Warehouse automation technologies: what they are and where they work

When considering warehouse automation, the first step is understanding which technologies to even look at, given how your warehouse runs today and where it is heading.

This article covers the seven categories that matter for most UK and European operations: automated storage and retrieval systems, autonomous mobile robots, automated guided vehicles, conveyors, sortation, goods-to-person systems, and pick-to-light. For each one, we describe what it does, the operational profile it tends to suit, and the conditions under which it usually fails to pay back.

Where we give throughput ranges, SKU counts, or building requirements, these reflect broad industry consensus from sources like MHI and Gartner, and from vendor specifications that have been in the public domain long enough to be cross-checked. They are starting points for a conversation, not thresholds you should base capital decisions on. The right answer for your operation depends on your data (which we would of course be very happy to help you with).

Automated storage and retrieval systems (ASRS)

An ASRS is a fixed system of cranes, shuttles, or robotic arms that store and retrieve goods from high-density racking with no human entry into the storage cube. The category covers a wide span, from unit-load cranes handling full pallets in 40-metre-high aisles, to mini-load shuttles handling totes, to cube-based systems like AutoStore where robots run on a grid above stacked bins.

The appeal of an ASRS is space is being able to store up to four times the SKUs of a conventional racked warehouse in the same footprint, and in some configurations even more.

ASRS tends to suit operations with stable SKU profiles, high throughput on a predictable range of products, and either expensive land or a building they cannot extend. It works best when the product range does not churn aggressively, because the system is optimised around the products it was designed for. If your top 200 SKUs in two years’ time look nothing like your top 200 today, an ASRS is a harder business case.

The trade-off is that once installed the storage cube is the storage cube, reconfiguring it is slow and expensive. It tends not to suit operations with high SKU churn, irregular product dimensions that fall outside the system’s handling envelope, or short lease horizons where the payback period exceeds the time you will be in the building. Certain systems can be relocated but often transport exceeds the asset value, so it’s a solution that requires long-term thinking.

Autonomous mobile robots (AMRs)

AMRs are wheeled robots that move around the warehouse floor using onboard sensors and mapping, without fixed infrastructure like wires or magnetic strips. They carry goods from point to point, either following a person doing the picking, or moving racks and shelves to a stationary picker, or transporting totes between zones.

The defining feature is flexibility. Because AMRs navigate from a software map rather than fixed guidance, you can add robots, change routes, and reconfigure the layout without civil works. They are also incremental, meaning that you can start with five robots and scale to fifty as throughput grows, which is harder to do with most other automation.

AMRs tend to suit operations with variable or growing throughput, layouts that change with the season or the contract, and businesses that want to automate without committing to a fixed system they will live with for fifteen years. They work particularly well in e-commerce and 3PL environments where the next contract or the next channel might change the operational profile entirely.

They tend not to suit operations where the bottleneck is not transport but something else, like picking density or packing throughput. If your problem is that pickers can’t find SKUs fast enough, or that packing can’t keep up with what the pickers produce, AMRs will move the bottleneck rather than remove it.

Automated guided vehicles (AGVs)

AGVs are the older relatives of AMRs. They follow fixed paths defined by wires, magnetic tape, reflectors, or painted lines, and they have been moving pallets and goods around warehouses since the 1950s.

The distinction between AGV and AMR matters less than vendors sometimes claim. The practical difference is that AGVs are cheaper per vehicle, more proven, and less flexible. If you know exactly what route the vehicle needs to run, and that route won’t change, an AGV will often do the job for less money than an AMR. If the route might change, or you want the system to route around obstacles, AMRs are the better fit.

AGVs tend to suit high-volume, repetitive flows in stable layouts, such as moving pallets from production into a finished goods warehouse, or shuttling between fixed pick and pack zones. Because many modern warehouse design tends to put a premium on flexibility, the use of these machines has reduced in recent years.

Conveyors

Conveyors are the workhorse of warehouse automation (and the least glamorous category in this article). Belt, roller, chain, and overhead conveyors move goods along fixed paths, usually between zones that need to be connected reliably and at high volume.

The reason conveyors deserve attention in a piece about automation is that they are often the right answer when more expensive technology is the wrong one. A well-designed conveyor system, integrated with sortation and labelling, handles enormous throughput at low cost per unit. The hardware is mature, the failure modes are well understood, and the capex per metre of fixed flow is far below what mobile or robotic systems cost.

Conveyors tend to suit operations with high-volume, repetitive flows between fixed points. They tend not to suit operations where flows change frequently, where the product mix includes items that cannot be conveyed reliably, or where the warehouse footprint can’t accommodate the fixed runs without sacrificing usable space.

Sortation systems

Sortation systems take a mixed flow of goods and route each item to the correct destination, whether that is a packing station, a despatch lane, a vehicle loading bay, or a returns zone. Cross-belt sorters, tilt-tray sorters, sliding shoe sorters, and bomb-bay sorters all do variations of the same job at different speeds and for different item profiles.

Sortation is throughput-driven. The case for a sorter usually rests on volume above a threshold where manual sortation becomes a bottleneck, and below a threshold where the sorter still pays back against the alternative of running multiple manual lanes. The threshold varies by item type, peak ratios, and labour cost, so understanding the range at which these machines can thrive is a crucial analysis.

Sortation tends to suit operations with high item-level throughput, a clear set of destinations, and a peak that justifies the fixed investment. It tends not to suit operations where the destination logic changes frequently, or where volumes do not justify the capital and floor space the sorter consumes.

Goods-to-person systems

Goods-to-person describes a category of system rather than a single technology. The principle is that instead of the picker walking to the goods, the goods are brought to the picker by an automated system. The picker stays at a workstation and items arrive in a sequence the system determines.

Goods-to-person can be delivered by several underlying technologies. A shuttle-based ASRS feeding picking stations is goods-to-person. AMRs that carry racks to pickers (these are sometimes known as ‘Harry Potter racking’, as latent robot’s racking appears to levitate as it moves).

The reason the category is worth treating separately is that the operational logic is the same regardless of the hardware underneath.

The advantage of goods-to-person systems is pick rate. A picker at a workstation can typically pick at two to four times the rate of a picker walking the aisles, depending on the system and the product. The advantage is also that the picker stops walking, reducing fatigue and injury risk, and the workstation can be designed around the person rather than around the racking.

Goods-to-person tends to suit operations with a high proportion of small-item picking, particularly e-commerce and parts distribution. It tends not to suit operations dominated by full-case or full-pallet picking, where the goods-to-person logic does not apply, and it struggles with very large product ranges where the storage cube required to hold every SKU live in the system becomes uneconomic.

Pick-to-light and put-to-light

Pick-to-light is not automation in the same sense as the others on this list. There are no robots, no conveyors, no moving parts beyond a button and an indicator. A light tells the picker which location to pick from and how many to take, and the picker confirms with a button press.

It is on the list because it is often the right answer when an operation is reaching for something more expensive. Pick-to-light removes the most common cause of pick errors, which is misreading a paper pick list or a handheld screen, and it increases pick rate by removing the cognitive load of locating the next SKU. The capex is low, the implementation is fast, and the technology is mature.

Pick-to-light tends to suit operations where pick accuracy is a problem, where pickers handle a large number of SKUs in a small area, and can work really well combined with goods-to-person. It tends not to suit operations where the bottleneck is travel rather than pick decision, or where the product range is too large for every SKU to have a dedicated light.

What about humanoid robots?

Perhaps the category that gets the most press (and the least floor time) is humanoid robotics. Figure, Apptronik, Agility, Boston Dynamics, Tesla and several others have promoted demos of bipedal or near-bipedal robots picking, sorting, and moving goods in warehouse environments.

The appeal makes sense: warehouses are built for humans therefore a robot with a human form factor can, in principle, drop into the existing operation without the racking changes or building modifications that other automation requires. The same robot could pick, pack, load a trailer, and handle exceptions, switching tasks the way a human operative does.

The reality is that the technology is not (yet) ready for production deployment at scale. The demonstrations are real, but they are narrow. Cycle times are well below those of either humans or established automation. Humanoid robotics will likely become a serious category for warehouse operations at some point in the next decade. However, it’s not a category that should appear on a shortlist for an automation decision being made today.

Operations that are evaluating automation now should make the decision against the technologies that are mature and deployed, and revisit humanoid options when the cycle times, costs, and reliability figures support a credible case.

How to think about warehouse automation choice

The technologies above are not alternatives in a straight line. Most real warehouses that automate end up with a combination, because different parts of the operation have different problems.

The question is not which technology is best. It is which combination, in what sequence, and with what phasing gives the best return for your specific operations. Throughput, SKU profile, growth trajectory, and building constraints all factor into which solution will work best. It is a question that data answers, not a vendor brochure.

For help better understanding your supply chain and how automation might play a part in it, or for assistance optimising existing solutions, please get in touch.