84
For the right leg of the right turning perturbation, the initial foot rotation is performed during the

swing phase of the right leg, as shown in state 5 of Figure 5.5 (c).

During the right leg stance

phase, this rotation is eliminated.

The return rotation is done in the second stance state (S3),

thereby ensuring that the foot is planted firmly on the ground (in S2) before attempting to turn the

body. Ground friction keeps the foot position and orientation fixed, while the torso rotates about

the leg.

The left foot performs a similar motion in its stance phase to distribute the complete

rotation evenly over both feet. A similar and symmetric set of perturbations is used for turning to

the left. Figure 5.6 shows the typical action of the turning perturbation.

the motion is illustrated for the case of a stationary stepping motion.

For the sake of clarity,

Figure 5.7 shows the hip plots from a set of turning trials performed using a base PCG similar to

that of Figure 3.7. The trials use up vector RVs with Q^d= [.25,0].

As can be seen, the turning

control works well for each of the three sampling strategies.

Each trial uses a different scaling of

the turning perturbation in Figure 5.5.

The largest scaling factor used which still yields forward

motion corresponds

to

a

hip

yaw

of

8

degrees

for

each

leg,

giving

a

turning

radius

of

approximately 2.5 meters.

All of the walks face slightly into the centre of the circle, an effect

which is more pronounced for tighter curves.

S3

S4

S5

S6

S6

- swing foot

S2

S3

S3

- stance foot

The largest scaling-factor trials attempted, corresponding to 16 degrees of right and left hip yaw,

demonstrates an interesting behaviour.

In such trials, the tendency to "lead" the turn with the