Can Smart Vision Disrupt Barcode Technology?
Meet the Experts
⇨ RFID and barcodes are both used extensively for inventory management in warehouses but there are certain challenges associated with them.
⇨ Smart vision technologies, enabled by AIoT are increasingly being used in warehouse operations and may have the potential to replace barcode technology.
⇨ However, a lot of work still needs to happen to develop a smart vision enabled solution that can deliver everything that barcodes can, in the warehousing world.
In an age where effective and strategic management of supply chains has become so crucial that every hyperscaler has launched a supply chain solution on their cloud platform, what will eventually differentiate these solutions? If you look for the high-level features in these solutions, most of the features are shared across the supply chain platforms by the top three hyperscalers. So, in the current state, the differentiating aspect may not come from the solution itself but from how organizations that embrace these solutions leverage it to create differentiating capabilities. However, in the future, differentiation will come from enhanced solutions that will work in tandem with these platforms. An example is leveraging supply chain operations platforms in tandem with these planning and visibility platforms. And this is where advanced AI-enabled algorithms like deep learning algorithms can come into play. In this article, we will discuss an example of how deep learning can help transform identification and visibility processes in warehouse operations.
The genesis of this article is a research article I read on the Amazon science website. The article “How Amazon robotics is working on new ways to eliminate the need for barcodes” touches upon how smart identification technology, powered by deep learning, can help replace barcode technology. You can read the article here. If you are familiar with barcode technology, there are many aspects that you will start thinking about after reading the article, and you may argue that the technology, in its current form, will not be able to eliminate barcodes. Let us explore where the solution currently stands.
As identified in the article, specific challenges do exist when it comes to leveraging barcode technology in warehouses. Despite many attempts to automate the process as much as possible, barcode scanning, still in many warehouses or most warehouses, exists as a manual process. While different types of on-the-floor robotics have been leveraged and experimented with to automate the process, ranging from mobile robots to flying drones, the accuracy of these automaton attempts is unsatisfactory.
The article on the Amazon science website proposes leveraging smart identification or deep learning to develop a process, which they have labeled as “MMID(multi-modal identification), to replace barcodes. In simple terms, To understand how it works, you can imagine a camera placed above a conveyor belt scanning items on the belt or scanning items in a tote and identifying them. Multiple warehouses across the Amazon fulfillment network are experimenting with this technology.
From a data scientist’s perspective, you may start wondering how difficult it will be to train such an algorithm because of the sheer number of SKUs that Amazon warehouses carry. The answer to that is in the article. This algorithm is very much in sync with the other fulfillment systems leveraged in the Amazon warehouse. When a tote with a certain number of boxes of different items lands in front of the camera, the algorithm already knows which items are in the tote. When it does the matching exercise, the number of possible candidates it has to match an image with is significantly lower (only the items in the tote).
What I find beautiful in this solution is that multiple analytics and fulfillment systems exchange data, which the algorithm leverages. This concept ties in with a suggestion I have made in a few of my articles on linking algorithms. The core idea is to connect many planning systems and algorithms to create a complete end-to-end warehouse or manufacturing solution. Capabilities like MMID can also be tied with the visibility or the control tower aspect of the supply chain planning platform that Amazon has or connect something like this with a solution that organizations may already have. And this is precisely how algorithms can be leveraged in the applied sense, beyond research papers and academia, into an industrial context.
As you may have already imagined, there are many different challenges associated with where the technology is in its current form. While it can identify what item the image belongs to, a typical barcode carries more comprehensive information. So obviously, this is still a work in progress since the goal is to leverage this technology to replace barcodes. There is no doubt that this technology if it can be perfected to a point where it can capture all the aspects that a barcode can capture, will eliminate a significant amount of waste in many different parts of warehousing processes. Though, the challenges that need to be addressed to get the solution to that point are not very straightforward.
An example is the unique serial number for each item captured in a barcode. So, while the camera might look at a package and identify it as a particular SKU ID, say a computer mouse, each mouse in that batch may have a unique device serial number assigned to each product and captured on a barcode. So how are you going to make MMID capture that kind of information? There are a few ways to do that, and I’m pretty sure that Amazon engineers are already experimenting. This is just one example, and many more aspects need to be incorporated to fulfill the vision of this technology replacing barcodes.
It is not rocket science that when we have perfected this technology to a point where it can indeed replace barcodes, that evolution will happen in tandem with the evolution of warehouse layout designs, storage rack designs, and product packaging designs. While a camera placed above conveyor belts can capture products flowing in the warehouse, tracking inventory stored is also critical. And if we intend to use smart cameras for that, there need to be associated changes in warehouse layout design, racking systems, and product primary packaging design. The gist is that much work still needs to be done to take this to a level that can disrupt barcode technology.
The role of barcodes in the world of supply chain tracking and visibility has been significant. While technologies like MMID may show the potential to disrupt the domain of barcode technology, a significant amount of work still needs to be done to make the potential of disruption a reality. Barcode technology companies need to explore the enhancements they can make to this technology that will make it difficult to replace by technologies like MMID.