Woohoo. The serial chain dynamics simulator successfully followed a straight line in the space, at constant speed, using the kinematic and dynamic model of a PUMA 560 manipulator and a computed torque controller using Recursive Newton-Euler inverse dynamics computation. The forward dynamics can be computed using the Composite Rigid-Body Method (CRBM), or Walker-Orin method, and Roy Featherstone's Articulated Body Method. In the CRBM case we get the accelerations using a Cholesky decomposition followed by a second-order Runge-Kutta-Nyström integration. See the screenshots!
In both screenshots the path is marked by the yellow line (they're more visible in the full size images: 1 2). Ok, they're not perfectly straight, but that's caused by dynamic effects. At the end of the trajectory, for example, the robot can't stop immediately -- instead, it goes ahead for a few centimeters before stopping. Having the simulator working is so great. My advisor doesn't want to kill me anymore.
By the way: new pokey strip released. New characters. Lots of action. Don't miss it!
I want a lava lamp. And SMAC.