EGOCENTRIC DATA CAPTURE & MANIPULATION LAB

The world,
from the
point of action.

Practical tools to capture real-world interactions from the task’s point of view, and turn them into data for robot learning.

PROJECT 001 / TRASH COLLECTION

01 / THE LAB

Real interactions.
A closer view.

We explore how to record physical tasks from the point of action: what a tool sees, how it moves, and what happens when it meets an object.

Our work connects low-cost capture hardware with video review and dataset development for perception and robotic manipulation research.

Explore our first project

02 / FIRST PROJECT: TRASH COLLECTION

A grabber. A phone.
A view of every pickup.

We’re developing an inexpensive, 3D-printable phone harness and trigger system for a reacher grabber. An iPhone near the claw is intended to record litter collection while a person uses the grabber and a separate bucket.

01 / SEE

See the moment

A removable phone harness is designed to capture video from the grabber’s point of view, including the approach to the litter.

02 / MARK

Mark the attempt

A mechanical cable is designed to link the squeeze trigger to the phone’s volume button, marking the attempt for later review.

03 / LEARN

Learn from the result

Deposits, misses, drops, and uncertainty can all inform careful human review.

CURRENT STAGE / PROTOTYPE DEVELOPMENT

From field capture
to training data.

The first project aims to connect pickup video and attempt markers with review, segmentation, classification, and imitation-learning research.

The reacher retrofit is currently a CAD fit prototype. Grip, cable motion, and field performance still need physical validation.

Explore the project