16

While pose controllers can be automatically synthesized with reasonable efficiency for relatively

simple creatures and motions [vKF94], they work best for naturally stable motions.

The base

pose controllers presented in this thesis are reasonably complex and have all been designed by

hand.

The possibility of using automatic synthesis with the control techniques described is

discussed later as future work.

To achieve the types of motions we are interested in, it is necessary to adjust our PCG to perform

appropriate corrective actions on each cycle. An approach to the selection of these control actions

is one of the key contributions of this thesis.

Adjustments to the PCG are accomplished by

applying linearly scaled perturbations to the PCG joint parameters during each cycle of motion.

2. 3. 1

Linear Parametric PCG Perturbations

To distinguish between the PCG providing our basic motion and the perturbations we apply to it,

we will introduce the notion of a base PCG. A base PCG is a pose control graph which provides

the fundamental cyclic motion of the creature we are trying to animate.

For example, in the case

of walking, the base PCG might consist of a left step followed by a right step.

Ideally, the

execution of a base PCG from a suitable initial state would result in the desired motion (e.g. a

walk).

However, the creatures we are interested in animating are unstable.

The periodic open-

loop actions of the base PCG invariably results in the creature falling.

In order to generate

balanced locomotion, additional control must be provided.

Chapter 3 introduces one approach to

providing such additional control, which uses the notion of linear parametric

(LPPs), which we define below.

perturbations

We begin by defining a relativePCG, which describes a change in the pose control to be applied

to a creature, typically used to effect a desired change a motion.

Consider a base PCG, B, to

which we add a relative PCG, [!]P, scaled by an arbitrary scalar constant, k:

C = B + k [!]P.