Augmenting Human Balance with Generic Supernumerary Robotic Limbs

TL;DR

This work tackles safe, versatile balance augmentation with generic supernumerary robotic limbs by introducing a three-layer hierarchical framework that predicts trunk and CoM dynamics, plans CoM trajectories to counteract movement, and executes SL commands in real time. The approach relies on a 48-dimensional state vector and an LQE-based prediction layer, a gradient-descent–driven CoM planner feeding an MPC-based tracker, and real-time SL actuation, achieving stability improvements in 10 participants during forward and lateral bending tasks. Key contributions include a practically implementable architecture, real-time optimization and control at 1 kHz, and empirical evidence that augmented balance is achievable with generic SLs, though hardware weight remains a key limitation. The framework advances safe, versatile human-SL interactions and lays groundwork for integrating balance augmentation with more advanced interactive control methods in industrial and healthcare contexts.

Abstract

Supernumerary robotic limbs (SLs) have the potential to transform a wide range of human activities, yet their usability remains limited by key technical challenges, particularly in ensuring safety and achieving versatile control. Here, we address the critical problem of maintaining balance in the human-SLs system, a prerequisite for safe and comfortable augmentation tasks. Unlike previous approaches that developed SLs specifically for stability support, we propose a general framework for preserving balance with SLs designed for generic use. Our hierarchical three-layer architecture consists of: (i) a prediction layer that estimates human trunk and center of mass (CoM) dynamics, (ii) a planning layer that generates optimal CoM trajectories to counteract trunk movements and computes the corresponding SL control inputs, and (iii) a control layer that executes these inputs on the SL hardware. We evaluated the framework with ten participants performing forward and lateral bending tasks. The results show a clear reduction in stance instability, demonstrating the framework's effectiveness in enhancing balance. This work paves the path towards safe and versatile human-SLs interactions. [This paper has been submitted for publication to IEEE.]

Figures (4)

• Figure 1: Hierarchical control structure for human-SLs balance. The top layer estimates the human-SLs system state and predicts future states using LQE. The middle layer plans (i) a trajectory that minimizes the shift of the human-SLs center of mass projected in a horizontal plane, and (ii) the future torques required to follow this trajectory. The bottom layer generates motor commands for the SLs.
• Figure 2: Experimental setup.A. MUVE system integrating motion capture, an orientable split treadmill, up to four SLs and two cobots. B. Definition of the planes used to control the SLs and positions of the motion capture markers used to extract these planes. C. Nomenclature of one SL's segments and joints. D. Protocol used to systematically induce instability in the human-SLs system. E,F. Photos of a representative participant with SLs providing active stabilization during a frontal bow trial (E) and a lateral bow trial (F).
• Figure 3: Results in frontal bow trials (top) and lateral bow trials (bottom).A-C. Evolution of the CoM-SUP distance for each individual (thin lines) and the population average (thick line), in the HOnly, NoComp and Comp conditions respectively. D. Average CoM-SUP distance across the three conditions for all ten participants. E. Ground reaction force (GRF) in the three conditions. Ellipses indicate the maximum GRF of the population average, and the dots show the force distribution for a representative participant.
• Figure 4: CoP-SUP distance. Average distance across each trial for every participant in A frontal bow trials and B lateral bow trials.