Impact of elastic ankle exoskeleton stiffness on neuromechanics and energetics of human walking across multiple speeds

BackgroundElastic ankle exoskeletons with intermediate stiffness springs in parallel with the human plantarflexors can reduce the metabolic cost of walking by similar to 7% at 1.25ms(-1). In a move toward 'real-world' application, we examined whether the unpowered approach has metabolic benefit across a range of walking speeds, and if so, whether the optimal exoskeleton stiffness was speed dependent. We hypothesized that, for any walking speed, there would be an optimal ankle exoskeleton stiffness - not too compliant and not too stiff - that minimizes the user's metabolic cost. In addition, we expected the optimal stiffness to increase with walking speed.MethodsEleven participants walked on a level treadmill at 1.25, 1.50, and 1.75ms(-1) while we used a state-of-the-art exoskeleton emulator to apply bilateral ankle exoskeleton assistance at five controlled rotational stiffnesses (k(exo)=0, 50, 100, 150, 250Nmrad(-1)). We measured metabolic cost, lower-limb joint mechanics, and EMG of muscles crossing the ankle, knee, and hip.ResultsMetabolic cost was significantly reduced at the lowest exoskeleton stiffness (50Nmrad(-1)) for assisted walking at both 1.25 (4.2%; p=0.0162) and 1.75ms(-1) (4.7%; p=0.0045). At these speeds, the metabolically optimal exoskeleton stiffness provided peak assistive torques of similar to 0.20Nmkg(-1) that resulted in reduced biological ankle moment of similar to 12% and reduced soleus muscle activity of similar to 10%. We found no stiffness that could reduce the metabolic cost of walking at 1.5ms(-1). Across all speeds, the non-weighted sum of soleus and tibialis anterior activation rate explained the change in metabolic rate due to exoskeleton assistance (p<0.05; R-2>0.56).ConclusionsElastic ankle exoskeletons with low rotational stiffness reduce users' metabolic cost of walking at slow and fast but not intermediate walking speed. The relationship between the non-weighted sum of soleus and tibialis activation rate and metabolic cost (R-2>0.56) indicates that muscle activation may drive metabolic demand. Future work using simulations and ultrasound imaging will get 'under the skin' and examine the interaction between exoskeleton stiffness and plantarflexor muscle dynamics to better inform stiffness selection in human-machine systems.