Verified, then logged.
Sun 22 Nov 2026 · 00:17 KST / 15:17 UTC한국어 | ENSearch /

Breaking01 / AI

DOCK solves 15 robot object placement tasks in a simulation

The researchers trained DOCK using estimates of how far objects held by a robot gripper moved, based on the robot’s joint angles.

Three thermal views accompany a three-dimensional reconstruction of an empty industrial room, showing temperature differences between pipes and walls. AI-generated image

Researchers reported on arXiv on 19 Nov 2026 that their DOCK method solved 15 object placement tasks in simulation.

The paper’s abstract lists 21 evaluation tasks. The researchers derived the movement estimates used for training from 9,951 trajectories of objects carried in simulated kitchens.

The researchers used the robot’s joint angles to estimate how far an object held by its gripper had moved. They trained DOCK so that changes in the object’s estimated position in images matched that movement.

According to the abstract, a baseline with direct access to the object’s true position also solved 15 tasks. A method based on the shape of the object’s region in the image solved 10, and a method based on distances between image features solved 9.

The researchers did not disclose in the abstract how many placements succeeded on a physical robot.