Industrial robots powered by AI improve manufacturing

Robotics and artificial intelligence (AI) are coming together and can enhance manufacturing by going beyond traditional routes. AI-enabled robotics can act more like humans and automate tasks. See AI-enabled robot video.

By Chris Vavra May 24, 2023
Courtesy: Chris Vavra, CFE Media and Technology

Robotics AI insights

  • AI-enabled robotics can help robots learn like humans, which could have a massive impact on how traditional manufacturing operates.
  • Robots are inflexible by design, but AI-enabled robotics that use sensors, data-driven computation and more can enhance their capabilities.

The combination of artificial intelligence (AI), machine learning (ML), deep learning (DL) and robotics are beginning to transition to industrial manufacturing facilities. These advanced robotic systems enable the automation of industrial processes that were impossible to automate up until now. A video of an AI-enabled robot shows how.

This can have a massive impact on how traditional manufacturing works, said Eugen Solowjow, head of research group, Siemens, in his presentation “Industrial Robotics AI and Its Applications” at Automate 2023 in Detroit.

Solowjow said traditional manufacturing has been very successful for a long time and with good reason. Robotics have played a key role in this automation evolution. The problem, he said, is robots cannot react to sensory environments and are not flexible.

Eugen Solowjow, head of research group, Siemens, in his presentation “Industrial Robotics AI and Its Applications” at Automate 2023 in Detroit. Courtesy: Chris Vavra, CFE Media and Technology

Eugen Solowjow, head of research group, Siemens, in his presentation “Industrial Robotics AI and Its Applications” at Automate 2023 in Detroit. Courtesy: Chris Vavra, CFE Media and Technology

“Classic industrial robotics is a very successful paradigm. However, for the next wave of robots, to unlock the next growth cycle, we will need to enable robots to deal with flexibility,” Solowjow said.

This is where AI-enabled robotics comes in. By combining AI, ML and DL into robotics, particularly through machine vision, robots can take the next step. “These technologies can help robotic systems that are able to automate automates generic tasks in unstructured and dynamically changing environments,” Solowjow said.

Robotic intelligence faces an interesting paradox. Hard problems, such as mastering chess, are easy for robots to achieve. Picking up an arbitrary object like a human would is not something a robot can perform very easily.

AI-enabled robotics comes in is the combination of robots in industrial automation applications and merging with AI, ML and DL to improve their flexibility. Courtesy: Chris Vavra, CFE Media and Technology

AI-enabled robotics comes in is the combination of robots in industrial automation applications and merging with AI, ML and DL to improve their flexibility. Courtesy: Chris Vavra, CFE Media and Technology

Achieving cognitive behavior in robots

Robots by themselves cannot learn human behavior, but through technology, they can achieve capabilities to do so. Solowjow said there are three ways manufacturers can do this.

  1. Perceive the environment using sensors, camera, torque sensor, tactile sensing and similar feedback.

  2. Data-driven computation. Situations need to be processed to something meaningful so the robot can compute the observation.

  3. Augmenting a robot arm with tool. Solowjow said a lot of what makes robots hard is the action and the process and being able to use the sensor and computation technology and integrating the automation systems.

More than anything, though, Solowjow said the barriers won’t come down until AI-enabled robotics are readily available. The democratization of AI-enabled robotics technology by encapsulating solutions for complex in easy-to-use software, he said, will help make robotics more accessible.

“They need to be offered in a way so they can be used for automation engineers in the field and not just the few experts,” he said.

Three robotics AI application examples

While AI-enabled robotics are not in the mainstream yet, Solowjow said there have been many examples of where they are being used and tested to help move the concept and make robots smarter. He cited three examples of ways where they are being used.

1. Piece picking. Solowjow said this is the most mature of the three examples, but even this has a long way to go. The goal is to have robots be able to pick up random objects without having any training or advanced programming. Like a human, it will just “know.”

“Grasping arbitrary objects can have a major impact,” he said. “Having a robot that wasn’t trained or developed to pick the object but can anyway is a major development. This is among the most basic concepts we teach ourselves.”

In keeping with the paradox theme at distribution centers and production facilities, piece picking is among the most expensive tasks and is mostly manual, but it is among the simplest human tasks. AI-enabled robotics can help change that. See video example.

2. Apparel manufacturing. Automation has transformed many industries including automotive and electronics, but apparel manufacturing has been mostly manual. Some difficult-to-automate industries left the United States and Europe some time ago because of lower labor costs elsewhere.

Solowjow said AI-enabled techniques such as vision and tactile sensing can help automate some of the tasks, but the challenge is for robots to provide a worthwhile return on investment (ROI) to make it pay off.

“After all,” he said, “no one wants to pay $200 for a t-shirt.”

3. Industrial assembly. AI-enabled robotics can become smarter and more efficient through reinforcement learning for insertions and assembly. Like a dog, the programmer or user can relay information to the robot when it performs a task correctly. The robot will pick up on this information and be more consistent when it does the task the next time.

The downside is it isn’t easy to apply those techniques, which is why democratizing those skills to engineers and programmers who are not experts is key. The simpler the work behind the scenes, the bigger the payoff.

Solowjow is optimistic about the future of AI-enabled robotics.

“With further advances in ML and reduced hardware costs, we will see further applications, but it is a long journey.”

Chris Vavra, web content manager, CFE Media and Technology, cvavra@cfemedia.com.


Author Bio: Chris Vavra is senior editor for WTWH Media LLC.