Reducing the metabolic cost of walking with an ankle exoskeleton: interaction between actuation timing and power

Stroke use and clinical testing needs a followup on this. Assisting dorsiflexion and plantatflexion(push off) would be a massive win for those survivors with drop foot. 11 years later and I still have no push off.
http://jneuroengrehab.biomedcentral.com/articles/10.1186/s12984-017-0235-0

• Samuel Galle†Email authorView ORCID ID profile,
• Philippe Malcolm†,
• Steven Hartley Collins and
• Dirk De Clercq

†Contributed equally

Journal of NeuroEngineering and Rehabilitation201714:35

DOI: 10.1186/s12984-017-0235-0

©  The Author(s). 2017

Received: 10 September 2016

Accepted: 17 March 2017

Published: 27 April 2017

Abstract

Background

Powered ankle-foot exoskeletons can reduce the metabolic cost of human walking to below normal levels, but optimal assistance properties remain unclear. The purpose of this study was to test the effects of different assistance timing and power characteristics in an experiment with a tethered ankle-foot exoskeleton.

Methods

Ten healthy female subjects walked on a treadmill with bilateral ankle-foot exoskeletons in 10 different assistance conditions. Artificial pneumatic muscles assisted plantarflexion during ankle push-off using one of four actuation onset timings (36, 42, 48 and 54% of the stride) and three power levels (average positive exoskeleton power over a stride, summed for both legs, of 0.2, 0.4 and 0.5 W∙kg⁻¹). We compared metabolic rate, kinematics and electromyography (EMG) between conditions.

Results

Optimal assistance was achieved with an onset of 42% stride and average power of 0.4 W∙kg⁻¹, leading to 21% reduction in metabolic cost compared to walking with the exoskeleton deactivated and 12% reduction compared to normal walking without the exoskeleton. With suboptimal timing or power, the exoskeleton still reduced metabolic cost, but substantially less so. The relationship between timing, power and metabolic rate was well-characterized by a two-dimensional quadratic function. The assistive mechanisms leading to these improvements included reducing muscular activity in the ankle plantarflexors and assisting leg swing initiation.

Conclusions

These results emphasize the importance of optimizing exoskeleton actuation properties when assisting or augmenting human locomotion. Our optimal assistance onset timing and average power levels could be used for other exoskeletons to improve assistance and resulting benefits.

Keywords

Human locomotion Augmentation Lower-limb exoskeletons Metabolic cost Optimal assistance

Background

Walking is the most frequent means of human locomotion [1]. While humans use many strategies to reduce energy expenditure [2], walking still requires a considerable amount of metabolic energy, sometimes referred to as the ‘metabolic cost’ of walking. Assisting the ankle joint with an exoskeleton can reduce the metabolic cost of walking to below the cost of normal walking [3, 4, 5, 6]. This shows that it is possible to reduce metabolic cost through robotic assistance.

Reductions in the metabolic cost of walking with ankle-foot exoskeletons result from two competing factors. A benefit can be derived from the exoskeleton when it acts to assist gait, expressed as the difference between powered exoskeleton¹ walking and walking in zero-work mode¹. However, wearing the exoskeleton in zero-work mode typically results in a metabolic penalty, expressed as the difference between normal walking¹ without an exoskeleton and walking in zero-work mode. Some full-body exoskeletons have resulted in large metabolic penalties (e.g. [7]) while lightweight ankle-foot exoskeletons have resulted in penalties of less than 3% for active autonomous¹ exoskeletons [4] and even close to zero for passive autonomous¹ exoskeletons [5]. Reducing the penalty of wearing an exoskeleton in zero-work mode is mainly a design challenge, while increasing the difference between the zero-work condition and powered exoskeleton conditions is mainly a biomechanics challenge.

In order to solve the latter human-exoskeleton interaction challenge, optimal assistance properties (e.g. actuation timing, assistance magnitude, etc.) are crucial to further reduce the metabolic energy cost of walking. Malcolm et al. [3] showed that the timing of exoskeleton actuation onset (referred to as actuation timing) is an important exoskeleton property that influences the metabolic cost of walking with active exoskeletons. They found a convex landscape in metabolic cost versus actuation timing with an optimum around 40% of the stride. Studies that have found the highest reductions in metabolic energy cost have also used an actuation timing around 40% of the stride [4, 6].

Of course, actuation timing is not the only determinant of metabolic cost when walking with ankle-exoskeleton assistance. Assistance magnitude also seems to have a strong effect [8, 9]. The average positive mechanical exoskeleton power per stride summed for both ankles (referred to here as exoskeleton power) can be as high as 0.38 W∙kg⁻¹ resulting in reductions in net metabolic cost of between 10 and 22% for powered exoskeleton conditions compared to zero-work conditions [3, 4, 6, 8, 10, 11, 12]. However, comparing these studies does not result in a clear relationship between exoskeleton power magnitude and metabolic cost, likely because many factors differ between studies (e.g. design, exoskeleton mass, actuation profile, etc.), confounding comparisons. The simplest walking model [13] would suggest that increasing exoskeleton power will reduce the mechanical energy requirements for walking until subjects walk with zero metabolic cost. Indeed, a recent study, in which both ankle and hip joints were assisted with a soft exo-suit [8] indicated that metabolic energy cost reduces linearly with increasing exoskeleton assistance magnitude, similar to some findings with active prostheses [9]. On the other hand, a study on unilateral exoskeleton assistance suggested an exponential relationship between device power and metabolic cost [14]. Experiments and simulation studies with exoskeletons have similarly suggested that under some conditions “more is not always better” [5, 15, 16]. Interpretations have been made more difficult by the limited range of attainable levels of exoskeleton power, which has often been between 50 and 80% of biological ankle power [4, 10, 11].

In order to study if and when the reduction in the metabolic cost of walking begins to level-off with increasing exoskeleton power during bilateral exoskeleton assistance, it seems necessary to deliver more power than in current studies. To identify the influence of exoskeleton power magnitude on the metabolic cost of walking, as well as the interaction with actuation timing, there is a need for a parametric study of actuation timing and exoskeleton power over a larger range. A study of both actuation characteristics is also expected to contribute to an improved understanding of the assistive mechanisms of ankle-foot exoskeletons. Several studies have indicated that other joints besides the ankle joint are involved in the reduction in metabolic cost experienced when using an ankle-foot exoskeleton [3, 4, 5, 8, 12, 17, 18] but the exact mechanisms are unclear. Exploring different assistance parameters over a broad range would help to identify the relationship between biomechanical changes and the resulting changes in metabolic cost.

The overall goal of this study was to characterize the relationship between ankle exoskeleton power, actuation timing, and metabolic cost during walking over a broad range. We used a tethered and powered plantarflexion-assisting exoskeleton to vary actuation onset timing and average exoskeleton power independently and over a broad range and studied the influence of these characteristics on the metabolic energy cost of walking. We expected a second-order effect of actuation timing on metabolic energy cost [3] and explored several candidate relationships between exoskeleton power and metabolic energy cost to evaluate the interaction between timing, power and metabolic cost. A secondary goal was to use the best relationship to define optimal assistance parameters. Finally, we analyzed muscle activation, exoskeleton kinetics and walking kinematics that describe the neuromechanical interaction between the exoskeleton and the human, with the goal of explaining the reduction in metabolic cost and improving our understanding of human-exoskeleton interaction.

More at link.