Gaussian Mixture Models for Control of Quasi-Passive Spinal Exoskeletons

Abstract

Research and development of active and passive exoskeletons for preventing work related injuries has steadily increased in the last decade. Recently, new types of quasi-passive designs have been emerging. These exoskeletons use passive viscoelastic elements, such as springs and dampers, to provide support to the user, while using small actuators only to change the level of support or to disengage the passive elements. Control of such devices is still largely unexplored, especially the algorithms that predict the movement of the user, to take maximum advantage of the passive viscoelastic elements. To address this issue, we developed a new control scheme consisting of Gaussian mixture models (GMM) in combination with a state machine controller to identify and classify the movement of the user as early as possible and thus provide a timely control output for the quasi-passive spinal exoskeleton. In a leave-one-out cross-validation procedure, the overall accuracy for providing support to the user was 86 . 72 ± 0 . 86 % (mean ± s.d.) with a sensitivity and specificity of 97 . 46 ± 2 . 09 % and 83 . 15 ± 0 . 85 % respectively. The results of this study indicate that our approach is a promising tool for the control of quasi-passive spinal exoskeletons.

Applicability of an Active Back-Support Exoskeleton to Carrying Activities

“…Passive exoskeletons, generally lighter, simpler, and cheaper than active ones, can avoid the lower limb hindrance found in walking activities (Baltrusch et al, 2019 ). This is achieved by resorting to manual clutches, spring offsets, and automatic engage or dis-engage of passive elements, like in the commercial products by Laevo 2 and Ottobock 3 , or in research prototypes (Jamšek et al, 2020 ). On the other hand, due to mechanical design limitations, passive devices cannot provide support in carrying activities.…”

“…Recent works on exoskeletons have discussed about the opportunity of exploiting human activity recognition to discriminate between different tasks such as lifting, walking, carrying, or sitting (Chen et al, 2018 , 2019 ; Poliero et al, 2019a ; Jamšek et al, 2020 ). For passive exoskeletons, this implies that, by using clutches for the engagement and disengagement of passive elements, as in Endo et al ( 2006 ), Walsh et al ( 2007 ), Ortiz et al ( 2018 ), Jamšek et al ( 2020 ), and Di Natali et al ( 2020a ), it is possible to assist only when needed, i.e., deactivate the passive elements when they create a restriction such as in the walking case. Active exoskeletons, on the other hand, thanks to their actuators versatility, could implement specific controllers for any of the previous tasks.…”

A Wearable Lower Limb Exoskeleton: Reducing the Energy Cost of Human Movement

“…In eliminating cross-validation, the total improvement in the overall accuracy of the supported users was 86.72 ± 0.86% (mean ± standard deviation), with sensitivities and specificities of 97.46 ± 2.09% and 83.15 ± 0.85%, respectively. The findings of this research suggest that the method is a highly promising instrument for the control of quasi-passive spinal exoskeletons [126].…”

Challenges and solutions for application and wider adoption of wearable robots

“…In the real world, however, nonlifting behaviors can include a variety of actions, many of which may appear similar to lifting (e.g., squatting or stretching the arms). One possible solution is to first classify a common denominator movement, such as a pre-lift and in later stages distinguish lifting versus squatting (Jamšek et al, 2020). Alternatively, a multilevel sensor fusion algorithm is needed to first categorize the wearer's behavior into broad classes, then more specific actions-for example, the three-level sensor fusion and control system proposed by Lazzaroni et al (2020) for back exoskeletons or the multilevel stumble detection system proposed by Zhang et al (2011) for artificial legs.…”