Physical interface dynamics alter how robotic exosuits augment human movement: implications for optimizing wearable assistive devices

Background Wearable assistive devices have demonstrated the potential to improve mobility outcomes for individuals with disabilities, and to augment healthy human performance; however, these benefits depend on how effectively power is transmitted from the device to the human user. Quantifying and understanding this power transmission is challenging due to complex human-device interface dynamics that occur as biological tissues and physical interface materials deform and displace under load, absorbing and returning power. Methods Here we introduce a new methodology for quickly estimating interface power dynamics during movement tasks using common motion capture and force measurements, and then apply this method to quantify how a soft robotic ankle exosuit interacts with and transfers power to the human body during walking. We partition exosuit end-effector power (i.e., power output from the device) into power that augments ankle plantarflexion (termed augmentation power) vs. power that goes into deformation and motion of interface materials and underlying soft tissues (termed interface power). Results We provide empirical evidence of how human-exosuit interfaces absorb and return energy, reshaping exosuit-to-human power flow and resulting in three key consequences: (i) During exosuit loading (as applied forces increased), about 55% of exosuit end-effector power was absorbed into the interfaces. (ii) However, during subsequent exosuit unloading (as applied forces decreased) most of the absorbed interface power was returned viscoelastically. Consequently, the majority (about 75%) of exosuit end-effector work over each stride contributed to augmenting ankle plantarflexion. (iii) Ankle augmentation power (and work) was delayed relative to exosuit end-effector power, due to these interface energy absorption and return dynamics. Conclusions Our findings elucidate the complexities of human-exosuit interface dynamics during transmission of power from assistive devices to the human body, and provide insight into improving the design and control of wearable robots. We conclude that in order to optimize the performance of wearable assistive devices it is important, throughout design and evaluation phases, to account for human-device interface dynamics that affect power transmission and thus human augmentation benefits. Electronic supplementary material The online version of this article (doi:10.1186/s12984-017-0247-9) contains supplementary material, which is available to authorized users.

The length estimate defined by _ assumes that the cable end-effector is taut (straight-line connection), which is true in this exosuit when non-negligible forces are applied. The measured load cell force magnitude borne by the cable, , is directed along the line of action of the cable end-effector, yielding the force vector in 3D space, , which we defined as oriented in the opposite direction of _ .
• _ ( S3 ) Velocity of cable lengthening/shortening along the line of action of the cable end-effector, _ , can then be computed as: With this convention, increasing negative values of _ signify increasing shortening velocity of the cable.
Finally, we can compute power due to length changes of the cable end-effector, _ , by computing the dot product of cable force and velocity vectors, yielding Eqn. 1 in the main text.

Direct Augmentation Power and Indirect Interface Power
In the main text, ankle augmentation power, _ , was estimated using what we term an indirect approach (adding interface power to cable end-effector power). An alternative way to compute augmentation power, which we term the direct method and denote as , is by taking the dot product of the applied force and the velocity due to joint rotation [5], [11], [22], [31]- [33]. In this study, this velocity was calculated by taking the cross product of the ankle angular velocity, , based on a rigid-body link-segment model, and the moment arm, , of the cable acting about the estimated ankle joint center. The moment arm was calculated as the perpendicular distance between the ankle joint center and the line of cable action.

• ( S5 )
Subtracting cable end-effector power, _ , from augmentation power, , then provides an indirect estimate of total interface power. We maintained the convention from the main text in which interface power absorption was negative. We refer to this estimate as indirect interface power, _ , because it represents what is left over after accounting for augmentation power contributions. This indirect estimate cannot localize or partition power contributions due to individual interfaces (e.g., proximal vs. distal).
A conceptual comparison of analysis methods is shown in Fig. S2: (A) starting from direct augmentation power estimates, then indirectly estimating total interface power, (B) starting from direct interface power estimates, then indirectly estimating augmentation power, or (C) using idealized analysis in which interface dynamics are assumed to be zero (non-existent). . The benefit of method (B) is that individual interface contributions can be partitioned. Idealized analysis, as shown in column (C), is expected to greatly overestimate augmentation power and underestimate biological power ( _ _ ), due to neglected interface dynamics. Indirect power estimates (i.e., computed by adding/subtracting power terms) are shown as dashed lines. Direct power estimates (i.e., computed by multiplying force by velocity, or torque by angular velocity) are shown as solid lines. No power units are shown because this is a conceptual/explanatory representation, not data.

Direct vs. Indirect Power Estimates
We found that direct and indirect estimates yielded similar interface power and similar augmentation power curves (Fig. S3). Both interface power estimates ( and _ ) followed a similar pattern of energy absorption and return. Ankle augmentation power was positive for both methods ( and _ ), with similar peak magnitudes and timing. In terms of work, direct vs. indirect augmentation work during exosuit loading phase was 4.7 ± 0.4 J vs. 4.7 ± 0.4 J, and during exosuit unloading 6.6 ± 0.7 J vs. 5.4 ± 0.5 J, respectively. Minor quantitative differences are to be expected due to measurement limitations and assumptions inherent in each estimation method. However, since these calculations are based on different underlying assumptions, the strong agreement between curves (Fig. S3) gives confidence in the general magnitudes and waveforms of augmentation and interface powers.

Work Values at Lower Peak Exosuit Force
Work values are provided in Table S1 based on data from the first minute of the walking trial, during which exosuit forces gradually ramped up.