Below is the list of papers that we covered in Spring 2019:
Course instructions:
Below is the list of papers that we will be covering in this course, along with some related background reading and some videos that can be interesting to show in your presentations. Note - you do not need to cover all of the material in the background reading; you should use it only if it will assist you in explaining the concepts contained in the assigned paper.
You can sign up for paper presentations here.
Outline:
1/15, 1/17
1/22
Robotics paper: Experiments in Balance with a 3D One-Legged Hopping Machine
Videos:
Deep RL paper: Continuous Control With Deep Reinforcement Learning
Video:
Background:
Deterministic Policy Gradient Algorithms
1/24
Video:
Background:
Efficient Bipedal Robots Based on Passive-Dynamic Walkers
Lecture by Russ Tedrake: Simple models of walking/running
Deep RL paper: Emergence of Locomotion Behaviours in Rich Environments
Background: Proximal Policy Optimization Algorithms
1/29
Background:
Learning to Walk in 20 Minutes
Actuating a Simple 3D Passive Dynamic Walker
Videos:
https://sites.google.com/view/soft-actor-critic
https://sites.google.com/view/minitaur-locomotion
https://sites.google.com/view/sac-and-applications
Background / Related Work:
Learning to Walk via Deep Reinforcement Learning
Soft Actor-Critic Algorithms and Applications
2/5
Background: Hierarchical Apprenticeship Learning, with Application to Quadruped Locomotion
2/7
Robotics paper: Policy Search via the Signed Derivative
Deep RL paper: Deep Reinforcement Learning from Human Preferences
2/12
Robotics paper: Control of a Biped Somersault in 3D
Deep RL paper: SFV: Reinforcement Learning of Physical Skills from Videos
2/14
Robotics paper: Coordinated Multi-Robot Exploration
Deep RL paper: Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments
2/19
Background: Universal robotic gripper based on the jamming of granular material
Deep RL paper: Learning 6-DoF Grasping and Pick-Place Using Attention Focus
2/21
2/26
Robotics paper: The Highly Adaptive SDM Hand: Design and Performance Evaluation
Followup work: A compliant, underactuated hand for robust manipulation
2/28
3/5
Background:
Previous conference paper: Precision Grasping and Manipulation of Small Objects from Flat Surfaces using Underactuated Fingers
3/7
Robotics paper: Shape Completion Enabled Robotic Grasping
Background: Graspit! a versatile simulator for robotic grasping
3/19
Deep RL paper: Learning Dexterous In-Hand Manipulation
https://blog.openai.com/learning-dexterity/
3/21
Deep RL paper: Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Control (unpublished)
Blog post: https://bair.berkeley.edu/blog/2018/11/30/visual-rl/
3/26
Robotics paper: Extrinsic dexterity: In-hand manipulation with external forces
Deep RL paper: Time-Contrastive Networks: Self-Supervised Learning from Video
3/28
Robotics paper: Prehensile Pushing: In-hand Manipulation with Push- Primitives
Deep RL paper: Hindsight Experience Replay
4/2
Robotics paper: Learning Robot In-Hand Manipulation with Tactile Features
Deep RL paper: Zero-Shot Visual Imitation
Background:
Combining Self-Supervised Learning and Imitation for Vision based Rope Manipulation
Learning to Poke by Poking: Experiential Learning of Intuitive Physics
4/4
Deep RL paper: Learning an Embedding Space for Transferable Robot Skills
4/9
Robotics paper: Open Loop Stable Control Strategies for Robot Juggling
Videos
Overview of blind and non-blind robot juggling by Chris Atkeson
3 Ball Blind Robot Paddle Juggling
Feedback Control - Swinging Blind Juggler - IDSC ETH Zurich
The Circular Cloverleaf Blind Juggler - IDSC ETH Zurich
The Blind Juggler - IDSC - ETH Zurich
Shannon (robot) juggling: 3 and 5 balls - blind juggling
Other related work;
Bouncing an Unconstrained Ball in Three Dimensions with a Blind Juggling Robot
Design and Analysis of a Blind Juggling Robot
Rhythmic Feedback Control of a Blind Planar Juggler
4/16
Robotics paper: Hierarchical Task and Motion Planning in the Now
Deep RL paper: Data-Efficient Hierarchical Reinforcement Learning
4/18
Robotics paper:
Prior work:
Learning vehicular dynamics with application to modeling helicopters.
Apprenticeship learning via inverse reinforcement learning.
Exploration and apprenticeship learning in reinforcement learning.
Followup work:
Autonomous Helicopter Aerobatics through Apprenticeship Learning
Related work:
4/23
Robotics paper: Experiments in Fixed-Wing UAV Perching
Video
Perching with a Simple Fixed-Wing Glider
Followup work:
Robust Post-Stall Perching with a Simple Fixed-Wing Glider using LQR-Trees
Related work:
Landing, perching and taking off from vertical surfaces
Deep RL paper: CAD2RL: Real Single-Image Flight Without a Single Real Image
4/25
4/30, 5/2