Does MultiSet replace my robot's SLAM stack?

No. SLAM keeps real-time tracking, obstacle avoidance and local path planning. MultiSet runs alongside it as the plant's reference, supplying absolute 6-DoF poses that reset accumulated drift and keep every robot aligned to one facility-wide coordinate system.

How do robots localize indoors without GPS?

Satellite signals don't survive a roof, and the partial fixes a receiver still reports are confidently wrong by tens of meters. SLAM handles frame-to-frame motion, but its frame starts wherever the robot started and drifts. A visual positioning system adds the absolute fix, in a shared map that already exists, so drift is corrected against the map.

Does this sit in my control loop or my safety case?

No. MultiSet works out how two maps line up, at commissioning, and re-checks it when the building changes. It does not run at control-loop rates, does not need a live network, and does not belong in a safety case. If it goes offline, every robot keeps running as it does today.

What scan data can I use to build the map?

Any of it. E57, LiDAR point clouds, Matterport exports, 3D Gaussian Splats, 360 video, CCTV footage and CAD or BIM all convert into localization maps. The capture you already own becomes the map, and you keep it.

Can MultiSet run on-premises or on-device?

Yes. Public cloud, private cloud, self-hosted and on-device. In self-hosted and on-device modes, no maps, images or spatial queries leave your infrastructure.

Does MultiSet work with ROS 2?

Yes, ROS 2 is a supported SDK target. The absolute pose arrives in the transform tree as a standard frame, alongside odometry and SLAM. Query the VPS on a timer or when confidence drops, and gate on the confidence value each response carries.

Spatial Ground Truth for
Every Robot, Device, and Human

Speak with our Robotics Team

Join the Developer Discord

Centimeter-Scale 6-DoF Localization for Robot Fleets

One unified map for the facility

One platform for spatial understanding.
Live 3D maps. Centimeter-scale 6-DoF localization.
The layer fleet coordination runs on, any vendor mix.
Scales from one zone to the whole site.

A construction site where a worker in a hard hat and headset, an overhead drone, a quadruped robot, a humanoid robot and a wheeled rover are all tracked in one shared coordinate frame

Why positioning is the bottleneck

A robot that cannot say where it is cannot act. Not roughly which room, but the exact position and orientation, in a frame something else can also read.

Cold Start, Drift and Map Fragmentation

Every system ships with its own map. That's the problem.

A mobile robot in a dark industrial aisle projecting a camera light cone down the corridor as it wakes and recovers its position

Two mobile robots tracing separate winding paths through wireframe maps of the same space, one purple and one blue, showing two coordinate frames drifting apart

A technician holding a headset visor between two humanoid robots, each holding a different floor plan of the same building

Cold start, every start.

SLAM starts from zero. Every robot explores before it works. With MultiSet, one camera frame on wake returns an exact 6-DoF pose. No wandering. No map-building delay.

Drift is inevitable. Misalignment isn't.

Sensor noise and wheel slip accumulate. Each robot's map drifts from reality, and from every other robot's. Absolute localization resets drift and keeps fleets aligned to ground truth.

Headsets and agents are not on the map.

No robot vendor ships the headset your technician wears, or the model that reads a label and decides whether the batch proceeds. Put people and agents on the same map as the fleet and the aisle can be shared instead of fenced.

Human-Robot Coordination on One Coordinate System

One shared map, and the handoffs stop being improvised.

A technician in smart glasses watching an industrial robot arm, with the assembly station's running status and task checklist overlaid in place in his field of view

Human-Robot Collaboration

Workers see robot position, intent and stop zones in place. Robots read waypoints people set.

A worker with a tablet on a factory floor while an autonomous mobile robot follows a marked path around a robot cell

Mixed-Fleet Interoperability

Fleet standards cover orders and status, not maps. MultiSet is the map underneath, so every vendor's robots agree on one position.

A tablet on a workbench showing a 3D map of the facility, with a mixed reality headset, an AMR carrying a tote and a humanoid robot working in the same space

Brownfield Commissioning

Most of the deployment cost is working out where everything is. Survey once. Every AMR, drone, humanoid and headset onboards against it.

Object Tracking for Docking and Inspection

Object Tracking recognizes a specific machine, pallet or assembly from its scan model and localizes against the asset instead of the room. For docking, inspection and manipulation.

Ingest

Build the map from capture you already own: E57, LiDAR, 3D Gaussian Splats, 360 video, CAD/BIM.

Localize

Every robot, drone, headset and phone gets its position and orientation from the same map.

Stay true

Periodic relocalization corrects drift; map versioning tracks facility changes as a living asset.

An autonomous mobile robot scanning warehouse racking, the capture rendered as a purple point cloud across the pallets beside it

An autonomous mobile robot in a warehouse aisle localizing against a ceiling-mounted camera reference point

An autonomous mobile robot following a marked lane down a warehouse aisle lined with racking

Scan-Agnostic Robot Localization, Built as Infrastructure

Scan-agnostic

Build localization maps from existing visual data. No proprietary scanner, no re-capture.

Cold-start localization

Absolute 6-DoF pose from a single camera frame on wake. No exploration run.

Two primitives, one SDK

VPS for the environment, Object Tracking for the equipment. Same map, same SDK.

Embed it under your platform

Robotics OEMs and platforms ship MultiSet as their localization layer. Your platform, your customers, our infrastructure.

Your map, your infrastructure

Public cloud, private cloud, self-hosted, or fully on-device. Maps, images or spatial queries never leave your infrastructure.

Fleet-wide drift correction

Periodic absolute relocalization bounds SLAM drift, keeping fleets aligned.

Every platform

ROS 2, REST APIs, Unity, iOS, Android, WebXR, AndroidXR, HMDs and wearables. Any camera.

On-Prem, Private Cloud, and On-Device Robot Localization

Deploys where your compliance team says it can live.

Public cloud, private cloud, self-hosted, fully on-device.

Every tier runs the same API on the same map. An air-gapped floor is not a cut-down version of the cloud product, it is the same product.

Diagram showing private cloud, on-prem and on-device deployment options connecting to one MultiSet core, which serves warehouse, factory, construction, healthcare and tunnel environments

Map-Owner and Scan-Agnostic Models Compared

	Map-owner model	Scan-agnostic model
Where the map comes from	The vendor's own coverage	Your existing scan, in any format
Works on a private site	Only where coverage already exists	Anywhere you have captured
Who owns the capture	The vendor	You
Capture hardware	Tied to the vendor's pipeline	Any scanner you already own
Deployment	Cloud, tied to the map	Cloud, private VPC, self-hosted, on-device