Wow, I haven't updated since July!
Well, work continues at USC. The motion capture suit project is entering testing and verification. Currently, we are testing three sensors (1 for elbow tracking, 1 for shoulder tracking, and 1 for reference.) A few months were burned designing an algorithm that fused the sensor information using floating point quaternions that could update at 100 Hz on a 8 bit, 16Mhz Atmel Mega32 processor with 32k of flash! I tried 2 algorithms from literature, Linear Gauss-Newton Iteration and a Quaternion-based Extended Kalman Filter. Both failed, the first because of a terribly slow 4x4 matrix inversion and the second because of size limitations. I scraped them and developed my own linear filter that works pretty well, although we are still tuning the parameters. Since this project is funded by DARPA, all of the info will be public domain, which can be VERY GOOD NEWS for anyone looking for a very small IMU with 3 axis gyro, 3 axis accelerometer, and 3 axis magnetometer for about 260 dollars in parts! If you are intersted let me know, I can always let you see what I got.
The humanoid is coming along. I have one arm and the body rapid prototyped and mostly constructed. I am most likely going to modify the servos so that I can just talk to them over a serial line using Atmel's cool addressing function. This way, I can get positional information back from the servo, and also program cool little functions like going slack when the joint hits something. I should have pictures up soon, I am waiting on the rapid prototyping company to send a piece that they had accidently left off of the last order, then I will have both arms and the body constructed. I feel this will give a better representation of the project, so I am holding out to take the pictures. It is cool though! Work continues on the head design to make it interactive and appealing, yet not scary to children or us for that matter. It is bigger than everyone expected, which wasn't very big since we are using digital servos. The arms are heavy enough to have to worry about inertia, etc. I think the size helps to distinguish this robot from the others that look like toys. I do want to invest a little bit of time over the summer developing a smaller version that may be purchased for at or under $3000 using rapid prototyping, or under $1000 for bulk plastic orders. I'll have to see if there is any interest for this though. Plus, you never know what is going to happen since I'll be graduating in June. There may be some good news, but I can't really talk about it now :-)