Robot 101 · Chapter 11

Keeping robots running: field maintenance and spare parts

Why uptime is the real product

A robot that performs beautifully in a demo but spends one day in five offline for repair has not solved the problem it was built for. For any operator running a fleet — in a warehouse, a factory, or a care facility — the commercial reality is that every hour a unit is down is an hour of lost service, and every unplanned technician visit adds labour and logistics cost on top of the part itself. That shifts the engineering question from "does it work" to "how long does it keep working, and how quickly can it be brought back online when it doesn't." Maintainability is a product decision as much as a mechanical one, and it has to be made early, because retrofitting it after a design is frozen is far more expensive than designing for it from the first prototype.

Wear items and typical lifetimes

A small number of components account for most service events on a deployed robot. Joint lubrication is one: gearboxes and harmonic drives depend on grease that degrades with heat and cycle count, and periodic relubrication (or a sealed-for-life unit rated for a specified number of operating hours) is what keeps a joint from becoming the failure point. Cables and connectors at moving joints see continuous flex fatigue, so robots built for real duty cycles typically use cabling rated for a high number of bending cycles rather than off-the-shelf wire, and connectors that are keyed so a cable can only be reattached the right way. Battery packs degrade gradually through charge-discharge cycling; a pack is typically rated to retain a defined capacity threshold (commonly on the order of 80%) after a stated number of cycles, after which range and runtime noticeably shorten even though the battery still functions. Casters and wheels wear through direct contact with the floor and are usually specified as solid tyres rather than pneumatic, precisely because a punctured tyre is an unplanned failure while tyre wear is a predictable, schedulable one.

Preventive maintenance vs run-to-failure

Not every part deserves the same maintenance strategy. The choice usually comes down to two questions: how expensive is it to replace on a fixed schedule regardless of actual wear, and how bad is it if the part fails unexpectedly instead. Cheap, fast-to-swap consumables — gripper pads, tyres — are often left closer to run-to-failure, or replaced on a simple calendar interval, because an operator would rather absorb the occasional early swap than build a monitoring system for a ten-dollar part. Components whose failure strands a robot mid-task, creates a safety issue, or is expensive to diagnose after the fact — batteries, structural fasteners, joint lubrication — are usually put on a preventive schedule tied to elapsed operating hours or cycle counts, serviced before they fail rather than after. The right split is specific to each fleet's economics, but the underlying logic generalizes across robot types.

The spare-parts kit: what to stock, what to order

A fleet operator's spare-parts strategy usually separates parts into two tiers. On-site stock covers the handful of components that fail often enough, and are cheap and fast enough to swap, that waiting for a shipment would cause unacceptable downtime — typically gripper pads, a couple of wheel assemblies, connectors, and a spare battery pack or two, sized to the fleet's failure rate. Everything else — structural parts, less common sensor modules, whole joint assemblies — is ordered against forecast demand rather than held on every site, since holding a deep inventory of every possible part at every location is rarely worth the carrying cost. Getting this split right depends on having reliable part numbers, lead times, and documentation for the long tail of components a robot is built from — which is as much a sourcing problem as a maintenance one.

Remote diagnostics and OTA: fewer site visits

The most effective lever for reducing service cost is often not a better part — it is catching a problem before it causes a failure at all. A telemetry pipeline that reports battery state of charge and state of health, joint temperatures, torque utilization, and task-success rates back to a fleet dashboard lets an operator watch for degradation trends rather than wait for an outright fault. Simple anomaly detection — a temperature trend drifting upward, a success rate dropping over a rolling window, battery health crossing a threshold — can flag a robot for scheduled service before it fails in the field, so a technician arrives with the right part already in hand instead of running a diagnostic session on site. Over-the-air (OTA) updates extend this further: many issues that would once have needed a site visit — calibration drift, a software bug, a sensor recalibration — can be pushed as a remote update instead, leaving physical visits for the wear items that genuinely require hands-on replacement.

Maintainability by design

Everything above is easier or harder depending on decisions made long before a robot ships. Connectorized wiring — each cable assembly terminating in a keyed connector rather than being soldered or spliced in place — turns a rewiring job into a swap. Modular joints, built as a single unit containing the motor, reducer, encoder, and driver, let a technician replace a whole joint as one part rather than repairing it in place. Accessible fasteners and tool-free access panels for consumable components mean a field replacement takes minutes with a standard driver set instead of requiring a partial teardown. None of this is free at the design stage — connectorized, modular hardware typically costs more up front than a tightly integrated equivalent — but for any robot that will be deployed at scale rather than built once, that tradeoff usually pays for itself many times over across the life of the fleet.