Events

CILVR Seminar: Feedback Systems of Search and Deep Learning for Robot Autonomy

Speaker: Ben Riviere

Location: 60 Fifth Avenue, Room 7th floor open space
Videoconference link: https://nyu.zoom.us/j/94926687156

Date: Wednesday, November 5, 2025

From self-driving cars to space exploration, autonomous robots are poised to transform industry and science. A promising strategy for autonomous decision-making is to combine search-based predictive reasoning and data-driven learning in a feedback system, which has empirically demonstrated superior performance and interpretability, e.g., the Go-playing AlphaZero algorithm. However, applying this combination to robotics is challenging because robots navigate a continuous, unstructured and high-dimensional physical world that is difficult to traverse efficiently with search trees in real-time. In this talk, I present two main results towards feedback systems of search and learning for autonomy in complex physical systems: First, to enable tree search on continuous systems, we use the spectral decomposition of the local controllability Grammian to construct trees that capture rich dynamical information like contact modes and actuator degradation. Second, to mathematically understand the robot’s real-time thinking process, we present our optimality convergence result of Monte Carlo Tree Search. Our solutions enable new capabilities in complex hardware experiments spanning active space debris removal, fault diagnosis isolation and recovery, swarm vs swarm drone games, wind-aware navigation, planning for self-driving cars (simulated) and dexterous manipulation. This work builds towards a sophisticated model of robot autonomy that blends intelligence modalities and real-time and offline computation.Bio: Benjamin (Ben) Riviere is an Assistant Professor at New York University (NYU) in the Department of Mechanical and Aerospace Engineering at the Tandon School of Engineering and the Department of Computer Science at the Courant Institute of Mathematical Sciences. Ben was previously a postdoctoral researcher with Joel Burdick and Yisong Yue at the California Institute of Technology (Caltech). He received his M.S. and Ph.D. at Caltech in Aeronautics advised by Soon-Jo Chung and his B.S. in Mechanical Engineering from Stanford University.