Excited to share THREE amazing projects from my group: T-Rex, a SOTA model/dataset for tactile-reactive dexterous manipulation; CAIP, a new plug in visual encoder that respects visual affordances, not just object semantics; and RATs, curiousity-driven agentic skill learning! ...
No Digg Deeper questions have been answered for this story yet.
3 of 3: Kids can learn how to generalize via play (vs rote repetition of goal tasks) to learn skills that are useful for the future; we think agentic robotics should do so as well. We revisit curiosity-based intrinsic learning for agentic robotics:
Children learn from play. Can robots do the same?
We propose 𝐏𝐥𝐚𝐲𝐟𝐮𝐥 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐑𝐨𝐛𝐨𝐭 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠, a paradigm that gives embodied coding agents a play stage before downstream tasks arrive, and instantiate it with 𝐑𝐀𝐓𝐬 (Robotics Agent Teams), where robots discover reusable skills through curious play.
Co-led with @jiaxin_ge_