Self Driving Car using Deep Q-Learning Networks

Car's target is to do round-trip from left-top corner to right-bottom corners

Demo - Level 1 (no obstacles)

Car took almost 60 seconds of training to understand source and target destination.

Demo - Level 2 (training with obstacles)

Car took almost 180 seconds of training to avoid obstacles in journey between source & destination.

Installation

Step 1: conda env create --name dqn -f=environments.yml

Step 2: source activate dqn

Step 3: python3 map.py

Step 4: Click load button to use already trained model.

Rewards Graph - Explore vs Exploitation