Files in this item



application/pdfECE499-Sp2017-fu-Kaiyi.pdf (1MB)
(no description provided)PDF


Title:A Hybrid Planning and Control Model for Biped Feet Rotation
Author(s):Fu, Kaiyi
Contributor(s):Hutchinson, Seth
Subject(s):Biped locomotion
feet rotation
preview control
Abstract:This thesis proposes methods for biped walking locomotion with feet rotation. The chief objective of this work is to first generate a guide trajectory based on designing a zero moment point (ZMP) trajectory within the support polygon and obtain linear controlling methods to stabilize the walking procedure with feet rotation. With feet rotation, the walking procedure will be more humanlike, more flexible and possible saving energy. However, when the feet are rotating around their edge, either toe or heel, the entire robot is under-actuated which are more difficult to control. By using preview control, a dynamic model of the system can be derived to control the robot. This thesis is based upon a simplified model of the Reemc Robot by PAL Robotics. The simplified model has fixed arms, since only leg motions are considered, and two legs. Each leg has three degrees of freedom. The robot is presented as a three mass model. A guided gait trajectory is first generated as the boundary condition for the ZMP. Interpolation methods are used to generate a ZMP trajectory from a set of discrete points that stay inside the boundary condition. By designing the transition model from single support phase and double support phase, a general schema can be achieved. Following the assumptions of a linear inverted pendulum, trajectories of all three masses can be solved. Inverse kinematics can now give the reference joint trajectories, which, together with the reference ZMP trajectory, is used in control methods to minimize the error between the reference trajectory and actual trajectory in simulation. Control methods are used to stabilize the motion of the walking procedure. Preview control is used for the single support phase where the behavior of all three masses is linear. A proper input can be obtained through optimization. During the double support phase, the feet rotations are nonlinear and under-actuated since the feet are rotating around their edge where no torque can be produced from the ground. By using preview control, an input can be applied to the robot so that the robot can maintain dynamic stability.
Issue Date:2017-05
Date Available in IDEALS:2017-08-18

This item appears in the following Collection(s)

Item Statistics