Developing AI attacks/defenses

• Main Author: Pang, Malcolm Qing Han
• Other Authors: Jun Zhao
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/162849
• Institution: Nanyang Technological University

Reinforcement Learning has numerous applications in the real world thanks to its ability to achieve high performance in a range of environments with little manual oversight. Reinforcement Learning can interact with multiple agents in a shared environment called Multi-Agent Reinforcement Learning. Multi-Agent Reinforcement Learning allows the interaction between agents. However, Multi-Agent Reinforcement Learning becomes problematic in asynchronous environment. Hence, in our work, we considered an environment with users with their user devices (UDs), downloading information data from the base station and uploading information data to the base station asynchronously via wireless communications. We designed an environment with multiple base station, where user devices (UDs) would be able to download and upload information data asynchronously using 2 agents. Our goal for both agent is to allocate system resources to minimize the total time taken for users to download information data from the base stations and to optimize power output for users uploading information data We utilize a deep reinforcement learning approach and evaluate the performance of the algorithms under a certain configuration.