Manipulation

Moving one object from A to B sounds trivial. It's the unsolved problem most of robotics research is still chipping at. Here are the five steps that hide under the phrase.

~12 min

Why one grasp holds and another slips. The classical theory of stable contacts — form closure, force closure, friction cones — that still defines the vocabulary every modern grasping system speaks.

~15 min

Dex-Net, Contact-GraspNet, and the model families that ate the grasp-planning world. How they work, what they need, and the workflow that gets you a working grasper in a weekend.

~15 min

Beyond the framework's setup — the manipulation patterns that actually ship. Attached collision objects, grasp constraints, dual-arm coordination, and the real-world workflow.

~16 min

A planned path is geometry. A trajectory is geometry plus timing. Time-optimal vs jerk-limited, trapezoidal vs S-curve, and the practical patterns for safe motion near humans.

~13 min

Why peg-in-hole needs compliance — and how to add it to any arm. Compliance directions per task, search strategies, tactile feedback, and the integration with vision that turns a 90% pipeline into 99%.

~14 min

Moving objects without grasping. The math of pushing under friction, the patterns of flipping, and why the latest VLAs are starting to handle the kind of dexterity that used to need a careful human hand.

~13 min

Multi-finger hands, in-hand reorientation, and why this is still mostly a research frontier in 2026. The classical theory, the 2018–24 deep-learning revolution, and the open questions.

~14 min

An arm on wheels. Whole-body planning, the redundancy that buys you reach, the pitfalls of moving while carrying. The architecture behind every household robot in 2026.

~14 min

GelSight, DIGIT, BioTac, e-skins — the sensors that finally give robots hands with feeling. What they measure, what to do with the data, and the workflow that turns 90% grasps into 99%.

~13 min