Abstract:
This report outlines my work in continuation to last semester on understanding human behavior when it comes to driving on Indian streets. Since traffic in India is not governed by lane-markings, zebra crossings and proper curbs autonomous cars will have to learn directly from human drivers. Here I explore a Learning from Demonstration technique namely Inverse Reinforcement Learning to learn the implicit rewards that a human driver maybe sub-consciously optimizing to decide his actions (policy) to successfully, safely and smoothly drive on our chaotic and unorganized roads.