Listen to this article

As development continues on humanoid robots, one mobile robot is already at work in hospitals. Diligent Robotics Inc. today announced that its Moxi robot has completed 110,000 autonomous elevator rides at health systems across the U.S.

The mobile manipulator has a single arm for opening doors and pushing buttons to operate elevators. Moxi’s achievement marks a milestone in artificial intelligence-driven automation for unstructured healthcare environments, said the Austin, Texas-based company.

“Achieving autonomy in robotics, particularly in health care environments, is an incredible challenge,” stated Andrea Thomaz, CEO of Diligent Robotics and a 20-year AI veteran. “Navigating elevators seems simple, but the unpredictable nature of shared spaces, real-time changes, and the need for accuracy make it one of many hard tasks that humanoids deployed in human environments need to solve.”

“With Moxi, we’ve demonstrated the ability to integrate AI into environments where collaboration between people and robots is vital for success,” she added. “As of today, we are completing over 20,000 fully autonomous elevator rides each month, something none of our competitors are doing.”

---

Register now

---

Moxi moves to greater autonomy

Founded in 2017, Diligent Robotics noted that it has integrated Moxi into existing healthcare workflows, such as delivering supplies and transporting laboratory specimens around the clock. The company said its robot has helped improve operational efficiency and enabled busy staffers to focus on patient care rather than routine transport tasks.

Diligent Robotics used “humans in the loop” to develop Moxi’s autonomy and ability to interact with elevators.

“There are two approaches: Waymo, which used its own drivers and did R&D until its product was fully autonomous, and Tesla, which got its product out in the wild with real customers and has increased autonomy with supervision over time,” Thomaz told The Robot Report. “We took the latter approach.”

“This milestone means we no longer need close human supervision, which is a significant one for mobile manipulation,” she said. “The number of rides per day really shows that we’ve gotten past R&D and are working in production.”

When it first deployed its robots, Diligent staffers supervised operations on site. They obtained labeled data for AI models, explained Thomaz. The company can now supervise its fleet of 100 robots in 20 sites remotely.

“A lot of our early partner hospitals got used to our staffers being around to do everything the robots would need,” Thomaz said. “In fact, it has been easier for hospitals taken live in the past few months, because they had fully autonomous robots from the start. For sites that were previously under human supervision, we went through a data-collection phase and asked the staff not to do anything.”

Moxi opens doors, a model for other actions 

How hard was it to get Moxi to operate elevators?

“They became a roadmap for autonomy,” replied Thomaz. “Nearly every delivery pass involves an elevator req, from the pharmacy or lab in the basement up to the patient wards. For the simplest elevator, you push a button and get on.”

“The most complex one in operation today, you have to scan a badge to activate the buttons and then push them,” she told The Robot Report. “There are a lot of patient floors that are secure. It’s a complex manipulation skill for a dual device, primarily because of the speed of swiping.”

By solving the problems of manipulating a variety of doors and elevators, Diligent Robotics is developing end-to-end action models.

“Our ability to develop models that are specific to these small skills is creating an infrastructure of training models that could then be applied to other skills,” Thomaz said. “They’re not large, general-purpose models, but we’re excited to have a fleet we can leverage to build foundation models.”

Complex environments still pose challenges

As environments with trained but busy personnel, a high degree of safety regulations, and the general public — some of whom are not well — hospitals are particularly challenging for robots.

“That’s why we treated this with white gloves; we’re not just dropping robots in to learn on their own,” noted Thomaz. “We spent two years deploying robots with people because of the sensitive environment and to get the robots to operate efficiently.”

For instance, she cited interventions where healthcare staffers push the emergency stop button and can manually move a robot out of the way for something like an urgent gurney. They sometimes forget to turn the robot back on so it can continue its mission.

“We’ve released a feature where the robot can ask on its screen for someone to un-e-stop it,” Thomaz said. “There are other environments, such as an elevator bay with six different cars and patients coming in and out, that are still complex.”

“Moxi also uses data to avoid routes or elevators that are always busy,” she said. “We collect that data and have preferred elevator bays for at night versus during the day.”

Diligent Robotics works to normalize robots 

With aging populations and workforce challenges, demand for automation is likely to grow, said Diligent Robotics. Moxi provides an example of how robots can address human needs, it said.

“I’ve been talking about Moxi as a ‘humanoid for healthcare.’ We’re doing the things that people are talking about what humanoids could do,” Thomaz said. “Bipedal locomotion isn’t the hardest part — it’s applicability of mobile manipulation. Most humanoids are still proofs of concept.”

“When I walk through hospitals where Moxi is deployed, it’s the first robot that many people have encountered,” she added. “Patients are getting used to seeing a future with robots.”

“Achieving full autonomy to enable hospital-wide transport tasks is just scratching the surface of what humanoid robots like Moxi will do in hospitals and beyond,” said Thomaz. “The knowledge and trust that we gain from healthcare settings will inform future product developments. We look forward to building humanoid social robots to collaborate and assist with caring for people in many different settings.”