Human Intelligence Assisted Robot Learning

Hello! You've reached the info page of Human Intelligence Assisted Robot Learning (HIARL), the PhD thesis to be finished by Zi Wang. In this page, you will find information relevant to HIARL, including proposals, papers, slides, videos etc. If you have any question, you can reach Zi by email (ziw 'at' mit.edu). Notice that this page is under constant update. Zi's thesis is planned to be finished around Feb, 2020.

Thesis proposal

thesis_proposal_online.pdf

Slides

talk_June2019.pdf

Poster for RSS Pioneers Workshop 2019

poster_v3.pdf


Demonstrations of PR2 completing long-horizon tasks using our Learning for Task and Motion Planning framework

Learning models of skills in the real world

Related publications

  1. Learning sparse relational transition models (Victoria Xia, Zi Wang, Kelsey Allen, Tom Silver and Leslie Pack Kaelbling), In International Conference on Learning Representations (ICLR), 2019. [bibtex][pdf]
  2. Beomjoon Kim, Zi Wang, Leslie Pack Kaelbling, Tomas Lozano-Perez. Learning to guide task and motion planning using score-space representation, International Journal of Robotics Research (IJRR), 2019. [doi] [arXiv]
  3. Regret bounds for meta Bayesian optimization with an unknown Gaussian process prior (Zi Wang, Beomjoon Kim and Leslie Pack Kaelbling), In Neural Information Processing Systems (NeurIPS), 2018. [bibtex] [pdf]
  4. Active model learning and diverse action sampling for task and motion planning (Zi Wang, Caelan Reed Garrett, Leslie Pack Kaelbling and Tomas Lozano-Perez), In International Conference on Intelligent Robots and Systems (IROS), 2018. [bibtex] [pdf]
  5. Sampling-based methods for factored task and motion planning (Caelan Reed Garrett, Tomas Lozano-Perez, Leslie Pack Kaelbling), In The International Journal of Robotics Research, 2018. [bibtex][pdf] [doi]
  6. Focused Model-Learning and Planning for Non-Gaussian Continuous State-Action Systems(Zi Wang, Stefanie Jegelka, Leslie Pack Kaelbling, Tomas Lozano-Perez), In IEEE Conference on Robotics and Automation (ICRA), 2017. [bibtex] [pdf]
  7. Sample-Based Methods for Factored Task and Motion Planning (Caelan Reed Garrett, Tomas Lozano-Perez and Leslie Pack Kaelbling), In Robotics: Science and Systems (RSS), 2017. [bibtex] [pdf]
  8. Max-value Entropy Search for Efficient Bayesian Optimization (Zi Wang and Stefanie Jegelka), In International Conference on Machine Learning (ICML), 2017. [bibtex] [pdf]

About the author

Zi Wang is a PhDc at MIT CSAIL, working on Bayesian optimization, robot learning and Human Intelligence Assisted Artificial Intelligence. She received M.S. from MIT in 2016 and B.Eng. from Tsinghua University in 2014. Zi is a recipient of MIT Graduate Women of Excellence Award, Rising Star in EECS and Google Anita Borg Scholarship. While at MIT, she served as co-president of Graduate Women in Course 6 (EECS), co-organizer of the first ML Across MIT Retreat, and volunteer lecturer at the US-China Cultural and Education Foundation. During her PhD, Zi is advised by Prof. Leslie Pack Kaelbling and Prof. Tomas Lozano-Perez, and she has mentored at least 8 undergrad/MEng students.

Acknowledgment

The project on Learning for Task and Motion Planning (an important part of my thesis; code to be released soon) has received tremendous help from a number of people from the LIS group. In particular, Caelan Garrett is the key contributor to our code base on the PR2 robot and made embedding learning possible by building the infrastructure where we are able to embed learning. The codebase has gone through several generations of evolution, with contributions from, in no particular order, Alex LaGrassa, Skye Thompson, Kevin Chen, Nishad Gothoskar, Ivan Jutamulia, Jingxi Xu, Jiayuan Mao, Leslie Kaelbling and Tomás Lozano-Pérez.

Thesis committee: Leslie Kaelbling (supervisor), Tomás Lozano-Pérez (supervisor) and Marc Toussaint (reader).