Variable walking trajectory generation method for biped robots based on redundancy analysis

Abstract

This paper proposes a variable walking trajectory generation method for biped robots based on redundancy analysis. In designing a bipedal trajectory for locomotion, maintaining locomotion stability without falling down and natural patterns of the locomotion are important. First, the positions and orientations of the feet and body of a robot are chosen as the reference factors, whose trajectories are predetermined. The augmented Jacobian method is used at the acceleration level with differential equation of the reference factors and the ZMP equation to secure stable walking motions. A virtual spring-damper system is implemented as an additional task in the null-space of the augmented Jacobian to make a balanced cyclic configuration of the robot. For a more stable trajectory generation, the condition to avoid kinematic singularity is also used. Computer simulations were conducted for various situations, for example, with locomotion speed changes or an asymmetrical mass distribution, which showed that the proposed trajectory generation method was very effective in generating various walking motions for biped robots.