{"version":1,"type":"story","url":"https://digestai.news/story/nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations","json":"https://digestai.news/story/nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations.json","markdown":"https://digestai.news/story/nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations.md","slug":"nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations","headline":"Nvidia Warp and MJWarp enable up to 2,048 parallel robot simulations","summary":"It walks through validating a single world, converting it with mjw.putmodel and mjw.makedata, and then scaling the batch by tiling the initial state across worlds. The guide details required parameters such as nconmax=128 and njmax=300 for the default profile and nconmax=256 and njmax=500 for the optional reBot variant, and explains how to capture a CUDA graph for repeated stepping.\n\nKey steps include installing warp‑lang (≥ 1.15 for optional deterministic execution) and mujoco‑warp, using wp.ScopedCapture to record the step graph, and timing the throughput with a warm‑up of ten launches followed by 200 timed launches. The article also lists batch‑size experiments (1, 64, 1024, 2048, 8192 worlds) and provides command‑line examples for running the SO‑101 pick‑and‑place task at 50 Hz control frames with 10 physics substeps (0.002 s timestep).","keyPoints":["MJWarp can batch up to 2,048 independent worlds on a single NVIDIA GPU.","Warp 1.15 adds optional deterministic execution for reproducible simulations.","Default batch limits are nconmax=128 and njmax=300; reBot variant uses nconmax=256 and njmax=500."],"whyItMatters":"Scaling robot simulation to thousands of parallel worlds speeds up reinforcement‑learning data collection and design optimization without changing model code.","category":{"slug":"agents","name":"Agents & Tools","url":"https://digestai.news/category/agents"},"entities":{"companies":["NVIDIA"],"models":[],"people":[]},"firstPublishedAt":"2026-09-23T18:41:40Z","updatedAt":"2026-09-23T18:41:40Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"Hugging Face","title":"How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows","url":"https://huggingface.co/blog/nvidia/how-to-use-nvidia-warp-and-mjwarp","publishedAt":"2026-09-23T18:41:40Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Nvidia Warp and MJWarp enable up to 2,048 parallel robot simulations\", 23 September 2026, https://digestai.news/story/nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations","publisher":"Digest AI","title":"Nvidia Warp and MJWarp enable up to 2,048 parallel robot simulations","datePublished":"2026-09-23T18:41:40Z","url":"https://digestai.news/story/nvidia-warp-and-mjwarp-enable-up-to-2-048-parallel-robot-simulations"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}