Robotics snapshot - interview with VP Dematic Robotics Centre of Excellence

This interview took place in early March, LogiMAT had been cancelled but it was still pre-coronavirus lockdown. We don’t quite know what the new normal will be post-pandemic, but it is likely many of the issues discussed will remain.

Growth

eCommerce was identified as a key motivator for the use of warehouse robotics.

“We all know the labour pool is shrinking and yet we are in an expanding market. Therefore, the primary focus is to alleviate the labour availability issue and allow clients to expand without increasing headcount,” says Crystal.

“Robotics can provide this. Further, a lot of robotics solutions are portable and modular, they don’t need big sub-systems to support them.”

But retail is changing fast as it grows, with volumes making returns handling more strategic. 

“In many cases, retailers consider returns are ‘not worth the effort’, and they are re-directed as job lots or destroyed, but increasingly retailers will have to learn how to handle them and robotics and automation plays a part in getting products back to the pick face quickly.”

Micro-fulfillment

One area that is booming due to the pandemic is e-grocery. This is typically a manual, add-on to supermarket operations, so micro-fulfillment gives these traditional retailers the option to deploy automation and potentially head off competition from the likes of Ocado and Amazon.

Crystal explains: “Robotics fits the way logistics is morphing, with an emphasis not just on huge DCs but also micro-fulfilment. We’re currently working on our engine for grocery. Dematic is deploying these solutions, but with human picking only at the moment. We’re adding the robotics component as it is available. It’s a little bit different for robotics in a grocery space as it is a small footprint, so the robot has to be able to handle all the SKUs in that scenario, or you have to have slightly bigger footprint, or multiple pick stations, and allow human and robot collaboration without compromising rates.

“Current micro-fulfillment opportunities are handling products into end use bags, and that provides more of a challenge on the robotics end, because it is no longer the picking function that’s the challenge, it’s the placing function as well, and the human has the dexterity for this.”

Challenges

Grasping products seems to be the greatest challenge at the moment, but another is the extent to which people should be deployed alongside robots, and the trade-offs that involves.

“The big challenge is on the grasping and handling side. If you’re not trying to manipulate the product after you handle it, that’s perfect for a robot,” says Crystal.

“If there is one or two items per order, a robot can beat a person, but when you get to 5 or 6 items per order, for example, in apparel, e.g. a number of shirts in plastic bags, a person can grab 5 or 6 things much faster than a robot.

“In terms of gripping, a lot of solutions are vacuum-based and they are hard to turn 90 degrees. Some graspers, if you need to re-orient a big cereal box, you can’t lay it down, you have to set it up, so if it comes laid down in the tote, how do you manipulate it to a standing position?

“You can get people to do the parts they are best at, and leave robots to do the bits they are best at. You need human solution intelligence, and you need a method for intervention because you can’t plan for every outcome to be automated.

“There are different ways of dealing with the robot errors. You can have a person remotely observing and telling the robot how to pick the item it is stuck on. Or you can have people alongside the robots, which will mean they are speed-limited. In all scenarios you have to assess the recovery component for when an error happens.”

Deep learning

A relatively new development is the extent to which automation can be assisted by AI and deep learning. 

Crystal says: “This is helpful in addressing a wide SKU range (2,000 to 100,000), and allows faster new product implementation, learning as it goes. It doesn’t have to be crafted to suit a product. The techniques are not new, what is new is that computational capacity is now so great you can carry out analysis dynamically.

“With the increasing maturity of the algorithms and open access to the data available, we are seeing AI not just on the picking front but along the entire process.

“You have to have something that can handle dynamics shifts in the fulfilment load. The warehouse changes on an hourly, daily, and seasonal basis -  it is much more dynamic than say automotive manufacturing.

“AI starts on a small subset, then adds, and builds up. Soon we will optimise picking, the bigger challenge is handling the product best, bearing in mind the downstream process.”

Finding a solution

Another key challenge is fitting the robotic system or automation installation into a coherent whole.

Crystal explains: “Dematic has the advantage of being able to look at it holistically, with the ability to produce fully automated warehouse solutions, as well as solutions involving AGVs, automated forklifts, and AMRs. We have both up and downstream solutions so we can map out the whole project. We will test and evaluate the efficiency gains that can be delivered. The key is to make it flexible and modular so that if the fulfilment requirement changes, you can mitigate that with operational processes unaffected.”