Image
user avatar
Jason Peng
@xbpeng4
Assistant Prof at @SFU and Research Scientist at @NVIDIA
Vancouver, Canada
xbpeng.github.io
Joined April 2018
48
Following
6,241
Followers
Posts
  • Pinned
    user avatar
    Jason Peng
    @xbpeng4
    Oct 8, 2025
    Implementing motion imitation methods involves lots of nuisances. Not many codebases get all the details right. So, we're excited to release MimicKit! github.com/xbpeng/MimicKit A framework with high quality implementations of our methods: DeepMimic, AMP, ASE, ADD, and more to come!
    Image
    00:00
    169K
  • user avatar
    Jason Peng
    @xbpeng4
    Apr 6, 2021
    Our paper on adversarial motion priors has been accepted into @siggraph 2021! Adversarial imitation learning is finally starting to work for complex simulated characters. Paper + code: xbpeng.github.io/projects/AMP/ Thanks to my collaborators: Ze Ma, @pabbeel @svlevine @akanazawa
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Oct 15, 2018
    We've just released the DeepMimic code + mocap data + pre-trained policies: github.com/xbpeng/DeepMim… Everything you need to train a simulated humanoid to do this:
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Oct 9, 2018
    Video imitation paper accepted into @SIGGRAPHAsia! Blog: bair.berkeley.edu/blog/2018/10/0… Project page: xbpeng.github.io/projects/SFV/i… Big thanks to all my collaborators for making this work possible, @akanazawa, Jitendra Malik, @pabbeel, @svlevine
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    May 24, 2019
    Ever wanted to train a T-Rex to play soccer? Check out our new paper on learning composable motor skills with multiplicative compositional policies (MCP): xbpeng.github.io/projects/MCP/ Big thanks to all my collaborators: @mmmbchang, Grace Zhang, @pabbeel, @svlevine
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Apr 3, 2020
    Our recent work on robots learning to move by imitating animals: xbpeng.github.io/projects/Robot… Huge thanks to all my collaborators at Google for making this project possible!
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    May 4, 2022
    Our paper on learning reusable skill embeddings for simulated characters has been accepted into #SIGGRAPH2022! Paper: nv-tlabs.github.io/ASE/ Thanks to all my collaborators: Kelly Guo, Lina Halper, @svlevine, @FidlerSanja, and all the folks at NVIDIA
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Jul 6, 2019
    The humanoid is practicing some kicks.
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Jul 6, 2022
    I'll be joining SFU as an Assistant Prof and NVIDIA as a Research Scientist! Currently looking for MSc/PhD students. If you are interested in working together on topics in RL, animation, and legged robots, apply through @SFU.
  • user avatar
    Jason Peng
    @xbpeng4
    Oct 2, 2019
    Checkout our new paper on advantage-weighted regression: xbpeng.github.io/projects/AWR an extremely simple off-policy RL algorithm that just uses supervised learning for policy updates. Code is available here: github.com/xbpeng/awr We can use AWR to train this little guy:
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Oct 2, 2018
    New paper on regularizing adversarial learning with a variational discriminator bottleneck. Our method (VAIL) can imitate challenging skills from mocap data, and can even learn to run directly from a video clip: xbpeng.github.io/projects/VDB/i…
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Mar 22, 2022
    We have released a super fast implementation of AMP using @nvidia's Isaac Gym. Models that used to take days to train can now be trained in minutes! github.com/NVIDIA-Omniver… There's also a retargeting script for loading new motion clips. Thanks to @viktor_m81, @gavrielstate.
    Image
    00:00
  • user avatar
    Jason Peng
    @xbpeng4
    Nov 7, 2025
    MimicKit now has support for motion retargeting with GMR. We also released a bunch of parkour motions recorded from a professional athlete, used in ADD and PARC. Anyone brave enough to deploy a double kong on a G1? 😉
    Image
    00:00
    38K
  • user avatar
    Jason Peng
    @xbpeng4
    Apr 29, 2020
    We've released the simulation environments + training code for the robotic imitation project: github.com/google-researc… Everything you need to train your own Laikago:
    Image
    00:00