Human Motion Intention Research Articles

Learn to Adapt to Human Walking: A Model-Based Reinforcement Learning Approach for a Robotic Assistant Rollator

In this letter, we tackle the problem of adapting the motion of a robotic assistant rollator to patients with different mobility status. The goal is to achieve a coupled human–robot motion in a front-following setting as if the patient was pushing the rollator himself/herself. To this end, we propose a novel approach using model-based reinforcement learning (MBRL) for adapting the control policy of the robotic assistant. This approach encapsulates our previous work on human tracking and gait analysis from RGB-D and laser streams into a human-in-the-loop decision making strategy. We use long short-term memory (LSTM) networks for designing a human motion intention model and a coupling parameters forecast model, leveraging on the outcome of human gait analysis. An initial LSTM-based policy network was trained via imitation learning from human demonstrations in a motion capture setup. This policy is then fine-tuned with the MBRL framework using tracking data from real patients. A thorough evaluation analysis proves the efficiency of the MBRL approach as a user-adaptive controller.

Neural-network-enhanced torque estimation control of a soft wearable exoskeleton for elbow assistance

Soft wearable exoskeletons are a new approach for the applications of power assistance and rehabilitation training. In the present work, a neural-network-enhanced torque estimation controller (NNETEC) is proposed for a soft wearable elbow assistance exoskeleton with compliant tendon-sheath actuator. A comprehensive overview for the major components of the soft exoskeleton is introduced. The locations of anchor points are optimized via the maximum-stiffness principle. The NNETEC strategy is developed by fusing the feedback signals from surface electromyography (sEMG) sensors, inertial measurement units, force sensors, and motor encoder. It consists of a joint torque estimation module to identify the elbow torque of wearer based on Kalman filter, a neural-network adjustment module to recognize human motion intention, and a proportional-integral-derivative controller with hybrid position/torque feedbacks. Further experimental investigations are carried out by five volunteers to validate the effectiveness of the proposed soft elbow exoskeleton and control strategy. The results of the dumbbell-lifting experiments with various weights and frequencies demonstrate that, when compared with the proportional control strategy and the sEMG-based assistive control strategy without neural-network adjustment, the developed NNETEC method can achieve higher power assistance efficiency.

Real-Time Human Intention Recognition of Multi-Joints Based on MYO

Hill musculoskeletal model (HMM) is commonly used to estimate human motion intentions. HMM utilizes electromyography (EMG) signals as the nonlinear model input to obtain muscle forces or torques. However, due to the fact that it contains many physiological parameters that are difficult to measure, HMM is generally applied in simple continuous intention estimation of a single joint. In this work, we aimed at recognizing shoulder and elbow joints angles and their angular velocities continuously in real time. Firstly, we used MYO armband as the EMG sensor. Then, a reasonable prediction model was deduced based on HMM and human dynamics to realize online continuous recognition of the four angles and angular velocities of shoulder and elbow joints. Nonlinear autoregressive with external input neural network (NARX) replaced the prediction equation. In addition, the framework of state space model was completed by constructing an observation equation. Thus, the closed-loop characteristic was realized to eliminate the influence of cumulative error and ensure good estimation performance. Experimental results verified the feasibility and accuracy of the algorithm. For predefined trajectory and random trajectory separately, the RMSE were 0.955 and 1.15 (degree) for angles estimation and 2.8, 3.40 for angular velocities (degree/s). Compared with the normally used back-propagation neural network (BPNN), the method proposed in this paper obviously got more accurate and smooth results.

A Multi-Mode Rehabilitation Robot With Magnetorheological Actuators Based on Human Motion Intention Estimation.

Lower extremity paralysis has become common in recent years, and robots have been developed to help patients recover from it. This paper presents such a robotic system that allows for two working modes, the robot-active mode and human-active mode. The robot is designed to be equipped with magnetorheological (MR) actuators that have the advantages of high torque, fast response, flexible controllability, low power consumption and safety guarantee. The design and characteristics of the MR actuator are introduced. In the robot-active mode, the MR actuator works as a clutch to transfer the torque to the robotic joint safely. In the human-active mode, the MR actuator functions as a brake to provide resistance to help strengthen muscles. The working mode is determined by the human motion intention, which is detected via the skin surface electromyography (EMG) signals. The human-robot interaction torques are estimated using the EMG-driven impedance model. The biomechanical analysis based on AnyBody Modeling System (AMS) is used to help optimization. Then, an adaptive control method is proposed to realize the assist-as-needed (AAN) training strategy, where the robot can switch between these two modes. Experiments are conducted to validate the proposed design.

Sequential Decision Fusion for Environmental Classification in Assistive Walking.

Powered prostheses are effective for helping amputees walk in a single environment, but these devices are inconvenient to use in complex environments. In order to help amputees walk in complex environments, prostheses need to understand the motion intent of amputees. Recently, researchers have found that vision sensors can be utilized to classify environments and predict the motion intent of amputees. Although previous studies have been able to classify environments accurately in offline analysis, the corresponding time delay has not been considered. To increase the accuracy and decrease the time delay of environmental classification, the present paper proposes a new decision fusion method. In this method, the sequential decisions of environmental classification are fused by constructing a hidden Markov model and designing a transition probability matrix. The developed method is evaluated by inviting five able-bodied subjects and three amputees to perform indoor and outdoor walking experiments. The results indicate that the proposed method can classify environments with accuracy improvements of 1.01% (indoor) and 2.48% (outdoor) over the previous voting method when a delay of only one frame is incorporated. The present method also achieves higher classification accuracy than with the methods of recurrent neural network (RNN), long-short term memory (LSTM), and gated recurrent unit (GRU). When achieving the same classification accuracy, the method of the present paper can decrease the time delay by 67 ms (indoor) and 733 ms (outdoor) in comparison to the previous voting method. Besides classifying environments, the proposed decision fusion method may be able to optimize the sequential predictions of the human motion intent.

SEMG Based Human Motion Intention Recognition

Human motion intention recognition is a key to achieve perfect human-machine coordination and wearing comfort of wearable robots. Surface electromyography (sEMG), as a bioelectrical signal, generates prior to the corresponding motion and reflects the human motion intention directly. Thus, a better human-machine interaction can be achieved by using sEMG based motion intention recognition. In this paper, we review and discuss the state of the art of the sEMG based motion intention recognition that is mainly used in detail. According to the method adopted, motion intention recognition is divided into two groups: sEMG-driven musculoskeletal (MS) model based motion intention recognition and machine learning (ML) model based motion intention recognition. The specific models and recognition effects of each study are analyzed and systematically compared. Finally, a discussion of the existing problems in the current studies, major advances, and future challenges is presented.

Sensorless force estimation of end-effect upper limb rehabilitation robot system with friction compensation

In human robot interaction systems, human intent detection plays an important role to improve the interactive performances and then the rehabilitation effects. A study is proposed to estimate the interactive forces that indirectly detect the human motion intent. A disturbance observer is designed to estimate interactive torques and friction forces without force sensors, and then a friction force model is constructed to estimate the friction force in the robot system. To detect the human–robot interaction force, we subtract the friction force from disturbance observer estimation result. Several experiments were performed to test the performances of the proposed methods. Those methods were applied in an end-effect upper limb rehabilitation robot system. The results show that the precision of the estimated sensor force can increase 5% than the force sensor. The senseless force estimation method we proposed in this article can be an alternative option in force control tasks when force sensors are not suitable.

Adaptive Hybrid Classifier for Myoelectric Pattern Recognition Against the Interferences of Outlier Motion, Muscle Fatigue, and Electrode Doffing.

Traditional myoelectric prostheses that employ a static pattern recognition model to identify human movement intention from surface electromyography (sEMG) signals hardly adapt to the changes in the sEMG characteristics caused by interferences from daily activities, which hinders the clinical applications of such prostheses. In this paper, we focus on methods to reduce or eliminate the impacts of three types of daily interferences on myoelectric pattern recognition (MPR), i.e., outlier motion, muscle fatigue, and electrode doffing/donning. We constructed an adaptive incremental hybrid classifier (AIHC) by combining one-class support vector data description and multi-class linear discriminant analysis in conjunction with two specific update schemes. We developed an AIHC-based MPR strategy to improve the robustness of MPR against the three interferences. Extensive experiments on hand-motion recognition were conducted to demonstrate the performance of the proposed method. Experimental results show that the AIHC has significant advantages over non-adaptive classifiers under various interferences, with improvements in the classification accuracy ranging from 7.1% to 39% ( ). The additional evaluations on data deviations demonstrate that the AIHC can accommodate large-scale changes in the sEMG characteristics, revealing the potential of the AIHC-based MPR strategy in the development of clinical myoelectric prostheses.

IMU-Based Gait Phase Recognition for Stroke Survivors

SummaryUsing wearable robots is an effective means of rehabilitation for stroke survivors, and effective recognition of human motion intentions is a key premise in controlling wearable robots. In this paper, we propose an inertial measurement unit (IMU)-based gait phase detection system. The system consists of two IMUs that are tied on the thigh and on the shank, respectively, for collecting acceleration and angular velocity. Features were extracted using a sliding window of 150 ms in length, which was then fed into a quadratic discriminant analysis (QDA) classifier for classification. We recruited five stroke survivors to test our system. They walked at their own preferred speed on the level ground. Experimental results show that our proposed system has the ability of recognizing the gait phase of stroke survivors. All recognition accuracy results are above 96.5%, and detections are about 5–15 ms in advance of time. In addition, using only one IMU can also give reliable recognition results. This paper proposes an idea about the further research on human–computer interaction for the control of wearable robots.

The Recognition of Motion Intention of Knee Joint Based on Piezoelectric Signals

Aiming at the human body exoskeleton motion intention recognition process, based on the pressure signals of muscle groups of control, building joint motion information acquisition system, the application of multi-sensor information fusion technique based on neural network, based on the piezoelectric signal muscle as input and output of the joint angle trajectory motion intention model, calculate the joint exercise intention and angle track. Finally, the knee flexion and extension movement as the main object of study, through calculation and practical test, the ideal results are obtained. It shows that the motion signals of the knee joint can be reflected by the piezoelectric signals collected during the motion of the knee joint.

Biomechanics, actuation, and multi-level control strategies of power-augmentation lower extremity exoskeletons: an overview

Improper manipulation of heavy objects can result in hard stresses (tension, compression and shear) throughout the human body parts, especially in the low-back spine. Biomechanics specialists state that injuries that occur in this area may address muscle tissue, joint tissues and intervertebral disc tissues. The effect of a carried load on kinematics and kinetics of body’s lower extremity is significant. Besides of labor protection rules and interventions designed to reduce and avoid injuries, powered wearable exoskeletons have been proposed to amplify human capabilities. The paper regards three significant issues related to exoskeletons: biomechanical modeling, actuation, and multi-level control strategies. Three modalities to get optimal performance of wearable robots are hereby summarized: (i) minimization of interaction force wrench by using direct/indirect force control strategies, (ii) modification of reference trajectory to compensate for unwanted interaction force wrench, and (iii) adding the power assist rate such that zero impedance at interaction attachments is guaranteed. To accomplish these points, most proposed control strategies consist of three levels of control: high-level control, responsible for capturing human movement intention; mid-level control for regulation of divisions of the gait cycle; and low-level control for stabilization of the coupled motion.

Proportional myoelectric and compensating control of a cable-conduit mechanism-driven upper limb exoskeleton

Recent studies have indicated that human motion recognition based on surface electromyography (sEMG) is a reliable and natural method for achieving motion intention. However, achieving accurate estimates of intended motion using a low computational cost is the main challenge in this scenario. In this study, a proportional myoelectric and compensating control method for estimating and assisting human motion intention with a cable-conduit mechanism-driven upper limb exoskeleton was proposed. The integral signal of sEMG and its time-delayed signals were applied as a new feature vector to represent the role of sEMG, which ensured the accuracy and real-time performance of motion estimation. An integrated circuit was used to reduce time of feature extraction. A feed-forward compensator was designed to compensate for the effect of the hysteresis problem in the exoskeleton, which is inevitable when the cable-conduit mechanism was applied to reduce the exoskeleton weight. The model-free control method based on PID method and least squares support vector machine were applied to avoid calculating the complex biomechanical model of human upper limb and the dynamic model of exoskeleton. Experimental results validated the proposed method. The average values of the root-mean-square difference (RMSD) for motion estimation were 0.0579 ± 0.0085 [motion with constant pace (CP)] and 0.0845 ± 0.0137 [motion with variable pace (VP)]. The Bland–Altman analysis results showed that the estimated angle of the proposed method was consistent with the actual angle. The performance of the control method was good, and the accuracies were 98.5608% ± 0.4485% (motion with CP) and 96.6119% ± 0.6628% (motion with VP).

A Novel Weight-Bearing Lower Limb Exoskeleton Based on Motion Intention Prediction and Locomotion State Identification

A variable magnification ratio transmission structure powered by the electric actuators is proposed to improve the flexibility and portability of the exoskeleton under heavy load carrying condition. The parameters of connecting rod size and hanging position are optimized to ensure that the output torque of active joints can fully envelope the demand load area. The control strategy based on intrinsic sensing is designed to realize the automatic human motion intention prediction and flexible trajectory tracking. The newly developed split embedded connecting rod can accurately measure the human-robot interaction (HRI) force applied to the exoskeleton and extract the human motion intention without being affected by the differences in wearing status. The force tracking control based on the zero-force following is modified by feedforward compensation with extreme learning machine (ELM), which enhances the response speed to human motion intention and reduces the HRI force by 70.6%. Based on multi-sensor information, stacked autoencoder deep neural networks (DNNs) are utilized to realize the automatic locomotion transition and the corresponding control parameters' switching. After optimization by a hybrid algorithm of genetic algorithm and particle swarm optimization (GA_PSO), the identification accuracy is enhanced from 96.2% to 99.7%. The adaptive neural-fuzzy inference system (ANFIS) is used to analyze the plantar pressure to achieve flexible switching between the swing phase and the stance phase. The experiments under various gait motion trajectories assisted by novel weight-bearing exoskeleton are carried out for evaluation, and the performance of the proposed control strategy based on motion intention prediction, locomotion mode identification, and gait phase switching is effectively verified.

Development and Hybrid Control of an Electrically Actuated Lower Limb Exoskeleton for Motion Assistance

This paper describes a system design and hybrid control algorithm of an electrically actuated lower limb exoskeleton (LLE). The system design mainly includes three parts: mechanical structure design, actuation system design and sensor system design. According to the initial state of the joint angle, LLE can be divided into Non-anthropomorphic state (NAS) and anthropomorphic state (AS). The human motion intention (HMI) estimation can be divided into gait phase classification and reference trajectory estimation. The fuzzy logic is used to detect different phases in the gait phase classification. In the reference trajectory estimation, the kinematic model of the LLE is utilized to obtain a continuous joint trajectory, which is used as input of the control law. To make the LLE accurately follow the movement of people and remain stable, a hybrid dual-mode control strategy is proposed in this paper, i.e., the adaptive impedance control (AIC) method is used to improve the stability and resistance to shock in stance phase, and the active disturbance rejection control with the fast terminal sliding mode control (ADRC-FTSMC) method is employed to improve the response speed and the tracking precision in swing phase. Furthermore, in order to solve the torque discontinuity in the switching process, a smoothing method is proposed during the transition. Finally, the prototype experiments were set up to verify the tracking performance and power-assisted effect of the proposed exoskeleton. The experiments results show the LLE can achieve excellent tracking performance and power-assisted effect based on the proposed HMI methodology and hybrid dual-mode control strategy.

Deep Learning-based Human Motion Prediction considering Context Awareness for Human-Robot Collaboration in Manufacturing

The interest of human-robot collaboration (HRC) for intelligent manufacturing service system is gradually increasing. Fluent human-robot coexistence in manufacturing requires accurate estimation of the human motion intention so that the efficiency and safety of HRC can be guaranteed. Human motion is mainly defined as the sequential positions of the joints of human skeletons among traditional motion prediction solutions, which lead to a deficiency of tools or product components holding in hand. Context awareness based temporal processing is the key to evaluating human motion before the accomplishment of it, so as to save time as well as recognize the intention of the human. In this paper, a deep learning system combing convolutional neural network (CNN) and long short-term memory network (LSTM) towards vision signals is explored to predict human motion accurately. Creatively, this paper utilizes LSTM to extract temporal patterns of human motion automatically outputting the prediction result before motion takes place. Not only does it avoid complex feature extraction due to its end-to-end characteristic, but provide a natural interaction between human and robot without wearable devices or tags that may become a burden for the former. A case study of desktop computer product disassembly is executed to demonstrate the feasibility of the recommended method. Experimental performance proves that our method outperforms the other three optimization algorithms on the prediction accuracy.

Enhanced Transparency for Physical Human-Robot Interaction Using Human Hand Impedance Compensation

In a physical human-robot interaction (pHRi) system, improving transparency that allows humans to move as if there is no robot is a challenging topic. In general pHRi, usage of the multiaxial force sensor for robot control is the norm. However, the signal measured from the force sensor contains not only the force applied to the robot by the human motion intention but also the influence of the natural force feedback between the robot and the human hand produced by the robot motion. Therefore, in order to improve the transparency, it is necessary to characterize the dynamics of the human hand as well as the dynamics of the robot. In this paper, an algorithm to improve the transparency is proposed. The proposed algorithm uses only a multiaxial force sensor and compensates the natural force feedback by using the human hand impedance. And it further improves the transparency of the pHRi system, which has a limitation adjusting admittance parameters. In order to verify the algorithm, a device that can measure the impedance of the human hand is introduced, and a system model analysis and experiments using the hydraulic upper limb exoskeleton are carried out.

A Novel Coordinated Motion Fusion-Based Walking-Aid Robot System.

Human locomotion is a coordinated motion between the upper and lower limbs, which should be considered in terms of both the user’s normal walking state and abnormal walking state for a walking-aid robot system. Therefore, a novel coordinated motion fusion-based walking-aid robot system was proposed. To develop the accurate human motion intention (HMI) of such robots when the user is in normal walking state, force-sensing resistor (FSR) sensors and a laser range finder (LRF) are used to detect the two HMIs expressed by the user’s upper and lower limbs. Then, a fuzzy logic control (FLC)-Kalman filter (LF)-based coordinated motion fusion algorithm is proposed to synthesize these two segmental HMIs to obtain an accurate HMI. A support vector machine (SVM)-based fall detection algorithm is used to detect whether the user is going to fall and to distinguish the user’s falling mode when he/she is in an abnormal walking state. The experimental results verify the effectiveness of the proposed algorithms.

Mechanical Design and Control Strategy for Hip Joint Power Assisting.

The basic requirements for mechanical design and control strategy are adapting to human joint movements and building an interaction model between human and robot. In this paper, a 3-UPS parallel mechanism is adopted to realize that the instantaneous rotation center of the assistive system coincides with human joint movement center, and a force sensory system is used to detect human movement intention and build the modeling of control strategy based on the interactive force. Then, based on the constructed experimental platform, the feasibility of movement intention detection and power assisting are verified through the experimental results.

On facilitating method for skill acquisition of robot arm manipulation using surface myoelectric signals

In these days, various control systems for robot arms using surface myoelectric signals have been developed. Abundant pattern recognition techniques have been proposed to predict human motion intent based on these signals. However, it may become burdensome for users to train voluntary control of myoelectric signals with these systems. In this study, we are seeking to develop schemes to evoke surface myoelectric signals and facilitate skill acquisition of robot arm manipulation. In this paper, we construct a simple 1-link robot arm which is controlled by estimating the wrist motion from the surface myoelectric signals on the forearm. We make several training experiments for skill acquisition of robot arm manipulation with and without physical constraints of the wrist, or with and without mechanical vibration stimulation. Then, we investigate the differences in facilitation effects of skill acquisition comparing with the results from these training experiments.

A novel hybrid closed-loop control approach for dexterous prosthetic hand based on myoelectric control and electrical stimulation

PurposeThe purpose of this study is to present a novel hybrid closed-loop control method together with its performance validation for the dexterous prosthetic hand.Design/methodology/approachThe hybrid closed-loop control is composed of a high-level closed-loop control with the user in the closed loop and a low-level closed-loop control for the direct robot motion control. The authors construct the high-level control loop by using electromyography (EMG)-based human motion intent decoding and electrical stimulation (ES)-based sensory feedback. The human motion intent is decoded by a finite state machine, which can achieve both the patterned motion control and the proportional force control. The sensory feedback is in the form of transcutaneous electrical nerve stimulation (TENS) with spatial-frequency modulation. To suppress the TENS interfering noise, the authors propose biphasic TENS to concentrate the stimulation current and the variable step-size least mean square adaptive filter to cancel the noise. Eight subjects participated in the validation experiments, including pattern selection and egg grasping tasks, to investigate the feasibility of the hybrid closed-loop control in clinical use.FindingsThe proposed noise cancellation method largely reduces the ES noise artifacts in the EMG electrodes by 18.5 dB on average. Compared with the open-loop control, the proposed hybrid closed-loop control method significantly improves both the pattern selection efficiency and the egg grasping success rate, both in blind operating scenarios (improved by 1.86 s, p < 0.001, and 63.7 per cent, p < 0.001) or in common operating scenarios (improved by 0.49 s, p = 0.008, and 41.3 per cent, p < 0.001).Practical implicationsThe proposed hybrid closed-loop control method can be implemented on a prosthetic hand to improve the operation efficiency and accuracy for fragile objects such as eggs.Originality/valueThe primary contribution is the proposal of the hybrid closed-loop control, the spatial-frequency modulation method for the sensory feedback and the noise cancellation method for the integrating of the myoelectric control and the ES-based sensory feedback.