
Research Scientist @merl_news / PhD @UCLA / @RoMeLaUCLA / Amazon Fellow / Robotics / Motion Planning / RL / Controls / State Estimation
Cambridge, MA
Joined April 2020
- Tweets729
- Following1.7K
- Followers2K
- Likes2.8K
Alexander Schperberg retweeted
Introducing MuJoCo Playground 1.0!
We now ship with full sim-to-real examples, not just RL in sim, bringing MuJoCo ever closer to a robot near you. SO101 and YAM are supported out of the box in sim and real.
See the thread below or playground.mujoco.org for more info
Alexander Schperberg retweeted
The first Humanoid IKEA Assembly Challenge, co-organized with @UnitreeRobotics, @LightwheelAI and @singaporetech, took place at @ieeeiros.
9 global teams tested their loco-manipulation policies to assemble an IKEA UTTER children’s table. Full results ↓
Alexander Schperberg retweeted
🏆三菱電機のフィジカルAIが、ロボット国際会議IROS 2026で優勝!
当社とMERLの合同チーム「MEL-Craft」が「Humanoid IKEA Assembly Challenge」において、ヒューマノイドによる家具の自律組立で最高得点を獲得しました🎉
関連技術はCEATEC 2026でご紹介します✨
#IROS2026 #CEATEC2026 #フィジカルAI #ヒューマノイド
It was an incredibly exciting moment for us at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026 in Pittsburgh! We won the 1st prize 🏆 at the IKEA Humanoid Assembly Challenge. Exciting time to research humanoid robotics at @merl_news!
2
24
1,261
Alexander Schperberg retweeted
One simple fold for robot, one giant flight for robotkind.
Everyone was clapping when the robot finally did it 👏
Come check out Robotic Origami Challenege if u r here at IROS2026 !
Looking for an internship at @merl_news , and work on foundation models for Humanoid robots? Apply now, we are currently looking to fill the position as soon as possible: linkedin.com/jobs/view/44503…
1
6
848
Alexander Schperberg retweeted
#RSS2026 paper PDFs are now available!
Explore the latest work in robot learning, control, perception, planning, HRI, embodied AI, and more.
Start building your reading list before the conference.
Looking forward to seeing many of you in Sydney!
Attending ICRA 2026 to present, “Safe Whole-Body Loco-Manipulation via Combined Model and Learning-based Control” (with Yeping Wang and Stefano Di Cairano) on Tuesday, 15:00–16:30, Paper TuI2I.117.
Paper: arxiv.org/abs/2603.02443
Video: youtube.com/watch?v=3HwqI6DK…
Excited to share our paper “MOBIUS: A Multi-Modal Bipedal Robot that can Walk, Crawl, Climb, and Roll” has been accepted to RSS 2026 in Sydney!
MOBIUS opens many research directions in loco-manipulation, force control, and multi-modal optimization.
youtu.be/RbzRl3m7t4c
#RSS2026 Call for Papers ✨
Workshop on Whole-Body Control & Bimanual Manipulation: Applications in Humanoids & Beyond
Share your latest work by Jun 8 &
Join us with our amazing lineup of speakers from academia & industry!
wcbm-workshop.github.io
@RoboticsSciSys
Happy to share our latest ICRA 2026 work on Safe Whole-Body Loco-Manipulation for Human-Robot Collaboration 😃
paper: arxiv.org/abs/2603.02443
video: youtube.com/watch?v=3HwqI6DK…
Looking forward to presenting it in Vienna, Austria, June 1st - 5th.
Alexander Schperberg retweeted
🚀 AURA (Autonomous Upskilling with Retrieval-Augmented Agents) is accepted to ICRA 2026! 🦾
LLMs + RAG + static schemas to autonomously generate full RL curricula (rewards, randomization, hyperparameters) for embodied robot learning.
Website: aura-research.org/
Our paper was accepted to #ICRA, and I get to visit my home country of Austria! Looking forward to sharing the paper once we make it available next month, it’s on Safe Whole-Body Force Control for Loco-Manipulation ☺️.
Alexander Schperberg retweeted
The table-to-dishwasher task is the classic nightmare scenario for roboticists:
Long-horizon, highly dexterous, precise, whole-body manipulation combined with delicate, transparent, reflective, and deformable objects.
Yet Memo handles it so naturally and elegantly.
Alexander Schperberg retweeted
There can be no useful home robots without maps
Maps will always be important for deploying robots in the world! Very interesting to see @sundayrobotics condition their policies on maps.
One of the cool works I spoke of during my thesis talk was our way of doing this kind of navigation in service of long-horizon manipulation without needing to learn but still inheriting the open world priors of Foundation models.
Excited to see more of this in the space!
Wrapped up my IROS 2025 presentation! Thanks to those that joined, and if you missed it and are interested in legged locomotion, see paper here: arxiv.org/abs/2503.06050
8
528























