Visit our aamas2022 paper "Decoupled Reinforcement Learning to Stabilise Intrinsically-Motivated Exploration" with Lukas Schäfer Filippos Christianos Josiah Hanna. Talk & Q&As on 11 May 18:00-19:00 (1A5-3) and 13 May 9:00-10:00 (3C1-2); all BST. ➡️ Paper: arxiv.org/abs/2107.08966

Low-cost teleop systems have democratized robot data collection, but they lack any force feedback, making it challenging to teleoperate contact-rich tasks. Many robot arms provide force information — a critical yet underutilized modality in robot learning. We introduce: 1. 🦾A

Check out our latest work, "Actor-Critic Model Predictive Control: Differentiable Optimization meets Reinforcement Learning for Agile Flight," published in the IEEE Transactions on Robotics, where we reconcile #OptimalControl and #ReinforcementLearning, achieving the same

I still remember this vividly. We got it working with RL. #mujoco #sim2real

arxiv.org/pdf/2601.17895 Masked Depth Modeling for Spatial Perception Get high-quality metric depth images with reconstruction model, amazing!

Video coming soon

Guess the controller

This is simple walking gait, much remain to be optimised. #quadruped #rumi #sim2real

Can we bridge the Sim-to-Real gap in complex manipulation without explicit system ID? 🤖 Presenting Contact-Aware Neural Dynamics — a diffusion-based framework that grounds simulation with real-world touch. Implicit Alignment: No tedious parameter tuning. Tactile-Driven:

#sim2real checkout our gravity compensation on kinova arm

New project! Flow Policy Gradients for Robot Control tldr; a simple online RL recipe for training and fine-tuning flow policies for robots co-led w/ Hongsuk Benjamin Choi: hongsukchoi.github.io/fpo-control

Real progress looks slow and boring #sim2real

🚀 Introducing CHIP: Adaptive Compliance for Humanoid Control through Hindsight Perturbation! Current humanoids face a trade-off: they are either Agile & Stiff OR Slow & Soft. CHIP breaks this barrier. We enable on-the-fly switching between Compliant (wiping 🧼,

imitate and then distill into a goal-reaching policy then finetune. It's kinda becoming a standard way to repurpose skills.

I've been saying for years that the biggest challenge for simulation in robotics is not actually the physics engine (although you do have to get that right). The real challenge is capturing the *diversity* of the real world. There was no doubt that generative AI had the potential

Introducing MolmoSpaces, a large-scale, fully open platform + benchmark for embodied AI research. 🤖 230k+ indoor scenes, 130k+ object models, & 42M annotated robotic grasps—all in one ecosystem.

Finally we achived locomotion in sim, main culprit was stiffness. If your policy is not working try lowering your stiffness just enough to support weight. Soon going to transfer to real robot #sim2real