Thank you to my co-authors @maulik__bhatt , Aayushi Shrivastava, Lasse Peters, and @negar_mehr !
Code: https://t.co/EI311DD9hs
Models: https://t.co/iVmxmSfMz7
Data: https://t.co/VAZlcS0KPx
How can we 𝘢𝘭𝘵𝘦𝘳 😉 pretrained robot policies to coordinate without forgetting how to work alone?
Introducing 𝗔𝗟𝗧𝗘𝗥: 𝗔daptation from 𝗟imited demonstrations for 𝗧eam coordination with 𝗘xisting-skill 𝗥etention.
Website: https://t.co/36NGkopgWl
On two XArm7 manipulators, ALTER achieves 90% coordination success while retaining 98% success (59/60 trials) on the original single-agent tasks.
The same adapted policy can coordinate with a partner and still work alone.
How can agents learn to coordinate without communicating when faced with multiple choices?
In our new work, MIMIC-D, we use decentralized diffusion models to achieve multi-modal, multi-agent coordination!
Paper: https://t.co/Mil1tiJRxu
Website: https://t.co/xLi9OtpKKa
The results:
✅ Our decentralized policies utilize only local observations while still achieving implicit coordination. ✅ We demonstrate MIMIC-D’s capabilities across simulation benchmarks and real robot hardware.
Congrats to @ethan_dong_ for getting an honorable mention in the CRA/URA award! Also congrats to all @YaleEngineering students who also received runner-up and honorable mentions!
https://t.co/vSJMaE9wLw
Grabbing a set of keys, walking over ice – humans learn such tasks fairly easily. Prof. Ian Abraham wants robots to be just as adaptable. It’s a project that won him an NSF CAREER Award – congrats! https://t.co/FgcPTGowFP #YaleEngineering
@YaleSEAS @Yale@ethan_dong Thank you! I am honored and excited to receive this award with Professor Abraham and Henry Berger! Huge thanks to @YaleSEAS and @ianabraha as well of course! FYI my account is @ethan_dong_
Congrats to roboticist Ian Abraham and his research team which includes @Yale undergraduates @ethan_dong and Henry Berger. Amazing work with much more to come!
Thank you to the RSS Conference and its organizers as well as @ianabraha for his mentorship and guidance throughout this whole process! I had an amazing time at the conference and learned so much! Check out this work at https://t.co/uk3NG5dHpb #RSS2023
Honored to have received the best paper award at #RSS2023@RoboticsSciSys with @ethan_dong_ and Henry Berger for our work on Time Optimal Ergodic search https://t.co/IZeUqSfUy9 !!! Stay tuned for more exciting research from my group in the near future at @YaleSEAS !!