Traditional integrators for robotic workcells experience significant difficulties trying to  implement industrial robots in high-mix environments. However, recent technological advancements are overcoming these challenges and getting robots into operations previously considered impossible. 

First, what do we mean by high-mix manufacturing? By some estimates, 75% of manufacturers in the US are considered high-mix operations, producing over 100 SKUs typically at quantities under 1000. The various parts might have different sizes, materials, or shapes. And robots hate variation. Typically, system integrators program them to do the same movements over and over again, very efficiently and very quickly.

Robot programming experts use a teach pendant, and every robot OEM has their own pendant interface and their own language, and they painstakingly program out the movements of that robot and other components in the workcell to achieve the project goals and cycle times. Depending on the complexity of the program, this can take days, weeks or even months of programming and testing. If the parts change, or the process changes, or the layout changes, chances are that they’re reprogramming something and keeping that system down while they do it.

And that’s really been the case for the past 50 years.

For high-mix operations, we hate everything about this. With a lot of different parts we need different programs and increased complexity of our system. We can’t be flexible and we can’t be nimble to handle our customers’ needs. And all that programming time and complexity leads to a terrible ROI. So we either have to carve out parts of our production that can fit into this small box of capabilities, or we just find other ways to get the work done as best we can.

The great news is that there’s a lot of technological advancement in key areas that are changing this bad ROI calculus. Let’s take a look at four ways this is all changing:

1. Collaborative Robots

Let’s start with collaborative robots, or cobots. Cobots have only been around for the past 15 years, with much of the popularity uptick occurring in the past 5 years. These are power and force limited systems that allow close proximity to humans without the need for fencing and other guards. That doesn’t mean fencing and barriers aren’t needed for certain applications. The end-of-arm-tooling (EOAT), such as a plasma cutter, might warrant the need for guards. But the robot itself is safe. That flexibility has made cobots really attractive, with companies like Universal Robots offering an easy-to-use experience and hand-guided programming.

In fact, cobot shipments are expected to grow 20-30% each year for the foreseeable future, with recent estimates in one report already outperforming their projections from last year. This growth is leading to more companies entering the market, which ultimately is providing manufacturers and fabricators with more options and features while keeping pricing in check.

This chart from Universal Robots offers a great comparison breakdown between the strengths of industrial robots and where cobots tend to shine. Not surprisingly, many of these also align quite well with the needs and limitations of high-mix manufacturing. Two key ones are outlined in red. It’s important to remember with cobots that you gain a lot of flexibility in terms of space utilization and repurposing to different processes, but you will sacrifice the speed and payload efficiencies of larger industrial robots. So keep that in mind when considering a cobot solution.

2. Simulation

Offline programming allows you to setup your system in a 3D simulation environment and import your parts and other components to program the robot. So instead of doing all this with the physical robot, in what’s called an online scenario, most of the programming can be done sitting in your office, without the hardware. And these tools also generate the robot program that you can then export to the robot controller. So that’s very powerful. It has been shown to dramatically speed up commissioning times, and even in high-mix scenarios it can be a very useful tool when integrated with your existing digital design workflows. Of course, these simulations have some limitations. They don’t have the advanced graphics and physics you see in modern game engines for example, so they won’t be able to help much with testing and training newer vision and force control technologies.

A much more recent advancement is the idea of photorealistic digital twins. Digital twins are used throughout the automation world and allow us to build a virtual representation of a machine or device and update it with real-time data to improve performance and future designs. NVIDIA has taken that concept and leveraged its extensive experience in graphics and AI tools to help engineers create what we see here, which are photorealistic digital twins. With this, companies can more easily test robot solutions that use cameras and touch. One of the exciting things here is that, since we’re concerned with high-mix manufacturing, this type of technology accelerates training new control algorithms with less real-world data. We can generate a bunch more data synthetically in a simulation environment and ultimately on the shop floor that translates to greater capabilities, flexibility and robustness for what the robot can do. 

We use these tools at Cohesive Robotics as a NVIDIA Inception partner, and while manufacturing automation engineers and operators probably won’t be using these tools directly, it’s important to know this exists, it’s maturing, and you may want to ask your vendors if and how they’re leveraging this technology to improve their offerings.

Scan and Plan

Just like it sounds, “scan & plan” uses a 3D camera sensor that is mounted on a robot or overlooking a work area, to scan an area, reconstruct the surfaces of the part or parts that it sees. Using that information along with all the process details you are trying to do (for example, polishing), it then plans the robot’s motions. And that planning happens in seconds or minutes, basically doing offline programming as part of this whole process. From there the robot program can get exported to the robot controller, or more sophisticated setups perform the robot control directly. This is just perfect for a high-mix environment. With this approach you don’t even need a CAD file, but it typically performs better with it. And it fits great with a batch workflow where you may have a bunch of different SKUs for a specific project that need to get all processed at the same time.

Since we are scanning a part’s surface, this technology is best suited for operations that happen on the surface, such as welding or finishing operations.

Here is a real-world example of one of our scan and plan setups, with the robot moving the camera around to perform the scan, and then after the operator selects options it plans and performs a polishing operation on a stainless steel part. We at Cohesive Robotics are excited about bringing these capabilities to smaller manufacturers with a pre-engineered turnkey solution.

How to prioritize automation goals?

Conduct an application audit

First, conduct an application audit, or update the one that you have. Here’s a good example, again from our friends at Universal Robots.

Review each of your processes and now consider the difficulty level for automation, anticipated payback, and the amount of pain it’s currently giving you. That’ll help prioritize your potential projects.

Determine your technology risk tolerance

Second, determine your organization’s risk tolerance level. Some of these technologies are newer than others, and so you might have to be an early adopter. With that comes high risk but also high reward. As an early adopter you can usually get better terms, and a partner that will provide a special kind of personal attention and dedication to your success. Or do you want to play it safe, but potentially lose a competitive edge to someone else?

Research and leverage partners and vendors

Finally, with both of these steps, be sure to research your potential partners and vendors, and leverage their expertise. This is an exciting time for robotics adoption in high-mix environments, and as the saying goes, a rising tide lifts all boats.