Simulating the Concept of Self-Driving Cars Using Deep-Q Learning

Abstract

Prior to the deployment of a critical software or functional hardware equipment, it is always suitable to test the working of the equipment in an environment that tests the capabilities of the functionality to the fullest. The equipment can be deployed in the real-world only on successful clearance granted to the test scenarios. Thus, an efficient foolproof simulated environment for any type of hardware is customary in a product development cycle. Thus, through this research, we aim at creating a suitable environment for a self-driving car which consists of simple test cases the agent will have to pass before it can be deployed in the real world. The outcome developed as the end-product of this research will help serve companies that are involved in the development of technologies for the purpose of building a full-fledged vehicle with autonomous capabilities. This product will not only help them to run the algorithms designed for technology but also help in designing suitable test case scenarios. The simulating agent can be visualized as a game-playing agent where an environment similar to that of a typical game comprises of the environment where the agent overcomes certain obstacles and challenges to complete a particular task.