Sim2Robot - Training Robots for the Real-World with Imperfect Simulators

Explore the cutting-edge research on training robots for real-world applications using imperfect simulators in this 46-minute talk by Joanne Truong. Delve into the challenges of real-world robot training and discover innovative methods to leverage simulations for scalable, safe, and reproducible robot learning. Learn about large-scale learning techniques in simulation for developing high and low-level robot skills through hierarchical control and abstracted physics. Examine the integration of robot skills with foundation models to enable open vocabulary mobile pick and place operations. Investigate strategies for generalizing learned navigation skills to out-of-distribution environments using additional context information. Gain insights into the intersection of machine learning and robotics, and understand how these advancements contribute to the development of useful intelligent systems like home assistant robots and autonomous delivery vehicles.