Do humans walk like robots when crossing an obstacle without visual information?

摘要

This work analyses the obstacle crossing during gait and compares the behavior of humans with a robot model. The aim of the study is to compare the obstacle crossing task between human subjects, performing the task both with vision and blindfolded, and ZMP biped robots. It was hypothesized that the trajectories of the hip joint and the foot of the blindfolded subjects would resemble those of the robot. Seven subjects walked on a flat surface with an obstacle of 0.26 m height and crossed the obstacle successfully 30 times under two conditions: blindfolded and with normal vision. The motion of the leading limb was recorded by video at 60 Hz. There were markers placed on the subject's hip, knee, ankle, rear foot, and forefoot. The following parameters were calculated: critical time, vertical foot position and average step velocity. A robot model with inertial parameters matched with the subjects and a controller based on the ZMP criterion was developed. The hip joint and foot trajectories of humans and robots were assessed and compared. Unavailability of visual information resulted in different strategies to cross the obstacle, like a higher toe clearance or lower step speed. Without vision, the crossing pattern seems to be more cautious and slower than with vision, thus resembling that of the robot with ZMP. The hip was kept behind until the foot has overtaken the obstacle as a possible mechanism to maintain a safe base of support if a trip occurs. This is also supported by the data showing that blindfolded behaved with an intermediate pattern between vision and robot in the hip antero-posterior trajectory. In this context, it is possible to extract some conclusions to improve the ZMP stability criterion of biped robots.