96

6.CONCLUSIONS AND FUTURE WORK

The initial goal of this thesis was to provide a general technique for the control of complex,

statically-unstable 3D bipedal creatures for use in physically-based animation.

For simplicity,

only cyclic motions were considered.

The proposed control formulation meets this goal by

discretizing the periodic motion into cycles and simplifying the control through the use of a limited

set of control perturbations which are used to stabilize a small set of observed variables.

The approach is general in the sense that the same control technique can be used to control a wide

variety of walking gaits for a wide range of control parameter choices.

Controlof speed and

direction for a biped has been demonstrated, as well as parametric variations of a number of

walking characteristics.

By using simple feedback to drive the parameterized controllers, it is

possible to give the animated creature a greater degree of autonomy, providing the animator with

high level control.

The approach has been demonstrated for walking and limited forms of

running, using both a human model and a bird-like robot model.

We believe that the technique

will also prove to be a suitable approach to animating many other types of periodic behaviours for

a wide variety of articulated figures.

While the bipeds are not of full human complexity (which would require about 200 DOFs), they

are sufficiently complex that many other forms of control would prove computationally infeasible.

Many of the DOFs in the human body cannot act independently, for example, the vertebrae of the

spine.

The tens of DOFs in our models are sufficient to capture the gross motions of natural

bipedal locomotion with reasonable fidelity.

The primary contribution of this thesis is to illustrate that control techniques can be successfully

applied to animate creatures of high complexity.