Reinforcement Learning on Atari Games

Project Title

Reinforcement Learning on Atari Games

Project Area of Specialization Artificial IntelligenceProject Summary

Exploration of deep reinforcement learning on open AI Gym and universe platforms. Along with some help of image processing, pixcels and scores as well as state variable are taken as input and deep Q-networking exploring algorithms are applied for findings results that human level play/results. Our first basic aim is to explore then reproduction of that are similar to a research paper calls "Human level control through deep reinforcement learning" published by "Volodymyr Mnih & vice versa".

Project Objectives

We set out to create an Agent that would be able to develop a wide range of competencies on a varied range of challenging tasks—a central goal of general artificial intelligence that has eluded previous efforts. To achieve this, we will develop a novel agent, a deep Q-network (DQN), which will be able to combine reinforcement learning with a class of artificial neural network known as deep neural networks. The agent will approach the best human player in context to game scores using deep Q-Network.

Project Implementation Method

Project will be implemented on Open AI Gym, universe environment using Python programming language.

Benefits of the Project

We will create an agent that can use artificial intelligence, Machine Learning in such a way that it can learn and improve upon Atari Games. Agent will be using Open AI Gym/Universe environment. Our agent will be able to learn by playing a specific Atari game again and again. Every time the success rate to learn a specific Atari game will increase as more outcomes will be found out. Once the agent has perfected itself on an Atari game, it will be capable of learning and training on other Atari games using same methodology (i.e. The Q Learning algorithms). In the end result our agent will be able to play a specific Atari game at full perfection. “The Agent can be alter used in larger applications like self-driving cars and Autonomous robots”.  
 

Technical Details of Final Deliverable

We will Train a model on environment such as OpenAI Gym/Universe which will provide us with a good and efficient performance.

Final Deliverable of the Project Software SystemType of Industry IT Technologies Artificial Intelligence(AI)Sustainable Development Goals Quality EducationRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60000
A Graphical Processing Unit (GPU) with at-least 8GB of V-RAM	Equipment	1	60000	60000