Abstract
This study aims at developing and evaluating a new practical method for the real-time estimate of joint torques and external wrenches during multi-contact sit-to-stand task using kinematics data only. The proposed method allows also identifying subject specific body inertial segment parameters that are required to perform inverse dynamics. The identification phase is performed using simple and repeatable motions. Thanks to an accurately identified model the estimate of the total external wrench can be used as an input to solve an under-determined multi-con-tact problem. It is solved using a constrained quadratic optimization process minimising a hybrid human-like energetic criterion. The weights of this hybrid cost function are adjusted and a sensitivity analysis is performed in order to reproduce robustly human external wrench distribution. The results showed that the pro-posed method could successfully estimate the external wrenches under buttocks, feet and hands during sit-to stand tasks (RMS error lower than 20 N and 6 N.m). The simplicity and generalisation abilities of the proposed method allow paving the way of future diagnosis solutions and rehabilitation applications, including in-home use.