RAI Institute

New tricks loading...

Reinforcement learning is used to speed the production of behavior for the Atlas humanoid robot. At the heart of the learning process is a physics-based simulator that generates training data for a variety of maneuvers. The control policy tracks and controls retargeted human motion data. Each maneuver is created with data from about 150 million runs of the simulator and transferred zero-shot to the hardware. This work is part of a collaboration between the RAI Institute and Boston Dynamics.

Using reinforcement learning, we trained policies for Boston Dynamics Spot that allow the robot to achieve record running speeds of 11.5 mph (5.2 m/s) — over three times faster than Spot's default max speed.

In this demo, Ultra Mobile Vehicle (UMV) drives, turns, jumps, tricks, and comes to a sudden stop called a track-stand. All of the driving, landings, balance, and track-stands are done using reinforcement learning.

“The key of RL in all of this is to discover new behavior and make this robust and reliable under conditions that are very hard to model. That’s where RL really, really shines.” Marco Hutter and Farbod Farshidian join Evan Ackerman from IEEE Spectrum to discuss some of the latest progress from the Institute. https://lnkd.in/e3c9Rh2t

Announcing our new partnership with Boston Dynamics! The Robotics & AI Institute is teaming up with Boston Dynamics to advance humanoid robotics through reinforcement learning. We're excited to bring our research teams together to produce new, dynamic and generalizable capabilities on Atlas. Learn more — https://lnkd.in/etMgKSQH

The AI Institute is now the Robotics and AI Institute, or RAI Institute! Our name and branding have evolved to better reflect who we are and what we're building. Stay tuned for more exciting things ahead.

Earlier this month, our executive director Marc Raibert participated in the New York Times DealBook Summit. He joined 10 of the leading voices in AI to discuss the future of AI capabilities, safety, investing, and policy. Watch the full panel discussion here — https://lnkd.in/em-qRfaa

Introducing Theia: a vision foundation model for robotics developed by our team at the Institute. Theia enables robots to learn and adapt more quickly and effectively by refining knowledge from multiple off-the-shelf vision models into compact representations for classification, segmentation, depth prediction, and other modalities—all at a lower computation cost. Learn more at https://lnkd.in/enccz9GE.

Our team is presenting work at the Conference on Robot Learning (CoRL) in Munich, Germany this week! Learn more about our accepted research — https://lnkd.in/dwNu2k92 #CoRL #CoRL2024