Real-time lumbosacral joint loading estimation in exoskeleton-assisted lifting conditions via electromyography-driven musculoskeletal models

Towards the adoption of wearable exoskeletons in occupational workspaces: model-based assessment and control of back-support exoskeletons

“…For instance, given similar lifting kinematics, kinematic-based controllers will typically provide similar assistive forces when lifting 5 or 15 kg weights. Nevertheless, lumbosacral joint compression forces greatly differ between these two external loading conditions [19,40], which suggests that modulation of assistive forces to different lifting conditions is crucial factor for back-support exoskeletons. Another limitation of state of the art active exoskeletons stems from state machine implementations, which rely on specific pre-defined kinematic trajectories to determine assistive forces [54,58,59,61,63].…”

“…Sensors used to measure electromyographic signals (EMG) and joint angles (Inertial Measurement Units) are represented. (B) In the musculoskeletal modeling stage, our previously validated EMG-driven musculoskeletal modeling framework [19,40,120] employed MTU lengths, moment arms and activations to obtain MTU forces, which were used to derive real-time L5/S1 flexion-extension joint moments. To do so, a multidimensional cubic B-spline block was used to estimate, in real-time, muscle-tendon unit (MTU) lengths and three-dimensional moment arms, using joint angles as input.…”

“…A simplified version of the OpenSim lifting full-body model [118] (earlier described in [40]), was used to compute ID L5/S1 joint moments, which were used for EMG-model calibration. First, 3D marker data from a static recording was used to linearly scale the generic model to match the subject-specific anthropometry, using OpenSim 4.1 [119].…”

“…Relying on the musculoskeletal geometry of the scaled models, we created subjectspecific real-time EMG-driven musculoskeletal models of the trunk, using our previously developed CEINMS toolbox (Calibrated EMG-informed Neuromusculoskeletal) [40,120]. EMG-driven models enabled the estimation of muscletendon forces using experimentally measured joint angles and EMGs, which can in turn be translated into joint kinetics via subject-specific geometrical models (see section 5.2.2).…”

“…In the present study, we employed our previously validated large-scale (164 musculo-tendon units) real-time EMG-driven musculoskeletal modeling framework [40] to develop a neuromechanical model-based control strategy (NMBC) for a back-support cable-driven soft exosuit. This novel HMI utilizes the active component of model-based lumbosacral joint moments to determine the magnitude of the assistance provided by the soft exosuit during lifting tasks involving different weights.…”