Trevor Ablett

Ph.D. Student
Department: ,

Personal Wesbsite Google Scholar LinkedIn Twitter

As we get closer to applying robotics to unstructured environments filled with humans, such as homes, roads, and hospitals, our traditional sense-plan-act approach to robotics has started to hit barriers. Reinforcement and imitation learning offer an appealing option for circumventing these limitations, in the same way that machine learning has completely revolutionized our capabilities in computer vision and language processing. Unfortunately, these techniques continue to be challenging to apply directly to robots due to their high data requirements and inherent lack of focus on safety and robustness.

 

Trevor Ablett is a PhD candidate in the STARS Laboratory. During his studies, he completed a year-long internship at Samsung AI Center in Montreal as an Applied Reinforcement Learning Researcher. He previously completed both a B. Eng in Mechatronics Engineering and a B. A. in Psychology at McMaster University in Hamilton, Ontario, Canada.


 
 

Force-Matched Imitation Learning with a See-Through Visuotactile Sensor


We learn to complete contact-rich cabinet opening and closing tasks with a two-mode, see-through sensor.

Multimodal and Force-Matched Imitation Learning with a See-Through Visuotactile Sensor
Trevor Ablett, Oliver Limoyo, Adam Sigal, Affan Jilani, Jonathan Kelly, Kaleem Siddiqi, Francois Hogan, Gregory Dudek
Submitted to Transactions on Robotics (T-RO): Special Section on Tactile Robotics

Learning from Guided Play


We resolve a deceptive reward problem in inverse reinforcement learning through the use of auxiliary tasks.

Learning from Guided Play: Improving Exploration for Adversarial Imitation Learning with Simple Auxiliary Tasks
Trevor Ablett, Bryan Chan, Jonathan Kelly
Robotics and Automation Letters (RA-L) and IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS’23)

Multiview Manipulation from Demonstrations


We show how imitation learning with teleoperated demonstrations can be applied to a mobile manipulator through initial pose randomization.

Seeing All the Angles: Learning Multiview Manipulation Policies for Contact-Rich Tasks from Demonstrations
Trevor Ablett, Daniel (Yifan) Zhai, Jonathan Kelly
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS’21)

Failure Identification in Intervention-based Learning


We show how identifying failures in intervention-based learning can improve learning efficiency for manipulation tasks.

Fighting Failures with FIRE: Failure Identification to Reduce Expert Burden in Intervention-Based Learning
Trevor Ablett, Daniel (Yifan) Zhai, Jonathan Kelly
Technical Report on arXiv (2020)

Self-Calibration of Mobile Manipulator Kinematic and Sensor Extrinsic Parameters Through Contact-Based Interaction


Our contact-based self-calibration procedure exclusively uses its immediate environment and on-board sensors.

Self-Calibration of Mobile Manipulator Kinematic and Sensor Extrinsic Parameters Through Contact-Based Interaction
Oliver Limoyo, Trevor Ablett, Filip Marić, Luke Volpatti and Jonathan Kelly
International Conference on Robotics and Automation (ICRA’18).