Multi-Phase Joint-Angle Trajectory Generation Inspired by Dog Motion for Control of Quadruped Robot.

Abstract

Quadruped robots are receiving great attention as a new means of transportation for various purposes, such as military, welfare, and rehabilitation systems. The use of four legs enables a robustly stable gait; compared to the humanoid robots, the quadruped robots are particularly advantageous in improving the locomotion speed, the maximum payload, and the robustness toward disturbances. However, the more demanding conditions robots are exposed to, the more challenging the trajectory generation of robotic legs becomes. Although various trajectory generation methods (e.x., central pattern generator, finite states machine) have been developed for this purpose, these methods have limited degrees of freedom with respect to the gait transition. The conventional methods do not consider the transition of the gait phase (i.e., walk, amble, trot, canter, and gallop) or use a pre-determined fixed gait phase. Additionally, some research teams have developed locomotion algorithms that take into account the transition of the gait phase. Still, the transition of the gait phase is limited (mostly from walking to trot), and the transition according to gait speed is not considered. In this paper, a multi-phase joint-angle trajectory generation algorithm is proposed for the quadruped robot. The joint-angles of an animal are expressed as a cyclic basis function, and an input to the basis function is manipulated to realize the joint-angle trajectories in multiple gait phases as desired. To control the desired input of a cyclic basis function, a synchronization function is formulated, by which the motions of legs are designed to have proper ground contact sequences with each other. In the gait of animals, each gait phase is optimal for a certain speed, and thus transition of the gait phases is necessary for effective increase or decrease in the locomotion speed. The classification of the gait phases, however, is discrete, and thus the resultant joint-angle trajectories may be discontinuous due to the transition. For the smooth and continuous transition of gait phases, fuzzy logic is utilized in the proposed algorithm. The proposed methods are verified by simulation studies.