CLeAR

Robot Learning

Below, you’ll find our work on robot learning methods, with a focus on learning from humans, e.g., for imitation learning, and inverse reinforcement learning.


★ CLeAR Group Members


The Dynamics of Q-learning in Population Games: a Physics-Inspired Continuity Equation Model

C22-1The Dynamics of Q-learning in Population Games: a Physics-Inspired Continuity Equation Model
Shuyue Hu★, Chin-Wing Leung, Ho-fung Leung, and Harold Soh★
International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2022 (Accepted)
TL;DR: We develop an accurate physics-inspired model for describing how a population of Q-learning agents adapt as they interact.

    More Links: [BibTeX]
    @inproceedings{hu2022,
    author    = {Shuyue Hu and Chin-Wing Leung and Ho-fung Leung and Harold Soh},
    title     = {The Dynamics of Q-learning in Population Games: a Physics-Inspired Continuity Equation Model},
    year      = {2022},
    booktitle = {International Conference on Autonomous Agents and Multiagent Systems (AAMAS)}}

Embedding Symbolic Temporal Knowledge into Deep Sequential Models

C21-2Embedding Symbolic Temporal Knowledge into Deep Sequential Models [PDF] [Blog]
Yaqi Xie ★, Fan Zhou ★, and Harold Soh ★
IEEE International Conference on Robotics and Automation (ICRA), 2021
TL;DR: We show how to improve flexible deep models using prior knowledge expressed as linear temporal logic.

    More Links: [Video] [BibTeX]
    @inproceedings{xie20templogic,
    title={Embedding Symbolic Temporal Knowledge into Deep Sequential Models},
    author={Yaqi Xie and Fan Zhou and Harold Soh},
    year={2021},
    booktitle={IEEE International Conference on Robotics and Automation (ICRA)}}

Efficient Exploration of Reward Functions in Inverse Reinforcement Learning via Bayesian Optimization

C20-7Efficient Exploration of Reward Functions in Inverse Reinforcement Learning via Bayesian Optimization [PDF] [Blog]
Sreejith Balakrishnan ★, Quoc Phong Nguyen, Bryan Kian Hsiang Low, and Harold Soh ★
Neural Information Processing Systems (NeurIPS), 2020
TL;DR: We propose new IRL methods that efficiently learns the *space* of plausible reward functions.

    More Links: [BibTeX]
    @misc{balakrishnan20,
    title={Efficient Exploration of Reward Functions in Inverse Reinforcement Learning via Bayesian Optimization},
    author={Sreejith Balakrishnan and Quoc Phong Nguyen and Bryan Kian Hsiang Low and Harold Soh},
    year={2020},
    booktitle={Advances in Neural Information Processing Systems}}

Embedding Symbolic Knowledge into Deep Networks

C19-6Embedding Symbolic Knowledge into Deep Networks [PDF]
Yaqi Xie★, Ziwei Xu, Mohan Kankanhalli, Kuldeep S. Meel, and Harold Soh ★
Neural Information Processing Systems (NeurIPS), 2019
TL;DR: We show certain forms of logic are more amendable to embedding and can be used to improve flexible deep models.

    More Links: [BibTeX]
    @inproceedings{yaqi2019embedding,
    title={Embedding Symbolic Knowledge into Deep Networks},
    author={Yaqi, Xie and Xu, Ziwei and Meel, Kuldeep S and Kankanhalli, Mohan and Soh, Harold},
    booktitle={Advances in Neural Information Processing Systems},
    pages={4235--4245},
    year={2019}}

Learning Assistance by Demonstration: Smart Mobility with Shared Control and Paired Haptic Controllers

P17-1J15-1Learning Assistance by Demonstration: Smart Mobility with Shared Control and Paired Haptic Controllers [PDF]
Harold Soh and Yiannis Demiris
Journal of Human-Robot Interaction, vol 4, No 3, 2015.

    More Links: [Publisher Link] [BibTeX]
    @article{soh2015learning,
    title={Learning assistance by demonstration: Smart mobility with shared control and paired haptic controllers},
    author={Soh, Harold and Demiris, Yiannis},
    journal={Journal of Human-Robot Interaction},
    volume={4},
    number={3},
    pages={76--100},
    year={2015},
    publisher={Journal of Human-Robot Interaction Steering Committee}}

Spatio-Temporal Learning with the Online Finite and Infinite Echo-state Gaussian Processes

J14-2Spatio-Temporal Learning with the Online Finite and Infinite Echo-state Gaussian Processes [PDF]
Harold Soh and Yiannis Demiris
IEEE Transactions on Neural Networks and Learning Systems, 2014.

    More Links: [Publisher Link] [BibTeX]
    @article{DBLP:journals/tnn/SohD15,
    author    = {Harold Soh and Yiannis Demiris},
    title     = {Spatio-Temporal Learning With the Online Finite and Infinite Echo-State Gaussian Processes},
    journal   = {{IEEE} Trans. Neural Networks Learn. Syst.},
    volume    = {26},
    number    = {3},
    pages     = {522--536},
    year      = {2015},
    url       = {https://doi.org/10.1109/TNNLS.2014.2316291},
    doi       = {10.1109/TNNLS.2014.2316291},
    timestamp = {Mon, 09 Mar 2020 15:52:04 +0100},
    biburl    = {https://dblp.org/rec/journals/tnn/SohD15.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}}