Imitation, Reinforcement, and Transfer Learning for Robots in the Wild

Robots in the wild are those that are operating in unseen, complex, challenging environments under new environmental conditions. Most of the time, robot policies and controllers, and even self-sensing (i.e., what is my current status) fail when taken from a controlled environment into the real world.

​ This research focuses on a fundamentally learning skills either on the fly (online reinforcement), or by leveraging past experiences (imitation and transfer) to as-quickly-as-possible respond to new, unseen environments. This includes robots that create a model of their motions as they move, to ones that solve for controllers or policies online as more data is collected in their new environments. This research also involves transferring knowledge between reality and simulation, where reinforcement learning can be used to guide exploration of the robot's capabilities while exploiting what they have already learned about themselves and their environments. ​

This research topic is generally open-ended by nature, as robots that monitor their own behavior, that explore new motions, and that adapt to environment variables has broad applications.

Students & Collaborators

  • Ahmed Qureshi
  • Jacob Johnson
  • Jingpei Lu
  • Zihyun Chiu
  • Nikhil Shinde

Publications

Publications

​

​

Bimanual Regrasping for Suture Needles using Reinforcement Learning for Rapid Motion Planning

Z.Y. Chiu, F. Richter, E.K. Funk, R.K. Orosco, M.C. Yip

IEEE Conference on Robotics and Automation (Accepted). Xi'an, China (2021). [arxiv][video]

​

Robust Keypoint Detection and Pose Estimation of Robot Manipulators with Self-Occlusions via Sim-to-Real Transfer

J. Lu, F. Richter, M. C. Yip

arXiv preprint arXiv:2010.08054 [arxiv] [video]

​

SOLAR-GP: Sparse, Online, Locally Adaptive Regression using Gaussian Processes for Bayesian Robot Model Learning and Control
B. P. Wilcox, M.C. Yip
IEEE Robotics and Automation Letters. vol. 5, no.2, pp. 2832-2839, 2020. [website]

​

Active Continual Learning for Planning and Navigation

A.H. Qureshi, Y.L. Miao, M.C. Yip

ICML 2020 Workshop on Real World Experiment Design and Active Learning. [pdf]

​

Composing Task-Agnostic Policies with Deep Reinforcement Learning

A.H.Qureshi, J.J. Johnson, Y. Qin, T. Henderson, B. Boots, M.C. Yip

International Conference on Learning Representations- ICLR2020. April 26-30, 2020. Addis Ababa, Ethiopia. (Accepted). [arxiv][website]

​

Adversarial Imitation via Variational Inverse Reinforcement Learning

A.H.Qureshi, B. Boots, M.C. Yip

Int. Conference on Learning Representations- ICLR2019. May 6-9, 2019. New Orleans, LA. [pdf,website,vid]

​

Open-Sourced Reinforcement Learning Environments for Surgical Robotics 

F. Richter, R. K. Orosco, M.C. Yip

arXiv preprint arXiv:1903.02090, 2019. [arxiv][vid][git]

Flexible Attention-Based Multi-Policy Fusion for Efficient Deep Reinforcement Learning

Advances in Neural Information Processing Systems (NeurIPS), 36, pp. 13590-13612 (2024).

Forward Kinematics Kernel for Improved Proxy Collision Checking

N Das, MC Yip

Composing Task-Agnostic Policies with Deep Reinforcement Learning

A.H.Qureshi, J.J. Johnson, Y. Qin, T. Henderson, B. Boots, M.C. Yip

SOLAR-GP: Sparse, Online, Locally Adaptive Regression using Gaussian Processes for Bayesian Robot Model Learning and Control

B. P. Wilcox, M.C. Yip

Adversarial Imitation via Variational Inverse Reinforcement Learning

A.H.Qureshi, B. Boots, M.C. Yip

Open-Sourced Reinforcement Learning Environments for Surgical Robotics

F. Richter, R. K. Orosco, M.C. Yip

Adversarial Reward and Policy learning Via Variational Inverse Optimal Control

A. Qureshi, M.C. Yip