Unpowered exoskeletons with springs in parallel to human plantar flexor muscle-tendons can reduce the metabolic cost of walking. We used ultrasound imaging to look ‘under the skin’ and measure how exoskeleton stiffness alters soleus muscle contractile dynamics and shapes the user’s metabolic rate during walking. Eleven participants (4F, 7M; age: 27.7 ± 3.3 years) walked on a treadmill at 1.25 m s^-1 and 0% grade with elastic ankle exoskeletons (rotational stiffness: 0-250 Nm rad^-1) in one training and two testing days. Metabolic savings were maximized (4.2%) at a stiffness of 50 Nm rad^-1. As exoskeleton stiffness increased, the soleus muscle operated at longer lengths and improved economy (force/activation) during early stance, but this benefit was offset by faster shortening velocity and poorer economy in late stance. Changes in soleus activation rate correlated with changes in users’ metabolic rate (p = 0.038, R² = 0.44), highlighting a crucial link between muscle neuromechanics and exoskeleton performance; perhaps informing future ‘muscle-in-the loop’ exoskeleton controllers designed to steer contractile dynamics toward more economical force production.

Exoskeletons that improve locomotion economy typically are engineered to reduce users’ limb joint mechanical work or moments. Yet, limb joint dynamics do not necessarily reflect muscle dynamics, which dictate whole-body metabolic energy expenditure. Here, we hypothesize that exoskeletons primarily reduce user metabolic energy expenditure across locomotion conditions by reducing active muscle volume.

Key Points

• Center of mass, limb joint, and muscle mechanical do not explain well how exoskeletons alter locomotion economy.
• Limb joint dynamics do not necessarily reflect the underlying muscle dynamics across locomotion conditions.
• Active muscles are the primary drivers of whole-body metabolic energy expenditure during locomotion. Consequently, exoskeletons likely need to consider muscle dynamics to optimize locomotion economy.
• During walking and hopping with an exoskeleton, muscle force generation is a better correlate to locomotion economy than previously measured mechanical work parameters.
• Tracking muscle length changes in vivo may help provide reasonably accurate active muscle volume calculations.
• Future exoskeleton controllers may incorporate real-time muscle physiology measures to update device characteristics and maintain minimal active muscle volume and metabolic energy expenditure across locomotion conditions.