How Many Robots Do You Actually Need?

Munro Footwear Group in Melbourne runs 205 Mushiny T-series robots against 5,239 shelf pods. Alloys runs 5 T6-800 robots against 100 pods. Each number is right for the site it sits in, and neither tells you what your count should be. That is the first thing to know about fleet sizing: the number of robots is an output. You do not pick it. It falls out of three other numbers, and this article shows you how to find them.

Every site in this article is on the reference sites page, and most of them run shelf-to-person.

Why is the robot count an output and not an input?

A robot does not produce anything on its own. It carries a pod to a picker, and the picker produces the lines. The fleet exists to keep the stations busy, so the question is how much work the stations need to do, and the robots follow from that.

Three inputs decide it:

1. Lines per hour at your peak hour.
2. Lines you get out of each pod presentation.
3. Cycle time for one trip.

Work those out and the fleet size is arithmetic. Skip them and any robot count you hear, including ours, is a guess.

What is your peak-hour line rate, and how do you get it?

It is the lines per hour at the busiest hour you have decided to serve. It is not daily volume and it is not the daily average.

A system sized on the average runs out of capacity at the moment orders are due out, and the queue appears at peak. So the peak hour sets the system, and the quiet hours only decide how much of it sits idle.

Get it from your order history. Take every order line with its time stamp, count the lines in each hour, and find the busiest ones. Count lines, not orders and not units, because a picker handles a line. Then make a decision: do you size for the single worst hour of the year, or for the hard hour you expect to see regularly? That is a trade between cost and service level, and it belongs to you.

How many lines do you get from one pod presentation?

This is the input nobody thinks about, and it moves the answer more than the robot count does.

A presentation is one trip. A robot delivers a pod, the picker works from it, and the robot takes it away. If the picker takes one line from that pod, the trip served one line. If the picker takes five lines for five different orders, the same trip, the same robot time and the same station time served five.

Batch more orders into one pass and the same fleet does far more work. Double the lines per presentation and you halve the presentations you need for the same peak hour. The ceiling is set by your stock and your orders: how many open orders want something from the same pod at the same time. The groundwork is in batch picking, including how to measure the overlap in your own orders.

Station rates make the effect concrete. One MIX 4.0 workstation presents up to 240 totes an hour. That is a ceiling on presentations, not on lines. The lines a station can support in an hour is up to 240 multiplied by the lines you get from each presentation.

What is cycle time, and why does a bigger grid need more robots?

Cycle time is how long one presentation takes a robot, start to finish. It has four parts: travel from the pod to the station, waiting in the queue at the station, the lift and lower, and travel back.

Every second you add to a trip lowers the trips a robot can complete in an hour, so you need more robots to deliver the same presentations. A bigger grid does exactly that. The average distance between a pod and a station is longer, so every trip takes longer. A bigger grid needs more robots for the same throughput, and the extra floor you add shows up again as fleet.

Congestion is the other piece. Mushiny's grid robots navigate by QR markers on the floor, inside a fenced area where people do not go. A marker tells a robot where it is, not where to go. Each robot still works out its own route and negotiates with the robots around it, so more robots in the same grid means more of that negotiation. How much it costs in cycle time depends on the layout and the order profile.

Cycle time, and so trips per hour, comes out of simulation on your own layout and order profile. A single published figure would not mean much, because the travel is set by the size and shape of your grid.

Are fleet size, station count and grid size three decisions or one?

One. You cannot fix two of them and solve for the third.

• More robots with too few stations gives you a queue at the station. The extra fleet buys nothing.
• More stations with too few robots gives you an idle picker, waiting for the next pod.
• A larger grid with the same stations and the same robots makes every trip longer, so throughput drops.

The same applies to pods and robots. Enough pods and too few robots means orders queue. Enough robots and too few pods means you run out of places to put stock. Size on only one and the system disappoints.

How do you get a first estimate?

Work the arithmetic in this order.

1. Take your peak-hour lines.
2. Divide by the lines you expect per presentation. That is the presentations you need per hour.
3. Divide by what one station can present in an hour. For a MIX 4.0 workstation the published ceiling is 240 totes an hour. That is the minimum number of stations.
4. Multiply presentations per hour by the cycle time of one trip, and divide by an hour. That is the number of robots busy at any moment, before anything else is counted.

You can do steps 1 to 3 today with your own data and the published station rate. Step 4 needs a cycle time, and the right one depends on your grid. Charging and congestion come on top, and a simulation exists to measure them.

This gets you to an order of magnitude, not a quote. The real answer comes from running your own order file through a simulation of your own layout.

What do fleets at named sites look like?

This is what Mushiny's reference sites list, as published on the references page. A dash means the page gives no figure.

Divide pods by robots and the answer moves a long way: about 7 pods per robot at Coca-Cola, 13 at Winc, 20 at Alloys, 26 at Munro and 28 at Full Beauty Brands. That is arithmetic on the published numbers, and it makes the point of this article. If robot count followed pod count by a fixed ratio, those figures would sit close together. They do not, because each site has a different peak, a different number of lines per presentation and a different grid.

Find the row closest to your operation, then ask what is different about yours. Station counts are not listed for these sites, so a row is a place to compare yourself, not a ratio to borrow.

Do you have to buy the final fleet on day one?

No.

The fleet is the part of the system that scales most easily. You add robots and pods as volume grows, so you buy for the business you have now. Alloys with 5 robots and Munro with 205 run the same system, sized to the operation.

Plan the timing, though. Robots run five to six months from order to running on site, so growth has to be seen coming. Adding to the fleet is a decision made two quarters out, not a response to a busy month.

The grid and the station count are different. They set the footprint, the layout and the workstation positions, which makes them the expensive decisions to change. Spend your design effort there. Choose a grid and a station count that can carry the business you expect, start the fleet at what today's peak needs, and add robots as the peak grows.

To size your own, enquire and bring your order file.

Frequently asked questions