Tuesday, April 18, 2006
11:00 AM
CSB 209
Xue Gu
University of Rochester
An Equilibrium Point Based Humanoids Control Model
Despite various models proposed for human movement control, few tackle the intricacies of the human musculoskeletal system itself. In human musculoskeleton system, a huge amount of energy can be stored passively in the biomechanics of the muscle system. Controlling such a system in a way that takes advantage of the stored energy has lead to the Equilibrium Point Hypothesis (EPH). This hypothesis holds that central nervous system computes the equilibrium points (EPs) for a task and movements are achieved by passively attracting the muscular system to those EPs. In the forty years since the EPH was initially proposed by Feldman, the focus of research has been on testing prediction of the EPH, but much less research has been directed to how EPs are calculated as part of a motor plan. The central motor planning issue is how to deploy the huge number degrees of freedom (DOFs) of the musculoskeletal system to achieve motor goals. This thesis proposes that human solve this problem by motor simulation. EPs are computed prior to motor execution during two phases of motor planning. During the motor simulation, gradient descent is used to steer points on an abstract musculoskeltal model to their targets to generate an initial inverse kinematics solution. The resultant configuration provides a set of EPs for an executable movement. In movement execution, damped springs are simulated as the abstraction of actual muscles. Given the EPs planned in first phase, spring natural lengths are configured and movements are generated as far as the spring actual lengths are deviated from the EPs. Demonstrations using our model show that it can unify the control of various motions, such as reaching, walking, grasping, object manipulation, sitting and rising. Furthermore, our model is simple and power enough to synthesize whole body movements in humanoids with high DOFs. We propose a 3 layered framework to autonomously compose complex movements from a repository of motor routines. Motor routine is a movement unit which implements a functional task and only involves those active joints participating in the task. We present three different ways of motor synergies over multiple motor routines to compose complex movements in a 33 DOF humanoid.