https://taosun.io/tags/robotic-assembly/Robotic AssemblySource Themes Academic (https://sourcethemes.com/academic/)en-us2026 © Tao Sun. All rights reserved.Wed, 21 Oct 2026 00:00:00 +0000https://taosun.io/img/icon-192.pnghttps://taosun.io/tags/robotic-assembly/https://taosun.io/code/assemblenist/Wed, 21 Oct 2026 00:00:00 +0000https://taosun.io/code/assemblenist/<p>AssembleNIST is an independent, open-source implementation of <a href="https://alpha-nist.github.io/">AlphaNIST</a> in <a href="https://mujoco.org/">MuJoCo</a>. A single reinforcement-learning policy, trained with PPO from binary success rewards, controls a Franka Panda arm and completes all 20 assembly tasks on the NIST Assembly Task Board 1. Code, pretrained policies, and instructions are in the <a href="https://github.com/suniique/assembly-nist">GitHub repository</a>. This project is not affiliated with or endorsed by the AlphaNIST authors.</p> <h2 id="results">Results</h2> <p>With 8,192 parallel environments, the policy reaches 98.4% success after 10K PPO iterations and 98.7% after 12.5K. AlphaNIST reports 86.9% after 25K iterations, but it uses a different simulator, asset set, and implementation, so the numbers are not directly comparable.</p> <p><img src="https://taosun.io/code/assemblenist/learning_curve.webp" alt="Success rate over training iterations. It rises from about 25% to above 98% by 5,500 iterations. AlphaNIST’s reported results are marked at 25K iterations."></p> <p><img src="https://taosun.io/code/assemblenist/board_start.webp" alt="The 20 parts at random initial poses on the board."> <img src="https://taosun.io/code/assemblenist/board_done.webp" alt="The board after the policy assembles every part."></p> <p>Left: the 20 parts at random initial poses. Right: the board after assembly.</p> <h2 id="acknowledgements">Acknowledgements</h2> <p>This project builds on <a href="https://alpha-nist.github.io/">AlphaNIST</a> and <a href="https://sgs-rl.github.io/">Success Guided Sampling</a>. Thanks to <a href="https://zoctipus.github.io">Octi Zhang</a> for guidance on training details and <a href="https://jingz6676.github.io/">Jing Zhang</a> for helpful discussions.</p>