Dyuman Aditya

PhDing @ USC in Robotics | Prev Research @crl_ethz, @ias_tudarmstadt

[Jin Cheng @catachiii]

(Maybe) the first deployable humanoid tracking policy trained via differentiable simulation! Happy to share our recent work on using bundled gradient to help stablizing the gradient information for stiff contact scenarios. Congrats! @dyumanaditya
#robotics #rl #diffsim #humanoid

[Dyuman Aditya @dyumanaditya]

7/7 Huge thanks to my co-authors @catachiii, Clemens Schwarke, @qnguyen_usc, @gauravsukhatme, @StelianCoros, and Gabriele Fadini for making this work possible! 🙌

@crl_ethz @uscresl

[Dyuman Aditya @dyumanaditya]

🤖 Ever wondered if you can train end-to-end humanoid policies entirely within a differentiable simulator?

🥁 Introducing 𝗕𝘂𝗻𝗱𝗹𝗲𝗱 𝗖𝗼𝗻𝘁𝗮𝗰𝘁 𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁𝘀 (𝗕𝗖𝗚)

🌐: https://t.co/h4i7NcBD0U
📄: https://t.co/b7TAjE8Yhh

🧵👇

[Dyuman Aditya @dyumanaditya]

6/7 Why care about differentiable simulation?

First-order learning can be more sample-efficient and faster than PPO.

Making these methods work with realistic contact opens the door to higher-dimensional problems like whole-body loco-manipulation and dexterous manipulation. 🦾

[Dyuman Aditya @dyumanaditya]

@davide_tateo @liu_puze All the best and good luck with the transition Davide!

[Dyuman Aditya @dyumanaditya]

It was a pleasure to present our work on behalf of the @ias_tudarmstadt at #ecmr2025!

It wouldn’t have been possible without the amazing supervisors: @davide_tateo, @goldleaf3, @Jan_R_Peters, @KrzysztofTWalas

And amazing co-authors: @WangNing6, @NicoBohlinger, @piotr_kicki

[Davide Tateo @davide_tateo]

Today, @dyumanaditya presented our paper "Robust Localization, Mapping, and Navigation for Quadruped Robots" at ECMR 2025. It's always a pleasure to see that students and interns of @ias_tudarmstadt succeed and disseminate their ideas and results at conferences!

[Dyuman Aditya @dyumanaditya]

@Jan_R_Peters Thanks Jan! Thanks and congrats to the amazing team of collaborators! It was fantastic working with everyone!

[Davide Tateo @davide_tateo]

Do you want to deploy a navigation stack on a cheap quadruped robot?

Take a look at our latest ECMR paper, "Robust Localization, Mapping, and Navigation for Quadruped Robots".
In our work, we tackle the issues arising from the combination of low-cost robots and RL policies