For decades, the story of robotics has been the same: You need high-volume production runs to see any return. In many areas, robotics has remained the game of large-scale manufacturers only, leaving any shop producing a variety of SKUs out in the cold.

Manufacturers with high product mixes have avoided using robots in many applications. The primary reason for this is not the cost of the robot itself, but the cost of adapting the environment for the robot (fixturing) and the cost of programming it.

Most high-mix environments are constantly changing. Movable fixtures, objects incorrectly hung on a conveyor belt, or any laterally placed device can render a traditionally programmed robot completely out of reach. A $50,000 robot might require $150,000 or more in integration costs for just one task, part, or operation. If the application requires complex part or batch changes, the costs can skyrocket.

This hasn’t stopped manufacturers from doing the work, but the cost of robotics, including programming costs, must come down rapidly as skilled labor faces ongoing challenges. Most robots are positioned on six axes and require programming instructions. This includes the position and mechanical limits of each robot joint. In most cases, instructions must be precise to the millimeter.

However, with artificial intelligence (AI), the cost-benefit equation for automation could change dramatically for high-mix manufacturers. That’s because some robots are now effectively programming themselves.

“What” vs. “How”

Let’s say you want a colleague to get you coffee on their way back from a break. For a human, it’s incredibly easy. But for a robot, you need specific instructions: “Walk through this door, go down the 90-foot hall, walk for a few minutes, and then return, get a cup of coffee from the far left side of the counter on the other side of the hall, walk back down the 90-foot hall, re-enter the workspace, and place the coffee in my hand.” And, if someone has moved the coffee machine since you last saw it, the operation will fail.

In recent years, manufacturers have programmed robots in one of three ways: sequential programming using a teach pendant, kinematic manipulation, and offline programming. Advances in teach pendants have attempted to make robot programming more intuitive. With kinematic manipulation, the operator can physically position the robot and record the positions, making things somewhat easier. While the program does need to account for speed and the robot’s mechanical limits, a path is derived from this. Offline programming and simulation can save shop-floor programming time, but operators spend time verifying, rewriting, and modifying programs to address the differences between the simulation environment and the real-world environment.

In all these cases, the programmer is telling the robot “how” to perform its task—like the detailed coffee-fetching instructions. Self-programming robotics takes a fundamentally different approach. There’s no need to tell them how. Instead, you only need to tell them “what” to do.

What are self-programming robots?

A self-programming robot automatically generates a task plan. It senses the environment, understands process constraints to the extent possible, and then executes the task without human intervention.

Unlike traditional industrial automation, self-programming robots are given a specific “goal,” such as a task description like “paint this surface.” The AI considers the robot’s specifications and limitations, such as the speed of allowed movement and the location of the object being handled; then, it creates an efficient plan for each part in real time.

A robot powder coating aerospace parts, with a simulation of the same process shown beside it—all automatically programmed in real time. Photo courtesy of Omnirobotic

A self-programming robot uses 3D vision to perceive various aspects in front of the object, such as the object’s shape, size, and relative position, in the same way a human eye would. The robot correlates all this information with the assigned task, then uses AI-based algorithms to interpret the information and figure out the most efficient way to achieve the goal. This might include situations where objects are moving, such as sheet metal parts hooked on a continuously moving conveyor line.

As the robot generates its own program for the task, it uses information from 3D vision (a “digital twin” of the environment) to verify each part of the plan. This AI operates similarly to how a skilled operator uses their eyes and brain. It adapts to changes in the position, orientation, and shape of the object in front of it.

The Potential of Coating Processes

High-mix manufacturers can deploy self-programming robots in a variety of operations. For example, these robots are now being used to coat aerospace parts.

Generally speaking, coating processes are the fastest route to market for self-programming robotics. Powder coating, liquid spraying, shot peening, sandblasting, thermal spraying, and dispensing are all applicable, primarily because visual sensors and robot precision can be relied upon within a few millimeters. For powder coating and liquid spraying in particular, self-programming robots are being deployed first because the need for a solution is so high.

For industrial painting and powder coating, the challenges are threefold: How to achieve high-quality output, how to limit costly rework, and how to protect the health and safety of workers? As the work progresses, workers not only become fatigued, but their vision and accuracy decline, and the consistency of the work suffers. Rework costs can be high, sometimes enough to make an entire batch unprofitable. Failure to follow proper safety and ergonomic guidelines can lead to respiratory illnesses, orthopedic challenges, and chronic pain with prolonged use in some industrial painting and powder coating environments.

With self-programming robots, once the vision sensors are set up, the precision will not degrade below its calibrated value as long as it receives regular maintenance, which is no more than the limited maintenance required by most industrial robots. Robots don’t need to wear masks or use air filtration systems, and they don’t develop the ailments that plague many skilled workers over time.

Today’s sensors can overcome challenges with lighting and reflectivity, and robots can operate at high line speeds and with parts placed in any orientation. Self-programming robots use heuristic methods to determine the amount of coating they spray and the coverage produced. The goal is to achieve as consistent a finish as possible with minimal powder waste.

Liquid painting does add some extra complexity. For example, if the robot pauses or repeatedly passes over a spot, the system must account for the risk of paint dripping, which ultimately becomes an added variable. However, once this variable is accounted for, the robot can achieve high consistency, precision, and repeatability.

Beyond spray-based applications, there’s no limit to what self-programming robots can do. There’s more work to be done. Sensors, fixtures, and other components must become cost-effective, and manufacturers need to be able to deploy these robotic systems quickly.

For example, contact processes such as welding, grinding, and polishing require higher precision than spray processes such as painting or powder coating. This simply means that other solutions need to be designed. Even so, the principle of self-programming robots remains.

Powder coating is a mature application for automation because it involves placing workers in environments that, after years of exposure, can cause health problems. Getty Images

AI for High-Volume/Low-Volume Production

By bringing AI into play, manufacturers can automate many of the most repetitive, lowest-engagement tasks that still require human-level skill, knowledge, and know-how. Paint and powder coating are just the first of many examples.

No one benefits from suppressing automation, but for its own sake, automation isn't necessarily cost-effective. The first step is to identify and validate whether the most tedious, repetitive, and dangerous jobs in an operation can actually be automated. By using robots and providing solutions in this context, manufacturers can grow their automation efforts to include entire cells and production lines. As this transition happens, companies will see where existing employees can fill gaps, contribute their ideas, and help produce better results.

In all of this, self-programming robotics can play a key role. This new type of automation will help free up resources and allow manufacturers to produce more efficiently, all without having to redesign facilities to install robots in the first place.