Linear inverted pendulum model and swing leg dynamics in biped robot walking trajectory generation

Official URL: http://risc01.sabanciuniv.edu/record=b1169198 (Table of Contents)

Expectations of people and researchers from robotics have changed in the last four decades. Although robots are used to play their roles in the industrial environment, they are anticipated to meet social demands of people in daily life. Therefore, the interest in humanoid robotics has been increasing day by day. Their use for elderly care, human assistance, rescue, hospital attendance and many other purposes is suggested due to their adaptability and human like structure. Biped reference trajectory generation is a challenging task as well as control owing to the instability trend, non-linear robot dynamics and high number of degrees of freedom. Hence, the generated reference trajectories have to be followed with minimum control interference. Linear Inverted Pendulum Model (LIPM) is used to meet this demand which assumes the body as a falling point mass connected to the ground with a massless rod. The Zero Moment Point (ZMP) is a stability criterion for legged robots which provides a more powerful, stable reference generation. With the assistance of this methodology, advanced Linear Inverted Pendulum Models are implemented. This thesis aims to improve the applicability of the versatile and computationally effective LIPM based reference generation approach for the robots with heavy legs. It proposes a swing-leg gravity compensation technique based on a two-mass linear inverted pendulum model which is simulated on a discrete state space model. LIPM modeling is implemented by switching between one-mass and two-mass models during double support and single support phases, respectively. The joint trajectories are then obtained by inverse kinematics and PID controllers are employed independently at joint level for locomotion. The effectiveness of the generated reference trajectories is verified by simulation. The reference generation and control algorithm is tested with a 3-D full dynamic simulator on the model of a 12 DOF biped robot. Results indicate better performance of the one-mass-twomass switching LIPM over the one-mass LIPM.