Pickface challenge

LinkedIn +

Order picking is possibly the single most challenging task in the warehouse, and choosing the right technological solution is critical to the efficiency and accuracy of the operation. Sam Tulip investigates.

According to the Office of National Statistics, the UK has 416,000 workers in ‘elementary storage occupations’: a high proportion of those will be picking orders (automation specialist Schaeffer reckons that 90 per cent of warehouse final picks are still done by hand). Meanwhile, productivity across industry has been flatlining for a decade.
Distribution centres typically have to handle many more skus than hitherto. Consumer e-commerce requires the fulfilment of many, often small, orders with quirky profiles, requiring goods for a single order to be picked from a wider range of locations, combining full cases and single items, and tending to defeat sophisticated picking strategies.
These trends are not confined to retail supply: as businesses focus on squeezing inventory, B2B is also about the faster fulfilment of greater numbers of smaller orders – but full cases and pallets are still in the mix.
The traditional response would be to throw more people at the problem, but that is no longer easy. Unemployment rates are low, there are fewer 18 year olds entering the workforce, even without Brexit, reliance on foreign workers may not be sustainable, and factors such as the minimum wage and the apprentice levy are raising labour costs.
An added problem – the UK is predicted to run out of warehouse space by 2020.
Alaster Purchase, corporate marketing director at Zetes, says: “The industry needs to look to automation, artificial intelligence and robotics to drive further efficiencies.
To keep up with the rate of change and the fast-paced nature of supply chain operations, the use of the cloud must be embraced to ensure that the traditional order picking solutions, such as voice-picking software can work seamlessly alongside newer technology and more complex picking processes.
“The right solution will enable businesses to rapidly deploy new technology, increase the visibility of events affecting operations and empower staff on the picking line.
It also allows the flexibility required to adopt a suite of solutions and these can include voice (with mobile screens for back up), pick-to-light, put-to-light, RFID, augmented reality, artificial intelligence and more.”

I, robot?
Surely, robotic order picking is the answer? In principle, order picking is simple – select an item and move it from one box or location to another. Repeat. In practice, there is an almost infinite range of targets that vary significantly in shape, size, rigidity, weight, material, presentation and orientation which have to be identified, picked, manipulated and placed with both speed and precision. As it happens, human beings are quite good at this.
However, there are difficulties to overcome. A robot has to identify the target, which may be in a random orientation, so it needs a vision processing system (in addition to any ID system).
All the problems can be solved. Amazon, no slouch when it comes to applying technology, has for some years run a Robotic Order Picking Challenge.
In 2015 the winner, from the Technical University of Berlin, managed to successfully pick and place 10 out of a possible 12 items in 20 minutes (or 30/hour). The 2016 winner (TU Delft) managed 100/hour but with a 16.7 per cent failure rate. Amazon would expect a human to manage around 400 successful picks in an hour.
The US start-up RightHand Robotics has systems in pilot programmes, enhanced by machine learning, that claim a pick rate of 500-600 an hour.
However, the firm admits that the robots are used mostly to pick items from cartons and that more complex tasks such as nestling items into a best fit in a box still need humans. The corporate video does look more ‘pick and drop’ than ‘pick and place’.
Progress is being made – Ocado has prototypes in its grocery customer fulfilment centres of robots that can pick items in any orientation in a mixed crate and place into a customer delivery box.
The system uses suction (so has limitations)and the system uses artificial intelligence and 3D vision to identify items and plan its approach, understanding the optimal grasp points of the items. Additionally it must search for free space in the delivery crate. Sensors avoid the risk of crushing the item and it also knows how not to crush items already in the delivery crate.
Items are only released for picking if they can be packed without protruding from the crate. Performance figures are not available and Ocado describes all this as ‘very challenging’.
There are of course special situations where robotic picking is used successfully – assembling kits of parts in automotive, for example, or situations where all products are boxed in a relatively limited range of shapes and sizes.
In such highly structured situations the major robotics companies like ABB or Hitachi will quite happily supply solutions based on standard industrial robots, but robotics for the less structured pick face still needs work.
Man riding
If we can’t yet replace human order pickers, can we use technology to improve productivity? Says Andrew Kirkwood, senior vice president and general manager – supply chain Execution at JDA: “The significant benefits of automation and robotics aren’t currently in picking – they are in everything else – moving goods, receiving, put-away, replenishment, sortation, perhaps packing”. The priority must be to reduce the estimated 60 per cent of a picker’s time that is spent walking around.
Where volumes allow, picking can be effectively supported by fixed automation in storage/retrieval, conveyor and sortation. But in many situations the ‘optimum’ is ever-changing and reconfiguring fixed automation is slow, expensive and disruptive, while a significant human element is often still required to balance flows and to pick exceptions.
And whereas if a truck (or a worker) fails it can be dragged out of the way, a failed conveyor system can mean a total shutdown. Flexibility is key.
As Steve Richmond, director, logistics systems, at Jungheinrich UK, points out, there are two approaches to the movement problem: bring the goods to the picker, or take the picker to the goods. In the latter case, man-riding order-picking trucks are commonplace and can take the form of automated guided vehicles (AGVs).
Mike Hawkins, head of logistics solutions at Linde Material Handling, notes that increasing internet orders mean greater demand for picking on two levels. “For multi-level picking applications, we offer order pickers with pick heights up to 1.2m or 2.4m as standard. These trucks can also connect to customers’ WMS Systems to give semi-automated navigated trucks helping make picking more efficient.
“Meanwhile the Optipick ground level order picker is connected via Bluetooth to a wrist band which allows the truck to move to the next pick location without the worker having to step back onto the truck. This system allows the worker to add products onto the truck quickly and efficiently.”
For greater flexibility, the AGV can become an intelligent AMR – Autonomous Mobile Robot. Freed from the constraints of fixed guidance paths, with live ‘maps’ of the warehouse and intelligence of its environment, an AMR can consistently deliver the picker to a location by the fastest current route.
Ditching the human picker and bolting on robotic picking technology would be the obvious next step but, for the reasons above, “this is some way off being a robust technology”.

Move goods, not people
A bigger impact on the people movement problem comes through goods to person strategies. Carousel-type solutions for small parts are familiar; at the other end of the spectrum are ‘heavy engineering’ solutions combining for example automated stacker cranes and conveyors to deliver a sequence of totes past the pick station, although AGVs may be preferred to conveyors.
Harry Chana, head of automotive automation at Daifuku, says that: “thanks to increasingly sophisticated technology, AGVs have moved up the pecking order and are now used within the picking and packing function, typically in FMCG”.
In automotive too, firms are increasingly choosing AGVs over conveyors as costs have come down.
“Our new generation of AGVs are simple to reprogram so any changes in production can be accommodated literally overnight [they use removable magnetic strips rather than embedded guidance wires]. Unlike traditional conveyor systems AGVs can be individually monitored so that potential performance issues can be spotted early”.
However, goods to person is where robotics is really coming into its own. We are all familiar with Amazon’s extensive use of Kiva robots (to adapt a slogan, ‘they liked the product so much they bought the company’) to slide under towers of shelving (themselves allowing great flexibility in how floor space is used) and move them to and from the pick station.
There are many variants on the theme. For example the Skypod system by French start-up Exotec Solutions features freely navigating robot carriers and a special narrow aisle racking system which allows the robot to ‘climb’ the racking to retrieve totes at heights of up to 10m. A fleet currently in use at the Bordeaux warehouse of e-commerce company Cdiscount is retrieving totes at a rate of 400-500 per hour.
Also attracting attention are ‘grid and bin’ systems. Swisslog’s AutoStore uses robots running on a grid above stacked bins to quickly process small parts orders.
Over time, the system automatically learns which products have a higher rotation, storing them on the top layer to ensure faster picking times. Ocado has a similar system.
One key feature of systems using such AMR robots is that capacity problems can be addressed simply by adding more bots to the fleet.
Additionally, an AMR system allows almost total flexibility in warehouse layout and picking strategy, because configuration is all through the software, not dictated by the hardware.

Not what you’ve got: how you manage it
Indeed, the rapid development of WMS capabilities is both enabling and driving automation. Tony Dobson, managing director at Snapfulfil, says: “Modern, best of breed WMS are always looking for ways to make picking more efficient – in manual picking that’ll look a lot like hybrid approaches: think batches with cellular picking, or wave management with subsets of batches.”
Although as Dobson says: “All the same things that help a manager determine where to put their people are still considerations for robotics”.
Eric Carter, solutions architect at Indigo Software, says: “Most of the warehouse managers we work with have a highly strategic role, but the daily reality of their working lives is somewhat different and many spend a large chunk of their day on firefighting activities.
“Artificial intelligence will change all this because the technical capability of the machines and the changes taking place in the warehouse, will enable warehouse managers to step back from operational work and focus on doing what they were actually employed for in the first place. It will mean that data management and analytics skills will be far more in demand in the future”.
Daryl Fisher, product director at WCS, agrees: “Management, supervision and planning will have to be much more skilled than before in support of the technologies, and flexing the systems to meet changing situations”.
Richmond adds: “WMS technology and material flow control (MFC) has come on massively in the last five years – for example, the ability to create sequencing logic for a fleet of 200 bots.
“MFC has to recognise that a load is being built with product from different areas using different methods, and has to sequence and optimise pick, pack and despatch.
“We can now analyse in detail pick rates, response times and much more – and how do you justify investment whether in voice picking or in full automation if you haven’t got benchmarks?”
Says JDA’s Kirkwood: “We still need to consider efficiency as if the robot was a person. Empty running is still wasted time – the same concepts need to be considered. You still have to avoid having all your pickers in the same location.
“The WMS is still relevant, solving the same problems. But now with cloud computing we can use mass computing to run optimisation algorithms and not just within the four walls of the warehouse: we can use artificial intelligence to predict demand and what requirements that will place on the warehouse, we can take edge data and the internet of things to replan and reschedule when things change.
“And this will be maximising the performance both of the automation and of the people, because there will still be people in the picking area for some time to come”.
Daryl Fisher at WCS points out that: “It used to be that you would do batch picking or wave picking. Now, especially with consumer business, orders come in but how they are picked may be dependent on the shipping option, so you have to have a very dynamic control.
“When an order comes in, the WMS has to look at the time it is required to go out so you need a much more flexible system for prioritising pick processes”.
So advanced WMS technologies are essential to improve productivity, regardless of whether people or machines are the agents.
Technology may augment human picking capability in other ways. In Japan, faced with an ageing population, exoskeletal devices are being developed for heavier lifting (one might note that in the US, the cost of personal injury litigation is often one of the drivers of automation).
The problem of nesting items efficiently was mentioned above: AI and special goggles can combine the computer’s ability to design a more efficient pack, and the human dexterity that can achieve it.
Voice and light direction are already commonplace but wearables may offer pickers further support in product identification and quality control.
Clearly there is no ‘one size fits all’ solution. In structured environments, it may be possible to automate or robotise the whole picking process.
Some pick lines may be susceptible to degrees of automation but always have some human presence working with the robots (and that mixed environment, as Jungheinrich’s Richmond points out, is now a lot safer – AGV/AMR vehicles know and sense their environment, and can stop a lot quicker).
AMR systems can increasingly give the flexibility of scale, layout and strategy previously only achievable by a human workforce. But simple aids for human pickers may be equally effective.
Jonathan Bellwood of PeopleVox says: “We are big fans of combining people and barcode scanning – for £200 you can equip someone with a wrist display and a finger scanner, bought off Amazon Prime and delivered in three hours”.
Or as Jungheinrich’s Richmond puts it “If the solution is a hand pallet truck on a mezzanine floor, so be it.”

This article first appeared in Logistics Manager, March 2018

Share this story: