Samsung · Filed Nov 17, 2025 · Published Jun 4, 2026 · verified — real USPTO data

Samsung Patents a Layer-by-Layer 3D Escape Route Finder for Robots

When a robot rolls into a crowded hallway, it doesn't just need to know what's blocking the floor — it needs to know what's blocking the space at knee height, waist height, and above. Samsung's new patent tackles exactly that problem.

Samsung Patent: Robot Navigation Through Crowded Spaces — figure from US 2026/0151909 A1
FIG. 1A — rendered from the official USPTO publication PDF.
Publication number US 2026/0151909 A1
Applicant SAMSUNG ELECTRONICS CO., LTD.
Filing date Nov 17, 2025
Publication date Jun 4, 2026
Inventors Youngil KOH, Woojeong KIM
CPC classification 700/245
Grant likelihood Medium
Examiner CENTRAL, DOCKET (Art Unit OPAP)
Status Docketed New Case - Ready for Examination (Dec 5, 2025)
Parent application is a Continuation of PCTKR2025014280 (filed 2025-09-12)
Document 20 claims

How Samsung's robot finds a way through crowded rooms

Imagine a delivery robot rolling through a busy office and hitting a bottleneck: a cluster of people, chairs, bags, and rolling carts all jammed together. A standard obstacle-avoidance system might just stop or try a single detour. Samsung's patented approach goes further.

When the robot detects a dense obstacle zone it can't just push through, it takes a full 3D scan of the area and then mentally slices that scan into horizontal layers — one at floor level, one at mid-height, one at the robot's full height, and so on. Any escape route that's clear on one layer but blocked on another gets thrown out.

What's left is a path that's actually clear all the way from the ground to the top of the robot's body — not just a gap that looks open from one angle. That final route is what the robot drives.

How the plane-slicing escape route system actually works

The system works in two sensing stages. First, while the robot is moving normally, it uses first sensing data (likely a forward-facing depth sensor or LiDAR) to flag when it's approaching an obstacle-dense area it can't navigate through.

Once that alert fires, the robot switches to a deeper scan: it collects 3D point cloud information (a dense map of millions of distance measurements that together form a 3D shape of the environment) using a second pass of its sensors.

The key step is what happens next. The robot divides that 3D point cloud into a plurality of horizontal planes based on its own physical height — essentially slicing the environment into stacked 2D maps at different elevations. For each slice, it calculates candidate escape routes.

The filter: any candidate route that is blocked in even one of those horizontal planes gets eliminated. The surviving routes — the ones clear across all height layers simultaneously — become the final escape routes the robot's driver is instructed to follow. This cross-plane validation is the core novelty: it ensures the robot doesn't try to squeeze through a gap that's open at floor level but blocked by a table overhang halfway up.

What this means for delivery and service robots in real spaces

Service and delivery robots are proliferating in spaces designed for humans — hospitals, hotels, warehouses, and office buildings — and those environments are full of obstacles that don't occupy a clean vertical column. A chair has legs at floor level but a seat and back jutting into space above. A person with a backpack might be passable at knee height but not at shoulder height. Traditional 2D occupancy-grid navigation doesn't capture any of that.

For Samsung, which has been building out its robotics lineup under the Samsung Bot umbrella, this kind of robust navigation in cluttered real-world spaces is a prerequisite for commercial viability. A robot that gets stuck or takes awkward detours every time a hallway gets busy isn't useful — and this patent is a direct attempt to solve that operational friction.

Editorial take

This is solid, practical robotics engineering — not a moonshot idea, but the kind of unglamorous navigation improvement that separates robots that work in the real world from ones that only work in demos. The multi-plane filtering approach is a clean solution to a genuine problem, and the fact that Samsung is patenting it suggests they're serious about deploying robots in messy, human-scale environments rather than just controlled warehouse settings.

Get one Big Tech patent every Sunday

Plain English, intelligent commentary, no hype. Free.

Source. Full patent text and figures from the official USPTO publication PDF.

Editorial commentary on a publicly published patent application. Not legal advice.