Unlocking Future Logistics: How Edge Machine Learning Boosts Performance

In recent years, the logistics industry has seen a significant shift towards automation and digitalization. With the rise of e-commerce and the increasing demand for faster, more efficient delivery, logistics companies are turning to advanced technologies such as edge machine learning to optimize their operations.

Edge machine learning, a subset of machine learning, is proving to be a game-changer in the logistics industry, enabling real-time data analysis, prediction, and decision-making "at the edge" of the network.

Meeting pandemic challenges with edge machine learning innovation

During the challenging times of the pandemic, logistics played a huge role in meeting high demands for production. An important concern during this time was to reduce the number of human touch points - or the amount of times a package was physically touched. When COVID was widespread, we provided fixed-mount barcode readers and handheld barcode scanners to help our customers and keep operators safe. By collaborating with our customers, and providing this innovative technology for their systems, this helped us make a significant impact when it mattered most.

Insights on the power of edge machine learning in logistics

We talked to Bart Piela, Product Marketing Manager, Logistics Vision Solutions, to get his thoughts on how edge machine learning is transforming logistics and the key benefits it offers to the industry.

Bart has worked at Cognex for six years, and in logistics for over two years. He's worked with a variety of logistics customers including small businesses just starting on their automation journey, OEMs, and highly experienced industry leaders. Thanks to this experience, Bart knows the ins-and-outs of how logistics operates along with its importance in production.

How does edge machine learning help the logistics industry?

Bart: In factory automation, environments are often controlled. Repetitive products that always look the same, like toothpaste or mechanical parts, are manufactured in these spaces. But in logistics no box or package is going to look the same.

When looking at retail, ecommerce distribution, or even parcel post distribution, there are so many differences from package to package being inspected day by day, hour by hour, and season by season.

The logistics industry is also newer to automation. About 80% of all warehousing today doesn't have any automation. This presents a huge opportunity to provide automated solutions that could benefit countless logistics providers. A solution is just a turnkey, off the shell, pre-built, packaged system from Cognex that's easy to set up.

Imagine purchasing a highly expensive piece of equipment that can be intimidating to use, compared to relying on humans to do automation tasks. Some customers don't feel comfortable investing in highly expensive solutions, so they rely on humans to do it the same old way.

It's far less overwhelming to install a simple customized solution that is packaged all together to meet a customer's needs. Slow integration of automation and Edge AI technology into manufacturing helps to start with one step at a time - and especially helps customers who are just setting out on their automation journey.

That's where edge machine learning comes into play.

What is edge machine learning?

Edge machine learning is an inspection technology that easily handles the variations of what a box, polybag, or jiffy could look like - which is particularly relevant in logistics.

Thanks to pre-trained algorithms, edge machine learning needs only a few training images to "learn from," which makes setup extremely easy. It optimizes the process by rapidly learning and adapting to the production line's needs. No prior machine vision or programming knowledge is required, either. It's perfect for both beginners and vision experts.

At the end of the day, we support vendors in all sorts of unique ways with edge machine learning technology. Our users are getting much better performance, accuracy, and speed out of their visual systems than ever before.

What type of applications does edge machine learning apply to?

Bart: For our edge machine learning tools, there's no limit to what we envisioned. We're targeting applications such as item detection, item sortation, and process issue detection.

Item detection is applications where we're looking to detect the presence or absence of an item on a sorter. Item sortation is where we're differentiating items by characteristic. For example, differentiating a box from a polybag.

Process issue detection is a category of applications where we're trying to perform, for example, preventative maintenance or inspection for hygiene issues, items that are stuck, or conveyor jams.

cgnx_pdfLogistics Solutions Guide

A box moving through a scanning tunnel conveyor belt.

Who is this technology going to benefit the most?

Bart: With the In-Sight 2800 detector we see some benefits in all warehouse areas but especially in areas where the customer wants to add process intelligence and replace old or poorly performing hardware like rule-based machine vision systems.

We really see this playing out well in process sortation - where the customer is trying to maintain the best operational efficiency and throughput to fulfill orders.
I also think this product can change the game when it comes to sortation accuracy and sorter performance. This technology can deliver on strong ROI, because improving sortation accuracy reduces extra costs associated with reworking, damaged, or manually handled items.

It can handle problems that otherwise would be cost prohibitive and require a lot of expertise. But now with edge machine learning, people can get power with simplicity at an affordable cost, making it accessible to anyone.

How do you set up edge machine learning on the In-Sight 2800?

Bart: It takes three quick and easy steps to set up the In-Sight 2800. After you take the device out of the box and apply power:

Step 1: Optimize your image, with only one click, to enhance focus and exposure to capture a good image.

Step 2: Train with a few images by collecting examples of what your process or application looks like.

For example, with an item detection application, you will need to collect a couple images of what an empty tray looks like and what an occupied tray looks like. After that, you can train the system in seconds.

With an item sortation application, collect a couple of images of what a carton, jiff, poly bag looks like, and train those in seconds.

Step 3: Deploy, set up your discrete IO or your industrial communications, and you're off and running!

The innovative pre-trained algorithms built into edge machine learning make it easy, straightforward, and quick to set up within 10 to 15 minutes - without needing a vision specialist or extra hardware.

How has edge machine learning enhanced the logistics industry in today's world compared to what was relied on in the past?

Bart: Over five years ago, we used rule-based algorithms - hard to design and brittle to changing conditions, including wear and tear on the conveyance system, plus changes in process flow or product design. Any of those challenges would require a lot of support to redesign the algorithm.

Three years ago, we introduced deep learning solutions into the marketplace which were extremely powerful in coping with variation. As mentioned before, they were difficult to deploy and challenging to commission. Deep learning requires a lot of overhead support, outside maintenance, and time invested into labeling or training.

For many customers, there is a barrier to entry, but edge machine learning helps open doors, especially for beginners just starting out on their journey. It brings deep learning into a convenient package that doesn't require the overhead that deep learning needs.

It offers a very flexible solution that can quickly adapt to new situations, including quick image tuning, that requires little outside maintenance too.

Can this technology help reduce costs?

Edge machine learning is an affordable option for manufacturers that may not have resources to invest in more expensive solutions. The best part about edge machine learning is that it can be retrained in the field without an external vision specialist. This leads to a decrease in costs for maintenance, operators, overhead, and other ownership costs too.

Handheld barcode reader scanning barcodes on boxes.

Looking forward, what do you think the next innovation with Edge machine learning within logistics will be?

Bart: I think in the next iterations of edge machine learning, we're going to build on our existing tool set to tackle more challenging applications while maintaining the ease of use. We plan to take our existing edge machine learning tool set and inject them into solutions where we have not used vision inspection before.

An example of this would involve our barcode readers - very efficient at reading barcodes in their current state. But with the integration of edge machine learning tools, we can inject more intelligence into our barcode readers to find even more information, such as inspecting barcodes for label damage. Currently, our scanners can identify damaged barcodes but cannot tell how they were damaged. If we integrate edge machine learning tools, we can create algorithms to find damage and identify if a label is torn, misprinted, or smeared. This would add tremendous value for our customers as they work to fix the root problem.

We also aim to take the original concept of our products and expand their abilities with edge machine learning to create more sophisticated and advanced tools that can work in less controlled environments like logistics. For example, creating an EL counting or segmentation tool suitable for demanding logistics applications.

The future of edge machine learning in logistics

The integration of edge machine learning technology is impacting the logistics industry in huge ways, providing many benefits in accuracy, efficiency, and cost savings. Bart shed light on the innovative power of edge machine learning, highlighting its ability to handle defect variation in logistics and deliver reliable performance compared to deep learning solutions or traditional rule-based algorithms.

Key takeaways from Bart's insights include:

1. Edge machine learning technology is a game-changing solution for logistics - providing accuracy, efficiency, and flexibility.

2. Edge machine learning easily handles variation in changing environments for accurate detections and inspections.

3. The future of edge machine learning in logistics shows much promise - from taking on challenging applications to maintaining its ease of use - paving the way for further advancement.

The advancements of edge machine learning technology will continue to shift how logistics operations are conducted, offering a reliable solution that addresses ongoing challenges within the industry - continuing its rapid growth and innovation.

To learn how this innovative technology could solve your greatest logistics challenges, get a copy of our logistics guide here.

keyboard_arrow_rightIn-Sight 2800 data sheet

Tags:ID - 2D Printed,Vision - Gauge Measure,Vision - Guide Align,Vision - Inspection,Vision - Absence Presence,Vision - OCR OCV,Electronic Hardware,Food and Beverage,Medical Devices,Other,Retail Distribution,Automotive Parts,Consumer Products,Electric Vehicles,Logistics,Life Sciences,Pharmaceutical,Robotics,Automotive Vehicles,Machine Vision,AI,Barcode Readers,3D Vision Systems,2D Vision Systems,Vision Software,Edge Learning,Getting Started,Vision Sensors,Fixed-mount Barcode Readers,Deep Learning