Learning the sit-to-stand human behavior: An inverse optimal control approach

The issue of understanding the underlying optimality criteria of a certain human movement behavior has received a considerable attention. Recently, an increased interest exists in understanding the Sit-to-Stand (STS) human movement behavior to facilitate the optimal design of assistive systems. Existing research of STS modeling merely depends on some of hard-coded optimality criteria, where the minimum torque change, minimum jerk or/and minimum effort are mainly adopted. In this paper, in the light of the Inverse Optimal Control (IOC) framework, the cost function underlying the STS kinematics is learned from the given human demonstrations. An Inverse Linear Quadratic Regulator (ILQR) algorithm is proposed to find out the unknown cost function that could perfectly reproduce the demonstrated STS data measured with respect to the human greater trochanter (hip) position. The retrieved STS cost function is reasonable and showing an acceptable fit between the simulated trajectories that are generated by the proposed ILQR approach and the given experimental data in terms of the hip position.