Predicting multiple step placements for human balance recovery tasks

Stepping is one of the predominant strategies to restore balance against an external perturbation. Although models have been proposed to estimate the recovery step placement for a given perturbation, they suffer from major limitations (step execution time usually neglected, no more than a single step recovery considered, etc.). The purpose of this study is to overcome these limitations and to develop a simple balance recovery model which can predict a complete multiple step recovery response. Inspired by the field of walking robots, we adapted a control scheme formerly proposed for biped robot locomotion. The scheme relies on a Linear Model Predictive Controller (LMPC) which estimates the best foot placements to zero the velocity of the Center of Mass (CoM), i.e. to reach a steady posture. The predicted step placements were compared against previously reported experimental data for tether-release conditions. They match correctly for various perturbation levels and both single step or multiple steps recovery. Although the current model still suffers from limitations (e.g., limited to the sagittal plane), these results demonstrate its ability to reproduce balance recovery reactions for different experimental scenarios.