Human-like Passing under Obstacles for Biped Robots

Passing under obstacles with biped robots is studied. Several experiments are conducted to analyze human motion while passing under obstacles. Kinematics of important joints of the body is obtained using motion capture data taken from the mentioned experiments. A pattern, observed when humans pass under obstacles, is introduced for simulating this kind of motion. Linear regression is used to generalize the proposed pattern. In order to investigate the mechanical characteristics of the normal and bent locomotion during the complete gait cycle, a two-dimensional biped dynamic model with seven links is developed. The foot-ground contact is simulated using a spring-damper contact model. Equations of motion of the model are derived using Lagrange method. The equations of motion, which are nonlinear, have been solved numerically. Optimization of the normal and bent human walking states provides constant coefficients for the driving torque equations that are applied to the joints. The results obtained show the effectiveness of the regression to predict the proposed pattern, leading to a successful human-like passing under motion.

Keywords:

Biped robot , Linear regression , Obstacleavoidance , Passing under motion