cuRoboV2-Fast: Fast Torque-Limited Collision-Free Motion Planning
cuRoboV2-Fast: Fast Torque-Limited Collision-Free Motion Planning
This paper introduces cuRoboV2-Fast, a GPU-accelerated planner designed for fast, collision-free motion planning under torque limits. By combining B-spline trajectory optimization with differentiable inverse dynamics, the method reduces iteration counts and refinement costs. On an RTX 5090 GPU, it achieves a mean planning time of 4.4 ms with a 3 kg payload, which is 14.7 times faster than cuRoboV2. The system also demonstrates improved pick-and-drop cycle times (5.93 s) on the YAM platform using stereo cameras.
The authors present cuRoboV2-Fast, a GPU-accelerated motion planning method that claims 14.7x faster planning latency (4.4 ms) on an RTX 5090 with a 3 kg payload compared to the prior cuRoboV2 version, though these results are based on their own implementation and specific hardware configuration.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
planning time
4.4
View original evidence
timizes traversal speed along the fixed path under sampled velocity, acceleration, jerk, and torque constraints, exploiting local spline support to solve banded Newton systems efficiently on the GPU. With a 3,kg payload, cuRoboV2-Fast achieves 4.4 ms mean planning time on RTX 5090, 14.7x faster than cuRoboV2. Our method on Jetson AGX Orin plans faster than cuRoboV1 + TOPPRA on RTX 5090. A stereo-camera-based deployme
Open source S3pick-and-drop cycles
5.93
Reported trials
View original evidence
nt on YAM achieves 5.93 s mean completed pick-and-drop cycles, 34.6% shorter than cuRoboV2. Code (Apache 2.0): www.github.com/NVlabs/curobo , Project Page: www.research.nvidia.com/labs/lpr/curobov2-fast .
Open source S4payload
3
View original evidence
timizes traversal speed along the fixed path under sampled velocity, acceleration, jerk, and torque constraints, exploiting local spline support to solve banded Newton systems efficiently on the GPU. With a 3,kg payload, cuRoboV2-Fast achieves 4.4 ms mean planning time on RTX 5090, 14.7x faster than cuRoboV2. Our method on Jetson AGX Orin plans faster than cuRoboV1 + TOPPRA on RTX 5090. A stereo-camera-based deployme
Open source S3planning time reduction factor
14.7
View original evidence
timizes traversal speed along the fixed path under sampled velocity, acceleration, jerk, and torque constraints, exploiting local spline support to solve banded Newton systems efficiently on the GPU. With a 3,kg payload, cuRoboV2-Fast achieves 4.4 ms mean planning time on RTX 5090, 14.7x faster than cuRoboV2. Our method on Jetson AGX Orin plans faster than cuRoboV1 + TOPPRA on RTX 5090. A stereo-camera-based deployme
Open source S3
- code: not_reported
- paper: not_reported
- dataset: not_reported
- weights: not_reported
- code: Reported pending Open source S4
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
timizes traversal speed along the fixed path under sampled velocity, acceleration, jerk, and torque constraints, exploiting local spline support to solve banded Newton systems efficiently on the GPU. With a 3,kg payload, cuRoboV2-Fast achieves 4.4 ms mean planning time on RTX 5090, 14.7x faster than cuRoboV2. Our method on Jetson AGX Orin plans faster than cuRoboV1 + TOPPRA on RTX 5090. A stereo-camera-based deployme
Open source S3
nt on YAM achieves 5.93 s mean completed pick-and-drop cycles, 34.6% shorter than cuRoboV2. Code (Apache 2.0): www.github.com/NVlabs/curobo , Project Page: www.research.nvidia.com/labs/lpr/curobov2-fast .
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org