Emesent GX1 SLAM Takes On the Hardest Test

Six minutes, one pole-mounted scanner, and a stretch of Montana with almost nothing to lock onto.

A 250 meter stretch of road outside Bozeman, Montana is not the kind of site that makes for a dramatic point cloud. Flat, straight, a couple of buildings at either end, and almost nothing in between. Which is exactly why Emesent picked it.

Featureless terrain is the classic failure mode for SLAM. If you have ever watched a mobile scan of a long corridor or an open field bend gently away from reality, you already know the problem. Emesent walked that road with a pole-mounted GX1, out and back, roughly 500 meters of travel, in about six minutes. The resulting dataset ships in two versions, one processed with RTK corrections and one without, which is a more useful thing to publish than a glossy render. It lets you see what the SLAM solution does on its own.

The choice of site was deliberate. “Emesent has always been about mapping the inaccessible, capturing data in challenging and GPS-denied environments,” says Dr. Stefan Hrabar, co-founder and chief strategy officer at Emesent. “Featureless roads and tunnels represent some of the most challenging, and it’s an area where Emesent’s mobile SLAM-based scanners deliver outstanding results.”

The Bozeman dataset provides a cogent entry point into what the GX1 actually is and why the surveying and mapping crowd has been paying attention to it.

Why Featureless Terrain Challenges SLAM Lidar Mapping

Worth a quick refresher on SLAM, or Simultaneous Localization and Mapping, because it explains most of the GX1’s design choices. This is the technology that allows a device to build a map of an unfamiliar environment while simultaneously tracking its own location within that growing map. It is the fundamental bridge between raw lidar returns and a coherent, georeferenced model.

SLAM works by matching what the scanner sees now against what it saw a moment ago, then using the overlap to figure out where the scanner moved. Corners, poles, tree trunks, and building edges all give the algorithm something to lock onto. Take those away and consecutive scans start to look nearly identical, so the position estimate loses confidence and small errors compound. Over a few hundred meters, that shows up as a road that curves when it should not, or an elevation profile that slowly climbs when it should be flat.

Emesent’s answer is a SLAM implementation that keeps re-optimizing the whole trajectory as the scan proceeds rather than just chaining one position estimate to the next. Add RTK positioning from a base station or CORS network, and the SLAM solution has an external anchor to keep it honest. For anyone generating DEMs or contour data from mobile scans, that combination is the part that matters. Drift in the vertical is what turns a nice-looking point cloud into an unusable surface model.

Emesent GX1 Lidar Scanner: At a Glance

The GX1 is Emesent’s follow-on to the Hovermap line, and the short version is that everything now lives in one enclosure. Lidar sensor, SLAM processing, RTK GNSS, four 20 MP cameras for 360 degree panoramic capture, and a removable internal battery. No external receiver, no GoPro strapped to a bracket, no cable harness, no post-hoc script to line imagery up with scan data. It mounts to a standard 5/8 inch survey pole, which means it drops into existing field kits without adapters, and it also runs handheld, on a backpack, or on a vehicle.

While it may lack glamour, this tight little package is precisely the heart of its appeal.

Specifications

All accuracy figures are Emesent’s, reported as 3D RMSE, and measured in the company’s own test environments. Source: Emesent GX1 product guide.

Field Performance: Emesent GX1 Lidar Scans Across Diverse Sites

Emesent’s published datasets are worth browsing because they cover a genuinely wide range of capture conditions.

At Mount Coot-tha Quarry in Brisbane, Australia the unit was vehicle-mounted and the whole site was captured in a little over 20 minutes, georeferenced with RTK against a local CORS network. Quarries are a natural fit for this workflow: repeatable surface capture, volumetric change detection between visits, and terrain that can change significantly on a weekly basis.

In Colorado Springs, the team scanned an intersection on foot with the GX1 on a backpack, walking at a deliberate pace and pausing at traffic lights. About 15 minutes for the intersection of East Cheyenne Mountain Boulevard and Geyser Drive. Road surveys in live traffic are exactly the situation where a one-piece unit beats a rig that needs setup time on a shoulder.

The catalog also includes the Colosseum in Rome, a Denver municipal building, an urban workspace and a bridge and creek crossing in Brisbane, and a lab facility in Atlanta. Several come with fly-through video, which is a faster way to assess coverage and noise than clunking through a data viewer.

Emesent GX1 Field Accuracy: Real-World Testing Results 

Note: All three firms quoted below were beta users or customer references supplied by Emesent, and the datasets described are Emesent’s own published captures.

Sonto, a Queensland survey firm running three Hovermap units, put the GX1 through six or seven residential projects during beta, mostly pole-mounted at 1.8 to 2 meters. Tested against ground control and total station check shots, they recorded 11 mm RMSE, matching their Hovermap STX results and comfortably inside their 20 mm pass/fail threshold for residential work.

Two smaller findings from Sonto are more interesting than the headline number. First, getting the sensor up to around three meters on an extended pole picked up roof pitch data that a lower scan position would have missed, with less shadowing on elevated features. Second, with the cameras sitting roughly 20 cm above the operator’s head, the operator largely disappears from the 360 imagery. Sonto delivers 600 to 700 photospheres per residential scan, and every one of those with a body in the middle of it is a frame someone downstream has to work around.

That pole-mounted workflow is one Emesent hears about often. “Emesent is known for the versatility of our platform, which reduces time in the field by simplifying the complexity of the kit,” says Hrabar. “That said, end-users often comment on a specific usage mode, which helps them the most. Such as using GX1 on a pole for AEC applications being a game changer for both speed and accuracy.”

Sonto reported zero field failures during the trial, saw a 20 percent reduction in training time for new staff, and expected a 30 percent decrease in annual field mistakes, mostly because the team can now avoid the hassle of managing multiple pieces of equipment.

Orion Spatial Solutions, a survey firm with 15 years of scanning experience across construction, marine, and mining work, framed it in terms of what they no longer carry. Director John Philipp described no longer needing to hunt down extra GPS receivers before a job. Orion is building a mixed fleet, keeping Hovermap alongside GX1 and choosing per job.

Gridvision, which inspects power line corridors across Australia, came at it from a different angle. Their team are electrical engineers, not surveyors, and the cabling and configuration on their previous setup took real effort to get right. Managing Director Mark Spinks noted that on runs of 20 to 30 minutes repeated across multiple days, setup time compounds. There is a safety argument too: less equipment fussing means the vehicle operator can concentrate on driving.

Gridvision runs vehicle-mounted corridor scans of 20 to 30 minutes, often several times a day. [Source: Emesent]

Summary: Is the Emesent GX1 Right for Your Survey Workflow?

The GX1 is not proposing a new measurement principle. It is consolidating hardware that surveyors were already assembling themselves, and the gains are perceived mostly in reliability, setup time, and imagery quality rather than in raw accuracy, which held steady against the previous generation.

Emesent’s own framing is more ambitious. “People who have worked with SLAM before are usually impressed by the performance of both Emesent Hovermap and GX1,” says Stefan Hrabar. “They tell us ours is much better than any other SLAM data they’ve seen, crossing the line between SLAM and terrestrial scanning.”

That is a claim readers are encouraged to test themselves. For terrain and DEM work specifically, the two things to watch are the RTK integration and the drift performance in low-feature environments. The Bozeman dataset, with and without corrections, is perhaps the quickest way to form your own opinion on both.

Emesent GX1 Wins Geo Week Pitch the Press Award