Construction Site Monitoring with UGVs: Use Cases, Challenges, and Early Results

TL;DR: Construction robot monitoring with unmanned ground vehicles (UGVs) automates progress tracking, BIM-to-as-built comparison, and safety inspection on active sites. Mobile robots like the Leo Rover carry LiDAR, RGB-D, and 360° cameras to capture repeatable route data. What works today: scheduled scan-to-BIM capture, thermal and PPE checks, and ROS 2-based autonomy on graded surfaces. What still challenges deployments: rough terrain, dust ingress, GNSS-denied localization inside structures, and integrating point clouds into BIM workflows without manual cleanup.

Why Construction Sites Need UGV-Based Monitoring

Construction sites are dynamic, hazardous, and data-hungry. Project managers need frequent updates on progress, safety compliance, and deviations from the Building Information Model (BIM). Manual walkthroughs are slow, inconsistent, and pull skilled staff away from higher-value work. Drones solve part of the problem outdoors, but they struggle inside structures, under decks, and in cluttered indoor spaces where safety constraints and payload endurance matter.

This is where construction robot monitoring with UGVs becomes practical. A ground robot can run the same route every day, capture identical viewpoints, and feed structured data into reality-capture pipelines. Unlike a handheld scanner operated by a technician, a UGV removes the human variability and frees staff for interpretation rather than acquisition. For an overview of Fictionlab’s work in this domain, see the construction applications page.

What UGV Construction Monitoring Actually Captures

A monitoring UGV is essentially a mobile sensor platform. The robot’s job is repeatable data acquisition; the value lies in the pipeline that processes that data. Typical payloads on a platform like the Leo Rover include:

• LiDAR (e.g., Livox Mid-360, Ouster OS0) for dense point clouds used in scan-to-BIM workflows.
• RGB and 360° cameras (e.g., Insta360, Ricoh Theta) for photogrammetry and visual progress logs.
• Depth cameras (Intel RealSense, Stereolabs ZED) for short-range obstacle data and SLAM.
• Thermal cameras for insulation checks, electrical hotspot detection, and concrete curing monitoring.
• RTK GNSS for outdoor centimeter-level positioning under good sky conditions, paired with IMU and wheel odometry for sensor fusion.

ROS 2 ties this together. Nodes for SLAM (LIO-SAM ports, FAST-LIO2), navigation (Nav2), and data logging (rosbag2) run on an onboard computer, while a base-station laptop handles teleoperation, mission planning, and offload.

How UGVs Support BIM Capture and Progress Tracking

BIM-to-as-built comparison is one of the clearest use cases. The workflow typically looks like this:

1. Define a route along key reference points using waypoints or a teach-and-repeat method.
2. Run the UGV on a daily or weekly schedule, capturing LiDAR scans and panoramic images at each stop.
3. Register point clouds into a common coordinate frame using RTK outdoors or fiducial markers indoors.
4. Push processed scans into reality-capture and point-cloud platforms (Autodesk ReCap, Leica Cyclone, OpenSpace, or open-source alternatives like CloudCompare).
5. Run automated deviation analysis against the federated BIM model.

Early results from academic and industry pilots show that scheduled UGV capture can detect geometric deviations in the 1-3 cm range when using survey-grade LiDAR with proper registration. The bottleneck is rarely the hardware; it is the time spent cleaning point clouds and matching them to evolving BIM revisions. Tools like NavVis IVION and Trimble Connect are helping to close that gap, and ROS 2 packages such as point_cloud_transport and nav2_collision_monitor make parts of the data and safety pipeline more reproducible.

Progress Tracking Without Full Autonomy

You do not need full autonomy to extract value. Many deployments use teleoperation or supervised autonomy: an operator walks the robot through a route once, then the platform replays it. This lowers the entry barrier and avoids the safety-case overhead of unattended operation on an active site. Fictionlab’s open ROS 2 stack on the Leo Rover supports teleoperation and can be extended toward supervised autonomy, which is one reason researchers adopt it for applied research projects.

How UGVs Improve Safety and Compliance Monitoring

Safety monitoring is the second mature use case. A UGV equipped with RGB and thermal cameras can:

• Detect missing PPE (helmets, vests) using onboard inference with YOLOv8 or similar models.
• Identify blocked emergency exits and missing or obstructed fire extinguisher locations against a known baseline.
• Spot thermal anomalies in electrical panels and curing concrete.
• Log environmental data (dust, noise, gas) along consistent transects.

The advantage over fixed CCTV is coverage flexibility: one robot can reach areas where infrastructure has not yet been installed and can reduce the number of temporary cameras needed for periodic checks. The advantage over inspectors is repeatability. The same route, the same lighting compensation, the same model running every shift.

What Still Limits UGV Deployment on Construction Sites

Field robotics on construction sites is hard. The honest picture includes several open challenges:

Terrain. Rebar, mud, cable trays, and 10 cm steps stop most wheeled platforms. Tracked robots handle more, but at the cost of mass, energy, and floor protection. A small UGV like the Leo Rover, with slope capability up to around 45° on dry surfaces and roughly 8-10 cm obstacle capability depending on configuration, fits indoor and graded outdoor scenarios well, while heavier tasks may push you toward larger platforms.

Localization. Inside reinforced concrete structures, GNSS is unreliable or absent. LiDAR-inertial SLAM works, but loop closures are difficult when the environment changes daily. Practical deployments often combine SLAM with fiducial markers (AprilTags) placed at known surveyed points.

Dust and weather. IP-rated enclosures matter. Concrete dust degrades optical sensors and bearings quickly. Many research-grade UGV payloads need protective covers or scheduled cleaning, which adds operational overhead.

Data integration. Capturing point clouds is well understood. Turning them into BIM-aligned, audit-ready deliverables without manual work is not. Expect to invest in pipeline engineering, not just hardware.

Site coordination. A robot on an active site needs a designated route, signage, and a responsible operator. This is a process change, not a technical one, and it often slows adoption more than any sensor limitation.

Early Results from Research and Pilot Deployments

Public reports from groups like ETH Zurich’s Robotic Systems Lab (ANYmal in construction-like and industrial environments), Boston Dynamics and Trimble (Spot with laser scanners on construction projects), and layout or drilling robots such as Hilti’s Jaibot show that legged and wheeled robots can deliver useful work on active sites. Smaller wheeled UGVs occupy a different niche: lower cost, easier maintenance, and a strong fit for research groups and integrators building custom stacks on ROS 2.

In some pilot programs, teams report scan acquisition times reduced by a factor of 3-5 compared to manual tripod-based scanning, with the caveat that registration and BIM integration still consume the majority of post-processing effort. These figures vary widely by site complexity and should be treated as indicative, not benchmarks.

How to Decide if a UGV Fits Your Construction Monitoring Workflow

A practical checklist before committing to a platform:

• Is your site mostly flat indoor floors, graded outdoor pads, or rough terrain? Match platform and payload to environment.
• Do you need raw data for your own pipeline, or a turnkey deliverable? Open ROS 2 platforms favor the former.
• What is your localization strategy in GNSS-denied zones?
• Who operates and maintains the robot? Budget for training, not just hardware.
• How will captured data flow into your BIM or safety platform?

For research groups, integrators, and engineering teams building their own construction robot monitoring stack, an open platform with a documented ROS 2 interface shortens the path from prototype to field test.

FAQ

What sensors do you need for UGV-based construction monitoring?

At minimum, a 3D LiDAR for geometry, an RGB or 360° camera for visual context, an IMU for sensor fusion, and either RTK GNSS outdoors or fiducial markers indoors for absolute positioning. Thermal cameras are optional but valuable for electrical and curing inspections.

Can a small UGV replace a Boston Dynamics Spot on a construction site?

Not for every task. Legged robots handle stairs and rebar that wheeled robots cannot. Small wheeled UGVs like the Leo Rover are better suited to graded surfaces, indoor floors, and research or integration projects where cost and openness matter more than maximum terrain capability.

How accurate is scan-to-BIM with a UGV-mounted LiDAR?

With survey-grade LiDAR and proper registration, deviations in the 1-3 cm range are reported in pilot studies. Accuracy depends heavily on sensor choice, registration method, and site reference points, not the robot platform itself.

Does the robot need full autonomy to be useful?

No. Teleoperation and teach-and-repeat workflows deliver most of the value with far less complexity. Full autonomy is attractive long-term but adds safety-case and localization overhead that many sites are not ready for.

How does ROS 2 help in construction monitoring?

ROS 2 provides standard interfaces for sensors, navigation (Nav2), SLAM, and data logging. This reduces vendor lock-in and lets teams reuse components across projects. Much current research in field robotics targets ROS 2 directly.

What is the biggest hidden cost in deploying a monitoring UGV?

Data pipeline engineering. Hardware and operation are relatively predictable; turning daily point clouds into BIM-aligned, deviation-flagged reports without manual cleanup is where most of the integration effort goes.

Can UGVs operate on active sites alongside workers?

Yes, with proper risk assessment, signage, and a designated operator. Unattended operation on active sites is still rare and typically requires additional safety validation or certification depending on jurisdiction.

Explore Construction Applications with Fictionlab

If you are evaluating UGV-based monitoring for a construction project, research program, or integration product, explore Fictionlab’s construction applications page for platform specifications, payload options, and reference deployments.