41

As in the case of RVs, the choice of perturbations should meet a minimal set of requirements to be

effective.

The criteria for good PCG perturbations are related to the criteria for suitable RVs,

discussed in Section 3.5. In general, the following should hold:

1. The scaled perturbation should be able to reach a wide range of desired RV values

2. Perturbations should provide smooth control over the RVs so that a suitable linear

model of the discrete system can be constructed.

The inverse of this model is used to

determine the appropriate scaling factors for each perturbation for each step.

For

3. Perturbations should be designed to be as independent as possible in order to reduce or

The control perturbations chosen in this thesis affect the stance hip pitch and stance hip roll DOFs.

They are shown in Figure 3.14. Items (c) and (d) show the pose table form of the perturbations

which are scaled and added to a base PCG pose table such as that of Figure 3.7 (b).

In our case,

each of the chosen PCG perturbations affects a single DOF.

Unity-valued perturbations are used

so that the scalar multiplier units are in degrees.

Each perturbation is applied to all poses (i.e. all

states) of the related step, left or right.

perturbation.

This provides a smoother and more effective control

Figure 3.14 illustrates the hip pitch and roll perturbations for a body in free space.

When applied

to a body in contact with the ground, the perturbations primarily affect the biped as shown in

Figure 3.15. Applying a stance hip pitch perturbation varies the torso angle in the sagittal plane.

Applying a stance hip roll perturbation varies the torso angle in the coronal plane.