A Modular Framework for Task-Agnostic, Energy Shaping Control of Lower Limb Exoskeletons

Area of Science:

• Robotics
• Biomechanics
• Control Systems

Background:

• Backdrivable lower-limb exoskeletons can assist able-bodied users and those with gait disorders.
• Existing control frameworks lack stability and versatility across diverse daily activities and exoskeleton configurations.

Purpose of the Study:

• To present the modular, multi-task optimal energy shaping (M-TOES) framework for stable, adaptable exoskeleton control.
• To develop a data-driven optimization method for training analytical control models.

Main Methods:

• Developed the M-TOES framework using convex, data-driven optimization.
• Trained analytical control models to determine assistive joint torques for various activities.
• Evaluated controllers on the modular backdrivable lower limb unloading exoskeleton (M-BLUE) with eight able-bodied users across sit-stand, stairs, ramps, and level walking.

Main Results:

• The M-TOES framework demonstrated a modular energy basis capable of fitting normative human joint torques.
• Unilateral M-TOES controller configurations significantly reduced overall muscle activation across all tasks (p < 0.001).
• Bilateral configurations showed minimal impact, potentially due to device weight and physical constraints.

Conclusions:

• The M-TOES framework offers a provably stable control solution for lower-limb exoskeletons across diverse activities.
• Unilateral configurations show promise for reducing user muscle effort in exoskeleton-assisted locomotion.