How to train your robot

Northern Lights' guard dog doesn’t have eyes, ears, a nose and fur. Instead, it has LIDAR, microphones, cameras and advanced sensors. Training it, however, isn’t all that different from working with a real dog, according to the robot trainer Svein Ivar.

Northern Lights is the world’s first international CO₂ transport and storage facility. For much of the time, it operates without staff on site.

This is where robot dogs like Roberta come in. They help collect data, carry out inspections and keep a watchful eye on the facility.

Before Roberta is allowed to roam freely, she needs plenty of training and rewards. Fortunately, training a robot dog is not so far removed from working with a living one, says Svein Ivar.

Curious about the similarities? Watch the video below or read on!

Dante is a 12-year-old springer spaniel who enjoys three walks a day, is cherished as “the family’s best friend” and is a model of good behaviour. But when Svein Ivar first brought Dante home 12 years ago, he wasn’t quite so well-mannered. It took years and countless walks before Dante was fully trained. The same goes for a robot dog.

The robot is first connected to a digital twin of the facility, taking its initial test runs virtually before being tested in the real environment. The most important part, though, is training the team who will use the robot.

“The robot is just a tool, albeit a highly sophisticated one, that helps our operators work more safely and efficiently,” Svein Ivar says. 

Not here to replace people

Roberta’s “brain” is certified, tested and quality assured before further training begins out in the field. She also interfaces with a digital twin, so her learning can happen both virtually and on site.

To get the most out of Roberta, we train both the robot and the staff who’ll use her. Testing is carried out in close collaboration between the robot team and the operations team at Northern Lights. 

The aim is always to expand the robot’s capabilities and find new ways to make work safer and more efficient. Looking ahead, Svein Ivar and the team envision even more intelligent solutions, especially when it comes to planning missions.

The goal is for the robot not only to patrol, but also to detect faults and plan tasks independently, through reinforced learning.

“This way, the robot will help us continually improve operations, while human skills and experience remain at the core,” Svein Ivar explains.

Roberta offers a glimpse into a future where technology and people work more closely together than ever before.