Search

Human dexterity far exceeds that of modern robots, despite a much slower neuromuscular system. Understanding how this is accomplished may lead to improved robot control. The slow neuromuscular system of humans implies that prediction based on some form of internal model plays a prominent role. However, the nature of the model itself remains unclear. To address this problem, we focused on one of the most complex and exotic tools humans can manipulate-a whip. We tested (in simulation) whether a distant target could be reached with a whip using a (small) number of dynamic primitives whose parameters could be learned through optimization. This approach was able to manage the complexity of an (extremely) high degree-of-freedom system and discovered five optimal parameters of a single movement that achieved the task. An internal model of the whip dynamics was not needed for this approach, thereby significantly relieving the computational burden of task representation and performance optimization. These results support our hypothesis that composing control using dynamic motor primitives may be a strategy which humans use to enable their remarkable dexterity. A similar approach may contribute to improved robot control.

Physical interaction with an object that has internal dynamics can be challenging, both for humans and robots. An example is carrying a cup of coffee, where the nonlinear dynamics between the cup and the liquid can be chaotic and unpredictable. This study examined how nonlinearity of an object’s dynamics contributed to the difficulty of a task and if linearization of the object dynamics facilitated performance. Human subjects did a task in a virtual set-up with a haptic interface using a robotic manipulandum. The task of transporting a cup of coffee was reduced to a 2D cart-and-pendulum model; subjects moved the cart and felt the dynamics of the pendulum representing sloshing coffee. Performance with the nonlinear system was compared to a linearized mass-spring version of the system. Subjects (n=16) executed continuous rhythmic, self-paced movements. In the linearized system subjects chose to move at frequencies close to the resonant frequencies and clearly avoided the anti-resonance frequency. In the nonlinear system subjects did not avoid the anti-resonance frequency. To evaluate performance, mutual information quantified predictability between the interaction force and the cup and object dynamics. Mutual information was lower in trials when the cup moved close to the anti-resonance frequency in both linear and nonlinear systems. The magnitudes of the interaction forces were higher in the linear system, especially at frequencies slightly below the anti-resonance. These results run counter to the expectation that linearization would simplify this task. These findings may be useful as design considerations for robot control and human-robot interaction: if humans interact with robots that exhibit complex dynamics in the frequency range of human actions, linearizing a nonlinear system may potentially disturb intuitive and low-effort cooperation.

Lower-limb exoskeletons have shown great potential to assist human locomotion. However, effective methods for modulating aspects of gait behavior beyond reducing metabolic effort, such as stride time and gait kinematics, are still needed, especially for individuals with impaired gait. Using the Samsung GEMS-H exoskeleton, we studied the effect on healthy gait behavior of applying hip stiffness during overground walking. We found that applying positive stiffness did not affect stride time, but it did reduce hip range of motion. We also found that applying negative stiffness increased both stride time and hip range of motion. Additional analyses showed that the effect of applied hip stiffness on hip range of motion during overground walking was similar to that observed during treadmill walking. The effect of positive stiffness on stride time was similar during overground and treadmill walking, but the increase in stride time in response to the application of negative stiffness was greater during treadmill walking than in overground walking. Lastly, we found no evidence to indicate that neural adaptation or learning occurred when hip stiffness was applied. These results suggest that applying joint stiffness may be a promising approach to restoring healthy gait kinematics.

Ten right-handed male subjects turned a crank (radius 10.29 cm) in two directions at three constant instructed speeds (fast, medium, very slow) with visual speed feedback. They completed 23 trials at each speed. With the hand constrained to move in a circle, non-zero forces against the constraint were measured. Assuming a plausible mathematical model of interactive dynamics, the peripheral neuromechanics could be ‘subtracted’, revealing an underlying motion that reflected neural control. We called this data-driven construct the zeroforce trajectory. The observed zero-force trajectory was approximately elliptical. Its major axis, estimated by the principal eigenvector of the covariance matrix, differed significantly for the two movement directions. As peripheral neuromuscular compliance (i.e. low mechanical impedance) mitigates the consequences of imperfect execution, the required precision of motion commands is reduced. An oscillatory zeroforce trajectory that leads hand motion suffices to produce circular hand motions. Due to non-isotropic peripheral dynamics, that lead differs between degrees of freedom, resulting in an elliptical zero-force trajectory. The ellipses’ orientations differ with direction of rotation, as observed in the experimental data. As elliptical motion is generated by two non-colinear sinusoids with non-zero phase difference, these results support the hypothesis that humans simplify this constrained-motion task by exploiting primitive dynamic actions, oscmations and impedance.

The scope and relevance of wearable robotics spans across a number of research fields with a variety of applications. A challenge across these research areas is improving user-interface control. One established approach is using neural control interfaces derived from surface electromyography (sEMG). Although there has been some success with sEMG controlled prosthetics, the coarse nature of traditional sEMG processing has limited the development of fully functional prosthetics and wearable robotics. To solve this problem, blind source separation (BSS) techniques have been implemented to extract the user's movement intent from high-density sEMG (HDsEMG) measurements; however, current methods have only been well validated during static, low-level muscle contractions, and it is unclear how they will perform during movement. In this paper we present a neural drive based method for predicting output torque during a constant force, concentric contraction. This was achieved by modifying an existing HDsEMG decomposition algorithm to decompose 1 sec. overlapping windows. The neural drive profile was computed using both rate coding and kernel smoothing. Neither rate coding nor kernel smoothing performed as well as HDsEMG amplitude estimation, indicating that there are still significant limitations in adapting current methods to decompose dynamic contractions, and that sEMG amplitude estimation methods still remain highly reliable estimators.

Relatively little work in human and robot control has examined the control of underactuated objects with internal dynamics, such as transporting a cup of coffee, a task that presents little problems for humans. This study examined how humans move a ‘cup of coffee’ with a view to identify principles that may be useful for robot control. The specific focus was on how humans choose initial conditions to safely reach a steady state. We hypothesized that subjects choose initial conditions that minimized the transient duration to reach the steady state faster, as it presented more predictable dynamics. In the experiment, the cup of coffee was reduced to a 2-D cup with a sliding ball inside which was simulated in a virtual environment. Human subjects interacted with this virtual object via a robotic manipulandum that provided haptic feedback. Participants moved the cup between two targets without losing the ball; they were instructed to explore different initial conditions before initiating the continuous interaction. Results showed that subjects converged to a small set of initial conditions that decreased their transient durations and achieved a predictable steady state faster. Simulations with a simple feedforward controller and inverse dynamics calculations confirmed that these initial conditions indeed led to shorter transients and less complex interaction forces. These results may inform robot control of objects with internal dynamics where the effects of initial conditions need further investigation.

Restoring healthy kinematics is a critical component of assisting and rehabilitating impaired locomotion. Here we tested whether spatiotemporal gait patterns can be modulated by applying mechanical impedance to hip joints. Using the Samsung GEMS-H exoskeleton, we emulated a virtual spring (positive and negative) between the user’s legs. We found that applying positive stiffness with the exoskeleton decreased stride time and hip range of motion for healthy subjects during treadmill walking. Conversely, the application of negative stiffness increased stride time and hip range of motion. These effects did not vary over long nor short repeated exposures to applied stiffness. In addition, minimal transient behavior was observed in spatiotemporal measures of gait when the stiffness controller transitioned between on and off states. These results suggest that changes in gait behavior induced by applying hip stiffness were purely a mechanical effect. Together, our findings indicate that applying mechanical impedance using lower-limb assistive devices may be an effective, minimally-encumbering intervention to restore healthy gait patterns.

While rehabilitation of upper-limb motor function with human-interactive robots has been met with success, robot-aided locomotor rehabilitation has proven challenging. To inform more effective approaches to robotic gait therapy, it is important to understand neuro-mechanical dynamics and control of unimpaired locomotion. Our previous studies reported that human gait entrained to periodic mechanical perturbations at the ankle when the perturbation period was close to preferred walking cadence. Moreover, entrainment was accompanied by synchronizing the perturbations to a constant gait phase, the same for all subjects, where they provided mechanical assistance. To test the generality of entrainment-based assistance, the present study evaluated the behavior of five unimpaired subjects who walked overground while wearing a hip exoskeleton robot. Periodic torque pulses were applied to the subjects’ hips, with a period different from, but close to, their preferred stride cadence. Results indicated that unimpaired subjects entrained their gait to periodic mechanical perturbations at the hip. Convergence of relative phase between gait and perturbations was observed, but clustered around two distinct gait phases, in contrast to the single converged phase observed in entrainment to periodic ankle torques. These entrainment studies quantify important aspects of the nonlinear neuro-mechanical dynamics underlying the control of walking, which will inform the development of effective approaches to robotic walking therapy.

The locomotion and balance capabilities of bipedal robots have greatly improved in recent years. However, maintaining balance on difficult terrain still poses a significant challenge. In this paper, we examined how humans maintain mediolateral balance when standing on a narrow beam with bare feet and wearing rigid soles. Our results show that foot-beam interaction dynamics critically influence balancing behavior. Importantly, this suggests that differences in human balancing behavior across different support surfaces may not solely result from changes in their neural control strategy. They may also result from changes in foot-ground interaction. Thus, the altered foot-ground interaction dynamics must be considered to accurately capture changes in the human controller across different support surfaces. A simplified model of foot-beam interaction was added to a double inverted pendulum model for human balancing. This extended model could replicate the change in human behavior across different foot contact conditions (bare feet vs. rigid feet). A better understanding of how humans coordinate whole-body behavior across a range of conditions may inform the development of balance controllers for bipedal robots.

This study examined strategies humans chose to manipulate an object with complex (nonlinear, underactuated) dynamics, such as liquid sloshing in a cup of coffee. The problem was simplified to the well-known cart-and-pendulum system moving on a horizontal line. This model was implemented in a virtual environment and human subjects manipulated the object via a robotic manipulandum. The task was to maneuver the system from rest to arrive at a target position such that no residual oscillations of the pendulum bob remained. Our goal was to test whether humans simplified control by employing dynamic primitives, specifically submovements. Experimental velocity profiles of the human movements were compared to those predicted by three different control models. Two models used continuous optimization-based control, the third control model was based on Input Shaping. Input Shaping is a method for controlling flexible objects by convolving a motion profile with impulses of appropriate amplitude and timing. To evaluate whether humans used Input Shaping, we decomposed the velocity profiles recorded from humans into submovements, as proxies for the convolved impulses. Comparing the motion profiles from the 3 models with the experimentally measured human profiles showed superior performance of the Input Shaping model. These initial results are consistent with our hypothesis that combining dynamic primitives, submovements, is a competent description of human performance and may provide a simpler alternative to computationally complex optimization-based methods of robot control.

Standing on a beam is a challenging motor skill that requires the regulation of upright balance and stability. In this paper, we analyzed the behavior of humans balancing on a narrow beam without footwear. The results revealed high anti-correlation between lumped upper- and lower-body angular momentum. Despite differences in gross measures of balance, interlimb coordination was consistent between the novice and expert subjects, suggesting that both performances could be described with the same balance controller. By simulating a double inverted pendulum model utilizing different balancing controllers described in the robotics literature, we identified that the whole behavior observed from humans standing on a beam was best replicated with controllers that predominantly utilized hip actuation.

For robotic systems to interact with or learn from the actions of surrounding humans, it is important that they can accurately interpret the intention driving human motor actions. Making such interpretations, however, requires the ability to perceive the relevant feature(s) from the observed human behavior. With visual sensing alone, robots are typically limited to perceiving only the human's overt motion in the form of joint angles and positions. Ideally, robots designed to interface with humans would also be able to infer information as to how the human is controlling itself from that overt motion. In this study, we investigated if and how humans might be able to visually sense changes in limb mechanical impedance of others. Results indicated that humans can visually perceive changes in joint stiffness from the motion of a two-link planar arm, suggesting that humans can extract information regarding how humans control limb impedance from kinematic information. These findings have important implications for applications where robots must interpret the motor actions of humans, such as during robot imitation learning and human-robot physical interaction.

Unlike upper-extremity robotic therapy, robotic therapy of lower extremities has not matched the effectiveness of human-administered approaches. We hypothesize that this may stem from inadvertent interference with natural movement control and investigated the oscillatory dynamics of human locomotion. Specifically, we assessed gait entrainment to periodic mechanical perturbations. Because the treadmills used in most studies necessarily interact with the dynamics of natural walking, we compared our experimental intervention during gait entrainment in treadmill and overground walking. Fourteen healthy subjects walked overground and on a treadmill while wearing an exoskeletal ankle robot which exerted short plantarflexion torque pulses at periods 50 ms shorter or longer than the subjects' preferred stride period. Entrainment to the periodic perturbation occurred in all conditions, however more readily in overground walking. In all cases, the stride period phase-locked with the torque pulse at ‘push-off’ such that it assisted propulsion. This entrainment of the stride period and its sensitivity to context indicate the subtlety and adaptability of human walking. Our observations suggest new avenues for gait rehabilitation and implications for exoskeleton design and legged locomotion research.

Energy efficient compliant actuation is the missing ingredient and key enabler of next-generation autonomous systems, domestic robots, prosthetic devices, orthotic devices, and wearable exoskeletons, to name a few. For all these devices, one would wish to develop actuators enabling wide range impedance modulation with low energy cost. Using conventional and biologically-inspired compliant actuation, previous research led to functional devices but with high energy cost. Here we introduce a minimalistic compliant actuator to realize impedance modulation with low energy cost. Using this actuator we demonstrate stiffness augmentation in human-machine collaboration. We argue that the non-biologically-inspired actuation concept presented here may effectively complement a biological system, by restoring or extending its functionality, with negligible energy cost.

Managing the trade-off between performance and stability is a crucial issue in physical human-robot interaction, and this has become more important than ever apace with growing needs for physically human-interactive robots in many fields, such as assistive robotics and rehabilitation robotics. In this paper, we present two essential considerations for design and control of robots physically interacting with humans: energetic passivity and mechanical impedance. Characterization of passive, dissipative, and active behavior of the human neuromuscular system is essential to ensure and control coupled stability in physical human-robot interaction. In addition, characterization of human mechanical impedance at the interaction port provides detailed quantitative information to describe interactive dynamics. The importance of these characterizations is demonstrated by simple examples and the authors' previous research on the human ankle. Implications for quantitative guidelines for robot design and control are discussed.

Dynamic behavior of legged robots is strongly affected by ground impedance. Empirical observation of robot hardware is needed because ground impedance and foot-ground interaction is challenging to predict in simulation. This paper presents experimental data of the MIT Super Mini Cheetah robot hopping on hard and soft ground. We show that controllers tuned for each surface perform better for each specific surface type, assessing performance using measurements of 1.) stability of the robot in response to self-disturbances applied by the robot onto itself and 2.) the peak accelerations of the robot that occur during ground impact, which should be minimized to reduce mechanical stress. To aid in controller selection on different ground types, we show that the robot can measure ground stiffness and friction in-situ by measuring its own interaction with the ground. To motivate future work in variable-terrain control and in-situ ground measurement, we show preliminary results of running gaits that transition between hard and soft ground.

The MIT Cheetah demonstrated a stable 6 m/s trot gait in the sagittal plane utilizing the self-stable characteristics of locomotion. This paper presents a numerical analysis of the behavior of a quadruped robot model with the proposed controller. We first demonstrate the existence of periodic trot gaits at various speeds and examine local orbital stability of each trajectory using Poincar`e map analysis. Beyond the local stability, we additionally demonstrate the stability of the model against large initial perturbations. Stability of trot gaits at a wide range of speed enables gradual acceleration demonstrated in this paper and a real machine. This simulation study also suggests the upper limit of the command speed that ensures stable steady-state running. As we increase the command speed, we observe series of period-doubling bifurcations, which suggests presence of chaotic dynamics beyond a certain level of command speed. Extension of this simulation analysis will provide useful guidelines for searching control parameters to further improve the system performance., United States. Defense Advanced Research Projects Agency. Maximum Mobility and Manipulation (M3) Program

Analyzing the structure of motor variability is important to understand neural control of movement. Although many previous studies have suggested analysis methods to account for task-relevant variability, it has been revealed that even a simple linear transformation of the coordinate system causes qualitatively different interpretation of the variability analysis. The present study suggests a revised version of the Covariation cost analysis and shows that it has little coordinate sensitivity. The modified algorithm allows more flexible optimization, but does not change mean and standard deviation of the execution variables. We also show that the cost analysis takes anisotropy of execution variables into account regardless of two tested coordinate systems. While not resolving the cause of the coordinate sensitivity, the present study is a step forward to coordinate invariance of variability analysis. Together with Tolerance and Noise cost analyses, which is already known to be coordinate insensitive, it is expected that the analysis will elucidate neural mechanism in goal-directed motor behavior in contexts of motor control, learning and rehabilitation.

The Floquet multiplier has been widely used to assess orbital stability of human walking. Floquet introduced it to quantify the stability of periodic motion of deterministic linear systems, but human walking involves highly nonlinear dynamics and inevitable noise. This preliminary study addresses the validity of Floquet multiplier estimation of a nonlinear system with stochastic noise. Noise was added to a simplified but nonlinear model of human walking so that the variability of model walking was comparable to that of human walking. Results show that linear regression yields a biased measure of the Floquet multiplier though its accuracy improves with a sufficient number of strides.

This paper presents a new method to characterize human ankle mechanical impedance during treadmill locomotion with a wearable ankle robot, Anklebot. An ensemble-based system identification method was used to investigate the time-varying behavior of ankle mechanical impedance in two degrees of freedom, both in the sagittal and frontal planes. We also provide solutions to overcome the limitations of original ensemble-based methods in practical applications. A pilot study of three human subjects demonstrated the efficacy of our approach. Analysis results showed clear time-varying behaviors of ankle impedance across the gait cycle except in the mid- and terminal-stance phases, and these behaviors were accurately approximated as a second-order model with stiffness, damping, and inertia components. Interestingly, all three subjects showed similar time-varying behaviors in both degrees of freedom: impedance increased around heel-strike and decreased significantly at the end of the stance phase.

Recent decades have witnessed a growing interest in the use of robotics to aid in neurosurgery. Purported advantages over currently available, image-guided micro-neurosurgery have been said to include an increase in precision in positioning needles used for the biopsy of tumors, and electrodes used for electrical stimulation or tissue ablation. Although such goals have been met, monetary costs have been high and real benefits, in terms of patient care, have been questioned. Numerous problems persist in neurosurgery that are not satisfactorily addressed either by conventional micro-neurosurgery or by current robot-assisted neurosurgery. Two critical areas include: (i) minimally invasive resection of hematomas and tumors deep to eloquent cortex, with minimal or no disturbance to overlying tissues; (ii) quasireal-time, intraoperative delineation of normal versus abnormal tissues, for the complete and safe resection of tumors such as gliomas. Successfully attacking these problems will require major technical advances that are uniquely suited to robotic engineering

This paper presents a novel method to characterize mechanical impedance of a human joint. We employed a non-parametric stochastic identification method and an interactive robot. This method allows us to determine several important properties including a) joint impedance in the direction of the joint principal axes, b) the spatial joint impedance structure, and c) the passivity properties of the joint. The method is applicable to studies of multivariable mechanical joint impedance, both for multi-joint impedance and single joint multi-DOF impedance. The method was successfully applied to the human ankle, and we showed that the determined ankle impedance precludes the utilization of a 2nd order model. In addition, the impedance structure exhibited a characteristic “peanut” shape, being weak in inversion-eversion, and the ankle has passive and dissipative behavior over wide range of frequency.

Humans interact with a variety of objects, many of which have dynamically complex properties, such as a cup of coffee. An efficient way to manipulate such objects is to drive them at a resonant frequency. This requires less effort due to dynamic amplification of control inputs. However, errors in control may also be amplified, increasing variability and making it difficult to achieve a kinematic goal. How humans control dynamically complex objects near a resonant frequency, while at the same time achieving kinematic goals, is unknown. To address this question, ten healthy subjects were asked to practice oscillating a cart and pendulum (simulated using a haptic display). Subjects had to move the cart between two spatial targets at the pendulum's resonant frequency. A visual display showed the pendulum bob moving in a semicircular cup (the cart was hidden), mimicking a ball rolling in a cup. Results showed that in early practice subjects moved below the resonant frequency and used an in-phase strategy — the cup and ball moved in the same direction. This was associated with large applied forces and high variability. With practice subjects moved above the resonant frequency and switched to an anti-phase strategy — the cart and pendulum moved in opposite directions. Concurrently, subjects' applied force decreased by half and the interaction force of the ball on the cup increased, which moved the cup to the spatial targets. Using this strategy, subjects became less variable and more accurate. Although the switch in phasing was in part dictated by the task dynamics, the direction of the shift is best explained by a controller that seeks to exploit the dynamics of an object to achieve task goals with small outlays of energy and low variability.

Robotic studies have suggested a contribution of limit-cycle oscillation of the neuro-mechanical periphery to human walking by demonstrating stable bipedal robotic gaits with minimal actuation and control. As behavioral evidence of limit-cycle oscillation in human walking, we recently reported entrainment of human gaits to mechanical perturbations. We observed synchronization of human walking with mechanical perturbation only when the perturbation period was close to the original walking period. In addition, the entrainment was always accompanied by phase locking at the end of double-stance. A highly-simplified state-determined walker reproduced these salient features: 1) entrainment to periodic perturbations with a narrow basin of entrainment and 2) phase-locking at the end of double stance. Importantly, the model required neither supra-spinal control nor an intrinsic self-sustaining neural oscillator (like a rhythmic central pattern generator), which suggests that prominent features of human walking may stem from simple afferent feedback processes that produce limit-cycle oscillation of the neuro-mechanical periphery without significant involvement of the brain or rhythmic central pattern generators. One limitation of that model was that it entrained only to perturbations faster than the unperturbed walking period. In the study reported here, we modified the model to have two independent steps per stride. The revised model reproduced entrainment to perturbations both slower and faster than the unperturbed cadence, as observed experimentally in human walking.

Quantitative characterization of ankle mechanical impedance is critical for understanding lower extremity function in persons with neurological disorders. In this paper, we examine the feasibility of employing an ankle robot and multivariable analysis to determine static ankle impedance in 4 patients: 1 with multiple sclerosis and 3 with stroke. We employed a scalar based vector field approximation method which was successful in identifying young healthy subjects' ankle impedance. It enabled clear interpretation of spatial ankle impedance structure and intermuscular feedback at the ankle for both affected and unaffected legs. Measured impedance of two patients was comparable to healthy young subjects, while the other two patients had significantly different static ankle impedance properties.

Entraining human gait with periodic torque from a robot may provide a novel approach to robot-aided walking therapy that is competent to exploit the natural oscillating dynamics of human walking. To test the feasibility of this strategy we applied a periodic ankle torque to neurologically impaired patients (one with stroke and one with multiple sclerosis). As observed in normal human walking, both patients adapted their gait periods to synchronize with the perturbation by phase-locking the robotic torque at terminal stance phase. In addition, their gait cadence became significantly faster due to the training with clear after effects when the perturbation ceased. These results support a new strategy for walking therapy that exploits an embedded neural oscillator interacting with peripheral mechanics and the resulting natural dynamics of walking, which are essential but hitherto neglected elements of walking therapy.

Dynamically entraining human gait with periodic torque from a robot may provide an approach to walking therapy that is uniquely supportive of normal biological function. To test the feasibility of this approach we perturbed the gait of unimpaired human subjects by applying a periodic impulsive ankle torque at various frequencies. Eighteen subjects out of 19 exhibited entrained gaits: for a small range of frequencies their gait periods adapted to synchronize with the mechanical perturbation. In addition, the phase when synchronized was such that the robot perturbation assisted propulsion. These results support a new strategy for walking therapy that exploits an embedded neural oscillator interacting with peripheral mechanics and the resulting natural dynamics of walking, which are essential but hitherto neglected elements of walking therapy.

In this paper we present initial results using a novel ankle robot to estimate human ankle stiffness. This lower-extremity robotic therapy module was developed at MIT to aid recovery of ankle function. Given the importance of the ankle during locomotion, an accurate estimate of ankle stiffness would be a valuable asset for locomotor rehabilitation, potentially providing a measure of recovery and a quantitative basis to design treatment protocols. We show that the ankle robot provides accurate measurement of ankle configuration and present a simple protocol to estimate ankle stiffness. Our initial ankle stiffness estimates compare favorably with previously published work, indicating that this method may serve as an efficient clinical measurement tool.

Robots for high-force interaction with humans face particular challenges to achieve performance and coupled stability. Because available actuators are unable to provide sufficiently high force density and low impedance, controllers for such machines often attempt to mask the robots' physical dynamics, though this threatens stability. Controlling for passivity, the state-of-the-art means of ensuring coupled stability, inherently limits performance to levels that are often unacceptable. A controller that imposes passivity is compared to a controller designed by a new method that uses limited knowledge of human dynamics to improve performance. Both controllers were implemented on a testbed, and coupled stability and performance were tested. Results show that the new controller can improve both stability and performance. The different structures of the controllers yield key differences in physical behavior, and guidelines are provided to assist in choosing the appropriate approach for specific applications.

In 1991, a novel robot named MIT-MANUS was introduced as a test bed to study the potential of using robots to assist in and quantify the neuro-rehabilitation of motor function. It introduced a new type of therapy, offering a highly back-drivable mechanism with a soft and stable feel for the user. MIT-MANUS proved an excellent fit for rehabilitation of the paretic shoulder and elbow in patients with stroke, with clinical trials showing a reduction of impairment at these joints. However, the desired outcome of rehabilitation services is reduced impairment in movement capacities, as well as improved functional abilities and participation in daily life tasks. While we have seen reductions in motor impairment, no specific attempts were made during these studies to help the person link the movements practiced during robotic therapy to motor actions during daily tasks. In this paper we are presenting our first attempt to link movements practiced in robotic therapy with distal training during functional tasks.

In 1991, a novel robot named MIT-MANUS was introduced as a test bed to study the potential of using robots to assist in and quantify the neuro-rehabilitation of motor function. It introduced a new type of therapy, offering a highly backdrivable mechanism with a soft and stable feel for the user. MIT-MANUS proved an excellent fit for the rehabilitation of shoulder and elbow of stroke patients with results in clinical trials showing a reduction of impairment in these joints. Notably, we found a greater reduction in motor impairment in the exercised muscle groups. This suggests a need for additional robots to rehabilitate other targeted joints and different muscle groups. The intent of this paper is to present data that supports the use of multiple robots to deliver focused therapy to different limb segments and joints.

In 1991, a novel robot named MIT-MANUS was introduced as a test bed to study the potential of using robots to assist in and quantify the neuro-rehabilitation of motor function. It introduced a new brand of therapy, offering a highly backdrivable mechanism with a soft and stable feel for the user. MIT-MANUS proved an excellent fit for shoulder and elbow rehabilitation in stroke patients, showing in clinical trials a reduction of impairment in these joints. The greater reduction in impairment was observed in the group of muscles exercised. This suggests a need for additional robots to rehabilitate other target areas of the body. The focus here is a robot for wrist rehabilitation designed to provide three rotational degrees of freedom. A previous paper at ICORR2003 and its companion book described the basic system design and characterization. In this paper we present clinical results from five (5) stroke patients. A comprehensive review of the wrist robot design, characterization, and initial clinical results are being submitted elsewhere (EEEE Transactions on Neural Systems and Rehabilitation Engineering).

The amplitude of the surface EMG was estimated with multiple channels of EMG by first temporally whitening individual EMG channels and then optimally combining multiple channels. Surface EMG was recorded for non-fatiguing, constantforce, isometric contractions. A signal-to-noise ratio (SNR) was computed (deviations about the mean value of the estimate were considered as noise). A moving average root mean square (MARMS) estimator (245 ms window) provided an average f standard deviation (AfSD) SNR of 10.7 f 3.3 for the individual channels. Temporal whitening improved the AfSD SNR to 17.6 i 6.0. Combining temporal whitening and spatial uncorrelation with four channels improved the AfSD SNR to 24.6 f 10.4. On one subject eight optimally combined channels were achieved, providing an AfSD SNR of 35.0 f 13.4.

Torque exerted about the elbow was estimated from flexor and extensor muscle surface EMG waveforms. An algebraic relation between joint torque and simultaneous flexor/ extensor EMG amplitudes was assumed, and a least squares method for identifying the parameters of this relation investigated. Initially, a simulation study investigated the performance of the least squares method and served as a tool to determine an appropriate experimental design. Then, multiple channels of flexor/ extensor EMG waveforms and joint torque were measured during non-fatiguing, slowly force-varying (quasi-isotonic), isometric contractions. Single/ multiple channel techniques for computing an EMG amplitude were used to identify an EMG to torque relation, and then estimate joint torque based on this relation.

Two experiments quantifying distortions of haptic spatial perception in humans are reported. Human subjects interacted with programmable "virtual objects" simulated mechanically using a planar robot under impedance control. The psychophysical functions for haptic perception of lengths and angles were determined. Comparison of these two measure shows that, unlike the typical geometrical model used in robot control, no consistent metric underlies human haptic spatial perception.

We review results of initial clinical trials with 76 stroke patients at the Burke Rehabilitation Hospital. They provide evidence that robot-aided training enhances recovery. We also discuss technology under development, including algorithms for evaluating patients' performance. To conclude, we summarize different modalities of robot-aided therapy. In our initial clinical trials we used a single modality of robot therapy, but since each lesion is unique there is no reason to believe that a "one-size-fits-all" optimal treatment exists. We argue that therapy should be tailored to each patient's needs.

In 1991, a novel robot named MIT-MANUS was introduced as a test bed to study the potential of using robots to assist in and quantify the neuro-rehabilitation of motor function. It proved an excellent fit for the rehabilitation of shoulder and elbow of stroke patients with results in clinical trials showing a reduction of impairment in these joints. The greater reduction in impairment was limited to the group of muscles exercised. This suggests a need for additional robots to rehabilitate other degrees of freedom. This paper outlines the mechanical design of a robot for wrist rehabilitation.

In this paper, we present considerations for a novel robot to rehabilitate the ankle following stroke. This module has been developed at MIT and should commence pilot testing with stroke patients at the Baltimore Veterans Administration Medical Center by the end of Summer 2004. The purpose of the module is to train stroke survivors to overcome the common foot-drop problem. Its design follows the same guidelines as our upper-extremity designs, i.e., it is a low friction, backdriveable device with intrinsically low mechanical impedance.

Robotics offers one solution to the rising problem of rehabilitating victims of neurological injury and disease. Rehabilitation robots have proven successful in speeding recovery for recent stroke victims, and in reducing impairment and pain for chronic victims who were thought to have little opportunity for improvement. Such robots require high force capability with a closely controlled "feel", requiring low endpoint impedance. For complex robot configurations, a combination of backdrivable hardware design and impedance-reducing controller design may offer the best solution. A novel therapy algorithm that exploits similarities between motor recovery and motor learning adapts robot impedance to patients as they recover. Results of therapy using this algorithm are a substantial improvement over the original robot therapy.

Quantitative observations are presented on how normal human subjects manipulate a loaded crank. Velocity patterns showed that the task was performed smoothly, despite kinematic singularities which required rapid, precisely timed reversals of joint motion. Significant forces were exerted normal to the crank such that the crank partially supported the weight of the arm. Kinematic data revealed that arm motion depended on the direction of crank motion. These observations suggest that humans use a task-oriented control strategy and may take advantage of motion constraints to reduce required effort. >

A paradigm is described in which subjects feel virtual objects, created using a two-link robot arm under computer control. Preliminary results show there are systematic distortions in how relative length and relative compliance (preliminary results) are perceived by the subjects, who use only their arm movements in making psychophysical judgments. These distortions depend on the location of the objects in the workspace, and they appear to be related to the dynamical properties of the arm itself. >

This paper presents some recent work on the development of a workstation for teaching and therapy in manual and manipulative skills. The experimental workstation, MANUS, as well as the overall concept are described. State-of-the-art aspects of the workstation under development are introduced. >

This paper describes a technique to investigate in hemiplegic subjects the effect of cerebral lesions on motor learning with the unimpaired arm. Robot-aided functional imaging experiments suggest that the ipsilateral hemisphere plays a role during learning of a motor task involving the ipsilateral arm. Hemiplegic subjects thus provide an opportunity to test: (a) the effect of cerebral lesions during motor learning, and (b) whether the measured significant changes in neural activity during imaging experiments are not artifacts. The authors' results demonstrate the role of selected ipsilateral brain structures during motor learning. For example, ipsilateral right temporal lesions affect motor performance and learning. Lesions in the motor strip area do not adversely affect motor learning with the unimpaired arm. These results have implications for general rehabilitation strategies that presently do not recognize this deficit.

Convergent force fields organized in the frog's spinal cord could be used as a new functional neuromuscular stimulation (FNS) technique to restore motor function using electrical stimulation of the spinal interneurons. We studied the recruitment modulation of these force fields to four parameters of the electrical microstimulation stimulus train. The parameters modulated were the stimulation frequency, the pulse duration, the pulse amplitude, and the train duration. We found that all four parameters can be used to modulate the force level, and that all the parameters but train duration also affect the force rate of rise. Furthermore, we found that the vectorial structure of the fields was not affected by the force modulation, as the normalized fields were nearly identical. These early results demonstrate the possibility of modulating the action of the fields while retaining their useful convergent property.

Amplitude estimates of simulated surface electromyograms (EMGs) were studied using both unwhitened and whitened EMG amplitude estimators. Constant-effort non-fatiguing EMGs were simulated as the superposition of simple synthetic motor unit action potential trains (MUAPTs). Each MUAPT was formed from random firings of randomly shaped motor unit action potentials. Two aspects of unwhitened vs. whitened amplitude estimators were studied: (1) the relationship between the mean value of the EMG amplitude estimate vs. the level of contraction and (2) the signal-to-noise-ratio (SNR) performance of the EMG amplitude estimate. It was found that the relationships between estimated EMG amplitude vs. MUAPT firing rate and the number of MUAPTs were not altered by whitening. However, the performance of whitened amplitude estimators was markedly better. These results suggest that whitening can provide a higher fidelity EMG amplitude estimate without distorting the relationship between EMG and the mechanical output of the muscle.