IMU position in Humanoid

[i-Bot]

Does anyone have any good experience on best position of the IMU in a humanoid ?

I have previously placed the Gyro and accelerometer close to the center of mass of the whole robot, within about 1cm. Looking at the postioning on the Darwin robot they do place theirs in a similar position:
http://darwin-op.springnote.com/pages/7239531

In my new design the easiest place for the IMU is on the contoller board which about 2 cm above the center of mass vertically, though in line left right and about 1 cm back.

Since I will most likely use Darwin code, will this give problems in conversion and operation ?

[billyzelsnack]

When you say center of mass you are talking about just the torso I assume? So the torso is a single rigidbody. An unconstrained rigidbody will always rotate about its center of mass, but a robot is always constrained in many ways so it is just about never going to rotate about its center of mass. So in theory I don't think it is going to matter. Stick an orientation frame of reference anywhere on the rigidbody it is still going to be the same orientation frame of reference for the rigidbody. Position and orientation are independent. So in theory I don't think it matters. However it's always that in-practice thing that's the bitch with robots! haha.

[SK]

As Billy said, gyros measure angular rate and this is the same regardless of position on the rigid body. Accelerometers are another story, normally you want them to measure acceleration by gravity to be able to estimate gyro biases. For this, mounting them close to the COG seems reasonable (provided the COG is somewhere close to the center of rotation the robot normally experiences).
What probably is more important than varying mounting by a few cm is to have them not vibrating. I've heard a few funny stories about gyros having the same natural frequency as the vibration of the quadrotors they were mounted in and putting out only garbage once the motors were switched on.