Behavior Cloning (Part 3): DAgger

In this video, I introduce the DAgger approach to Behavior Cloning. DAgger provides a targetted way for collecting data for training an autonomous policy. In the video, I give an overview of the DAgger concept, and show how to practically apply it to our real-world autonomous driving problem. It's actually not that straight forward, so you need to be a little creative :-) Towards the end, I run through one iteration of the DAgger process, train some policies, and see how they go at autonomous driving. This is the eleventh video in the overall project.