Meet Moxi, a humanoid robot that has become a valued member of medical teams at hospital networks across the US. The brainchild of two remarkable experts in robotics, Moxi may be the shape of things to come in assuming routine tasks that can free up medical professionals to focus more on direct patient care.
In a world where technical innovation may dazzle but fail to deliver, even in its relative infancy, Moxi is the real deal by demonstrating value today in the care delivery chain. It’s a model that suggests there’ll likely be many contexts where robots don’t displace workers but rather make the role of humans even more valuable.
The Genesis Of A Humanoid Robot For Healthcare
When Dr. Andrea Thomaz, a former robotics professor at UT Austin, suggested to her student, Dr. Vivian Chu, that rather than continuing in academia they should start a robotics company, little convincing was required. What sealed the deal was that they both recognized the opportunity, had the technical skills to make it happen, and shared a passion for impact. While acknowledging the essential nature of scholarship and research, they knew that there would be few things that could beat bringing their robot ideas into the real world and Diligent Robotics was born.
The central problem the team set out to solve was how a robot could assist the medical team in a meaningful way such that staff could spend more time interacting with patients and less time on lower value routine tasks. Thomaz and Chu understood that, done right, a solution would be in high demand and societally impactful.
The founders could have chosen to develop robotics for almost any industry, but they specifically identified healthcare. Their research expertise focused on humans and robots collaborating closely with each other—a space called human-robot interaction—and a hospital or clinic setting met that target. Rather than building robots that would do independent tasks away from people such as lights-out manufacturing, they aimed to incorporate robots into a team of humans.
They had one other big reason for choosing healthcare for their first robot fleets. In the long term, building for a complex and busy environment such as a hospital would also help the team develop the necessary AI models for their robots to eventually work in other industries where robots work alongside humans.
Building A Minimum Viable Humanoid
Moxi is an operationally independent humanoid robot. This means it has some features that are similar to that of the human body and can move around and interact untethered with physical objects.
Unlike humanoid robots from companies such as Unitree, Figure, and Tesla, that are in the form of a complete human body, Moxi doesn’t have all the limbs. The Diligent Robotics team coined the term, minimum viable humanoid, as a category to reflect Moxi’s current distinctive characteristics. Specifically, above the waist it has an arm and head, but the lower trunk is used for storage, and it moves on wheels. It’s also relatively short since even healthcare staff with modest stature need to be able to interact with it too.
One striking feature of Moxi is its large LED eyes, which beyond giving it a more approachable look without entering the uncanny valley, provides an essential body language function.
Through her academic research, Thomaz discovered that the eyes help a human know what the robot is planning to do next. For example, if a Moxi reaches the end of a corridor and has an option to turn left or right, the eyes will first dart right, followed by the head, which also hosts a camera, and then the body will follow in that direction. It’s the same thing humans do, and it communicates the next move to another person. These details form the basis for optimal human-robot interaction.
Moxi does have some minimal voice features today, such as announcing its arrival or alerting people in an elevator that it needs to get off at that floor. Feedback from the field suggests that additional vocal activation and interaction will be part of its future.
Ten years ago, investors questioned why Thomaz and Chu were including human features in their robot design. Today, they question why it doesn’t have more.
How Does The Moxi Robot Work?
When a hospital decides to implement a fleet of Moxi’s, each deployment must be designed to meet the specific needs and infrastructure at that facility.
During the design of the deployment, the team at Diligent Robotics works closely with hospital administrators and medical staff to decide on task priorities and their urgency designations, and what physical areas the robots will traverse as they go about their work.
Moxi is controlled from an app through which authorized staff provide it with instructions. Engagement with the robot is similar to how a person might order food delivery. A nurse, for example, will request that an item such as a medical instrument be picked up from one location and carried to another. The app keeps the requestor up to date on each step until the Moxi’s job is accomplished.
A cloud-based scheduling system receives the request, locates a Moxi, and the task is queued up to process as availability permits. Once the robot has its instructions, even if wireless connectively fails, it will still complete its mission; an important business continuity feature in a time-sensitive environment.
When the robot arrives at its location, the item it’s carrying is securely unloaded with an authorized staff members security badge.
The Challenges And Future Of Humanoid Robots
Today, Moxi is a minimum viable humanoid, but as the team receives feedback from their deployments, its form and function will evolve. Rather than continue to research and prototype, Diligent Robotics is responsibly deploying into the real world and learning as they go.
With a subscription-based model, they can swap out and upgrade Moxi’s as the hardware and software evolves. It’s an admirable quality since the pursuit of perfection can so often delay and even destroy valuable ideas.
The team, and anyone working in robotics, knows that there are still a lot of challenges ahead to overcome.
Specifically, for Diligent Robotics, they consider an area called vision language action models as their next major milestone. While robots have become better at perceiving and responding to the world, challenges still exist for them to independently reason and generate actions based on dynamic environments.
For example, if Moxi knows where certain items are in a storage closet, how will it deal with that room being completely reconfigured without it being told where everything is? Training models for its AI to achieve this is still a challenge.
With Moxi fleets deployed and conducting thousands of hours of actions right now, the team is ahead of the curve in building that potential model and also learning about the necessary physical design choices for the robots.
Thomaz and Chu also think deeply about the shared concerns that robots may ultimately displace large numbers of human workers. Like everyone, they find it difficult to look too far out, but in the medium term they’re probably more sanguine than most and think many humanoid robots will work and add value positively alongside humans.
Planning For A Future Of Humanoid Robots
When asked how leaders should think about planning for humanoid robots in the near future, the co-founders offer two areas of advice.
First, for the areas in which robots are anticipated to work in, leaders must ensure that workflows are well understood and documented. For example, within those workflows identify work that is complex, repetitive, high-volume, or error prone. In other words, be ready to plug robots into processes that are already well structured.
Second, organizations must prioritize change management so that all the right people, from leaders on down, are ready to embrace their robotic colleagues. They must make sure there is plenty of time put aside to gather input and have candid internal discussions. Leaders and staff need to believe that a future of teams made up of humans and robots is desired and trusted.
Thomaz and Chu are already delivering a version of human-robot interaction in the workplace and it’s not too early for organizations of all types to begin planning for their humanoid robot future.