GPU-Accelerated cuNRTO Boosts Robot Safety Planning 140x

jiqizhixin · x · 2026-08-01

Georgia Tech, UC San Diego, and Deemos Corp introduced cuNRTO, a GPU-powered framework that solves complex nonlinear robust trajectory optimization using parallel computation.

Tested on unicycles, quadcopters, and robotic arms, cuNRTO achieves speedups of up to 139.6x compared to previous methods, drastically accelerating safe motion planning for robots.

Original post →

More from Embodied

Embodied channel →