Kinematic Coordination of Reach and Balance

Abstract
In the present article, the authors review inverse kinematic research across a number of fields to highlight the similarities among modeling approaches to the problems of reaching and maintenance of posture. Two particular techniques are described: the Berkinblit algorithm, used to model the leg-wiping reflexes of the frog, and the parallel control scheme of Hinton, in which simple stick-man simulations maintain a balanced posture while reaching. Those 2 algorithms are shown to be equivalent to one another and are described in terms of transpose jacobian control. A controller based on those methods is compared with one based on resolved motion rate control or pseudoinverse control, commonly used in robotics applications. Both controllers are applied to the problem of coordinating reaching and balancing behavior in a simulated 3-dimensional anthropomorphic robot. It is demonstrated that both of those techniques can produce postures that are surprisingly natural.