Which inspection rounds should a maintenance robot take on?

A mobile inspection robot is worth considering when a maintenance team needs repeatable observations at several known places and can explain what a useful observation looks like. Start by choosing those observations. A route that the robot can walk is only a candidate; it still has to return evidence a reliability lead can use.

This guide helps a maintenance manager, technician and reliability engineer shortlist routine rounds for a supplier assessment. It is our procurement framework, not a robot operating procedure, fault-diagnosis method or authorization to enter a hazardous area. We have not run a site trial. The worksheet below separates the observation, the route and the person who must review the result.

What the current product evidence supports

Boston Dynamics describes Spot inspections using visual images and gauge reading, an optional thermal camera, and acoustic imagers. These are different sensing tasks. A buyer should specify the required observation before selecting a payload, rather than treating every Spot configuration as an equivalent inspection system.

The current Spot payload list, reviewed November 4, 2026, includes Spot Cam 2 and Sorama L642, alongside other equipment. This establishes current listed product names, not that every sensor is included in a base purchase or suitable for your equipment. Record the proposed camera, acoustic sensor, software and hardware revision in the quotation. Do not silently apply older payload specifications to a new configuration.

ANYbotics' partner documentation on Siemens' developer site describes ANYmal and the separate ANYmal X platform for routine condition monitoring. It documents mission management, inspection-data transfer, and time-stamped, position-referenced records through an API and SDK. This is useful integration evidence, but it does not settle the latest robot revision, certification scope or availability at your site. Keep those as supplier evidence requests, especially for ANYmal X; this article makes no hazardous-area suitability judgment.

Spot's product page describes Orbit for managing robots and industrial inspections, plus an API and payload options. Neither a fleet dashboard nor an API name proves that the proposed system will produce your required asset record. Ask to see that record from the specific proposed configuration.

Read a deployment account at the right scale

In the vendor-hosted Michelin case-study transcript, personnel at the US5 tire plant in Lexington, South Carolina describe an initiative that began as a pilot, then report programmed inspection missions in use. Reliability staff describe selecting equipment they already knew, reviewing results in Orbit, and sending a technician to verify a flagged condition. They also describe further routes and expansion as future work.

That is a customer account of operating rounds published by the robot supplier. It is more specific than a demonstration video, but it is not an independent performance study or proof that every planned route was deployed. We use it to identify the human work around collection: task selection, mission setup, record review and verification. We do not transfer its results, staffing assumptions or economics to another plant.

Shortlist observations before combining them into a round

The following selection rules are our analysis. Start with existing rounds and ask the technicians which entries produce useful evidence today. Keep each candidate observation small enough to describe in one sentence, such as an image of a named instrument or a thermal record of a particular asset under a recorded operating condition. Avoid a brief that merely says "inspect the compressor room."

For every entry, identify the decision owner and the missing evidence. An inspection point with no recipient should wait. An ambiguous asset identifier needs correction before automation makes the ambiguity repeatable. If the current task requires a person to investigate an unfamiliar condition, divide routine collection from that investigation rather than assuming the robot can inherit both.

Then group points into a proposed route. Discuss access changes, stairs, doors, traffic, temporary storage, docking space and connectivity with the supplier and the site's responsible team. These are assessment questions, not permission to navigate a particular environment. A route should remain a proposal until the appropriate site and supplier reviews are complete.

Give fixed sensors and manual rounds a fair comparison

Use three alternatives in the brief: retain the current manual round, assess fixed sensing for a defined point, or assess mobile collection across several points. The comparison here is a planning method, not a claim that one architecture is cheaper or more reliable.

Ask whether the decision needs a continuous record, a periodic observation or a visit prompted by another event. If continuous evidence is required, establish how a proposed mobile round would address the intervals between visits; do not treat a scheduled visit as an uninterrupted measurement. For fixed sensing, request the exact measurement, installation boundary, maintenance work and data destination. For manual collection, record both the captured evidence and the judgment the worker contributes.

Mixed arrangements deserve a place on the shortlist. One route may retain specialist human checks while automating a repeatable image elsewhere. Leave those boundaries visible. Do not count a manual follow-up or an inaccessible asset as automated coverage simply because it appears on the same route map.

A worksheet for the first supplier discussion

Copy these fields once per inspection point, then add a route-level summary. This is an original Awetonomous worksheet. It sets no universal interval, diagnostic threshold or acceptance tolerance.

Observation and recipient. Name the asset, location, evidence requested, current collection method, and person who will review it. State the decision the evidence is intended to inform.
Timing and context. Record the proposed collection interval or trigger, the equipment state that gives the observation meaning, and how that state will be documented. Leave intervals for the responsible team to define.
Sensor and viewpoint. Name the proposed sensor, configuration, required field of view and reference evidence. Ask the supplier to demonstrate usable capture under the approved test conditions.
Evidence acceptance. Define how a reviewer distinguishes usable, obscured, incomplete and missing records. Identify who can accept the observation, request another collection or escalate it.
Route and dependencies. List access assumptions, docking and communication dependencies, and the owners responsible for assessing them. Mark unassessed conditions explicitly.
Exceptions and recovery. Name who receives an incomplete-mission notice, who decides on alternative collection, and who is authorized to arrange recovery. Request the supplier's approved process; do not improvise one.
Record ownership. Specify asset association, timestamp, original media, configuration reference, correction history, access and retention. Identify the integration owner and failed-transfer review.
Shortlist decision. Choose manual, fixed-sensing assessment, mobile assessment or hold. Record the reason, missing evidence and next owner rather than calculating an unsupported score.

A fictional shortlist, with the exclusions retained

Suppose a plant team brings three candidate tasks to a meeting. First is a recurring visual record of a clearly identified instrument. Second is periodic thermal capture of equipment whose operating state is available to the reviewer. Third is an intermittent observation currently requiring a technician to open an enclosure. This is a fictional planning example, not a reported deployment.

The team could request supplier evidence for the first two while holding the third outside the proposed mobile collection scope. It would still need to establish view quality and comparable context for each shortlisted point. It would also record any continuous-monitoring requirement separately. No diagnosis or safe-access conclusion follows from the shortlist.

That small exercise produces a better brief than a target number of route points. It shows what remains manual, which observation needs another sensing approach, and which supplier questions could change the decision. A hold is a useful result when it prevents unclear scope from becoming a purchase assumption.

Review the evidence and the work left behind

At an authorized supplier demonstration, review original records alongside mission completion. Count requested points, usable records, missing records, repeat collections and human interventions separately. Document the conditions and configuration. A completed route with unreadable evidence should not become a successful inspection in the acceptance record.

Assign a maintenance owner to review findings and a separate technical owner where needed for robot care, access, configuration changes and integration. Agree who notices a missed round and what happens while collection is unavailable. Obtain location-specific support and training commitments in writing; we have not verified regional service for either supplier.

The useful next step is a shortlist of bounded observations, each with an evidence request and an owner. Use our industrial robotics guide for the wider integration boundary. Once records arrive, our building-maintenance handoff guide addresses the separate question of getting an observation to the right asset and work order.

Sources & evidence

Official sources reviewed November 4, 2026. Vendor materials establish descriptions, not independently measured performance. Original evaluation criteria are our analysis.

Editorial status. Published November 4, 2026. This is source review and original analysis, not a hands-on test. Read our selection, corrections, and update policy ↗