Carnegie Mellon and NVIDIA Enable Humanoid Robots to Perform Complex Athletic Moves
- Carnegie Mellon University and NVIDIA developed the ASAP framework for humanoid robots.
- The framework allows robots to replicate complex athletic moves, such as Cristiano Ronaldo’s mid-air spin and Kobe Bryant’s fadeaway jump shot.
- The “delta action model” reduces tracking errors by up to 52.7% compared to previous methods.
- ASAP uses a two-stage process: pre-training in simulations and real-world deployment for data collection.
- Robots can perform forward and side jumps over one meter despite hardware limitations.
The ASAP framework developed by Carnegie Mellon University and NVIDIA enables humanoid robots to perform complex athletic movements, reducing tracking errors by up to 52.7% using the “delta action model.”
Source (2.6)https://decrypt.co/304403/these-robots-can-now-move-like-cristiano-ronaldo-and-lebron-james?rand=52368