How many warehouse robots do you need?
The number of autonomous mobile robots (AMRs) or automated guided vehicles (AGVs) a warehouse needs comes from two things: the work they must do in the peak hour, and the minutes each robot can really work in an hour once breakdowns, congestion and charging are taken out.
The calculator works out the task cycle time, the robots needed for your peak, the extra robots and the chargers that charging takes, and the tasks per hour a given fleet can deliver.
Why size the fleet before you sign?
- Too few robots: stations wait for racks or totes, and the pick rate in the business case is never reached.
- Too many robots: you pay for idle robots, and more robots on the same aisles means more congestion, which simulation studies show can cost time instead of saving it.
- Charging is part of the fleet: a robot on a charger is not working. A PhD study on AGV fleet sizing found that battery management significantly affects throughput and vehicle use.
- Vendor proposals differ: the same calculation for every proposal shows which assumptions about speed, travel and charging each one makes.
What inputs do you need?
| Input | Where to find it | Tip |
|---|---|---|
| Tasks per hour at peak | WMS order history: lines or totes per hour on your busiest days | White Friday, Ramadan and Eid are usually the peak in the Gulf. |
| Station time per task | Pilot, time study, vendor simulation | Include waiting at the station, not only the hand-over. |
| Loaded and empty travel | Layout drawing with the robot routes | Use average distances over all storage locations, not the nearest ones. |
| Average speed | Pilot data or the vendor | Average with turns and slowing near people, below the top speed. |
| Run time and charge time | Vendor datasheet at your payload | Check the charge window (for example 20% to 80%). |
| Availability and traffic factor | Vendor service level, simulation, your history | Ask the vendor what they assumed in their own fleet calculation. |
How is the fleet size calculated?
Task cycle and robots for the work
This is the analysis of vehicle-based material transport systems in Groover's textbook (delivery cycle time, workload, available time and number of vehicles), the same logic as in Mantel and Landeweerd: the number of vehicles is the loaded and empty travel time plus waiting time over a busy period, divided by the time a vehicle is available.
Charging
The charging step is an energy balance: every hour of work needs charge time ÷ run time of an hour on a charger. Opportunity charging, a few minutes between tasks, spreads that time across the shift but does not remove it.
Sources
- M. P. Groover, Automation, Production Systems, and Computer-Integrated Manufacturing, chapter 10 (material transport systems), section 10.6: delivery cycle time, workload, available time = 60 × availability × traffic factor × efficiency, number of vehicles (example with 95% availability and a 0.90 traffic factor), as taught in NIT Srinagar course MEC 405. Source: NIT Srinagar lecture notes.
- R. J. Mantel and H. R. A. Landeweerd, Design and operational control of an AGV system, International Journal of Production Economics 41, 1995, pp. 257–266. Source: University of Twente.
- C. R. Standridge, Beyond Lean: Simulation in Practice, section 17.3 (an analytic estimate that assumed no blocking gave 2 AGVs; simulation showed 3 were needed, with more congestion as vehicles were added). Source: LibreTexts.
- OMRON, Autonomous Mobile Robot MD series datasheet I885-E, 2024 (run time 10 h with no payload and 8 h at full payload; recharge 19.6 min from 20% to 80%). OTTO Motors, How do AMRs charge?, 15 December 2025 (opportunistic charging between jobs; charging speed affects fleet size and the number of chargers). Source: OMRON, OTTO Motors.
- A. Hamdy, Optimization of Automated Guided Vehicles (AGV) Fleet Size With Incorporation of Battery Management, PhD dissertation, Old Dominion University, 2019. Source: ODU Digital Commons. Last reviewed 21 October 2026.
Worked example: 600 rack moves an hour
The defaults are examples, not robot specifications; replace them with your own figures.
- Each task has 30 seconds at the station and 60 m loaded plus 60 m empty at 1.2 m/s: 0.5 + 0.83 + 0.83 = 2.17 minutes, 77% of it travel.
- 600 tasks an hour make a workload of 600 × 2.17 = 1,300 robot-minutes an hour. With 95% availability and a 0.90 traffic factor each robot gives 51.3 minutes an hour, so 1,300 ÷ 51.3 = 25.3, or 26 robots for the work.
- With 8 hours of work per charge and 60 minutes to recharge, the charge ratio is 0.125. Running 20 hours a day leaves too little time to charge between shifts, so robots charge during operation: 25.3 × 1.125 = 28.5, or 29 robots, three more for charging.
- On average 3.2 robots are charging at once: 4 chargers, still 4 chargers with 25% spare.
- A proposal with 25 robots delivers about 526 tasks an hour, 12% below the peak: four robots short.
- A single 9-hour shift with 10 hours of run time and 2 hours of charging would let the 26 robots charge after the shift, 7 robots per charger in the 15 hours left: 4 chargers and no extra robots.
How to read your results
- Exact working robots: the fraction tells you how close the rounded fleet is to its limit; 25.3 rounded to 26 has little spare.
- Travel share of the cycle: when travel dominates, slotting fast movers near the stations or adding stations shortens the cycle more than faster robots.
- Charge share: the share of fleet time spent on chargers. Faster charging or a longer run time lowers the extra robots.
- Capacity of a fleet: below 0% headroom, the fleet cannot keep up with the peak you entered.
How to need fewer robots
- Shorten travel: put fast-moving SKUs near the stations and review the layout of routes.
- Cut station waiting: balance the work between stations so robots do not queue.
- Reduce congestion: one-way routes, wider crossings and separated people and robot zones raise the traffic factor.
- Plan charging: charge in low-demand hours, and place chargers near the routes.
- Check the money: test the fleet in the Warehouse Robot ROI Calculator.
Assumptions and limits
The calculator uses averages: every task has the same cycle, tasks arrive evenly through the hour, and the peak rate runs for all operating hours, so real queues are longer when cycles vary. Congestion is a single factor you enter; it rises as robots are added, which only a simulation or a pilot can show, and Standridge's case study shows an analytic estimate that was one vehicle short. Charging assumes the energy used is proportional to working time. Battery ageing, heat in summer and the charge window in the datasheet change run and charge times. All defaults are examples, not robot specifications. Use the result to plan and to compare vendor proposals, then confirm it with the vendor's simulation and a pilot.