ZMP based reference generation for a bipedal humanoid robot

Official URL: http://192.168.1.20/record=b1293821 (Table of Contents)

Recent fifteen years witnessed fast improvements in the field of humanoid robotics. The human-like robot structure is more suitable to human environment with its supreme obstacle avoidance properties when compared with wheeled service robots. However, the walking control for bipedal robots is a challenging task due to their complex dynamics. Stable reference generation plays a very important role in control. Linear Inverted Pendulum Model (LIPM) and the Zero Moment Point (ZMP) criterion are applied in a number of studies for stable walking reference generation of biped walking robots. This thesis takes this main approach too. This thesis proposes a natural and continuous ZMP reference trajectory for a stable and human-like walk. Natural, human-like walking is obtained by ZMP trajectories which move forward under the sole of the support foot when the robot body is supported by a single leg. Robot center of mass trajectory is obtained from predefined ZMP reference trajectories by a Fourier series approximation method. The Gibbs phenomenon problem common with Fourier approximations of discontinuous functions is avoided by employing continuous ZMP references. Also, these ZMP reference trajectories, unlike many examples in the literature, possess pre-assigned single and double support phases, which are very useful in experimental tuning work. In this thesis, a method for generating a stepping sequence of finite number of steps is proposed too. The ZMP based reference generation strategy is applied on the full body humanoid robot SURALP designed at Sabanci University. Experimental results indicate that the proposed reference trajectory generation technique is successful.