Edinburgh, UK
Joined October 2021
- Tweets14
- Following139
- Followers63
- Likes49
Pinned Tweet
Excited to share our #NeurIPS2025 paper "Enhancing Tactile-based RL for Robotic Control"!
We show how incorporating self-supervision into tactile RL can lead to simulated superhuman dexterity🤏 By the way, this agent is fully blind.
Find out more: elle-miller.github.io/tactil…
🧵👇
We find that training RL agents with a multi-step forward dynamics objective in the latent space (aka a world model 🌍) leads to more robust representations that encode object information.
We demonstrate simulated superhuman dexterity using only proprioception + 17 contacts!
We release new envs as the Robot Tactile Olympiad "RoTO" benchmark: github.com/elle-miller/roto
Come and chat with me at #NeurIPS! My poster session is Wed 3 Dec, 11 a.m. - 2 p.m. at #2317.
Huge thanks to co-authors @trevormcinroe @dabelcs @oisinmacaodha @sethuvij
Elle Miller retweeted
Hello world! This is the RL & Agents Reading Group
We organise regular meetings to discuss recent papers in Reinforcement Learning (RL), Multi-Agent RL and related areas (open-ended learning, LLM agents, robotics, etc).
Meetings take place online and are open to everyone 😊
Elle Miller retweeted
Looking forward to presenting our study on representation learning for on-policy actor-critic algorithms in Singapore for #ICLR2025 !
Here's a picture to tease 3 key insights in our study...
1⃣ Decoupled actor-critic model architectures outperform their shared counterparts. This is because decoupling their representations allows the actor and critic to specialise in terms of what information gets extracted from the environment.
2⃣ The choice of representation learning task for the actor matters significantly less than increasing data diversity.
3⃣ Even a decoupled critic can still influence data collection to benefit its own learning objective.
This was a collaboration between @trevormcinroe, @pcastr, @dabelcs, @prakash127 , Chris Lucas, @s_albrecht and myself.
Come meet us at Poster#392 this Saturday 10-12:30!
This study was supported by @Mila_Quebec @InfAtEd @EdinburghUni @GoogleDeepMind @EDINrobotics @EPSRC @UKRI_News .
Elle Miller retweeted
📚🧵1/7 It is finally here!! Only one more week until the print release of our textbook “Multi-Agent Reinforcement Learning: Foundations and Modern Approaches” with @mitpress!
What you get, why you should be interested and more, all below in a short 🧵👇
Landed at #ICRA2024 in Japan with the wonderful @EdinRobotML group 🇯🇵
On Thursday I will present my research on unknown object grasping in assistive robotics, check out paper website ellemiller.info/assistive_gr… and talk details here ras.papercept.net/conference… ⭐️
Elle Miller retweeted
Hello all, we are here in Yokohama for #ICRA2024! As per tradition, we are handing over our social media to our ICRA Ambassadors. They are Vishva Bhate, @astroboticist, @breadli428, and @ElleAMiller. Welcome everyone to ICRA and please enjoy the conference!
Elle Miller retweeted
Congratulations to the EDAN team – our very first robot winners. Thanks to all our robot teams for taking part, we can’t wait to see you at our next event!
@anybotics @ETH_en @maxonglobal @bfh_hesb @DLR_en #Robotics #Innovation #AssistiveTechnology #cybathlon
Elle Miller retweeted
Very proud of our pilot Mattias and the whole EDAN Team from @DLR_de to get 1st Place in the Assistance Robot Race of the 2023 Cybathlon Challenge! Looking forward to the big event in 2024! @cybathlon
youtu.be/EoER_5vYZsU
Elle Miller retweeted
Introducing EVA, our new hybrid #robotic platform. EVA is an efficient and versatile robotic assistant. Using a mobile base for locomotion, we overcome the challenges of biped motion and achieve a safer and more robust platform.
Great work @ElleAMiller!
youtube.com/watch?v=nMkcBbof…
Elle Miller retweeted
If planetesimals form in rings in ppdisks, why exoKuiper belts are wide, while rings are narrow? In arxiv.org/abs/2110.04007 @ElleMiller0 shows that wide exoKuiper belts could be formed by migrating dust traps. If caused by planets undergoing type-II migration, then alpha>4e-4