AMR warehouse robots: how many robots and charging stations are needed?
Enter transport orders per hour and the distance — the calculator gives the number of robots running at the same time, then simulates the battery over a full day to get the total number of robots and charging stations. Below are five stages on dispatching, traffic and battery charging.
What the model calculates, and what it doesn't
The warehouse is a square grid and robots travel by Manhattan distance (horizontal + vertical). The list of pick-and-deliver orders is generated from a seed, so changing the seed changes the scenario; keeping the seed makes it repeatable. Time = cells ÷ speed + 2 handling times.
Two rules run on the same list: nearest robot picks the robot closest to the pickup point (on a tie, the robot that becomes free earlier); earliest finish picks the robot with the earliest expected completion time (on a tie, the robot with less work). Finishing the last order sooner is the main metric; total cells traveled and load imbalance are two secondary metrics.
Stage 4 includes traffic: a 17 × 11 cell grid warehouse, horizontal shelf rows, one-cell-wide aisles, handover station in the left column. Each cell holds only one robot; if the next cell is occupied, the robot waits; if it waits too long, it looks for a detour, and deadlocks are resolved by priority (the higher-numbered robot yields). Each fleet size runs 5 order layouts, 1 hour per layout; "jammed" means a layout in which robots spend more than 80% of the time waiting.
Stage 5 — battery and charging stations: MiR250 robot parameters taken from the manufacturer's spec sheet (PDF version): carrying maximum load it runs 13 hours (100% → 0%), idle 22 hours, charges 10% → 90% in 52 minutes, docking takes up to 44 seconds (plus 1 minute to travel to the station — an assumption). Simulated minute by minute over 24 hours, with 6 robots needed working at the same time; a working robot keeps its job and is replaced only when a robot leaves to charge. Two policies: run-to-empty (charge at 10%) and rotating charge (an idle robot below 70% charges; a working robot below 40% hands its work to a standby robot at ≥ 70%).
Not modeled (first three stages): aisle blocking and robots avoiding each other — only stage 4 has them. All four stages leave out: battery charging, priority for urgent orders, robot breakdowns mid-shift. In real warehouses, dispatching software uses far more complex rules — this simulation only shows why dispatching rules change the result.
Frequently asked questions
Why does the nearest-robot rule travel fewer cells yet finish later?
Because it loads many orders onto the nearby robot, which becomes a bottleneck while other robots wait. Fewer cells traveled saves battery; finishing early is productivity — two different goals.
Does it apply to AGVs in ports?
The principle is the same, but a port adds fixed routes and yard cranes as bottlenecks — see the automated container terminal dispatching simulation.
Have real project parameters? Send them to CGD and get feedback on the design scope
This simulation stops at the principle. For a specific project, CGD builds the model on real operating data and floor plan drawings.
Submit a consultation request