Título: Vision-based Perception and Control Methods for Autonomous Aerial Robots
Ponente: Eric Cristofalo
Lugar: Seminario del Departamento
Día: Viernes 8 de febrero a las 10:30
Abstract:
Oftentimes in robotics, we treat the problems of perception and control independently. Active perception is a method that couples the two components by computing a control input that actively improves the quality of the perception. In this talk, I will present an active vision-based feedback controller for a quadrotor with a monocular camera that is attempting to reconstruct the unknown scene structure. The controller maneuvers the quadrotor in a fashion that minimizes the future uncertainty of its 3D reconstruction. We apply this method in simulation and on real quadrotors with downward-facing cameras. I will also present recent experimental results for autonomous drone racing with vision-equipped quadrotors that uses an iterative best response game-theoretic planner. The competing quadrotors estimate the position of their opponents by coupling on-board images and the predicted opponent trajectories generated by the game planner.
Bio:
Eric Cristofalo is a Ph.D. candidate and National Defense Science and Engineering Graduate (NDSEG) fellow in the Department of Aeronautics and Astronautics at Stanford University in the Multi-robot Systems Lab. He received his B.S. degree in Mechanical Engineering from Drexel University in 2013 and his M.S. in Mechanical Engineering from Boston University in 2016. His research interests include vision-based control, active perception, and 3D reconstruction with multi-robot systems.
