Rachel Thomasson

SLIM: Symmetric, Low-Inertia Manipulator

Designing a simple but capable hand for reaching and grasping in constrained, cluttered household spaces.

Publication in preparation — more details to come!

Tactile Object Mobility Classification

For robots working in cluttered household settings, planning collision-free paths is not always an option. However, if collisions do happen, the robot should be able to evaluate whether a contact event is safe or unsafe.

We designed a custom, mutual capacitance-based tactile skin for the Allegro Hand which covers the front, back, and side surfaces of the hand. We then developed a classification algorithm which takes in time-series tactile data from incidental contact and predicts an object mobility category (immovable, sliding, or tipping) without requiring a model of the object we are in contact with.

Thomasson, R., Roberge, E., Cutkosky, M. R., & Roberge, J. P. (2022, October). Going in blind: Object motion classification using distributed tactile sensing for safe reaching in clutter. In 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (pp. 1440-1446). IEEE.

Finalist for Best Paper Award at IROS 2022.

Ocean One^K Hands

Ocean One^K is an underwater robotic avatar (developed by Oussama Khatib’s group at Stanford University) which extends our reach beyond depths which humans can safely go — up to 1km below sea level!

As we don’t know exactly what the robot will encounter during its exploration, we developed hands which are highly versatile, highly robust, but require only one motor each to actuate.

Publication in preperation.

Exploratory Hand

Avoiding contact is hard, but contact can be dangerous. We designed a robot gripper with a low-inertia exploratory finger which can quickly and safely make contact. A near direct drive transmission leads to high force transparency, turning the motor driving the hand into a sensor which can be to gain information about the environment!

Lin, M. A*., Thomasson, R.*, Uribe, G., Choi, H., & Cutkosky, M. R. (2021). Exploratory hand: Leveraging safe contact to facilitate manipulation in cluttered spaces. IEEE Robotics and Automation Letters, 6(3), 5159-5166. *indicates equal contribution

Distal Hyperextension
is Handy

We designed a hand for helping to clean up messes in the kitchen. An integrated high range-of-motion wrist joint expands the manipulator’s workspace and enables re-orientation of objects without re-grasping or in-hand manipulation. The distal joints of the fingers hyperextend for powerful pinches. An octopus-inspired particle jamming suction cup palm helps deal with difficult edge-cases, such as picking up a plate placed upside down in a sink.

Ruotolo, W., Thomasson, R., Herrera, J., Gruebele, A., & Cutkosky, M. (2020). Distal hyperextension is handy: High range of motion in cluttered environments. IEEE Robotics and Automation Letters, 5(2), 921-928.

Tactile Terrain Classification

Small legged robots can more efficiently locomote if they adapt their gait pattern based on the terrain. We use a 1D CNN to predict terrain from time-series tactile data, resulting in a +10% improvement from the SVM classifier baseline.

Figure from Wu, X. A., Huh, T. M., Sabin, A., Suresh, S. A., & Cutkosky, M. R. (2019). Tactile sensing and terrain-based gait control for small legged robots. IEEE Transactions on Robotics, 36(1), 15-27.

Safe Reaching in Clutter

We are exploring the use of control barrier functions (CBFs) for generating safe robot trajectories when navigating amidst movable obstacles. We design barriers based on object mobility (e.g. sliding vs immovable), which prior work has shown can be estimated from tactile data.