Michael Piseno

CV ยท Github

About Me

I am a Master's student in Computer Science at Stanford University advised by Karen Liu. I work on robotics and motion prediction in the The Movement Lab (TML). I am currently applying to PhD programs and am interested in joining labs relevant to my research goals discussed below. My research is supported by the National Defense Science and Engineering Graduate (NDSEG) Fellowship.

Prior to joining TML, I was a research assistant in the Intelligent and Interactive Autonomous Systems Group (ILIAD) at Stanford advised by Dorsa Sadigh, where I worked on language-conditioned imitation learning for robotics. I also spent time at Robotics @ Google (now DeepMind) working on biped locomotion under Jie Tan and Atil Iscen in collaboration with the Hybrid Robotics Lab at UC Berkeley under Koushil Sreenath.

Medium-Term Research Goals: I believe humanoids are the best morphology for studying interactive intelligence at scale. I believe this because human motion data is much more scalable than that of other robotic morphologies (e.g. through video-based pose estimation or leveraging gaming inputs for semi-supervision), affording researchers much more data to tackle The Bitter Lesson. Furthermore, motion data combined with other multi-modal data sources grounds semantic knowledge in the physical world, which Large Language Model (LLMs) planners struggle with on their own. Thus, I am interested in combining human motion data with other data sources at scale to create a Large Motion Model (LMM), and studying its emergent effects on general intelligence.


Papers

  1. Learning Stylized Humanoid Locomotion with Adversarial Motion Priors
    Michael Piseno, Zhongyu Li, Alejandro Escontrela, Xue Bin Peng, Atil Iscen. Full Author List Coming Soon...
    ArXiv Coming Soon...