Deformable physics sim with GPT-6 Astra: PET bottle crushing with plasticity and fluids
rocky_mountain12 · reddit · 2026-09-30
A demo of creating deformable simulation objects with GPT-6 Astra via RobotGym: an accurate physics squeeze of a water bottle. The bottle is modeled as a 0.3mm PET shell with real material values (3 GPa stiffness, 55 MPa yield), simulated on GPU with Newton's VBD solver and self-contact enabled so the wall buckles like plastic rather than squishing like rubber.
The hard part was making the dent persist: a custom plasticity kernel makes creases past PET's yield point become the new resting shape, so the bottle stays crushed when released. Water is 32,000 particles in a custom position-based fluid solver that preserves volume, forcing water up and out of the neck as the bottle compresses. Everything is path-traced in a full apartment scene with a real Franka arm performing the squeeze. The author notes this is highly useful for LLM-based robotics training in simulation.
Related event: GPT-6 Astra Simulates Physically Accurate PET Bottle Squeezing(2 posts)→
More from Embodied
- DexAgent learns bimanual dexterous manipulation from one human video, hitting 63.6% success — YunzhuLiYZ · 2026-09-30
- DexTaG uses human tactile demos as RL reward guidance to teach robots human-like tool use — gan_chuang · 2026-09-30
- Wayve's supervised robotaxi ride in London via Uber impresses, closing in on Waymo — Sethwinterroth · 2026-09-30
- DynaRobotics trains robust whole-body controller for wheeled robot via Isaac Sim RL — zhengyiluo · 2026-09-30
- Mark Cuban predicts humanoid robots will fail in 5–10 years, homes redesigned around task-specific robots — ericwdolan · 2026-09-30
- OpenAI's DOTS hardware device spotted in Geekbench 7 benchmarks — BenBajarin · 2026-09-30