by Brian Sheehan Brian Sheehan

Logistics is in a constant, drastic, and largely unpredictable state of flux. Predictions for the future of logistics span from the apocalyptic hit robotics are believed to make on employment numbers, to optimistic predictions that advanced robotics will be a soothing force in both business and job growth. More than likely, the truth lies somewhere between the two extremes, but we can only observe and report events as they occur.

With that said, we’ll look at which advancements are breaking through now, what are some of the expected advancements to come in development, and the changes these are both already starting to make and expected to make in the industry at large. Here are methods that are used to reinvent order fulfillment operations.

Existing Robots in Logistics Operations

In 2012, the Seattle-based behemoth Amazon purchased Kiva Systems, now known as Amazon Robotics. They continued selling to other companies for a time but now serve as a robotic division exclusive to Amazon, producing machines to optimize their own distribution centers.

There are a few alternatives to Kiva Systems now, but by-and-large they serve the same purpose in similar ways. The challenge in implementing robotics in logistics has been multifold but has been minimized when approached with a new angle, just shrunk a bit.

Wired explains that at first robots were only effective at brute force tasks such as welding or lifting heavy objects. However, while some products like weightlifting barbells are tough and sturdy, you wouldn’t want the same machine to handle a package filled with delicate dishes. When you consider all of the different types of packages that are shipped from warehouses, it quickly becomes clear how impossible it is to program for all of the thousands of different products being moved, stored, and shipped. Factors that would need to be considered in how to handle packages include how the package can be held, how much force needs to be applied to keep a steady grip on the package, where to safely apply that force without breaking the package, and even if the package can safely be tilted one direction or another.

That’s exhaustingly impractical, if not outright impossible to work with.

Yet, to an extent that is what we have had for quite some time, which is precisely why robots have been relegated to repetitive brute force tasks.

The Next Generation of Robots

Machines are now able to take over the work that before only humans could do, such as careful and articulate manipulation of packages and quickly filling boxes. While we’re not totally there yet, we’re on our way thanks to advancements in machine learning. This is only one of the ways we can reinvent order fulfillment operations

Amazon’s Picking Challenge has a large role to play in facilitating these advancements. This is a competition that weds academic robotic research to practical application in the industry. Team Delft won the championship round of this competition, called RoboCup 2016. This entailed a combination of stowing and picking tasks.

The stowing task involved moving 12 target items from a tote into their respective bins on a shelf. The items were arranged in the tote in such as way that they covered one another up either entirely or partially, presenting the machine with dynamic challenges that it would encounter in a real working environment.

The picking task was to take 12 target items from a shelf and place them into a tote. Every bin included anywhere between one and 10 items. The total of the 12 bins contained 50 items. There was a single target item for each bin, and the target item had to be picked from said bin. As with the stowing task, several target items were either partially or completely obscured.

Keep in mind that, in addition to recognizing the target items and manipulating them and their environment to complete the task, the robot had to be able to apply the appropriate amount of pressure and the right manner of manipulation for different sized and shaped objects, some requiring a gentler touch than others.

To accomplish this, many companies are utilizing a process called robotic imitation learning, also referred to as Robot Learning from Demonstration (LfD), or Robot Programming by Demonstration (PbD). This describes the effort of enabling robots to autonomously adapt to and perform new tasks as opposed to needing their users/programmers to break down and manually reprogram the desired behavior for each new task.

Success in this type of programming is springing up in different places. One of the more impressive feats in this field is the DeepMind AI. First, they developed AlphaGo Zero, which required three days of playing against itself in order to build up the proficiency to beat the best human in the world at the game of Go. It’s successor, Alpha Zero, needed only eight hours.

This was all without any prior knowledge of the game. Just load up the AI with the game, let it experiment on its own, and within 8 hours it will have taught itself enough to beat the best players in the world. When the team decided to try their new AI at chess, it took only four hours of self-training and self-teaching to achieve this.

Utilizing this methodology of self-improving programming in logistics is called reinforcement learning. That is when the robot takes what it has already learned, and by measuring the results via trial and error, further hones the skill to a greater level of skill and accuracy.

The day is not far off when these technologies will reach a more practical level, and robots will no longer be limited to the tasks for which they can be pre-programmed. Rather, they can and will be programmed to figure it out, to analyze the situation and apply and combine already existing information into new “ideas” in order to address new situations.

This sort of abstract Cartesian methodology of combining known ideas into new, never before seen ideas is a fascinating prospect when you consider that it is one of, if not the defining, concept of creativity itself.

The Changing Workforce – The Worst Predictions

A pressing question that comes from all of this is to ask what that means for those who are currently doing these jobs. Will they just be replaced as has been feared in other industries? Will their positions simply change, or become less strenuous?

We can and will explore the predictions leaning towards both extremes, but with that said, speculation isn’t something that can be measured for a definitive answer. For that, we’ll just have to wait and see.

First, let’s speak to the elephant in the room; the perceived but uncertain threat robotics poses to the logistics industry, which has helped to employ nearly one million Americans in recent years. This has offset the loss in jobs that brick-and-mortar retail locations have lost to drop shipping and e-commerce.

The National Bureau of Economic Research reports that for every robot added to the workforce, there are 5.6 workers replaced. In addition, for every robot added per 1,000 human workers, there is a drop in wages of up to 5 percent. That’s plenty of reason for people who depend on the industry to fear for their jobs. Also, note that this is not counting the effect that AI will likely have on the workforce once it is wed to robotics in a practical setting.

According to the International Data Committee, this radical employment change will not be limited to labor jobs but extends to virtually any position when you consider the artificial creativity and abstract reconstruction discussed in AI advancements above.

Gizmodo speculates that one of the only reasons more jobs have not been lost already is “because there are relatively few robots in the US economy, the number of jobs lost due to robots has been limited so far (ranging between 360,000 and 670,000 jobs).” The only exception they found in their line of reasoning was in management.

The most extreme “doomsday” predictions that show a nearly complete replacement of the human workforce is relieved by a basic universal income. In the coming decades, we may very well be in a discussion of whether this should be implemented, and how to do so if it is found necessary.

Changing Jobs, Not Replacing Them

When the telegraph fell out of use, those who worked on them as a career didn’t remain a forever unemployed segment of the population; they went on to the radio that replaced it, they became writers, farmers, sales specialists, and more. Work wasn’t eliminated, it changed and it evolved.

CNN notes in a similar vein that 100 years ago, the United States horse population reached its peak at roughly 26 million. Stables have since been replaced with auto shops, and mechanics are performing oil changes daily. A new workforce and group of career options appeared.

There are some positions that require a human touch. You won’t have a robot teaching your kids how to play baseball anytime soon, and the same goes for musicians, nurses, and physicians. It can become a bit muddier looking at logistics and warehouses, however.

Thankfully, this is one field where we can report on the results of the already existing robotization of the industry.

Boxed is the poster child for automation optimism thanks to their recent move: 75 percent of the workforce at their single largest fulfillment center was replaced with machines, and not one person was left unemployed afterward. This was done to facilitate an increase in productivity of 600 percent.

The change they made was from human workers retrieving each package, to human workers telling robots which packages to retrieve. While Boxed reportedly spent several million retraining its employees, it’s made waves in positive feedback and respect for its brand image, in addition to an anticipated 600 percent increase in facility efficiency.

The United States Treasury Secretary Steven Mnuchin echoes the same optimistic perspective when he said that, “Technology has made the American worker more productive. In terms of artificial intelligence taking over American jobs, I think we’re so far away from that, that it’s not even on my radar screen.” Expanding upon that, he stated that there is a significant difference between cars being driven from point A to point B by a computer, and having “R2-D2” come into the office to do your job for you. If anything, it provides an incentive for learning new skills to move higher up on the career ladder.

Should more companies follow the example laid out by Boxed in providing training for its employees to work with robotics, this could be a significant gain for both big businesses and the general workforce, as everyone sees a greater return for their efforts.

Returning to the e-commerce giant Amazon, they have done similarly to Boxed, and arguably paved the way. Amazon has roughly 100,000 working robots in its fleet. And even with that expansion of an automatic labor force, they still continue to hire — almost insatiably.

Robots may be excellent at performing some of the tasks that humans currently perform, but a task does not constitute a job. Tractors may have replaced manual plowing in agriculture, but we still have farmers. The task has become more automated than it was in the past, but the job still exists, it has only changed.

At the End of the Rabbit Hole

What waits for us at the end of the robotic renaissance ‘rabbit hole’? Who knows? While the extreme and more parody reports predict distant robot uprisings branded with Arnold Schwarzenegger’s face, it’s much more likely we’re looking at additional compliments to automation and augmented reality efforts that do not replace workers but streamline their efforts. We are more likely looking at tasks that are taken over by machines, and transitioning jobs into managing and maintaining the robots performing the tasks.

In another fifty years, however, that may change and we will be back to the same questions as we look into the future. As stated above, speculation cannot be measured. So, the same as with every technological innovation, we will wait, adapt, and continue to improve the industry one step at a time.