NVlabs/curobo

Health v2 · maintenance only

Adoption not part of the score

1865 stars · 354 forks observed · 2026-09-23

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

cuRobo is a CUDA-accelerated Python library for robot motion generation built on PyTorch, CUDA, and Warp. It provides GPU-parallel algorithms for forward/inverse kinematics, collision checking, trajectory optimization, geometric planning, GPU-native ESDF perception, and whole-body motion generation for manipulators and high-DoF humanoids.

Use cases

• generate collision-free motion plans for robot arms on GPU
• compute inverse kinematics in parallel with CUDA
• optimize smooth dynamics-aware trajectories respecting torque limits
• build signed distance fields from depth images for collision avoidance
• plan whole-body motions for high-DoF humanoid robots
• check self-collisions and environment collisions quickly during planning

When to choose

• you need fast GPU-parallel motion planning or IK for manipulators or humanoids
• your stack already uses PyTorch and NVIDIA GPUs
• you need depth-image-based distance fields for perception-aware planning

When to avoid

• you have no NVIDIA CUDA-capable GPU
• you need a lightweight CPU-only planner
• you depend on the cuRobo v1 API without pinning to v0.7.8

Facets

library · maturity active

simulation machine-learning graphics math robotics gpu-computing autonomous-vehicles simulation python cuda motion-planning inverse-kinematics trajectory-optimization pytorch warp collision-checking esdf tsdf humanoids manipulators perception linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVlabs/curobo")