UGVs on Construction Sites: Progress Monitoring, Scanning and Site Safety

TL;DR: A construction site robot for monitoring is a small unmanned ground vehicle (UGV) that autonomously or semi-autonomously traverses build sites to capture photos, LiDAR scans, thermal data, and safety observations. On construction sites, UGVs can support three concrete workflows: repeatable progress documentation, dense 3D scanning without tying up a surveyor for every capture, and off-hours safety checks. The hard parts are dust ingress, uneven terrain, GNSS denial inside structures, and site access rules. This article explains how to deploy a UGV like the Leo Rover in these three workflows, what sensor payloads make sense, and how to process the data after each mission.

Why construction sites need a UGV for monitoring in the first place

Construction sites generate enormous documentation demand. General contractors need weekly progress reports for owners. BIM coordinators need up-to-date reality captures to compare against the model. Safety officers need evidence that housekeeping, fall protection, and fire lanes meet applicable standards. Most of this work is still done by people walking the site with a phone, a 360 camera, or a tripod-mounted scanner.

A construction site robot for monitoring can change the economics of that data collection. Instead of a superintendent spending two hours a day on documentation walks, a UGV can run the same route on a schedule and return consistently captured, timestamped data. The robot does not replace the superintendent’s judgment. It can reduce the repetitive capture step so humans can focus on decisions.

Fictionlab offers the Leo Rover as an open ROS 2 platform for field robotics work: a 4×4 chassis that can be extended with application-specific sensors, compute, and autonomy software. That extensibility matters on construction sites, where no two workflows use the same sensor combination.

What conditions a construction UGV actually faces

Before discussing applications, it helps to be precise about the environment. A construction site robot for monitoring has to handle:

• Concrete dust and silica particulates that can clog fans, bearings, and optical sensors. Environmental sealing, filtration, and a cleaning and maintenance plan are important; required IP ratings depend on the actual exposure.
• Uneven terrain: rebar, cable trays, expansion joints, mud, gravel, and step obstacles that may exceed the robot’s capability depending on floor state. Ground clearance and wheel torque matter more than top speed.
• GNSS denial once you move inside a structure or between concrete cores. You may need visual-inertial odometry, LiDAR SLAM, surveyed control points, or fiducial markers to localize.
• Restricted access windows. Robots often operate during off-hours (nights, weekends) or in cordoned zones. Battery runtime and remote supervision drive schedule feasibility.
• Site safety regulations. Depending on jurisdiction and the site risk assessment, robots may require exclusion zones, high-visibility markings, defined operating procedures, and human supervision.

The Leo Rover’s ground clearance, ingress protection, and payload capacity depend on the version and configuration. Verify current manufacturer specifications, including the protection rating of the complete system and installed sensor enclosures, against the conditions on your specific site. A small wheeled UGV may be suitable for mapped slab-deck or finished-floor environments, but its step-climbing capability should be evaluated before deployment.

How UGVs handle construction progress monitoring

Progress monitoring is a common construction site robot application. The workflow is simple: the robot follows a repeatable route, captures 360-degree imagery and depth data at defined waypoints, and uploads the results to a reality-capture platform (such as OpenSpace or Reconstruct) or a custom pipeline.

A typical progress-monitoring payload on a Leo Rover includes:

• A 360 camera (such as an Insta360 or Ricoh Theta) mounted on a mast approximately 1.5-2 m above the ground to approximate a walking superintendent’s perspective.
• An RGB-D or stereo camera for local obstacle detection and avoidance.
• An onboard computer running ROS 2 for waypoint navigation and time-synchronized capture triggers.

The robot navigates using a prebuilt map of the floor, updated as the structure changes. In ROS 2, Nav2 can handle path planning, while AMCL or SLAM Toolbox can support localization and mapping when used with appropriate sensors and maps. When the environment has changed enough that localization becomes unreliable, you may need to remap, use surveyed reference targets, or use fiducial markers such as ArUco tags as localization aids.

After each run, the raw imagery can be uploaded and associated with floor plans, site coordinates, or a BIM model. The output may be a floor-plan view where users can compare current photos with earlier captures and, where the platform supports it, with the model. That is often what BIM coordinators and owners want to see.

How UGVs perform 3D scanning without tying up a surveyor

Terrestrial laser scanning has traditionally required a survey-grade tripod scanner and an operator to reposition it repeatedly. Survey-grade static scanners often cost tens of thousands of dollars, although prices vary substantially by model and capability. A UGV with a mobile LiDAR payload can change that workflow: the robot drives a route and produces a registered point cloud through SLAM.

Common mobile scanning payloads include Livox MID-360, Ouster OS0, or Velodyne VLP-16 combined with an IMU. On the software side, LIO-SAM, FAST-LIO2, or Point-LIO can produce real-time odometry and a point cloud. Post-processing tools such as Open3D and CloudCompare can help with registration, downsampling, quality control, and conversion or export to formats such as E57 or LAS, depending on the selected toolchain.

Accuracy is the honest question here. Mobile LiDAR SLAM can provide centimeter-scale relative accuracy over short, well-constrained areas, but absolute accuracy and accumulated drift depend heavily on the sensor, trajectory, loop closures, control points, and site geometry. Unconstrained runs can accumulate errors of several centimeters or more. This can be sufficient for progress documentation, indicative scan-to-BIM comparison, and some volumetric workflows when validated against control. It is not sufficient for structural setout or precision fabrication references, where survey control and survey-grade instruments remain necessary.

The value of the UGV workflow is scan frequency. Weekly full-floor scans may become feasible because the marginal cost per scan can fall to the labor needed to launch the robot, maintain it, and process and validate the cloud. That frequency can turn 3D scanning from a milestone activity into a continuous data stream that can be compared week over week.

How UGVs contribute to site safety inspection

Safety inspection is the third practical application and can be easier to justify when it supplements, rather than replaces, competent human inspection. A UGV running an off-hours patrol can check for:

• Fire lane and egress path obstructions using RGB imagery and object detection, with human review of reported exceptions.
• Heat anomalies near electrical equipment or hot-work areas using a thermal camera (such as a FLIR Boson or Lepton), subject to appropriate calibration and interpretation.
• Localized air-quality indicators using onboard CO, VOC, or particulate sensors. These sensors require suitable placement, calibration, and validation for the intended use.
• Perimeter security through scheduled patrols with recorded video.

A standard thermal camera does not reliably detect gas leaks in general. Detecting specific gases requires a sensor designed for that gas, such as an appropriate gas detector or optical gas-imaging system.

The ROS 2 stack for this application can pair Nav2 for navigation with a detection model running on an onboard GPU, such as a Jetson Orin Nano or Orin NX. Detections can be logged with timestamps, map coordinates, and reference images. The safety officer can then review an exception report rather than reviewing hours of footage.

Fictionlab’s work on inspection applications may be relevant when selecting sensor integration patterns and ROS 2 components for construction safety use cases.

What to do with the data after each mission

Collecting data is the easy part. Making it useful is where most deployments succeed or fail. A construction site robot for monitoring should feed into a defined pipeline:

1. Ingest: The robot uploads rosbag2 files or structured data (images, point clouds, detections) to a local NAS or cloud bucket at the end of each run. Wi-Fi coverage on active sites is inconsistent, so budget for offline buffering and manual sync.
2. Align: Point clouds are registered to site coordinates or project control. Images are tagged with map locations. This can be automated when localization is reliable, but may require manual quality control or correction when it drifts.
3. Compare: The current capture is compared against the BIM model for scan-to-BIM workflows or against the previous run for progress and safety review.
4. Report: Exceptions, approved deltas, and any progress metrics supported by the workflow can be pushed to the tools stakeholders already use: Procore, Autodesk Construction Cloud, or a project dashboard.
5. Archive: Raw data is retained according to project contracts, owner requirements, and applicable legal or regulatory retention rules.

The pipeline design decisions (which reality-capture platform, which BIM tool, which alerting channel) usually matter more to project success than the robot hardware choice.

What integration with existing site processes looks like

A UGV is a piece of site equipment. It needs a risk assessment, a designated responsible operator or supervisor, and a place in the site logistics plan. Practical questions to answer before the first deployment include where the robot charges (often the site office trailer), who launches missions and reviews outputs, how the robot is transported between floors (elevator or hoist), and how it interacts with workers during shift overlaps.

Most successful deployments start with a single floor or a single trade coordination scope, prove the data pipeline works, and then scale. The Leo Rover’s open architecture can support that incremental approach because you can start with a basic imaging payload and add LiDAR or thermal sensors as new use cases emerge.

For a deeper look at how Fictionlab supports construction deployments, see the construction applications page.

Frequently asked questions

Can a UGV replace a total station or survey-grade scanner on a construction site?

No. Mobile UGV LiDAR can provide useful relative measurements and point clouds for progress tracking and some scan-to-BIM workflows, but its absolute accuracy depends on control, calibration, trajectory, and SLAM quality. Structural setout, control point establishment, and precision fabrication references still require appropriate survey control and survey-grade instruments.

How does a construction site robot handle GNSS-denied environments inside structures?

It can use visual-inertial odometry, LiDAR SLAM, surveyed reference targets, or fiducial markers such as ArUco tags placed at known coordinates. SLAM Toolbox supports 2D LiDAR mapping and localization, FAST-LIO2 can provide LiDAR-inertial odometry, and Nav2 uses the resulting localization estimate for navigation. Accuracy depends on sensor quality, site geometry, control points, and loop-closure opportunities.

What IP rating does a construction UGV need for dust and moisture?

There is no single required IP rating. The appropriate rating depends on whether the robot is used indoors, exposed to wind-driven dust, operated in rain, or cleaned with water. IP ratings apply to specific enclosures, so assess the chassis, compute enclosure, connectors, cameras, and LiDAR separately. Verify the rating of the complete configured system rather than relying on the rating of an individual component.

Can the robot operate autonomously during active work hours?

Technically yes, but whether it may do so depends on the site risk assessment, applicable regulations, worker traffic, and the robot’s safety functions. Many sites limit autonomous operation to off-hours or cordoned exclusion zones. Semi-autonomous operation with human supervision is common during active hours.

What is the typical battery runtime for a construction monitoring mission?

Battery runtime depends on battery configuration, sensor payload, drive profile, terrain, compute load, and ambient temperature. A progress-monitoring route may take 30-60 minutes on some floors, but route duration and usable runtime should be measured on the actual configured robot before operational scheduling.

How does data from the UGV integrate with BIM tools like Revit or Navisworks?

Point clouds are commonly processed in Autodesk ReCap or similar software and converted to formats such as RCP/RCS for use in Revit or Navisworks. E57 and LAS are widely used interchange formats, but direct import support varies by application and version. Reality-capture platforms such as OpenSpace and Reconstruct can associate imagery with drawings, locations, and models; available integrations should be confirmed for the selected platform.

What ROS 2 distribution should you use for a construction UGV project?

Use a ROS 2 distribution supported by the robot vendor, the operating system image, and the required sensor drivers and navigation packages. ROS 2 Humble is an LTS release supported upstream through May 2027. For multi-year projects, plan upgrade testing before the selected distribution reaches end of life.

Start planning your construction deployment

If you are evaluating a UGV for progress monitoring, 3D scanning, or safety inspection on your projects, review Fictionlab’s construction applications page for reference configurations, payload examples, and integration guidance for the Leo Rover platform.