A 3D generation startup reaches RSS with GPU-accelerated robust trajectory optimization
机器之心 · wechat · 2026-07-23
A 3D generation company lands an RSS robotics award nomination with GPU-accelerated motion planning
Machine Heart reports that Imagery/3D generation company Shadowmind Technology was nominated for RSS 2026's Outstanding Paper Award for its robotics paper, cuNRTO: GPU-Accelerated Nonlinear Robust Trajectory Optimization.
- The work addresses the sim-to-real gap by making trajectories robust to bounded uncertainty instead of trying to perfectly simulate reality.
- It reformulates nonlinear robust trajectory optimization so the inner loop can run efficiently on GPU, replacing an interior-point-heavy pipeline with DR-style splitting and then a full ADMM version.
- Reported speedups are dramatic: a unicycle task drops from 30,423 s to 218 s; a quadrotor task from 13,374 s to 200 s; and a Franka arm task from 3,949 s to 152 s.
- The paper claims 100% constraint satisfaction across 2,000 Monte Carlo rollouts and includes a real-robot validation on Robotarium hardware.
- The article frames this as part of a broader research arc from 3D generation and Sim-Ready assets to physical-world-ready robot control.
More from Embodied
- Overcoming RGB Sim-to-Real Gap: Gaussian Splatting Enables Zero-Shot Robot Transfer — ZeYanjie · 2026-07-23
- Palo Alto robot builders are planning a Sunday show-and-tell meetup — ZeYanjie · 2026-07-23
- ICLR 2024 paper uses LLMs to accelerate RL on 25+ long-horizon robot tasks — rsalakhu · 2026-07-23
- Self-driving electric trucks in China are reshaping logistics — TansuYegen · 2026-07-23
- ByteDance reportedly has a Vision Pro-style headset one-fifth the weight of Apple’s — Scobleizer · 2026-07-23
- MolmoAct2 runs text-to-action on a $150 robot arm and RTX 4060 GPU — DJiafei · 2026-07-23