Date Published: July 12, 2017
Publisher: Public Library of Science
Author(s): Christopher L. Dembia, Amy Silder, Thomas K. Uchida, Jennifer L. Hicks, Scott L. Delp, Øyvind Sandbakk.
Wearable robotic devices can restore and enhance mobility. There is growing interest in designing devices that reduce the metabolic cost of walking; however, designers lack guidelines for which joints to assist and when to provide the assistance. To help address this problem, we used musculoskeletal simulation to predict how hypothetical devices affect muscle activity and metabolic cost when walking with heavy loads. We explored 7 massless devices, each providing unrestricted torque at one degree of freedom in one direction (hip abduction, hip flexion, hip extension, knee flexion, knee extension, ankle plantarflexion, or ankle dorsiflexion). We used the Computed Muscle Control algorithm in OpenSim to find device torque profiles that minimized the sum of squared muscle activations while tracking measured kinematics of loaded walking without assistance. We then examined the metabolic savings provided by each device, the corresponding device torque profiles, and the resulting changes in muscle activity. We found that the hip flexion, knee flexion, and hip abduction devices provided greater metabolic savings than the ankle plantarflexion device. The hip abduction device had the greatest ratio of metabolic savings to peak instantaneous positive device power, suggesting that frontal-plane hip assistance may be an efficient way to reduce metabolic cost. Overall, the device torque profiles generally differed from the corresponding net joint moment generated by muscles without assistance, and occasionally exceeded the net joint moment to reduce muscle activity at other degrees of freedom. Many devices affected the activity of muscles elsewhere in the limb; for example, the hip flexion device affected muscles that span the ankle joint. Our results may help experimentalists decide which joint motions to target when building devices and can provide intuition for how devices may interact with the musculoskeletal system. The simulations are freely available online, allowing others to reproduce and extend our work.
Wearable robotic devices are currently used to help restore mobility to individuals following a stroke, a spinal cord injury, or the loss of a limb [1–4]. Other potential uses for assistive devices are to reduce injury risk for those carrying heavy loads, such as firefighters , laborers , and soldiers . A common goal of assistive device designers is to reduce the metabolic cost of walking. Yet, reducing the metabolic cost of walking using a device is difficult—despite decades of effort [8–11], this has been accomplished only recently [9,12–19]. Designers are making progress on overcoming the challenges of large subject-to-subject variability in performance , minimizing the metabolic penalty of carrying a device , and designing effective training protocols [20,21]. One of the largest challenges is understanding the complex neuromusculoskeletal adaptations (short- and long-term) that occur when the body is augmented with assistive devices. For example, even if a device reduces muscle activity, the device may not reduce metabolic cost or muscle fiber power [22,23].
We simulated 7 hypothetical ideal devices and found that three of them yielded greater metabolic savings than our simulated ankle plantarflexion device (Fig 1). This is noteworthy given the current popularity of experimental ankle plantarflexion devices [9,13,14,22,50,56]. Because we directly estimated the metabolic savings achieved with different device locations, our study is an important step away from relying on indirect and coarse measures like positive joint power to decide where to assist. Part of the focus on ankle devices comes from the ankle’s large share of positive power output in walking [46,57,58]. Our results suggest that a device at a joint with high positive work, such as the ankle in loaded walking , does not necessarily yield the highest metabolic savings . Nevertheless, all the devices we explored warrant consideration from device designers: reducing metabolic rate by even 5%—as is possible with a hip extension device —is tantamount to removing 4 kg from a torso load  and would markedly help load carriers.
In this study, we used musculoskeletal simulation to evaluate how 7 hypothetical, ideal, bilateral assistive devices affected muscle activity and metabolic cost when walking with heavy loads. This work provides a foundation for understanding the musculoskeletal factors that may affect device performance. We also provided suggestions to device designers, which can serve as a springboard for deciding which devices to create next. In particular, we are excited for designers to create hip abduction devices that incorporate passive components, and to explore devices that actuate multiple degrees of freedom.