Human-like gait generation from a reduced set of tasks using the hierarchical control framework from robotics

Abstract

This paper shows that controlling only a small set of adequately selected tasks is sufficient to closely reproduce the human gait kinematics. To this aim, a hierarchical controller is applied to a whole-body model including 42 degrees of freedom with 3 hierarchical tasks. The analysis of the simulated gaits shows the emergence of significant human-like properties in walking. In order to validate our results, a comparison between joint rotations in the simulated motion and in human reference motions is conducted. Finally, a discussion is given to illustrate the interest of the approach in view of related works.

Anthropomorphic Gait Generation using Differential Dynamic Programming with a Reduced Number of Cost Criteria

“…As a result we observed that the torques values remain within the physiological limits of the human movement all along the gait cycle. In our previous work [20], in which the task-hierarchy framework was used with a kinematic model, it was necessary to give the highest priority to the foot tracking task. Here, as the full body dynamics is considered, a good quality of the trajectories is obtained by putting a higher weight on the CoM cost than on the foot cost.…”

“…Our objective is to demonstrate that human-like walking can be closely reproduced by optimizing a small number of key costs functions. Based on the biomechanical literature and the observations of the human gait throughout experimental analysis [20], an appropriate reduced set of optimization criteria is identified. A specific optimization method, Differential Dynamic Programming (DDP), is used to simulate the gait from the minimization of the identified criteria.…”

“…The whole-body skeletal model proposed by [24] was considered for this study as it was proven to be sufficient to reproduce the main features of human walking [20] or parkour jumps [25]. It is a 3D whole-body skeletal dynamic model involving 42 DoFs and 19 segments.…”

Prediction of plausible locomotion using nonlinear kinematic optimization

“…Dynamic simulation of bipedal gait has mostly been investigated in robotics using optimal control (Boukheddimi et al 2019). It generally requires knowledge on the motion pattern (named walking pattern generator) and may lead to unnatural (i.e., robotic) motion.…”