Apple researchers have developed MoMo, a two-stage imitation-learning framework designed to enable robots to adapt their manipulation behaviors to different contexts. The system uses a spatiotemporal action tokenizer and a transformer that takes task and a continuous motion-mode condition as input. Experiments on six real-world robot manipulation tasks demonstrated that varying this condition could produce distinct steady, dynamic, and intermediate behaviors, and MoMo successfully transferred unseen motion modes while maintaining task success. AI
IMPACT Enhances robot adaptability and generalization in manipulation tasks, potentially improving efficiency in robotics research and application.
RANK_REASON The cluster contains a research paper detailing a new framework for robot manipulation. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Apple Machine Learning Research →
- Apple Inc.
- Arto Kivila
- Benoît Landry
- Efficient ConvBN Blocks
- Hugues Thomas
- Humanoid Policy
- International Conference on Learning Representations
- Momo
- Mouli Sivapurapu
- Peide Huang
- Yuhan Hu
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
