Toward Humanlike Passive Dynamic Walking With Physical and Dynamic Gait Resemblance

Abstract

In this study, we present the first passive walking model on a slope to have exact anthropomorphic physical parameters in addition to the ability to closely reproduce the dynamic gait of humans. To simulate the stance phase, we utilized knee joints with variable stiffness instead of the locking mechanism that is prevalent among passive models with knees. Additionally, the model has flat feet with ankle compliance to include the distinctive double stance phase of human walking in the gait pattern. The stiffness and damping parameters of the joints are then adjusted to increase the resemblance between the model and gait data of human walking reported in literature. Following the comparison between the joint trajectories of the model and humans, it is demonstrated that the presented model displays a gait that is substantially more comparable with human walking than previous attempts at modeling humanlike passive walkers. The resulting model shows great potential, as semi-passive and efficient active walkers can be developed based on its natural humanlike dynamics. Furthermore, it may be possible to utilize this walker to simulate the complex behaviors of the human gait under different conditions and facilitate the design stages of rehabilitative or assistive devices by decreasing the need for expensive human trials and experiments.

Keywords

• Biomimetics
• humanoid and bipedal locomotion
• legged robots
• passive walking