How AI is making industrial robots smarter

In the future, machine learning and AI will enable robots to self-learn and self-adjust, enabling improved performance. Through machine learning and AI, we see opportunities to further develop human-robot collaboration and to make robots more autonomous, within set parameters.

The robot should be self-learning or self-optimizing in the long term. It's not about copying human abilities. We want to enable robots to work in unstructured environments. For this, it is essential that the robots recognize specific patterns, for example, labels on bottles, and allow them to correct errors independently. In the future, robots can learn new tasks from other robots as well.

Swaminathan Ramamurthy, GM of OMRON Automation Centre & Robotics at OMRON Asia Pacific, listed how AI makes industrial robots more efficient and smarter.
 

1. Ability to sense and respond: The AI-enabled robots are well equipped with "Part Agnostic" know-how. They can pick objects for which there is no predetermined trajectory and a specified spatial location in the working space. Machine learning makes them do so. This also leads to a better and informed incoming parts inspection during an assembly process.
2. Ability to move around autonomously: The modern robots equipped with AI are more autonomous and precise in their navigation around complex and unpredictable environments. They can sense the obstacles in their path and re-plan the route accordingly. 
3. Ability to adapt to the changes around: Embedded AI in the servo controller helps robots adapt to changes in environments and payloads seamlessly.
4. Optimization of processes: With AI, manufacturers can now take full control of repair costs and breakdowns by gaining access to real-time data from sensors delivering much better reliability. 

Expanding possibilities

Subrata Karmakar, President of Robotics and Discrete Automation at ABB India, said that his company already offers several applications that combine robotics and machine learning. These include using AI to enable robots to sense and respond to their environment, inspect and analyze defects, and optimize processes autonomously. For instance, robots equipped with vision sensors can use AI to identify objects regardless of their position.

In autonomous process optimization, solutions like ABB's robotic paint atomizer enable real-time smart diagnostics and paint quality optimization. By monitoring the condition of critical variables such as acceleration, pressure, vibration, and temperature, the atomizer reduces internal waste during color changes by 75 percent. It reduces compressed air consumption by 20 percent.

"ABB and Silicon Valley AI start-up Covariant have a partnership to bring AI-enabled robotics solutions to market, starting with a fully autonomous warehouse order fulfillment solution," Karmakar continued. "Today, warehouse operations are labor-intensive, and the industry struggles to find and retain employees for picking and packing. While robots are ideally suited to repetitive tasks, they lacked the intelligence to identify and handle tens of thousands of constantly changing products in a typical dynamic warehouse operation. ABB's partnership with Covariant is part of our strategy to expand into new growth sectors such as distribution and e-commerce fulfillment and leverage the scaling potential in these fields. We identified a significant opportunity for robotic solutions across a broad range of applications, including logistics, warehousing, and parcels &

mail sorting."

Better recognition

The ability to understand the environment with relatively inexpensive sensor technology is one of the evolving capabilities in this industry. And this is where AI is of paramount importance. Stefan Nusser, Chief Product Officer at Fetch Robotics, explained that this enables robots, like their own, to recognize things on a factory or warehouse floor and act accordingly.

"That is the area where AI and computer vision progress is helping us," Nusser said. "Similarly, for example, for a robotic arm, computer vision is what allows it to pick. Some companies in this space are looking at automating the picking process. The general trend here is just understanding the location and position of objects in the environment, which allows the robot to interact with it."