Self-Autonomous Car Simulation Using Deep Q-Learning Algorithm

Abstract

This paper has proposed a car game using the algorithm Deep Q-Learning for simulation of self-driving autonomous car. It creates an environment where the self-driving car moves in left right up down direction starting from source to destination. Q-Learning is a RL algorithm which is an of f-policy algorithm that is completely different from a Deep Q-Learning as it replaces a normal Q table with neural network. It maps neural network with input states as their actions and Q values. This Game is reward based where agent gets the reward when it reaches the destination and does not hit any obstacles else it gets penalty which is then kept as an experience for future plays.

Topics & Concepts