ArticlePDF Available
Analysing human-exoskeleton interaction: on the human adaptation to modified
gravito-inertial dynamics
S. Bastidea,b, N. Vignaisa,b*, F. Geffardc, and B. Berreta,b,d
aCIAMS, Univ. Paris-Sud, Université Paris-Saclay, 91405 Orsay Cedex, France; bCIAMS, Université d'Orléans, 45067,
Orléans, France; cCEA, LIST, Interactive Robotics Laboratory, 91191 Gif-sur-Yvette, France; dInstitut Universitaire de
France (IUF)
Keywords: gravito-inertial dynamics; motor control; adaptation; exoskeleton
1. Introduction
Robotic exoskeletons are promising devices to assist
operators in the industry sector and prevent
musculoskeletal disorders. With such devices, typical
load lifting tasks can be performed with less effort for
the operator. She/he would just have to guide the
motion of the exoskeleton, usually set in transparent
mode. In the most basic and robust implementation of
this mode, the exoskeleton compensates its own weight
(plus the weight of any attached load) and frictions.
From a human-centered point of view, such an
interaction leads to a quite unfamiliar dynamic
situation. Indeed, in daily life, our central nervous
system (CNS) has to deal with both static and dynamic
forces acting at the joints. During object manipulation,
gravitational torque (GT) scales with inertial torques
(IT) because gravitational and inertial masses are the
same. In contrast, with a transparent exoskeleton,
this implicit relationship is altered because the operator
feels the additional inertia of the robot (and potential
additional loads) without any associated weight
increment. Given that GT and IT are major constitutive
components of upper-limb dynamics, understanding
how humans adapt to such unfamiliar perturbations in
practice is critical. In motor control, adaptation to a
weight load at the hand (incrementing GT and IT
together) is quasi-instantaneous as it corresponds to a
familiar dynamic situation (Bock et al., 1990).
However, adaptation to unfamiliar dynamic conditions
can take tens to hundreds of trials (Ingram et al., 2011).
Previous studies on weight load perturbations (Bock et
al., 1992), force field adaptation (Kurtzer et al., 2005),
grasping forces (Zatsiorsky et al., 2005) or
weightlessness (Gaveau et al., 2016) suggest that the
CNS has internal models dissociating GT and IT. This
could allow conforming efficiently and rapidly to
novel gravito-inertial dynamics. However, moving
objects with inertia but no apparent weight, as in a
human/exoskeleton interaction, is quite uncommon on
earth and it may well disturb human motor control
(Bastide et al,. 2018). Given the unfamiliarity of such
a perturbation, it is therefore reasonable to expect that
the CNS would need few trials to adapt to its motor
controller. Thus, the aim of this study is to analyze the
adaptation to a “transparent” upper-limb exoskeleton
during simple elbow flexion/extension movements,
given that the exoskeleton induces a relative
modification of GT and IT during the task.
2. Methods
2.1 Participants
Twenty-one healthy right-handed young adults (7
females, 14 males) participated to this study. Mean
age, height and weight were 24.9±4.7, 175.7±8.4 cm
and 68.9±11.1 kg, respectively. Written informed
consent was obtained from each participant in the
study as required by the Helsinki declaration. The local
ethical committee for research (Univ. Paris Saclay)
approved the experimental protocol.
2.2 Materials
The ABLE upper-limb exoskeleton was used in this
experiment (Garrec et al., 2008). The mass and the
length to the center of mass of its forearm segment
were 2.32 kg and 11 cm. The elbow joint position was
recorded from the exoskeleton sensors. We also
recorded surface electromyographic (EMG) data of
two flexors muscles (biceps brachial, brachioradialis)
and two extensors (triceps brachial lateral head, triceps
brachial long head). EMG signals were collected using
wireless sensors (Biometrics Ltd, UK). Both kinematic
and EMG signals were sampled at 1000 Hz.
2.3 Procedure
Participants sat straight with their back leaning against
the rigid base of the exoskeleton. The participant’s
right forearm was attached to the exoskeleton at wrist
level. Alignment of elbow centers of rotation of the
participant and the exoskeleton was adjusted by
calibrating the height of the whole robotic device.
Participants were asked to perform point-to-point
reaching movements between two lighting targets
(LED), involving 60° elbow flexion/extension
movements. Exoskeleton joints other than the elbow
were frozen to ensure that movements could only be
performed with the forearm. Participants were
instructed to point in the direction of the target that lit
up. The target remained on for one second before it
turned off. To get discrete movements, a random pause
duration of 1 to 2.5 seconds was displayed between
each movement. Overall, each participant performed
50 flexions and 50 extensions of the elbow. The
exoskeleton control law was set to compensate its own
frictions and GT (but not its IT).
2.4 Data processing
Angular velocity and acceleration were obtained by
numerical differentiation of the recorded angular
positions. The movement was considered effective
when the elbow angular velocity exceeded 5% of the
maximum. Flexors and extensors muscles were
respectively grouped and averaged. Muscle activations
were expressed as a percentage of the maximal data
obtained during the experiment. The last 40 trials have
been used to define the 95% Confidence Interval of the
Plateau (CIP).
3. Results and discussion
Figure 1: Mean duration, velocity, maximal
acceleration and maximal elbow flexor activity (±SE)
during elbow flexion movements for all participants.
Red lines represent data exponential fitting.
3.1 Kinematics
Duration of the three firsts movements were
significatively different from the subsequent ones
according to the predefined CIP (see figure 1). The
mean velocity and the mean maximal acceleration are
also significantly lower for the three firsts flexions.
Same observations were made for extensions. Thus six
full movements, i.e. flexion and extension, were
necessary to conform to the new gravito-inertial
dynamics situation. Moreover, a higher overshoot
(6.83±3.4°) was found only for the very first flexion
compared to overshoot CIP ([1.6°, 2°]), showing a
rapid adaptation process to achieve the task.
3.2 Muscle activity
Agonists muscular activations followed the same
trend, with only the very first flexion being out of the
maximal activation CIP (see figure 1). Concerning
extensors, maximal activation (33±15.2 %) was also
out of the CIP ([51.2%, 53.6%]) only for the first
extension, suggesting that participants underestimated
the inertial mass being manipulated during the very
first movement of flexion/extension (likely because it
could not be inferred from the exoskeleton weight).
4. Conclusions
This study aimed at analyzing the human adaptation to
modified gravito-inertial dynamics induced by an
upper-limb exoskeleton. Participants needed one to
three flexion/extension movements to adapt to the
“transparent” mode of the exoskeleton. Thus, our
initial hypothesis was confirmed, i.e. adaptation to a
transparent exoskeleton is rapid but not as immediate
as in classical load lifting tasks. Future work will
investigate human adaptations to other control modes,
e.g. compensation of the user’s arm weight.
Bock O. 1992. The characteristics of arm movements
executed in unusual force environments. ASR.
12(1): 237-241.
Bastide S, Vignais N, Geffard F, Berret B. 2018.
Interacting with a “Transparent” Upper-Limb
Exoskeleton: A Human Motor Control Approach.
Proc IEEE IROS. 4661-4666.
Kurtzer I, DiZio P, Lackner JR. 2005. Adaptation to a
novel multi-force environment. Exp. Brain Res.,
164(1): 120-132.
Zatsiorsky VM, Gao F, Latash ML. 2005. Motor
control goes beyond physics: differential effects of
gravity and inertia on finger forces during
manipulation of hand-held objects. Exp. Brain Res.
162(3): 300-308.
Gaveau J, Berret B, Angelaki DE, Papaxanthis C.
2016. Direction-dependent arm kinematics reveal
optimal integration of gravity cues. eLife, 5.
Ingram JN, Howard IS, Flanagan JR, Wolpert DM.
2011. A Single-Rate Context-Dependent Learning
Process Underlies Rapid Adaptation to Familiar
Object Dynamics. PLoS Comput. Biol.
7(9) :e1002196, Sept.
*Corresponding author. Email:
... De nombreuses études, notamment issues du champ de l'adaptation motrice, ont employé une approche relativement similaire en se limitant souvent à des mouvements dans un plan horizontal, perturbés par un champ de force prédéfini (i.e., dépendant de la vitesse de l'effecteur) appliqué par un bras robotique ou un exosquelette (notamment le KinArm [226]), très similaire dans les différentes études [227][228][229][230][231][232][233][234]. Au cours de précédents travaux, Bastide & al. [6,85,235], ont commencé à introduire des notions de contrôle moteur pour évaluer l'impact d'un exosquelette transparent sur le mouvement humain. Notamment, la préservation d'invariants bien connus du mouvement humain lors de cette interaction a été vérifiée : asymétries directionnelles de mouvement (dues à la gravité, voir Section 2.2), relation linéaire entre amplitude et durée de mouvement ou encore forme des profils de vitesse. ...
... De surcroit, pour pouvoir tester et éventuellement étendre ces théories, plusieurs développements matériels et logiciels ont été réalisés, tant concernant la transparence que la compensation précise du poids. Ces travaux sont dans la continuité des travaux de Bastide & al. [6,85,235], et correspondront donc à la branche inférieure de notre démarche de recherche. ...
Full-text available
Les exosquelettes actifs sont des systèmes prometteurs, dont les nombreuses applications potentielles visent essentiellement à prévenir et traiter les déficiences motrices. Ce potentiel est, actuellement, sous-exploité. Les principales raisons de cette sous exploitation sont, d’abord, les limites des exosquelettes, tant en termes de conception mécanique que de conception de modes de contrôle, ensuite, les limites des connaissances quant à l’adaptation du mouvement humain lors de l’interaction avec un exosquelette. Ces travaux de thèse se basent sur l’hypothèse selon laquelle un développement conjoint des exosquelettes et des connaissances en contrôle moteur humain pourrait permettre d’améliorer la qualité de l’interaction. Nous montrerons notamment que, l’évaluation des exosquelettes par des métriques inspirées du contrôle moteur et la caractérisation de l’adaptation du mouvement humain, en réponse à des efforts appliqués par l’exosquelette, constituent un cercle vertueux. Au cours de ces travaux, un mode transparent (i.e., l’exosquelette suit le mouvement humain en le perturbant le moins possible) et un mode antigravitaire (i.e., l’exosquelette compense le poids du membre humain) ont été développés et évalués sur un exosquelette de membre supérieur. Ces développements ont d’abord permis d’améliorer la transparence de l’exosquelette, par la conception d’une loi de contrôle puis d’interfaces humain-exosquelette. Ensuite, la prise en compte des défauts d’alignement articulaires entre l’humain et l’exosquelette a permis la conception d’un mode antigravitaire efficace et individualisé. Ces développements ont ensuite été utilisés pour tester et étendre deux théories influentes en contrôle moteur : la théorie d’exploitation optimale de la gravité ambiante dans la planification du mouvement et la théorie du coût du temps. L’extension de ces deux théories pourrait avoir des implications intéressantes pour le développement futur des exosquelettes. D’abord, l’utilisation de champs gravitaires réduits est une méthode classique d’utilisation des exosquelettes en rééducation neuromotrice. Une meilleure compréhension de l’adaptation du mouvement humain à de tels champs pourrait donc, par exemple, permettre la conception de nouveaux protocoles de rééducation. Ensuite, l’extension de la théorie du coût du temps pourrait permettre de prédire la durée de mouvements arbitraires à partir d’une identification simple. Cela pourrait, notamment, permettre de mieux estimer les trajectoires autour desquelles le mouvement humain peut être assisté. Ces travaux de thèse ont d’abord permis de montrer que le mouvement humain s’adapte très rapidement et optimalement à des champs gravitaires arbitraires, y compris inversés (i.e., dirigés vers le haut). Ensuite, un compromis optimal entre effort et temps de mouvement, sous-tendu par un coût du temps, a pu être mis en évidence et quantifié. Cela a notamment permis de montrer que l’interaction avec un exosquelette qui serait plus lent que les mouvements planifiés par l’humain, peut conduire ce dernier à dépenser des quantités importantes d’énergie pour gagner du temps, ce qui conduirait nécessairement à une fatigue prématurée et une baisse de l’acceptabilité du système lors d’une utilisation prolongée. Ces travaux de thèse illustrent le cercle vertueux précédemment introduit et montrent ses effets bénéfiques, tant pour le développement et l’évaluation des exosquelettes que pour l’extension de théories sur le contrôle moteur humain. Ils ouvrent également de nombreuses perspectives pour le développement de nouveaux exosquelettes actifs, pour leur utilisation à large échelle en recherche sur le contrôle moteur humain et pour des applications en rééducation neuromotrice et en prévention des troubles musculo-squelettiques au travail.
... A similar conclusion had been drawn with the same exoskeleton when increasing the apparent inertial torque without affecting the participants' gravity torque. 49 However, such an adaptation within a dozen trials may contrast with other works showing a slower adaptation to microgravity for similar single-joint arm movements, which took about 75 movements during parabolic flights to remove directional asymmetries in velocity profiles. 10 The main difference between these studies seems to lie in the nature of the sensory changes induced by the novel environment. ...
... Muscle weakness due to prolonged use of exoskeletons or orthoses is often raised as a concern (Azadinia et al., 2017). But on a shorter timescale, motor adaptation related to familiarization with the exoskeleton may also happen and affect the mea-sured biomechanical effects (Bastide, Vignais, Geffard, & Berret, 2019). Though the participants in the current study did practice with the Laevo for a few minutes prior to the experiment, the duration of the familiarization process for such exoskeletons remains unclear, and might involve different time scales. ...
Low-back pain is a major concern among healthcare workers. One cause is the frequent adoption of repetitive forward bent postures in their daily activities. Occupational exoskeletons have the potential to assist workers in such situations. However, their efficacy is largely task-dependent, and their biomechanical benefit in the healthcare sector has rarely been evaluated. The present study investigates the effects of a passive back support exoskeleton in a simulated patient bed bathing task. Nine participants performed the task on a medical manikin, with and without the exoskeleton. Results show that working with the exoskeleton induced a significantly larger trunk forward flexion, by 13¼deg in average. Due to this postural change, using the exoskeleton did not affect substantially the muscular and cardiovascular demands nor the perceived effort. These results illustrate that postural changes induced by exoskeleton use, whether voluntary or not, should be considered carefully since they may cancel out biomechanical benefits expected from the assistance.
Full-text available
Gravity is a ubiquitous component of our environment that we learnt to optimally integrate in movement control. Yet, altered gravity conditions arise in numerous applications from space exploration to rehabilitation, thereby pressing the sensorimotor system to adapt. Here, we used a robotic exoskeleton to test whether humans can quickly reoptimize their motor patterns in arbitrary gravity fields, ranging from 1g to -1g and passing through Mars- and Moon-like gravities. By comparing the motor patterns of actual arm movements with those predicted by an optimal control model,we showthat our participants (N = 61) quickly and optimally adapted their motor patterns to each local gravity condition. These findings show that arbitrary gravity-like fields can be efficiently apprehended by humans, thus opening new perspectives in arm weight support training in manipulation tasks, whether it be for patients or astronauts.
Conference Paper
Full-text available
Establishing a symbiotic relationship between a human and a exoskeleton is the end goal in many applications in order to provide benefits to the user. However, the literature focusing on the human side of human-exoskeleton interaction has remained less exhaustive than the literature focusing on the design (hardware/software) of the exoskeleton device itself. It is, though, essential to understand how a human adapts his motor control when interacting with an exoskeleton. Motor adaptation is an implicit process carried out by the central nervous system when the body encounters a perturbation, a paradigm that has been extensively studied in the field of human motor control research. When wearing an exoskeleton, even “as-transparent as- possible”, contact/interaction forces may impact well-known motor control laws in a way that may be detrimental to the user, and even compromise usability in real applications. The present paper investigates how interaction with a backdrivable upper-limb exoskeleton (ABLE) set in “transparent” mode of control affects the kinematics/dynamics of human movement in a simple task. We find that important motor control features are preserved when moving with ABLE but an overall movement slowness occurs, likely as a response to increased inertia according to optimal control simulations. Such a human motor control approach illustrates one possible way to assess the degree of symbiosis between human and exoskeleton, i.e. by grounding on well-known findings in motor control research.
Full-text available
The brain has evolved an internal model of gravity to cope with life in the Earth's gravitational environment. How this internal model benefits the implementation of skilled movement has remained unsolved. One prevailing theory has assumed that this internal model is used to compensate for gravity's mechanical effects on the body, such as to maintain invariant motor trajectories. Alternatively, gravity force could be used purposely and efficiently for the planning and execution of voluntary movements, thereby resulting in direction-depending kinematics. Here we experimentally interrogate these two hypotheses by measuring arm kinematics while varying movement direction in normal and zero-G gravity conditions. By comparing experimental results with model predictions, we show that the brain uses the internal model to implement control policies that take advantage of gravity to minimize movement effort.
Full-text available
Motor learning has been extensively studied using dynamic (force-field) perturbations. These induce movement errors that result in adaptive changes to the motor commands. Several state-space models have been developed to explain how trial-by-trial errors drive the progressive adaptation observed in such studies. These models have been applied to adaptation involving novel dynamics, which typically occurs over tens to hundreds of trials, and which appears to be mediated by a dual-rate adaptation process. In contrast, when manipulating objects with familiar dynamics, subjects adapt rapidly within a few trials. Here, we apply state-space models to familiar dynamics, asking whether adaptation is mediated by a single-rate or dual-rate process. Previously, we reported a task in which subjects rotate an object with known dynamics. By presenting the object at different visual orientations, adaptation was shown to be context-specific, with limited generalization to novel orientations. Here we show that a multiple-context state-space model, with a generalization function tuned to visual object orientation, can reproduce the time-course of adaptation and de-adaptation as well as the observed context-dependent behavior. In contrast to the dual-rate process associated with novel dynamics, we show that a single-rate process mediates adaptation to familiar object dynamics. The model predicts that during exposure to the object across multiple orientations, there will be a degree of independence for adaptation and de-adaptation within each context, and that the states associated with all contexts will slowly de-adapt during exposure in one particular context. We confirm these predictions in two new experiments. Results of the current study thus highlight similarities and differences in the processes engaged during exposure to novel versus familiar dynamics. In both cases, adaptation is mediated by multiple context-specific representations. In the case of familiar object dynamics, however, the representations can be engaged based on visual context, and are updated by a single-rate process.
Full-text available
According to basic physics, the local effects induced by gravity and acceleration are identical and cannot be separated by any physical experiment. In contrast-as this study shows-people adjust the grip forces associated with gravitational and inertial forces differently. In the experiment, subjects oscillated a vertically-oriented handle loaded with five different weights (from 3.8 N to 13.8 N) at three different frequencies in the vertical plane: 1 Hz, 1.5 Hz and 2.0 Hz. Three contributions to the grip force-static, dynamic, and stato-dynamic fractions-were quantified. The static fraction reflects grip force related to holding a load statically. The stato-dynamic fraction reflects a steady change in the grip force when the same load is moved cyclically. The dynamic fraction is due to acceleration-related adjustments of the grip force during oscillation cycles. The slope of the relation between the grip force and the load force was steeper for the static fraction than for the dynamic fraction. The stato-dynamic fraction increased with the frequency and load. The slope of the dynamic grip force-load force relation decreased with frequency, and as a rule, increased with the load. Hence, when adjusting grip force to task requirements, the central controller takes into account not only the expected magnitude of the load force but also such factors as whether the force is gravitational or inertial and the contributions of the object mass and acceleration to the inertial force. As an auxiliary finding, a complex finger coordination pattern aimed at preserving the rotational equilibrium of the object during shaking movements was reported.
Full-text available
Humans display accurate limb behavior when they move in familiar environments composed of many simultaneously-acting forces. Little is known about how multi-force environments are represented and whether this process partitions between the underlying force components, reflects the net forces present, or is cued to the force-context. We tested between these three main alternatives by examining how reaching movements adapt to a novel multi-force field composed of a velocity-dependent force and a constant force. These hypotheses were dissociated first by making the constant force larger and oppositely-oriented to the velocity-dependent force; thereby, the net force was always opposite the velocity-dependent component. Second, we tested adaptation with all novel forces removed to eliminate any potential cues for the force-context. In two experiments that used forces perpendicular or parallel to the forward movement direction, we found adaptation aftereffects consistent with a mechanism that partitioned the velocity-dependent component from the net force field. Specifically, we found aftereffects opposite the rightward or resistive velocity-dependent component of the multi-force field, even though the net force imposed was leftward or assistive, respectively. An additional experiment suggested that the velocity-dependent component is partitioned relative to the background load in a limb-based coordinate frame.
Human subjects pointed at stationary visual targets without sight of their arm while the force environment was varied by applying weight or spring loads to the hand. The path travelled by the finger, pointing accuracy, and the shape of the finger velocity profile remained invariant across all force environments after a single practice trial. However, the magnitude and duration of the velocity profile depended consistently on the presence and size of a weight load. In contrast, velocity was not affected by spring loads. An analysis of movement dynamics in our study indicated that inertial and gravitational load components were compensated by separate mechanisms, the former employing time- and the latter magnitude scaling of muscle force profiles. The presence of such separate mechanisms led us to predict little problems for movement dynamics in weightlessness, which was indeed confirmed in a study on pointing movements aboard the KC-135 aircraft.