TD-MPC Explained, With Alexander Soare (Part 1 of 2)

In this video I explain the problem formulation of TD-MPC and how TD-MPC works at rollout. TD-MPC paper: https://arxiv.org/abs/2203.04955 Many thanks to Nicklas Hansen et. al. for publishing their research and open sourcing their code. Chapters: 0:00 - Intro 0:54 - Notation and problem formulation 6:20 - High level summary of MPC 11:15 - Why are we optimizing for a fixed horizon? 16:03 - Generalizing to a formulation for CEM 17:38 - CEM with a physics thought experiment 23:32 - CEM applied to action trajectories 25:30 - Summary