The majority of self-driving accidents are caused by unnatural behavior of the autonomous car. My research focuses on the human component of self-driving to make autonomous cars behave more like human drivers. My projects in perception determine the social cues of pedestrians, cyclists and other drivers providing a richer representation of the world that is relevant for social behavior. My control projects investigate end-to-end trainable policies that obey social norms, e.g., a robot should not try to pass between two people that are talking to each other and instead go around.
I am a postdoc at the Visual Intelligence for Transportation (VITA) lab at EPFL in Switzerland. Selected publications:
- MonoLoco: Monocular 3D Pedestrian Localization and Uncertainty Estimation.
Presented at ICCV2019 in Seoul. Lead by Lorenzo Bertoni.
- PifPaf: Composite Fields for Human Pose Estimation.
Presented at CVPR2019 in Los Angeles.
- Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning.
Presented at ICRA2019 in Montreal. Lead by Changan Chan and Yuejiang Liu.
I am also investigating new ideas for the Social Force model (presented at STRC2019 and hEART2019) and human path prediction in the context of self-driving cars.
Before returning to academia, I was the Senior Data Scientist at Sidewalk Labs (Alphabet) in New York City where I focused on machine learning for urban environments. My background is in particle physics where I hold a PhD from NYU and where I worked with Kyle Cranmer on physics and statistical analyses for ATLAS at CERN. I was on the core team that discovered the Higgs boson and the first person to see the 5σ threshold crossed.