When a mobile-robot deployment underdelivers, the battery is usually blamed first. In practice, the bottleneck is rarely the watt-hours on board — it is the charging infrastructure around the fleet: how many charging points there are, where they sit, and how long each robot dwells on them. A fleet with generous batteries but too few charging stations will still stall at peak; a fleet with modest batteries and well-placed opportunity charging can run three shifts. This article walks through how AMR charging actually works and offers a practical way to size the charging side of a fleet before you commit to a layout.

How AMR charging actually works

Modern autonomous mobile robots almost universally use lithium chemistries and opportunity charging: instead of a full recharge at end of shift, the robot tops up in short windows whenever it is idle — waiting for a pick, parked between missions, or dwelling at a station. The number that matters is not “how big is the battery” but “how much charge can the robot recover in the idle windows the process actually gives it.” Two robots with identical batteries can behave completely differently depending on how much idle time the workflow leaves them, and whether a charging point is free when that idle time occurs.

The four variables that decide it

• Mission rate — how many missions per hour the process demands from the fleet.
• Cycle time — how long a mission takes, including travel, waiting and handling.
• Charge window — the idle time available for charging per cycle, and the current the robot accepts.
• Number and placement of charging points — the physical capacity to convert idle time into energy.

Availability is the interplay of these four, not any one of them in isolation.

A practical sizing method

Start from throughput, not from the datasheet.

• Compute the number of robots the mission rate needs at the given cycle time (missions/hour × cycle time in hours = robots working concurrently, plus a margin).
• Estimate the energy each robot consumes per hour of operation and the charge it recovers per minute at a station.
• From those, derive the charging time each robot needs per hour of work — call it the robot’s “charge duty.”
• Multiply charge duty by fleet size to get the total station-minutes required per hour; divide by 60 to get how many charging points must be occupied at the same time. Then add headroom for peaks.

A simple illustration: a 10-robot fleet that each need roughly 6 minutes of charging for every 60 minutes of work requires 60 station-minutes per hour — one continuously occupied point on average, but realistically two or three, because robots do not arrive to charge in an orderly, evenly spaced way.

Figure 1 — Sizing the charging side of an AMR fleet.

Queueing is the real availability killer

Averages hide the problem. Robots reach charging points in bursts, not evenly. If two robots need to charge at the same moment and there is one point, one of them waits — and a waiting robot is a robot not running missions. This is why sizing on the average almost always under-provisions: you must size for the peak arrival pattern, not the mean. Near the knee of the queueing curve, a small increase in the number of charging points produces a large drop in waiting time — usually at a fraction of the cost of the alternative below.

Add charging points, not bigger batteries

When availability drops, the instinct is to specify larger batteries. But bigger batteries are heavier, more expensive, slower to charge, and they delay the problem rather than solving it: the robot still has to find a free charging point. Adding charging points is typically cheaper, faster to deploy, and it attacks the queueing bottleneck directly. Battery size sets how long a single robot can run between charges; charging-point capacity sets how reliably the fleet as a whole stays available. The two are not interchangeable.

Placement and electrical power

Where the charging points sit matters as much as how many there are. Points placed on natural dwell locations — near pick faces, staging areas, or the ends of frequent routes — turn existing idle time into charge with no added travel. Points banished to a remote “charging room” force robots to drive in order to charge, adding non-productive travel that eats into the very availability you are trying to protect. Electrical distribution is the constraint teams tend to discover late: a bank of fast chargers is a real load on the local panel, and the layout has to account for it from the start.

Behaviour on the peak shift

A fleet is not sized for the average hour; it is sized for the worst hour it must survive. Model the peak shift explicitly: the hour with the highest mission rate and the least idle time is exactly when charging capacity is most stressed and queueing is worst. If the fleet holds its availability through the peak, the rest of the day looks after itself. Fleet-orchestration software — interoperability frameworks such as VDA 5050 are increasingly common — decides which robot charges when, but it can only allocate the charging points that physically exist.

Key takeaways

• Size the charging side from throughput and from the peak, not from battery capacity.
• Treat queueing at the charging points as a first-class design variable.
• Prefer more, better-placed charging points over bigger batteries.
• Design electrical distribution and dwell-based placement into the layout from day one.