Department of Electrical Engineering
Control Robotics and Machine Learning Lab
Technion - Israel Institute of Technology
המעבדה לבקרה רובוטיקה ולמידה חישובית
Action Robust Reinforcement Learning
Background
Deep reinforcement learning (DRL) methods such as the Deep Q-Network (DQN) have achieved state-of-the-art results in a variety of challenging, high-dimensional domains. This success is mainly attributed to the power of deep neural networks to learn rich domain representations for approximating the value function or policy.
Current methods are unable to generalize to similar, yet previously unseen, problems. As such, we would want the algorithms to be ROBUST - this is the focus of the project.
Project Goal
Implement the Action Robust RL framework within the OpenAI baselines codebase and evaluate the approach across multiple algorithms.
Project steps
-
Understand the various RL algorithms in the OpenAI baselines codebase.
-
Understand the Action Robust RL paper.
-
Implement Action Robustness in the various baselines algorithms.
Required knowledge
-
Strong programming skills.
-
Knowledge in DL and/or RL.
-
Any knowledge in optimization is an advantage.
Environment
-
TensorFlow
Comments and links
-
See the baselines codebase by OpenAI.
-
See the Action Robust paper.
