21% improvement). Accuracy gains promoted by both encoding schemes came at a cost in terms of temporal efficiency; target capture times were considerably longer (1.5 s longer) when reaching with supplemental kinesthetic feedback than without. Furthermore, neither encoding scheme yielded movements that were particularly smooth, although movements made with joint angle encoding were smoother than movements with Cartesian encoding. Participant responses on user experience surveys indicate that both encoding schemes were motivating and that both yielded passable user satisfaction scores. However, only Cartesian endpoint encoding was found to have passable usability; participants felt more competent using Cartesian encoding than joint angle encoding. These results are expected to inform future efforts to develop wearable technology to enhance the accuracy and efficiency of goal-directed actions using continuous supplemental kinesthetic feedback.

Integrated control of non-motor and motor efforts during perceptual decision-making and action execution: a pilot study

Humans daily life is characterized by a succession of voluntary actions. Since energy resources are limited, the ability to invest the appropriate amount of effort for selecting and executing these actions is a hallmark of adapted behavior. Recent studies indicate that decisions and actions share important principles, including the optimization of their duration when the context requires it. In the present pilot study, we test the hypothesis that the management of effort-related energy resources is shared between decision and action too. Healthy human subjects performed a perceptual decision task where they had to choose between two levels of effort to invest in making the decision (i.e. two levels of perceptual difficulty), and report it with a reaching movement. Crucially, the movement accuracy requirement gradually increased from trial to trial depending on participants’ decision performance. Results indicate an overall moderate and non-significant impact of the increasing motor difficulty on the choice of the non-motor (decision) effort to invest in each trial and on decision performance. By contrast, motor performance strongly decreased depending on both the motor and decisional difficulties. Together, the results support the hypothesis of an integrated management of the effort-related energy resources between decision and action. They also suggest that in the present task, the mutualized resources are primarily allocated to the decision-making process to the detriment of movements.

Multimodal Fusion Convolutional Neural Network Based on sEMG and Accelerometer Signals for Intersubject Upper Limb Movement Classification

The variation in the distributions of recorded data between individuals leads to low classification accuracy. To address this issue, we introduce a multimodal fusion convolutional neural network (MFCNN). This network extracts common information from surface electromyography (sEMG) and accelerometer signals of different subjects using a two-stream CNN. To enhance the classification accuracy of a particular subject, a fine-tuning approach was implemented. The performance of the proposed method was assessed in four different scenarios, which include inter-subject classification, inter-subject classification when training data from multiple subjects, fine-tuned inter-subject classification, and fine-tuned inter-subject classification when training data from multiple subjects. The results demonstrate that in the inter-subject scenario, when multiple subjects are available for training, the MFCNN achieves higher classification accuracy (p < 0.05) than other neural networks and support vector machines (SVMs) that use sEMG signals (NN and SVM), accelerometer signals (accNN and accSVM), sEMG and accelerometer signals (MFNN and MFSVM) as inputs, as well as a CNN that uses sEMG signals as input after fine-tuning. Furthermore, compared with an MFCNN model trained with data from a single subject and an accCNN model trained with data from a single subject or multiple subjects, an MFCNN trained with multiple subjects demonstrated better performance on new subjects after fine-tuning (p < 0.05). This method can learn common features among different subjects and improve the performance of classification among subjects. Our proposed method demonstrates the innovation of using a multimodal fusion approach and two-stream CNN to improve inter-subject classification accuracy in upper limb movements.

Numerical simulation of Non-Lethal Kinetic Energy Projectile Impact on Human Torso Model

In order to analyse the dynamic response and damage characteristics of human tissues and organs under the non-lethal kinetic energy projectiles, a numerical simulation model of human torso is established based on the Chinese visual human (CVH) data sets. By the dummy established with CVH data sets, the calculation accuracy of movement of human torso impact by non-Lethal kinetic energy projectile of finite element model were validated, and the skin perforation performance were validated by animal test. The non-lethal kinetic energy projectile, which is rubber ball in diameter 8mm and 50 shores hardness, impact on human torso with different velocity were calculated, the dynamic responses of sternum, rib and lung were captured and analysed. The results shows that, if the projectile speed is near the probability perforation velocity, due to the hardness of projectile and structure of organs and ribs, the projectile will rebound no matter the skin is perforation or not, and the vulnerable organ lung may not be damaged. At the same time, the pressure of lungs without skin penetration were significantly higher than skin without penetration, the human damage risk under soft projectiles is greater than hard projectiles when skin with damage but without perforation.

Accuracy of Invisalign® on Upper Incisors: A Systematic Review.

The current systematic review appraises the effectiveness of the types of tooth movements performed with Invisalign® clear aligner on the maxillary incisors. An electronic literature search of published trials was performed through PubMed, LILACS, Scopus, Cochrane Library, and Web of Science databases, and selected journals, from 2009 to 2020. Out of 291 references, five relevant publications were identified for analysis: four studies were performed retrospectively and one prospectively, all non-randomized. Despite the limited set of selected articles, the sample size is significant, with 148 subjects included in the reviewed studies involving the orthodontic treatment of upper incisors. We concluded that movements with the Invisalign® clear aligner on the upper incisors present distinct accuracy, possibly related with movement complexity; intrusion of the incisors has low accuracy (in some cases, 0% of accuracy was reported when the tooth extruded), while incisor extrusion exhibit some of the highest accuracy values reported in the included studies (45%-142%, when the achieved movement was greater than the predicted). Besides, axial (i.e., torque and tip) and horizontal (i.e., translation and rotation) movements are usually effective, with accuracy values between 39%-156% and 42%-79%, respectively. Overall, we determined that the efficiency of aligner to reach the desired movements in the upper incisors was low, as often refinements were required in the included studies. The use of aligner features must be more often considered to improve movement accuracy.

Multi-Attention Feature Fusion Network for Accurate Estimation of Finger Kinematics From Surface Electromyographic Signals

Simultaneous and proportional control (SPC) based on surface electromyographic (sEMG) signals has led to a broad range of applications. However, due to the limitation in the generalization and stability of current machine learning algorithms, these methods can only estimate less than 15 simultaneuous and proportional (SP) categories of finger movement. In this article, a novel deep learning algorithm, named multiattention feature fusion network (MAFN), is proposed to estimate comprehensive finger movement (up to 28 categories SP movements) from sEMG signals. MAFN is based on the multihead attention mechanism, which adaptively extracts essential features for analyzing the joint angles from the extracted sEMG features. Furthermore, a real-time exponential smoothing algorithm is designed for further improvement of the prediction stability. MAFN was evaluated on 28 finger movements of 38 subjects in the Ninapro_db2 dataset, and benchmarked with the state-of-the-art methods, such as temporal convolutional network (TCN) and long short term memory network (LSTM). The results demonstrated that the average Pearson correlation coefficient, root mean squared error of MAFN (0.84 ± 0.03,0.09 ± 0.01) were significantly higher than those of TCN (0.52 ± 0.06, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</i> < 0.001;0.16 ± 0.02, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</i> < 0.001) and LSTM (0.62 ± 0.06, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</i> < 0.001;0.13 ± 0.01, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</i> < 0.001). These improvements led to more stable and accurate movement predictions. Additionally, the time delay and power consumption of MAFN when applied to sEMG signals on a portable device are only 83.4 ms and 3 W, which implies prospective commercial applications.

Adaptive changes to saccade amplitude and target localization do not require pre-saccadic target visibility

The accuracy of saccadic eye movements is maintained by saccadic adaptation, a learning mechanism that is proposed to rely on visual prediction error, i.e., a mismatch between the pre-saccadically predicted and post-saccadically experienced position of the saccade target. However, recent research indicates that saccadic adaptation might be driven by postdictive motor error, i.e., a retrospective estimation of the pre-saccadic target position based on the post-saccadic image. We investigated whether oculomotor behavior can be adapted based on post-saccadic target information alone. We measured eye movements and localization judgements as participants aimed saccades at an initially invisible target, which was always shown only after the saccade. Each such trial was followed by either a pre- or a post-saccadic localization trial. The target position was fixed for the first 100 trials of the experiment and, during the following 200 trials, successively shifted inward or outward. Saccade amplitude and the pre- and post-saccadic localization judgements adjusted to the changing target position. Our results suggest that post-saccadic information is sufficient to induce error-reducing adaptive changes in saccade amplitude and target localization, possibly reflecting continuous updating of the estimated pre-saccadic target location driven by postdictive motor error.

A Vision Comprehension-Driven Intelligent Recognition Approach for Actions of Tennis Players Based on Improved Convolution Neural Networks

In this paper, we focus on two tasks: semantic segmentation and target detection in the visual semantic understanding of tennis sports images and we optimize the network structure to achieve a more complete location contour information mining of the target. In detail, we focus on a weakly supervised image semantic segmentation method based on null convolution pixel relations. To address the problem of incomplete pixel-level pseudo-labeling, we introduce a cavity convolution unit with multiple cavity rates and a self-attentive mechanism in the classification model to adaptively enhance the target regions and suppress other irrelevant regions while expanding the perceptual field to generate high-quality pixel-level pseudo-labeling and then train the semantic segmentation model. The final experimental results show that the hierarchical fusion algorithm proposed in this paper significantly outperforms other algorithms, and the overall classification accuracy of the tandem cavity neural network algorithm reaches 81% with good overall classification results. The recognition accuracy of static movements is higher than that of dynamic movements.

Effect of Visual Feedback on Behavioral Control and Functional Activity During Bilateral Hand Movement.

Previous researches state vision as a vital source of information for movement control and more precisely for accurate hand movement. Further, fine bimanual motor activity may be associated with various oscillatory activities within distinct brain areas and inter-hemispheric interactions. However, neural coordination among the distinct brain areas responsible to enhance motor accuracy is still not adequate. In the current study, we investigated task-dependent modulation by simultaneously measuring high time resolution electroencephalogram (EEG), electromyogram (EMG) and force along with bi-manual and unimanual motor tasks. The errors were controlled using visual feedback. To complete the unimanual tasks, the participant was asked to grip the strain gauge using the index finger and thumb of the right hand thereby exerting force on the connected visual feedback system. Whereas the bi-manual task involved finger abduction of the left index finger in two contractions along with visual feedback system and at the same time the right hand gripped using definite force on two conditions that whether visual feedback existed or not for the right hand. Primarily, the existence of visual feedback for the right hand significantly decreased brain network global and local efficiency in theta and alpha bands when compared with the elimination of visual feedback using twenty participants. Brain network activity in theta and alpha bands coordinates to facilitate fine hand movement. The findings may provide new neurological insight on virtual reality auxiliary equipment and participants with neurological disorders that cause movement errors requiring accurate motor training. The current study investigates task-dependent modulation by simultaneously measuring high time resolution electroencephalogram, electromyogram and force along with bi-manual and unimanual motor tasks. The findings show that visual feedback for right hand decreases the force root mean square error of right hand. Visual feedback for right hand decreases local and global efficiency of brain network in theta and alpha bands.

Simulation and experiment on obstacle avoidance control of concrete pump truck boom based on improved danger field and gradient projection method

Aiming at the shortcomings of the traditional robot obstacle avoidance algorithm applied directly to the leader–follower task transformation of concrete pump truck boom, an algorithm combined with improved danger field and improved gradient projection method for obstacle avoidance control of the boom is proposed. In the method, a joint limit avoidance function is used to avoid the angle overrun of the boom joints. And the danger field expression is modified to improve its engineering suitability. Moreover, a smoothing adjustment factor is introduced to improve the smoothness of the leader–follower task transformation of the boom. The influence of key control parameters on the safety and accuracy of boom movement is discussed through the simulation, and the reasonable ranges of the parameters are given. Furthermore, experiments are carried out to verify the effectiveness of the algorithm. The method provides a solution for obstacle avoidance control of construction machinery with multijoint series boom.

Development of a New Eye Movement Measurement Device Using Eye-Tracking Analysis Technology

Smooth pursuit eye movements and saccadic eye movements are vital for precise vision. Therefore, tests for eye movement are important for assessing nervous or muscular diseases. However, objective measurements are not frequently performed due to the need for a polygraph system, electrodes, amplifier, and personal computer for data analysis. To address this, we developed an all-in-one eye-movement-measuring device that simultaneously presents visual stimuli, records eye positions, and examines its feasibility for evaluating eye movements. This device generates stimulus that induces eye movements and records those movements continuously. The horizontal or vertical eye movements of 16 participants were measured at various visual target speeds of 20–100 deg/s. The maximum cross-correlation coefficient (rho max) between the eye and visual target positions was used as an index of eye movement accuracy. A repeated-measures multi-way analysis of variance was performed, with the main effect being that rho max significantly decreased as the visual target speed increased. The average (±standard deviation) rho max values across all velocities were 0.995 ± 0.008 and 0.967 ± 0.062 in the horizontal and vertical directions, respectively, and were significantly higher for horizontal eye movements than for vertical eye movements. Moreover, rho max and saccadic frequency were significantly correlated for the slowest and fastest visual target motions. These suggest that our device enables accurate measurements of eye movements. We believe our new measurement device can be applied clinically for easily and objectively evaluating eye movements.

Design and Implementation of Iterative Learning Control for an Electro-Hydraulic Servo System

In order to produce an accurate movement of double-acting hydraulic cylinder of the Electro-Hydraulic Servo System (EHSS), an Iterative Learning Controller (ILC) approach is implemented in this work. Nonlinearity and imprecision occurs in the hydraulic systems because of friction. Traditional controllers are incapable of providing effective control performance throughout the entire operating range and insufficient to handle repetitive task.&nbsp;To manage the repetitive task, a memory-based learning control analysis is suitable. This research focuses to construct the ILC for governing the servo spool valve, which is responsible for the hydraulic cylinder displacement. The proposed ILC consists of learning filter, learning gain and robustness filter. Proportional, Integral and Derivative (PID) controller is devised to validate the outcome of ILC. Controllers are constructed based on the system modelling. The effectiveness of the suggested controller is shown through simulation and experimental findings. ILC provides an average of 50% minimal overshoot and settles 7 sec before the PID controller is designed. Due to model uncertainty, PID controller results 0.2 sec better rise time than ILC. In ILC's architecture objective function is designed to achieve minimal overshoot and quick settling of hydraulic cylinder. So, no consideration is given to the error indices. These findings will help hydraulic stamping application, which requires the accurate displacement of piston to avoid damage of work piece. These results show that the proposed controller achieves desired outcome displacement of hydraulic cylinder.

Parasitic Motions of 3-PRS Parallel Mechanisms with Two Different Branch Chain Arrangements

The kinematics of 3-PRS parallel mechanisms (PMs), an important category of 2R1T PMs, is very important for the motion control and error compensation of PMs. In order to produce a spindle head that could meet the requirements of various machining accuracy levels, a definite kinematics analysis of the PMs par should be taken with caution firstly. Here, the mathematical models of two 3-PRS PMs with different branch chain arrangements were established using the Euler angle. The kinematic equations of two 3-PRS PMs were obtained by establishing the closed-vector loop equation according to structural conditions. Parasitic motions that would cause a great effect were obtained, and their variations with structural and kinematic parameters were also obtained through MATLAB. Since the relationship between parasitic motions, and the given structural and kinematic parameters, is inextricable, an analysis is performed for the purpose of reducing or eliminating parasitic motion to ensure the accuracy of end-effector movement. The caused effect of parasitic motion can be improved with scale optimization and precision compensation. Lastly, two improved 3-PRS PMs without the parasitic motion of different branch chain arrangements used as the spindle head were obtained. Parasitic motion was related to the mechanism configuration and could be avoided in the configuration design stage with the comprehensive optimization of the structural and kinematic parameters, which provides a reference to obtain a sufficient workspace and improve the precision of traditional five-axis CNC machine tools.

Reaching Accuracy Assessment in Cerebellar Stroke using Virtual Reality (S20.010)

<h3>Objective:</h3> To investigate the application of virtual reality (VR) in the rapid quantification of reaching accuracy at the bedside for patients with cerebellar stroke (CS). <h3>Background:</h3> Dysmetria, the inability to measure distance in muscular tasks correctly, is a characteristic clinical feature of cerebellar injury. Even though dysmetria can be quickly detected during the neurological examination with the finger-to-nose test, objective quantification of reaching accuracy for clinical assessment is still lacking. Emerging VR technology, together with recent improvements in the hand-tracking feature, offers an opportunity to closely examine the speed, accuracy, and consistency of fine hand movements and proprioceptive function. <h3>Design/Methods:</h3> 29 individuals (10 CS patients and 19 age-matched not-disabled controls) performed a task measuring reaching accuracy on the VR headset (Oculus Quest 2). 50% of the trials displayed a visible rendering of the hand as the participant reached for the target (visible hand condition), while the remaining 50% only showed the target being extinguished (invisible hand condition). Reaching error was calculated as the difference in degrees between where the fingertip passed the arch and where the target was positioned. <h3>Results:</h3> Reaching error was higher in CS than in age-matched controls in both visible hand and invisible hand conditions. Reaching error was higher in the invisible hand condition than in the visible hand condition in both healthy controls and CS patients. Average time taken to perform each trial was higher in CS patients than in controls in both visible and invisible hand conditions. <h3>Conclusions:</h3> Reaching accuracy assessed by VR promises to be a non-invasive and rapid approach to quantifying fine motor functions in clinical settings. In addition, this device has the potential to be a useful supplemental technology in monitoring fine motor and proprioceptive functions during physical rehabilitation. Further studies are needed to examine quantitative changes in reaching accuracy during post-stroke progression. <b>Disclosure:</b> Mr. Du has nothing to disclose. Mr. Benavides has received personal compensation for serving as an employee of Stryker. Emily Isenstein has received research support from Autism Science Foundation. Prof. Tadin has received personal compensation in the range of $50,000-$99,999 for serving as a Consultant for NeuroTrainer. Prof. Tadin has received personal compensation in the range of $10,000-$49,999 for serving as an Expert Witness for Finkelstein and partners. The institution of Prof. Tadin has received research support from NIH. The institution of Prof. Tadin has received research support from The Brain &amp; Behavior Research Foundation. Prof. Tadin has received intellectual property interests from a discovery or technology relating to health care. Dr. Busza has received research support from NIH/NINDS. Dr. Busza has received personal compensation in the range of $500-$4,999 for serving as a Grant reviewer with NIH.

An Exploration of Robot-Mediated Tai Chi Exercise for Older Adults

In this fast-aging society, many older adults fail to meet the required level of exercise due to trainer shortages. Therefore, we developed a robot tutor to investigate the feasibility of robot-mediated exercise for older adults. Twenty older adults participated in an experimental study. A pre-exercise survey was used to assess their background. Each participant experienced a 30-min robot-led Tai Chi exercise followed by a post-exercise survey to evaluate the easiness of following the robot and expectations for future robot design. Participants’ Tai Chi performances were evaluated in terms of completion and accuracy. Associations between the surveys and the performance were also analyzed. All participants completed the study. Fifteen out of the twenty subjects had at least one chronic condition, and most practiced Tai Chi before the study but had never interacted with a robot. On average, the participants scored 93.09 and 85.21 out of 100 for movement completion and accuracy, respectively. Their initial movement accuracy was correlated with their attitude towards exercise. Most subjects reported that they could follow the robot’s movements and speeches well and were interested in using a robot tutor in the community. The study demonstrated the initial feasibility of robot-led Tai Chi exercise for older adults.