› AGV/AMR Warehouse Traffic Throughput Simulator Back
Warehouse Robotics

AGV/AMR Warehouse Traffic Throughput Simulator

Pick one of four warehouse types — e-commerce fulfillment, manufacturing flow, cold storage or cross-dock — then dial in floor area, AGV count, speed, charging cycle and order line profile. The tool returns items per hour, peak-day volume, utilization, fleet density and an annual ROI estimate, all in real time, for early sizing and investment screening.

Parameters
Warehouse type
Preset for cycleTime feel and density tolerance
Floor area
m²
AGV count
units
AGV average speed
m/s
Picks per AGV per hour
/hr
Reference value — used as a sanity check against cycleTime
Charging time
min
Battery runtime
hr
Lines per order
lines
Results
—
Avg travel distance (m)
—
Cycle time (s)
—
Per-AGV orders (orders/hr)
—
Total throughput (items/hr)
—
Utilization (%)
—
AGV density (AGV/1000m²)
—
Warehouse layout — AGVs, paths, charging, congestion

Gray bands are racks, blue dots are AGVs, green dots are charging, red haloes mark congestion. Paths and charging stations are drawn live.

Sensitivity — throughput vs AGV count
Throughput by warehouse type (current settings)
Theory & Key Formulas

$$\text{cycleTime} = \frac{\bar d}{v} + t_{\text{pick}}, \qquad \bar d = 0.5\sqrt{A}$$

Average travel distance $\bar d$ scales with the square root of floor area $A$. $v$ is AGV speed, $t_{\text{pick}}=30\,\text{s}$ is the pick/drop overhead.

$$Q_{\text{hr}} = \frac{3600}{\text{cycleTime}}\cdot N \cdot 0.85 \cdot \eta_{\text{batt}} \cdot L_{\text{order}}$$

$N$: AGV count, $0.85$: traffic-control loss, $\eta_{\text{batt}} = T_{\text{run}}/(T_{\text{run}}+T_{\text{chg}})$: utilization, $L_{\text{order}}$: lines per order.

$$\rho = \frac{N}{A}\times 1000\;[\text{AGV}/1000\text{m}^2], \quad \text{ROI}_{\text{yr}} = Q_{\text{day}}\cdot 250\cdot 0.5 - 0.2\cdot C_{\text{fleet}}$$

$\rho\gt 8$ triggers the overcrowded verdict. $C_{\text{fleet}} = 35{,}000\,\text{USD}\times N$ is the fleet capex; the annual gain assumes 250 working days at 0.5 USD/item with a 20% allowance for maintenance.

AGV/AMR Warehouse Throughput — Tuning Waits, Paths and Charging

🙋
I have been asked to automate an e-commerce DC, but I cannot tell how many AGVs we really need. The vendor says "about 50", but I would like to know where that number comes from.
🎓
Classic question. Break it down step by step. First, how many orders per hour does a single AGV handle? cycleTime = average travel distance / speed + pick overhead (about 30 s). For a 10,000 m² building the average travel is √10,000 × 0.5 = 50 m; at 1.5 m/s that is 33 s of motion plus 30 s of pick, so one cycle is 63 s. 3600/63 ≈ 57 orders/hr is the per-AGV ceiling.
🙋
So 57 × 50 = 2,850 orders/hr? That sounds lower than I expected.
🎓
Two reality factors then kick in. Traffic management eats about 15%, so multiply by 0.85. Charging shaves more off: 8 h runtime with 60 min charging gives 480/(480+60) = 0.89 utilization. 2850 × 0.85 × 0.89 ≈ 2156 orders/hr. With 6 lines per order typical for e-commerce, that is about 12,900 items/hr — exactly what the tool calls total throughput.
🙋
So can I keep adding AGVs and grow throughput linearly?
🎓
That is the trap. Once density (AGV/1000m²) crosses about 8, intersection locks dominate. 10,000 m² with 80 AGVs sits exactly at 8; 100 AGVs is already 10. The tool flips the verdict to red there. To grow further you redesign one-way lanes or rack layouts first, and only then add units. Throwing more AGVs at a saturated network usually slows the whole fleet down.
🙋
One last thing — why does the warehouse-type selector matter, beyond just the label?
🎓
Because the bottleneck shape changes. Cold storage costs you 20 to 30% battery capacity at −25°C, so try entering 0.7× of catalogue runtime. Cross-dock barely picks at all, so it is almost pure traffic — congestion bites earliest. Manufacturing routes are long inter-process moves with low density. The formulas stay the same, but the comparison bar chart lets you see, under the same parameters, which operation suffers most.

Frequently Asked Questions

Start from the peak items/hr you need. Per-AGV capacity is ordersPerAGVperHr = 3600 / cycleTime, with cycleTime = average travel distance / speed + a pick/drop overhead (30 s in this tool). Multiply by 0.85 (traffic-management losses) and utilization, then divide your target items/hr by that figure to get the AGV count. Finally check that density = numAGVs / area × 1000 stays under 8 AGV/1000m².
Utilization is modeled as runtime / (runtime + charging time). For 8 h runtime / 60 min charge that is 480/540 = 0.89. With LFP batteries and opportunity charging you can push 4 h runtime / 15 min charge up to 0.94, but ignoring the cooldown between fast charges shortens cell life by 30 to 50 percent. A realistic mix is fast-charging only during peaks and slow charging overnight.
Typical aisle widths are 2.5 to 3.5 m and the safe AGV separation is 1.5 to 2 m, so passing and intersection stops start to dominate empirically around 8 AGV/1000m². Above that, the traffic manager locks more often and real cycle time degrades by 20 to 40 percent. This tool does not embed that penalty into cycleTime, so it surfaces a density warning instead. Mitigations are zoning (one-way lanes), platoon control and rack-layout redesign.
Per-AGV cost is fixed at 35,000 USD including system integration and WMS hooks, so fleetSizeCostUsd = numAGVs × 35000. The annual gain is peakDayThroughput × 250 working days × 0.5 USD/item (labor savings and error reduction), minus 20 percent of capex for depreciation and maintenance. Real deployments add 30 to 60 percent for racking, communications and floor flatness work, so treat the tool figure as an optimistic upper bound.

Real-World Applications

E-commerce fulfillment centers: Operators like Amazon, JD.com and Ocado run 50 to 120 AGVs per 10,000 m² mixing Goods-to-Person shuttle racks with under-rack AMRs. Average lines per order are 3 to 8, and peak days are 3 to 5× a normal day. Use the tool to read peakDayThroughput and check whether the count needed for peak still keeps density under the 8 AGV/1000m² ceiling.

Manufacturing component delivery: Automotive and electronics plants use AMRs to deliver kits to assembly lines and to evacuate finished goods. Travel legs are long (100 to 200 m), so cycleTime dominates and density stays low at 3 to 5 AGV/1000m². The KPI is empty-travel reduction rather than congestion. Vary speed and utilization here to benchmark against manual carts on a TCO basis.

Cold and frozen storage: At −25°C lithium-ion cells deliver only 70 to 80% of nominal capacity and fast charging is throttled. Enter roughly 0.7× of the catalogue runtime to land in reality. Worker time limits (15 minutes of rest per hour in many countries) make robotics ROI attractive here, often more so than in ambient warehouses.

Cross-dock and logistics hubs: No storage step means pick overhead is negligible and congestion rules. Cycles of 30 to 60 s push density above the warning threshold quickly. Use the tool to read the density verdict while you partition lanes or stagger peak windows.

Common Misconceptions and Pitfalls

The biggest trap is "taking the catalogue pick rate at face value". The "100 picks/hr" vendors quote assumes ideal short loops with no waits. In a live warehouse, rack-ready delays, route arbitration and comms latency clip it down to 50 to 70%. That is why the tool computes throughput from cycleTime and treats pick rate as a reference only. Before signing off, run a pilot and back-fit the real ordersPerAGVperHr against the tool's value.

The second mistake is "assuming 95%+ utilization in your business case". Once you add WMS outages, picking jams, replenishment delays and AGV reboots, year-average utilization sits at 0.85 to 0.90 even for best-run sites. The tool's 0.89 is a design target. Plan year one at 0.70 to 0.80 to build a safety buffer — especially for e-commerce, where peak weeks generate 30% of annual revenue and a 10-point utilization gap can blow up service levels.

Finally, do not "compute ROI without WMS, racking and floor work". fleetSizeCostUsd only covers AGVs and core integration. Add Goods-to-Person racks, 5G or Wi-Fi 6 mesh, and the ±5 mm floor-flatness correction that AGVs require, and total capex grows to 1.3 to 1.6× the tool number. Maintenance contracts, spare units and software licenses add another 8 to 12% per year. Many projects show positive ROI on paper but negative cash flow — always rebuild capex, opex and savings on three separate sheets.

How to Use

  1. Select warehouse type (e-commerce fulfillment, manufacturing flow, cold storage, or cross-dock) to set baseline congestion and routing complexity.
  2. Enter warehouse floor area in m², number of active AGVs, average AGV speed in m/s (typically 1.2–1.8 m/s for industrial units), and pick rate in items/hour per station.
  3. Run simulation to calculate average travel distance, cycle time per order, per-AGV throughput, total system items/hour, AGV utilization percentage, and density ratio (AGVs per 1000 m²).

Worked Example

E-commerce fulfillment warehouse: 5000 m² facility, 12 AGVs, 1.5 m/s speed, 600 items/hour pick rate. Simulation returns: avg travel distance 78 m, cycle time 52 seconds, 5 orders/AGV/hour throughput, 12,000 total items/hour system output, 68% utilization, 2.4 AGV density. Increasing to 16 AGVs reduces cycle time to 38 seconds and pushes throughput to 16,000 items/hour but lowers per-unit utilization to 51%.

Practical Notes

  1. Cold storage (–20°C) reduces AGV speed 15–20%; set speed to 1.1 m/s vs. 1.5 m/s for ambient facilities to reflect ice-formation drag and safety margins.
  2. Cross-dock operations with 24-hour dock cycles benefit from 15+ AGVs in 10,000 m²; below 10 units, bottleneck occurs at conveyor handoff, capping throughput regardless of pick rate.
  3. Manufacturing flow (machine-to-line delivery) typically requires lower AGV density (1.2/1000 m²) than e-commerce (2.5+/1000 m²) due to predictable routes and smaller pick lists per trip.