Inside Logistics

Materials handling column: Robotic automation

Can robots do picking?


September 13, 2013
by Dave Luton

The impending demise of humans and their replacement by robots has long been rumored in warehouse operations, but the actual implementation has taken much longer than many robotics advocates have anticipated. One of the problems in devising a universal robotic material handling solution is the diversity of functions handled in an individual warehouse.

It’s impractical to examine all of the various functions and their constraints in a single page so we will look only at one function. Order selection is commonly recognized as being among the most labour-intensive of material handling operations, so we will focus on that function. To further simplify, we will deal with the traditional type of order selection that can be found in a corrugated case operation. The permutations and combinations to be found in a hardware or automotive aftermarket parts warehouse are too complex to cover here.

We will assume a limited number of SKUs, organized in the traditional manner, with fixed ground-level picking and a random unit-load reserve above, with an average of 40 cases per unit-load. Even in this type of warehouse operation there are a large number of different potential order-picking operations. Traditional types of order selection includes:

  • Full unit-load selection
  • Partial pallet (half- to three quarters of a pallet)
  • Layer picking (one quarter- to one half-pallet)
  • Case selection
  • Less-than-full case selection

It is not uncommon for one SKU to be picked by more than one picking method for an individual order. This outlines the first issue in designing a robotic picking operation because the order components must be divided among different operations and selection areas.

For a smaller operation, if one is willing to sacrifice efficiency, different types of picking operations can be combined. Case selection can also handle layer and partial pallet selection, at a loss in productivity.

The problem is as the number of cases per pallet selected increases, the inefficiency also extends to other support warehouse functions. Traditionally, full and partial pallet picking functions are selected from upper level reserve pallet locations because selecting them from a pick slot depletes the pick location too quickly and triggers frequent replenishments.

For handling unit loads, old-fashioned wooden pallets still retain their dominance, but there are a number of palletized and non palletized substitutes. Fork lift attachments form half of what is commonly referred to as palletless material handling systems. There are three main unit-load attachments for the most common palletless systems—basiloid attachments for top lift handling systems; clamp attachment systems which work through the application of side pressure (like a bear hug), push-pull attachments for bottom supported systems and used with slip sheets. Each requires different handling characteristics and skill sets.

Our poor robot has to drive a unit-load handling vehicle (of which there are several types) or at least interface with it. For full-pallet picks the robot can proceed directly to the loading dock, while partial-pallet picks force the robot to remove some pallets and place them elsewhere—logically in the pick slot.

To solve the unit-load problem we can have a robot drive the vehicle or have a driverless vehicle interface with a case-selecting robot. This range of skills is likely too challenging for one robot type to handle so now we must consider multiple types of robots.

At the other end of the case selection spectrum, less-than-full case operations (sometimes called broken-case operations) can be as complex to design. Consider a pick-and-pack operation for only a few SKUs—one that is organized so that (potentially multiple) fixed workstation robots can handle the function. For this to succeed, each work cell has to be carefully organized and instead of bringing the picker to the stock, the stock has to be brought and presented to the (robotic) picker. (Think of a carousel or flow shelving.) Exterior packing must be removed first.

Additionally, the correct number and size of empty packing cases have to be selected and the bottoms sealed. At the pick face the robot must select each item correctly and place it in the correct part of the case. The placing orientation must consider the three dimensions of the case (length, width and height) and allow for different case sizes and different types of items to be handled. This can be a complex logic to program into the picking robot because there are usually several layers of product in a case.

Multiple sizes and types of consumer packaging can greatly complicate the function. Next time you are in a supermarket look at the cosmetics or toiletries section and you will see what I mean. The simplest function is items of a similar size and type, such as cigarette cartons.

While the problems are complex, they are not unsolvable, and robotic automation is creeping into parts of warehouse operations. In future it will continue to grow, but good old Homo sapiens can still count on their versatility to make replacement by silicon some way off.

Dave Luton is a consultant in the greater Toronto area.