Human movement is produced resulting from synergetic combinations of multiple muscle contractions. The resultant joint movement can be estimated through the related multiple-muscle activities, which is formulated as the forward problem. Neuroprosthetic applications may benefit from cocontraction of agonist and antagonist muscle pairs to achieve more stable and robust joint movements. It is necessary to estimate the activation of each individual muscle from desired joint torque(s), which is the inverse problem. A synergy-based solution is presented for the inverse estimation of multiple muscle activations from joint movement, focusing on one degree-of-freedom tasks. The approach comprises muscle synergy extraction via the nonnegative matrix factorization algorithm. Cross validation is performed to evaluate the method for prediction accuracy based on experimental data from ten able-bodied subjects. The results demonstrate that the approach succeeds to inversely estimate the multiple muscle activities from the given joint torque sequence. In addition, the other one's averaged synergy ratio was applied for muscle activation estimation with leave-one-out cross-validation manner, which resulted in 9.3% estimation error over all the subjects. The obtained results support the common muscle synergy-based neuroprosthetics control concept.

Functional electrical stimulation (FES) is able to restore motor function of spinal cord injured (SCI) patients. To make adaptive FES control taking into account the actual muscle state with muscular feedback information, torque estimation and prediction are important to be provided beforehand. Evoked EMG (eEMG) has been found to be highly correlated with FES-induced torque under various muscle conditions, indicating that it can be an useful tool for torque/force prediction. To better construct the relationship between eEMG and stimulated muscular torque, nonlinear-arx-type (NARX-type) model is preferred. This paper presents and exploits a NARX-type recurrent neural network (NARX-RNN) model for identification and prediction of FES-induced muscular dynamics with eEMG. Such NARX-RNN model is with a novel architecture for prediction, with robust prediction performance. To make fast convergence for identification of such NARX-RNN, directly-learning pattern is exploited during the learning phase. Due to difficulty of choosing a proper forgetting factor of Kalman filter for predicting time-variant torque with eEMG, such NARX-RNN may be considered to be a better alternative as torque predictor. Data gathered from two SCI patients is used to evaluate the proposed NARX-RNN model. The NARX-RNN model shows promising estimation and prediction performance only based on eEMG.

Functional electrical stimulation (FES) is effective to restore movement in spinal cord injured (SCI) subjects. Unfortunately, muscle fatigue constrains the application of FES so that output torque feedback is interesting for fatigue compensation. Whereas, inadequacy of torque sensors is another challenge for FES control. Torque estimation is thereby essential in fatigue tracking task for practical FES employment. In this work, the Hammstein cascade with electromyography (EMG) as input is applied to model the myoelectrical mechanical behavior of the stimulated muscle. Kalman filter with forgetting factor is presented to estimate the muscle model and track fatigue. Fatigue inducing protocol was conducted on three SCI subjects through surface electrical stimulation. Assessment in simulation and with experimental data reveals that the muscle model properly fits the muscle behavior well. Moreover, the time-varying parameters tracking performance in simulation is efficient such that real time tracking is feasible with Kalman filter. The fatigue tracking with experimental data further demonstrates that the proposed method is suitable for fatigue tracking as well as adaptive torque prediction at different prediction horizons.

This paper concerns the whole physiological parameters identification of a musculoskeletal model of a human subject. The patient is equipped with an implanted Functional Electrical Stimulation (FES) system as part of the SUAW's European project. The biomechanical model represents the knee and its associated muscles. The identification protocol is noninvasive and based on the in-vivo experimental data acquisition of a Spinal Cord Injured (SCI) patient. However, the human noninvasive identification poses problems of inaccessibility to some data. The identification procedure consists of several steps, in order to identify: the anthropometrical parameters, the geometrical parameters, the joint mechanical parameters, the force-length relationship and the recruitment function. Up to now, only the quadriceps muscle is considered with the knee joint in the identification procedure. A cross-validation has been done using data set not used during the identification process. The identified model shows a satisfactory response comparing to the measured knee response, which is obtained by stimulating the quadriceps through the implanted FES system. In this work, knee model-parameters of the implanted subject were identified successfully using the noninvasive identification procedure.

EMG-to-force estimation for voluntary muscle contraction has many applications in human-machine interaction, motion analysis, and rehabilitation robotics for prosthetic limbs or exoskeletons. EMG-based model can account for a subject's individual activation patterns to estimate muscle force. For the estimation, so-called Hill-type model has been used in most of the cases. It already has shown its promising performance, but it is still known as a phenomenological model considering only macroscopic physiology. We have already developed the physiological based muscle model for the use of functional electrical stimulation (FES) which can render the myoelectrical property also in microscopic scale. In this paper we discuss EMG-to-force estimation based on this full physiological based muscle model in voluntary contraction. In addition to Hill macroscopic structure, a microscopic physiology originally designed by Huxley is integrated. It has significant meaning to realize the same kind of EMG-to-force estimation with a physiological based model not with a phenomenological model, because it brings the understanding of the internal biophysical dynamics and new insights about neuromuscular activations. Using same EMG data of isometric muscle contraction, the force estimation results are shown by classical approach and new physiological based approach. Its interpretation is also discussed.

A model-based FES would be very helpful for the adaptive movement synthesis of spinal-cord-injured patients. For the fulfillment, we need a precise skeletal muscle model to predict the force of each muscle. Thus, we have to estimate many unknown parameters in the nonlinear muscle system. The identification process is essential for the realistic force prediction. We previously proposed a mathematical muscle model of skeletal muscle which describes the complex physiological system of skeletal muscle based on the macroscopic Hill-Maxwell and microscopic Huxley concepts. It has an original skeletal muscle model to enable consideration for the muscular masses and the viscous frictions caused by the muscle-tendon complex. In this paper, we present an experimental identification method of biomechanical parameters using Sigma-Point Kalman Filter applied to the nonlinear skeletal muscle model. Result of the identification shows its effective performance. The evaluation is provided by comparing the estimated isometric force with experimental data with the stimulation of the rabbit medial gastrocnemius muscle. This approach has the advantage of fast and robust computation, that can be implemented for online application of FES control.

The work presented in this paper concerns the synthesis of Functional Electrical Stimulation (FES) patterns to generate movements of paralysed limbs for spinal cord injured patients. We propose an approach based on a nonlinear optimization formulation that may encounter physiological and technological constraints. The study considers a biomechanical knee model and the associated agonist/antagonist muscles. The goal of this method is to synthesize optimal patterns which minimize the muscular activities and/or tracking trajectory errors in order to reduce the muscular fatigue while achieving a desired movement. Different tests have been performed and the results compared with regard to the energetic balance. The approach is illustrated in simulation with: 1) sinusoidal desired knee joint trajectory, 2) optimal reference knee joint trajectory and 3) without explicit reference knee joint trajectory. The simulations have been performed with model parameters estimated from real subject data. We show that the trajectory tracking presents high energy consumption which demonstrates the inappropriateness of classical robotics methods for musculoskeletal system. Instead, minimization of muscle activation only gives better results with regard to energy consumption, still with a reasonnable trajectory tracking error.

In this paper, we deal with complementary aspects of an implanted functional electrical stimulation control architecture: distributed stimulation units, deterministic implanted network, architecture implantation (electronic devices) and control over this network. We argue from analogy with embedded control architecture, of robots for example, from distributed control, closed-loop control, supervisory control and fieldbus considerations. The global embedded FES architecture on which can rely a functional rehabilitation is presented. We give some details on the distributed entities of this architecture as they constitute the interface between the artificial and living ldquoworldsrdquo, and we expose their interconnection through an intrabody network which is similar to an implanted fieldbus.

This paper addresses the problem of restoring standing in paralegia via functional electrical stimulation (FES) and investigates the relashionship between body posture and voluntary upper-body movements. A methodology is presented for upper-body posture estimation in the sagittal plane from force and torque measurements exerted on handles during human standing, in the hypothesis of quasi-static equilibrium. The method consists in setting up constraints related to the geometric equations and the hand-handle interaction. All measured quantities are subject to an uncertainty assumed unknown but bounded. The set membership estimation problem is solved via interval analysis. Guaranteed uncertainty bounds are computed for the estimated postures. The methodology is validated experimentally with spinal cord injured patients with lesions between T5 and T12. Possible applications of the developed methodology are lower limbs function rehabilitation within clinical centers, walk assistance and independent mobility for spinal cord injured patients.

The main challenge that we face when applying functional electrical stimulation (FES) to paralyzed lower limbs is to avoid hyperstimulation and to defer the muscular fatigue as much as possible. FES is used to excite paralyzed muscles that are under lesions and consequently no more controlled by paraplegic patients. We aimed in this study to compute the needed patterns stimulation necessary to perform a desired given motion of the knee joint. We coupled the exact input output feedback linearization with a model predictive controller (MPC). This latter enables us to incorporate explicitly constraints on inputs, outputs and system states. Internal dynamics stability was mathematically proved and MPC performances were compared to a classical pole placement controller in terms of robustness, stability and finite time convergence.

The main goal when applying functional electrical stimulation (FES) to the paralyzed lower limbs of the paraplegic patients is to avoid hyperstimulation and to defer the muscular fatigue as much as possible. In this paper a closed loop position control of the knee joint actuated by the quadriceps muscle to perform flexion-extension has been presented. The feedback control consists of a model predictive control (MPC) technique which is also known by a receding horizon control or moving horizon control. This controller is applied to a complex physio-mathematical muscle model that is based on a macroscopic Hill and a microscopic Huxley concepts. An MPC constitutes an adequate controller with nonlinear multivariable systems. Furthermore it enables us to incorporate explicitly constraints on inputs, outputs and system states. The controller has shown a robustness against force perturbation and model mismatch as well as high capability of tracking a pre-defined reference trajectory

Functional electrical stimulation (FES) is used to excite paralysed muscles that would otherwise be uncontrollable by paraplegic patients. Consequently, the patient could recover partially some of lower limb functions improving the cardiovascular system, increasing oxygen uptake and bettering the whole quality of life. In this paper, we apply a control design based on a higher order sliding mode to a complex physio-mathematical muscle model. This model is based on macroscopic Hill and microscopic Huxley concepts. The main goal concerns the prediction of the needed pattern stimulation (current and pulse width), which will extend the overall performances and defer the muscle fatigue as much as possible. The controller is mathematically computed and shown to provide satisfactory stability and tracking errors. Its efficiency is illustrated with the control of the knee joint angle under a co-contraction approach.

In paraplegic patients with upper motor neuron lesions the signal path from the central nervous system to muscles is interrupted. Functional Electrical Stimulation (FES) applied to the lower motor neurons can replace the lacking signals. A neuroprosthesis may be used to restore motor function in paraplegic patients on the basis of FES. The neuroprosthesic implant allows muscles to be controlled with high accuracy, high selectivity and the repeatability of the muscle's response can be achieved. The SUAW project succeeded in the implantation of an advanced neuroprosthetic device on two patients, but the movement generation remains open loop and is tuned empirically. The system is thus insufficient to enhance significantly the daily-life of the patient, nevertheless, the good results obtained give us the opportunity to envisage the system evolves towards the automatic synthesis of the stimulation patterns generating the desired movement and closed loop control. To achieve this goal, some preliminary researches have to be carried out; starting with a specific modeling that can be used in the contest of FES. The main issues concern muscle modeling including FES parameters as inputs, fatigue, the interaction with the skeleton, and the identification of parameters. This paper describes the mathematical modeling of the skeletal muscle.

The new generation of implanted neuroprostheses allows muscles to be controlled with fine accuracy, high selectivity and the repeatability of the muscle's response can be achieved. Thus, the closed loop control of such systems becomes possible. The SUAW project succeeded in the implantation of an advanced neuroprosthetic device on two patients, but the movement generation remains open loop and is tuned empirically. Nevertheless, the good results obtained give us the opportunity to envisage the system evolves towards closed loop control and automatic synthesis of the stimulation patterns generating the desired movement. To achieve this goal, some preliminary researches have to be carried out, beginning with a specific modelling that can be used in the context of functional electrical stimulation (FES). The main issues concern muscle modelling including the interaction with the skeleton, fatigue, FES parameters as inputs, and the identification of dynamic parameters, and afterwards, the motion synthesis and the closed loop control based on this model. Besides, the scientific approach is the same as in robotics so that the theoretical tools used in the control theory are the same and directly applicable. This paper describes the results obtained in the previous project SUAW and how we attempt, through the new project DEMAR, to enhance the global performances of the system.