Dynamic simulation of multi-chain robotic systems, such as multiple manipulators for multilegged vehicles, is computationally complex involving a large number of repetitive matrix inversions and trigonometric function evaluations. "Super-realtime" computational rates for simulation (planning seconds of motion in milliseconds) are needed for predicting the response of the system in order to determine the coordination/control strategies to ensure safety and stability while optimizing the planned trajectory. It is believed that parallel algorithms for the dynamical equations can be developed which will significantly speed up simulation so that realtime prediction of system response will be possible. The Electrical Engineering Department of the Ohio State University will be purchasing a 32-node suitable for developing parallel algorithms. In fact, the dynamical equations in the recursive doubling form may map directly onto the hypercube. The dynamical equations are known, but the hypercube is a new computational system for which only limited software is available. This Expedited Award for Novel Research will take full advantage for this new parallel computer to test the viability of parallel algorithms for realtime system simulation. The institutional support is adequate and the PI is well qualified, I recommend support.
