Sanctuary AI expands robotics footprint into automotive manufacturing

In April, Sanctuary AI unveiled the seventh generation of its general-purpose robot, Phoenix, one of the most sophisticated human behavioural data capture technologies available today. (Credit: Sanctuary AI)

In April, Vancouver’s Sanctuary AI announced a strategic partnership with mobility technology company Magna to support the development, deployment and scaling of general-purpose robots in automotive manufacturing.

Magna—one of the largest automotive suppliers in the world based in Aurora, Ont., and a significant buyer of industrial robots—aims to leverage Sanctuary robots’ unique capabilities across multiple applications within automotive manufacturing processes. The partnership aims to catalyze the scaling of Sanctuary’s robots while maturing the technology for use in challenging manufacturing environments for Magna and other industrial and automotive customers.

The Magna collaboration supports Sanctuary’s goal to become the first organization in the world to create the world’s first human-like intelligence in general purpose robots. “World-changing goals like these require world-changing partners,” Sanctuary’s co-founder and CEO Geordie Rose says. “Magna’s position as a world leader in the use of robots today makes this partnership an essential advancement for our mission.”

The partnership features the development of general-purpose AI robots for deployment in Magna’s manufacturing operations; a multi-disciplinary assessment of improving cost and scalability of robots using Magna’s automotive product portfolio, engineering and manufacturing capabilities; and a strategic equity investment by Magna.

Sanctuary says experts estimate that only 10 per cent of automotive manufacturing is automated. “This is because all the robots currently working within automotive manufacturing are special-purpose tools with a narrowly defined set of tasks they can achieve,” Rose says. “The only way to significantly move the needle on automation in automotive manufacturing is to create ‘general purpose’ robots, which are versatile enough to do anything a person can do.”

This is where Sanctuary’s Phoenix robots come in. “General-purpose technologies can deal with tasks and environments that, by design, can’t be rigidly structured,” he says. For example, one of the hardest tasks to automate is called chaotic picking—or the picking of a variety of parts in an uncontrolled environment—which is what machines like Phoenix are built for.

“For a robot to act as a versatile tool in manufacturing processes, it must have human-like intelligence, form, function and senses,” Rose explains. “Adopting general-purpose robots into the workforce is much easier and cost-effective than re-structuring an entire workplace around different, special-purpose tools.”

Early on, Sanctuary realized the key to unlocking useful work is in the hands, with 98 per cent of all work requiring the dexterity of a person’s hand. 

“With 19 degrees of freedom and proprietary haptic technology, our Phoenix robots have both the dexterity and sense of touch essential to performing tasks successfully,” Rose says. “Our robots are designed to do any task a person reasonably could, with a focus first on addressing the dirty, dull and dangerous jobs in the manufacturing space.” 

But there is a key challenge in scaling robot technology for wide-spread use in industrial and automotive settings: data. 

“The type of data needed is much less readily available than data you can find on the internet,” Rose says. “This is because the AI models we’re building—large behaviour models—need high-quality, high-fidelity human behavioral data.” Sanctuary collects this by showing the robot how to do a task in a process called teleoperation, which involves pilots carrying out tasks via the robot, with the data from this being collected and used to teach the AI models. “We have been able to speed up that process, reducing the time it takes to automate a task from weeks to less than 24 hours,” he adds.