Introduction to Deep RL from a Robotics Perspective
Examine the application of Deep Reinforcement Learning (Deep RL) in robotics for motor skill learning, analyzing its fundamental factors and techniques.
Modules/Weeks
Weekly Effort
Discipline
School
Format
Cost
Course Description
- Acquire a solid understanding of fundamental concepts in deep reinforcement learning, starting from the basics even with no prior exposure.
- Gain the skills to apply deep RL techniques to robotics, specifically in the context of motor skill learning.
- Develop the ability to navigate through deep RL concepts, which although beneficial, are not a prerequisite for comprehending the course material.
- Build a comprehensive foundational knowledge of deep RL algorithms and their practical application within the field of robotics for motor skill learning.
What You Will Learn
By the end of this course, learners will be able to:
Demonstrate a strong understanding of essential concepts in reinforcement learning, including Markov decision processes, Q-learning, policy iteration, Non-deterministic MDPs, Deep Neural Networks, Deep Q-Networks, Policy Gradient, and Actor-Critic Methods.
Apply Deep RL in Robotics: Gain practical insights into applying Deep Reinforcement Learning techniques within the field of Robotics, specifically for enhancing motor skill learning.
Understand and utilize the concept of domain randomization to bridge the gap between simulated training and real-world applications of deep RL in robotics.
Explore emerging research directions and potential areas of advancement in the application of deep RL to real robots.
Course Outline
Module 1: Reinforcement learning basics
Module 2: Q-learning, your first RL algorithm
Module 3: Non-deterministic MDPs
Module 4: Policy Iteration
Module 5: Deep neural networks
Module 6: Deep Q-networks
Module 7: Policy gradient and actor-critic methods
Module 8: Domain randomization
Module 9: Current tools and future paths
Instructors
Matei Ciocarlie is a robotics researcher with a focus on building robots that can operate in unforeseen situations through physical interaction with their environment. His interests lie in building robotic manipulators that can handle clutter, wearable robotic rehabilitation devices for patients, and teleoperated robots that can simplify tasks by using their own sense of the surrounding environment. Ciocarlie’s work encompasses mechanism and sensor design, control, planning, and learning for robots aiming to perform complex tasks in the real world, spanning both hardware and software. His research focuses on robotic manipulation, including grasp and manipulation theory, tactile sensor development, robotic hand designs, and wearable hand orthoses for stroke patients. Matei has received multiple awards, including the Early Career Award by the IEEE Robotics and Automation Society, a Young Investigator Award by the Office of Naval Research, a CAREER Award by the National Science Foundation, and a Sloan Research Fellowship by the Alfred P. Sloan Foundation.
Please note that there are no instructors or course assistants actively monitoring this course.