Here's a probable explanation from the Yahoo newsgroup. Thanks to B. Farley for making sense of this information.

My question:
> I collected some current plots for the 12V and 6V lines. The peak
> current for the 12V line is around 4 amps. The data I collected
> showed that the current goes negative and positive on the 6V line.
> I don't understand this and will need to check my test equipment
> again.
>
> The triangular power board takes 12 volts in and outputs +5V and
> +3.6V. I know +3.6V is used for the processor and memory. I'm not
> sure about the +5V (perhaps power for the op amp circuits). This
> power board contains a power resistor, two voltage regulators, two
> drive transistors and some filtering caps.
>
> I think the +12V and +6V is used as part of the h-bridge circuitry
> just based on the board traces. Here's a clue- if you measure the
> voltage from ground to the motor pins, you get +6V on both pins
> when the motors are off. I don't know enough about the motor
> drivers to make sense of this information yet.

B. Farley's response:
OK, this all fits together well -- the +5 volts would be used for the
op-amps, which are LM324 types and as such would have voltage swings
from ground to +3.3 volts when powered from +5 volts (according to
the LM324 spec sheets), making it a perfect match for the CPU's 3.3-
volt logic levels. The +5 volts is also probably used for the IR
receiver, since most of them require at least 4 volts to run well
(and they have open-collector outputs so they can connect to 3-volt
or 5-volt logic).

The motors most likely have one terminal wired to the +6V line, and
then the CPU switches the other terminal between +12V (forward), +6V
(stopped) and Ground (reverse). Since i-Cybie is a walking machine,
all of the motors reverse directions continuously throughout the
walking cycle. Virtually every motion i-Cybie ever needs to make is
going to be reversed shortly afterwards, so this type of motor
connection would not produce an uneven drain on the batteries the way
it would in a wheeled vehicle or robot (which would spend most of
it's time going forward, so only 1/2 of the battery pack would
normally drain). This would explain the 6-volts on either side of a
stopped motor, and also why you saw both positive and negative
current on the 6-volt line.

Hopefully I will have more time this weekend and can start putting
together a detailed electrical schematic for this beastie -- I just
got 2 more dogs (cheap!) today, so now I can procede with reckless
abandon and not have to worry about frying one while probing around
in it -- I'll be cannibalizing it for spare parts soon enough anyway!

My  response:
It's starting to make more sense now. I've confirmed that one side of
the motors is hardwired to +6V. It makes sense that the other side is
switching between gnd and +12V for direction control. Sounds like an
ingenious idea. I bet this simplifies the motor drive circuitry and
to your point, it doesn't create a voltage imbalance in the batteries
due to the walking nature of the robot.