Harold Soh

Hello! I’m an Assistant Professor at the Department of Computer Science at the National University of Singapore, where I direct the Collaborative Learning and Adaptive Robots (CLeAR) group.

My research interests are in broadly in human-AI/robot interaction, machine learning, and robotics. I’m particularly interested in developing trustworthy collaborative robots. On this research journey, we’re building some fundamental building blocks, including computational trust models, methods for encoding prior knowledge or constraints into expressive (deep) models, and touch/tactile learning for physical interaction .

Latest News

4 August 2020: Congratulations to Tasbolat who has just won a Research Achievement Award!
1 July 2020: Congratulations to Eugene Lim who won the NUS Outstanding Undergraduate Research Award for his work on deep models for collaboration and Juiced!
30 June 2020: Three papers accepted at IROS 2020 on
- Capability, Intent and Trust Calibration in Human Collaboration by Joshua, Jeffrey and Bingcai
- Spiking Graph Neural Network for Event-based Tactile Object Recognition by Fuqiang, Weicong and Tasbolat (PDF Coming soon)
- Fast Tactile-based Texture Classification Using Spiking Neural Network also by Tasbolat in collaboration with Wu Yan at A*STAR! (PDF coming soon!)
6 May 2020: Our work on Visual + Tactile Event-Driven Perception was just accepted to RSS 2020!
28 Feb 2020: Awarded a Faculty Teaching Excellence Award for 2018/19!
24 Feb 2020: Our work on deep generative modeling via characteristic functions was just accepted at CVPR 2020 and selected for oral presentation!
12 Jan 2020: Congratulations to Yaqi who has just won a Research Achievement Award!
11 Jan 2020: Two Late Breaking Reports (one on Situational Awareness and Human-Robot Trust and the other on deep learning for cooperative tasks) accepted to HRI 2020 in Cambridge!

The Collaborative, Learning, and Adaptive Robots (CLeAR) group at NUS investigates the science and engineering of human-AI/robot teams. We conduct human-robot-interaction (HRI) experiments and develop novel machine-learning (ML) methods to enable robots to work effectively with people. We are interested in both cognitive and physical aspects of human-robot interaction. For example, we’ve recently developed new machine-learning (ML) models of human trust and enabled robots to identify objects by touch. As part of our research, we also develop general-purpose ML methods for integrating prior knowledge into deep-learning, yielding deep disentangled generative networks and deep models that leverage symbolic knowledge.

For more information, please check out our recent publications or get in touch.

Selected Recent Publications

Multi-Task Trust Transfer for Human Robot Interaction, Harold Soh, Yaqi Xie, Min Chen, and David Hsu, International Journal of Robotics Research (IJRR), 2019
[ Pre-print PDF | IJRR Link ]

Embedding Symbolic Knowledge into Deep Networks, Yaqi Xie, Ziwei Xu, Kuldeep Meel, Mohan Kankanhalli, and Harold Soh, Neural Information Processing Systems (NeurIPS), 2019
[ Pre-print PDF ]

Hyperprior Induced Unsupervised Disentanglement of Latent Representations, Abdul Fatir Ansari and Harold Soh, Thirty-Third AAAI Conference on Artificial Intelligence (AAAI), 2019
[ PDF | Github ]

Robot Capability and Intention in Trust-based Decisions across Tasks, Xie Yaqi, Indu Prasad, Desmond Ong, David Hsu and Harold Soh, ACM/IEEE Conference on Human Robot Interaction (HRI), 2019
[ PDF ]

Towards Effective Tactile Identification of Textures using a Hybrid Touch Approach, Tasbolat Taunyazov, Hui Fang Koh, Yan Wu, Caixia Cai and Harold Soh, IEEE International Conference on Robotics and Automation (ICRA), 2019
[ PDF | Data@Github ]

Generation Meets Recommendation: Proposing Novel Items for Groups of Users, Vo Vinh Thanh and Harold Soh, ACM Recommender Systems Conference (RecSys), 2018, (Best Long Paper Award Runner-up)
[ PDF | Poster | Slides ]

Latest News

Selected Recent Publications

Share this: