Robot Platforms for Infrastructure Inspection: What Buyers Get Wrong

TL;DR: Choosing an infrastructure inspection robot platform is rarely a hardware problem. Many failed deployments trace back to four mistakes: oversizing the platform for the terrain, having no plan for the data it produces, ignoring how the robot physically enters the structure, and relying on a single communications link with no fallback. This article breaks down each mistake, shows what a right-sized UGV looks like for bridges, tunnels, culverts, substations, and construction sites, and explains how ROS 2 platforms like Leo Rover fit into that picture.

Why the Infrastructure Inspection Robot Platform Discussion Goes Wrong Early

Infrastructure inspection is a broad label. A concrete bridge soffit, a 600 mm stormwater culvert, a live 110 kV substation, and a partially built parking structure share almost nothing in terms of access, lighting, connectivity, or safety constraints. Yet buyers often start the conversation with “we need an inspection robot” and jump straight to payload and top speed. That framing skips the questions that actually decide whether the robot returns usable data.

An infrastructure inspection robot platform is a mobile base plus sensors plus a data pipeline plus a connectivity plan. Miss any of the four, and the hardware may sit in a case. The mistakes below are the ones Fictionlab’s team sees most often when researchers, integrators, and inspection service providers evaluate mobile robots for field work.

Mistake 1: Oversizing the Platform

The default assumption is that bigger is safer. A 120 kg tracked UGV with a 40 kg payload sounds like it will handle anything. In practice, oversizing creates its own problems:

• It cannot enter culverts, crawl spaces, or narrow catwalks.
• It may require two operators and a trailer to move between sites.
• Battery swaps become a logistics task rather than a pocket operation.
• Site owners may refuse to let it near finished surfaces or live equipment.

For a large share of inspection tasks – visual documentation, thermal scanning, crack mapping, gas sensing – a compact 4-wheel platform in the 5 to 15 kg range can carry the required sensors and reach places a larger robot cannot. The Leo Rover, for example, offers a 5 kg payload, an integrated 5 V/12 V power rail, and a ROS 2-compatible stack you can extend without reverse engineering a closed controller. That can be enough for a GoPro-class camera, a small LiDAR, a thermal module, and an edge compute board, provided that power, mass, mounting, and environmental limits are verified for the specific configuration.

Right-sizing means matching the platform to the smallest opening, the tightest turn, and the steepest ramp on the route – not to the heaviest sensor you might theoretically mount.

Mistake 2: No Plan for the Data

A single inspection run can produce tens or hundreds of gigabytes of raw sensor data: 4K video, LiDAR scans, IMU logs, thermal frames, ROS 2 bags. Buyers focus on capturing it and forget three questions:

1. Where does the data live during the mission (onboard SSD, edge server, cloud upload)?
2. Who processes it, and with what tooling (photogrammetry, SLAM, defect classification)?
3. How does it reach the client in a format their asset management system accepts?

ROS 2 helps here because rosbag2, DDS-based communication, and standard message types (sensor_msgs/Image, sensor_msgs/PointCloud2) can support repeatable data capture. But the pipeline downstream – georeferencing, tiling, defect tagging – is where many projects stall. If you cannot answer “what does the deliverable PDF or 3D model look like” before buying the robot, the robot is premature.

A practical rule: prototype the full data path on a small platform first. Fictionlab’s approach with Leo Rover is to let teams validate the sensor stack and the ROS 2 pipeline on a lower-cost base, then scale hardware once the pipeline is proven.

Mistake 3: Ignoring How the Robot Enters the Structure

Access is the most underestimated variable in field robotics. A robot that clears a lab doorway can fail at:

• A 400 mm manhole with a vertical ladder.
• A construction site with rebar mats, mud, and 200 mm curbs.
• A bridge inspection walkway with a 150 mm expansion gap every 10 m.
• A substation with gravel ballast and cable trays at ankle height.

Before selecting a platform, walk the site with a tape measure and a camera. Log the narrowest passage, the highest step, the worst surface, and the deployment method (carried by hand, lowered by rope, driven off a van ramp). Compare those numbers against ground clearance, wheel diameter, approach angle, and total width. A 250 mm wheel and 130 mm clearance behave very differently on loose ballast than on a warehouse floor.

For construction environments specifically, Fictionlab documents typical deployment patterns on the construction applications page, where surface variability is the norm rather than the exception.

Mistake 4: No Redundancy in Connectivity

Single-radio deployments can cause aborted missions. Wi-Fi drops behind a concrete pier. LTE has no signal in a tunnel. A 2.4 GHz telemetry link can suffer interference from other equipment. When the link goes down mid-inspection, the operator may lose the robot’s live feed or need to trigger a preconfigured safe-stop or return procedure.

A resilient infrastructure inspection robot platform assumes the link will fail and plans for it:

• Primary link for teleoperation and live video (Wi-Fi mesh, private 5G, or dedicated 5 GHz).
• Secondary link for telemetry and safe-stop commands, where appropriate (cellular, sub-GHz radio, or a wired tether).
• Onboard autonomy for waypoint following and data logging when both links drop.
• Local recording of full-resolution data, so the mission can still produce a deliverable even if streaming fails.

ROS 2 can support systems with multiple network interfaces, but DDS discovery and routing across interfaces or subnets require deliberate network configuration. The network design is still an engineering task, not a checkbox.

What Matches What: Structure Type vs Platform and Sensor Requirements

The table below summarizes what Fictionlab’s team typically sees as workable combinations. Treat it as a starting point for site-specific evaluation, not a specification.

For a worked example of how these choices play out on a specific asset class, see the earlier Fictionlab discussion of bridge inspection workflows on the inspection applications page, where sensor placement and access constraints drive the entire platform choice.

How to Evaluate an Infrastructure Inspection Robot Platform Before You Buy

A short evaluation loop saves months. Before committing to a platform, run through this sequence:

1. Define the deliverable. What file, report, or model does the client actually receive?
2. Walk one representative site with measurements and photos.
3. Sketch the data pipeline end to end, including storage and processing time.
4. Prototype on a low-cost ROS 2 platform to validate sensors and software.
5. Design connectivity with at least one fallback path.
6. Scale to the mission-specific hardware only after the pipeline works.

This is why Fictionlab designed Leo Rover as an open, ROS 2-compatible base. It lets teams answer the pipeline and sensor questions at lower cost, then transfer the same software stack to a larger platform such as Raph Rover when the mission profile demands more payload, range, or environmental protection.

FAQ

What is the minimum sensor set for a general-purpose inspection robot?

An RGB camera with suitable low-light performance and onboard lighting cover many visual documentation tasks. An IMU is useful for motion data and sensor fusion. Thermal and LiDAR are added when the deliverable requires temperature data or 3D geometry.

Do you need ROS 2, or is a closed platform easier?

Closed platforms are often faster to deploy for a fixed use case. ROS 2 is a practical choice when you expect to change sensors, integrate site-specific software, or extend autonomy over time.

How much autonomy is realistic on an inspection UGV today?

Waypoint navigation with obstacle avoidance can be effective on structured, well-mapped sites, but performance depends on sensing, lighting, terrain, and communications. Automated defect detection can support inspectors, but human review remains standard for safety- and asset-critical decisions. Plan for teleoperation plus assisted navigation, not full autonomy.

Can a small platform like Leo Rover really handle field inspection?

Yes, within its envelope: up to its specified payload capacity, moderate terrain, and missions where compactness and cost matter more than range. For heavier sensor stacks or long-range outdoor missions, a larger platform is appropriate.

What about tethered vs untethered operation?

Tethered operation is often used in confined spaces such as culverts and tanks because it can provide reliable communications and, in some configurations, power. However, tether management and snag risks must be assessed. Untethered operation is common outdoors. The choice follows the structure and mission requirements, not preference.

How do you handle explosive or hazardous atmospheres?

Standard UGVs are not ATEX-certified. For hazardous locations, use equipment certified for the applicable zone and gas or dust classification under the relevant regulatory framework, such as ATEX in the EU or IECEx where applicable. Do not retrofit a general-purpose robot for hazardous locations without appropriate certification.

What is the biggest hidden cost in an inspection robot deployment?

Data processing labor can be a major hidden cost. Capturing footage is fast; turning it into a client-ready report is not. Budget for the pipeline, not just the platform.

Plan the Mission, Then the Robot

Infrastructure inspection rewards teams that treat the robot as one component in a larger system. Get the access analysis, the data pipeline, and the connectivity plan right, and a modest ROS 2 platform can outperform an oversized one on the same site. To see how Fictionlab structures these deployments across bridges, tunnels, substations, and construction environments, visit the inspection applications page.