Intelligent, Space-Native Robots: Jamie Palmer on Redefining Automation Beyond Earth

Space-native robots mark the next great leap in humanity’s journey beyond Earth, machines engineered not merely to survive in space, but to thrive within its extremes.

Unlike traditional machines, space-native robots evolve through experience, learn to anticipate human needs, and adapt dynamically to mission demands. They are not passive tools, but intelligent partners that extend human capability beyond planetary limits, enabling deeper, safer, and more efficient exploration of the cosmos.

Jamie Palmer, Co-founder and CTO of Icarus Robotics, whose work sits at the intersection of AI, robotics, and space engineering, explored this frontier with us. Drawing from his background in robotics research at Columbia University, Palmer explains how rapid prototyping, embodied AI, and human-in-the-loop training are shaping a new generation of robotic labor in the orbit.

The technology: Embodied AI

Palmer highlights that the real challenge isn’t just building smarter machines — it’s building trustworthy AI partners that amplify human potential in high-stakes environments. His broader vision frames AI-powered robots as the missing workforce that will make future space operations not only more scalable and efficient, but also sustainable and economically viable.

Q. Space is an unforgiving environment. What unique design and engineering challenges come with building robots that can survive and perform in orbit?

Palmer: Space brings extreme engineering constraints. Microgravity fundamentally alters how robots move, grasp, and balance, and systems that behave reliably on Earth can fail in the absence of gravity.

At the same time, there is communication latency from Earth, all of which complicates hardware resilience and real-time control. Unlike spacious, well-lit factory floors, in-orbit environments are confined, cluttered, and low-light, which creates a totally different perception and navigation problem. The most challenging task is testing robots before they are launched into space. We address this through a rigorous progression: starting in physics simulation, building numerous prototypes through rapid design cycles, testing on 2D microgravity test rigs such as air-bearing platforms and gyro rigs, conducting parabolic test flights, and finally deploying to space.

This iterative approach allows us to validate performance under increasingly realistic conditions before committing to orbital operations.

Q. How does human-in-the-loop training accelerate a robot’s ability to generalize and improve over time? Does it reduce the need for redesigning or reprogramming for every new mission?

Palmer: We build general-purpose robotic hardware designed to perform many tasks originally intended for humans, rather than single-purpose machines that must be physically redesigned for each new mission.

And because we use embodied AI, we don’t need to reprogram the robot from scratch every time it takes on a new task; we simply teleoperate the task, and the robot learns from those demonstrations. Through this method, human-in-the-loop training accelerates generalization: each teleoperated session becomes training data, showing the system how robots solve problems and enabling it to operate autonomously and improve over time.

This approach removes two major bottlenecks in traditional space robotics.

Teaming Astronauts and Robots

Q. How do you view the evolving relationship between astronauts and robots — are they primarily tools, teammates, or true extensions of human capability in space exploration?

Palmer: I view the evolving relationship between astronauts and robots as extensions of human capability.

Right now, astronauts are essentially “Amazon warehouse workers in space,” highly trained people who spend weeks unpacking cargo bags instead of conducting critical research. Our robots handle the work astronauts don’t want to do, freeing them for high-value activities only humans can perform.

We are not planning on replacing astronauts; we are multiplying them.

Q. What are the psychological or operational factors in getting astronauts comfortable with AI-assisted co-workers in space?

Palmer: Astronauts are already comfortable working alongside robotic systems—free-flying robots like Astrobee, Int-Ball, and CIMON have been operating aboard the ISS for years.

Our robot builds on this established trust, but extends capabilities from monitoring to physical labor assistance, taking on the repetitive, tedious, or physically demanding tasks that astronauts dread. We’ve heard from astronauts that certain tasks, like watering plants, can be therapeutic; those aren’t what we’d offload. Operationally, comfort comes from safety and control.

These are highly tested, robust systems designed for close human interaction.

Q. How does human-robot collaboration in orbit differ from Earth-based applications, such as in energy or logistics sectors?

Palmer:  The economics of space fundamentally change the human-robot collaboration model. On Earth, companies invest millions in expensive operators to pilot robots, primarily for the purpose of collecting data for robot learning.

In orbit, astronaut time is so valuable that we can deploy robots in teleoperated mode while immediately delivering customer value, collecting training data as a byproduct rather than a costly prerequisite. The other key difference is communication latency.

While space faces greater latency than Earth-based sectors, laser-based communications have made this human-in-the-loop approach feasible for orbital operations.

Leadership and Vision

Q. How has your experience shaped your approach to robotics innovation at Icarus?

Palmer:  My introduction to robotics was unusual – I began working on therapeutic robots for children in hospitals, then wrote my thesis on operationalising robots in human environments.

These gave me an appreciation for the realities of human-robot teaming. During the pandemic, I deployed autonomous mobile robots into live hospitals across Europe. That experience taught me that once a robot is designed, you’re only 30% of the way there; the real work is deployment, iteration, and adapting to unstructured human environments. I also spent time in F1 where the culture of extreme performance and fast iteration shaped my expectations for engineering pace. Later at Columbia, my research deepened my focus on dexterous manipulation and general-purpose embodied AI.

That’s why, at Icarus, we believe in deployment-first robotics, learning through real-world experience, iterating fast, and scaling general-purpose hardware.

Q. What’s the most complex technical or philosophical problem you’re tackling right now at Icarus?

Palmer:  Space operations are severely labor-constrained.

There are less than 400 NASA astronauts ever active, yet the space industry is rapidly expanding with commercial space stations, lunar missions, and Mars exploration, requiring massive infrastructure development and maintenance. AI-powered robotics may be the only way to close the orbital labor gap before it becomes a national security crisis. Here at Icarus, our approach enables an earth-based operator to initially supervise one robot, scaling to manage many robots over time, thereby multiplying labor in the most labor-scarce environment imaginable. Philosophically, we’re solving how humanity scales into space. Labor represents roughly 60% of Earth’s GDP; space will require the same foundation.

Humans don’t scale in space; that’s why we need to massively expand the labor force. We’re building the robotic workforce that makes that possible. We’re building the robotic labor force that makes this possible.

Looking ahead

Q. Icarus is pioneering embodied AI for space. What does that mean in practice? How does your system differ from traditional industrial or service robots?

Palmer: In a nutshell, when we say embodied AI for space, we mean we’re building robots that are space-native designed and can actually learn directly from human demonstration and develop general-purpose intelligence.

Our system differs from traditional robots, which rely on pre-programmed classical control systems and must be rebuilt for every new task. Instead, our robots can adapt and improve over time. Essentially, what differentiates us is that our approach can scale to the complexity and variability of real space operations in a way that legacy systems cannot.

Q. As space becomes increasingly commercial, what role will private robotics companies play alongside NASA and SpaceX, in the future?

Palmer: We’re watching space shift from”old space” to “new space,” with commercial stations coming online much faster than the government-led programs of the past.

Some of these new players are building entire space stations in under 5 years, versus the 25+ years it took for ISS. This pace opens the door for private robotics companies like ours to actually keep up. We’re not trying to compete with NASA or SpaceX; we’re complementing them.

Q. Moving ahead, what breakthrough—technical or cultural—needs to happen for autonomous robotics to become a true partner in human space exploration?

Palmer: I think there needs to be a cultural shift in the way that humans think about working alongside robots. On Earth, robots are still being treated as tools for narrow tasks.

But, in space, they need to be fully trusted by their human teammates in handling these repetitive, time-consuming tasks. This will require extensive training and testing to create a mindset shift; we are emphasizing that robots aren’t replacing crews, but rather extending the capabilities of what crews are capable of.

Q. If you could assign one “dream task” to an Icarus robot on a future Mars or lunar mission, what would it be, and why?

Palmer: I would love for our robots to be the construction workers in space – building and maintaining habitats.

To expand humanity into space in a meaningful way, we must have the necessary infrastructure to support us. If Icarus robots can be the labour that builds cities on the moon and Mars for us to arrive at and live in, I think this would be very exciting.

Key takeaways and what lies ahead for tech leaders

For future tech leaders, the story offers a glimpse into how AI-powered robotics and humanity can scale together.

It forces leaders to see automation as an evolving ecosystem, not a static product. Likewise, the principles described by Palmer — human-in-the-loop training, continuous learning, and rapid deployment cycles — highlight the next decade of technological progress.

The takeaway is clear: the future of technology isn’t about machines replacing humans, it’s about creating intelligent partnerships that push the boundaries of what’s possible.

Whether in orbit or on the factory floor, the most forward-looking leaders will be those who treat AI not as a workforce replacement, but as a force multiplier for creativity, safety, and progress.