Lucas Manuelli

I am a Senior Staff Research Scientist at Boston Dynamics and technical lead of the Large Behavior Models project. Before that I was a research scientist at NVIDIA. Before NVIDIA I did my PhD in Computer Science at MIT with Russ Tedrake in the Robot Locomotion Group within CSAIL. My research focuses on perception and control for robotic manipulation. These days I am specifically working on policies that can control robots in an end-to-end manner from pixels to actions leveraging human demonstrations as one of the sources of supervision.

Prior to starting the PhD I worked on the MIT Darpa Robotics Challenge Team doing planning and controls for the Atlas robot.

Contact / links: Email: lucas dot t dot manuelli at gmail dot com // Google Scholar // GitHub // X

Please reach out if you want to chat about research, possible roles/internships etc. Email is the best way to reach me.

News

Boston Dynamics

I was the technical lead for the Large Behavior Models team and was deeply involved in all areas ofthe project. Specifically I led work on the data pipelines, ML training and deployment infrastructure, model architecture and training loop along with data collection operations. This work produced the first end-to-end pixels to action neural network policy at Boston Dynamics. For more details see the blog post I co-wrote.
I worked on many parts of the perception stack that powers this video. In particular I wrote the pose estimator (based on Megapose6D) that allows Atlas to perform object pose estimation both when the object is on the ground or in a fixture, as well as when it is in hand. I also worked on much of the underlying ML infrastructure that powers our pipelines. For examples of all the different objects that can be manipulated using this ML driven approach see the blog post.
Fun video showcasing the capabilities of the OK-MPC controller. I worked on some of the perception and behavior authoring for this video, specifically the bag throw. [blog post]

Publications

* denotes co-lead authorship
  • Yann Labbé, Lucas Manuelli, Araslan Mousavian, Stephen Tyree, Stan Birchfield, Jonathan Tremblay, Justin Carptentier, Mathieu Aubry, Dieter Fox, Josef Sivic. Megapose: 6d pose estimation of novel objects via render & compare. Conference on Robot Learning, November 2022. [ pdf ] [ project page ]
  • Mohit Shridhar, Lucas Manuelli, Dieter Fox. CLIPort: What and Where Pathways for Robotic Manipulation. Conference on Robot Learning, November 2021. [ pdf ] [ project page ]
  • Lucas Manuelli, Yunzhu Li, Peter R. Florence, and Russ Tedrake. Keypoints into the Future: Self-Supervised Correspondence in Model-Based Reinforcement Learning. Conference on Robot Learning, November 2020. [ pdf ] [ project page ]
  • Peter R. Florence, Lucas Manuelli, and Russ Tedrake. Self-Supervised Correspondence in Visuomotor Policy Learning. IEEE Robotics and Automation Letters , April 2020.
    IEEE Robotics and Automation Letters Best Paper Award 2020 .
    [ pdf ] [ project page ]
  • Lucas Manuelli*, Wei Gao*, Peter R. Florence and Russ Tedrake. kPAM: KeyPoint Affordances for Category Level Manipulation. ISRR 2019[ pdf ] [ project page ]
  • Peter R. Florence*, Lucas Manuelli*, and Russ Tedrake. Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation. Conference on Robot Learning 2018.
    CoRL 2018 Best Paper Award.
    Winner Best Technical Paper, 2018 Amazon Robotics Best Paper Awards in Manipulation
    [ pdf ] [ video ] [ code ]
  • Lucas Manuelli, and Russ Tedrake. Localizing external contact using proprioceptive sensors: The contact particle filter. Masters Thesis , MIT Feb 2018
    [ pdf ]. [ hardware experiments ]
  • Pat Marion*, Peter R. Florence*, Lucas Manuelli*, and Russ Tedrake. LabelFusion: A Pipeline for Generating Ground Truth Labels for Real RGBD Data of Cluttered Scenes. In International Conference on Robotics and Automation (ICRA), Brisbane, Australia, May 2018.
    Best Vision Paper Nominee
    [ website with code, video, paper]
  • Lucas Manuelli and Russ Tedrake. Localizing external contact using proprioceptive sensors: The contact particle filter. In International Conference on Intelligent Robots and Systems (IROS), October 2016.
    [ pdf ][ hardware experiments ] [ simulation video ]
  • Pat Marion, Maurice Fallon, Robin Deits, Andrés Valenzuela, Claudia Pérez D'Arpino, Greg Izatt, Lucas Manuelli, Matt Antone, Hongkai Dai, Twan Koolen, John Carter, Scott Kuindersma, and Russ Tedrake. Director: A user interface designed for robot operation with shared autonomy. Journal of Field Robotics, 1556-4967, 2016.
    [ pdf ]

Other Projects

  • I wrote a walking controller for the Atlas humanoid robot built by Boston Dynamics.
  • I wrote a walking controller for the Valkyrie humanoid robot built by NASA.

Internships

Videos