Using portable physiological sensors to estimate energy cost for ‘body-in-the-loop’ optimization of assistive robotic devices

Lower-limb assistive robotic devices have the potential to restore ambulation in people with movement disorders. The assistance provided by these devices is governed by a large number of parameters that must be tuned on a subject-specific basis. Recently, our group developed ‘body-in-the-loop’ optimization algorithms, and demonstrated that they can be used to automatically determine the user's energetically optimal parameter setting. However, this algorithm relies on real-time estimates of energetic cost collected via indirect calorimetry, which is unsuited for long-term use. The purpose of this study was to estimate energy cost using data from portable, wearable sensors. We collected global signals (heart rate, electrodermal activity, skin temperature, oxygen saturation) and local signals (EMG, accelerometry) from 10 healthy subjects performing 6 different activities. We trained five multiple linear regression models with different subsets of the collected data, and concluded that the regression model trained with both global and local signals performed the best for all subjects (R
²
=0.94±0.02). This work has the potential to result in translational, clinically-relevant tuning algorithms for assistive robotic devices. This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. 1256260. This material is also based upon work supported by the National Science Foundation under Grant No. 1536188. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.