Creating an 'AI Researcher' for Space Experiments: Four Companies Jointly File Two Patents for an Autonomous Experiment Platform Using LLMs to Connect Observation, Judgment, and Operation

Space Seed Holdings, Inc., in collaboration with its subsidiary RegeneSome, Inc., IDDK, Inc., and Space NOME Institute, Inc., has filed two patents on June 10, 2026, for an apparatus and control technology that utilizes machine learning models, including Large Language Models (LLMs), to autonomously conduct cell cultivation and various scientific experiments.

The two patents filed are for an "Integrated Autonomous Experimental Device for Cell Experiments and its Control Method" and an "Experimental Module Integrated Control Device and its Control Method." Both are joint applications by the four companies.

The company views this filing not merely as the development of a single cell culture device, but as the first step in a vision to develop a common platform for autonomously designing, executing, and analyzing experiments across different scientific fields, such as microalgae, microorganisms, drug discovery screening, regenerative medicine, and material synthesis.

In Space, Researchers Cannot Stand Next to the Equipment

In a terrestrial laboratory, a researcher can check microscope images, judge the state of cells, add media or reagents, and adjust temperature, pH, and oxygen levels. In contrast, in environments like orbit or the lunar surface, researchers cannot remain beside the equipment. Available power, volume, and communication bandwidth are limited, and communication with the ground can experience delays or interruptions. To establish space experiments as a mainstream research method, the entire research process of "observing," "interpreting," and "deciding the next operation" must be integrated into the apparatus itself.

JAXA has also pointed out that in the post-ISS era, the primary users of microgravity experiments will shift from government to the private sector, and the costs, preparation time for developing dedicated equipment for each experiment, and astronaut work hours will become challenges. Our goal is not just to bring equipment from Earth to space. It is to create an "autonomous laboratory" by implementing the researcher's decision-making process into the device itself, so experiments do not stop even if communication is lost.

From Automation to "Autonomization"—LLMs Begin to Change Scientific Research

Globally, research on "Self-Driving Laboratories," which combine robotics, machine learning, and literature databases, is advancing. Representative examples include "Coscientist," an AI system that assists with chemical experiments, and "A-Lab," a robotic laboratory that autonomously searches for synthesis conditions for new inorganic materials. These cases show that AI is evolving from a mere respondent to researchers' questions into an entity connected to actual experimental equipment, repeating cycles of observation, judgment, and operation.

To deploy this concept in space, a mechanism is needed to coordinate multiple experimental instruments with a common control layer, rather than just automating specific experiments. The two new patent applications target two complementary technological layers: autonomous control specialized for cell experiments, and a higher-level integrated control that allows for the reconfiguration of equipment across different experimental fields.

Patented Technology 1: LLM Observes Cellular Changes and Decides the Next Operation

The "Integrated Autonomous Experimental Device for Cell Experiments" integrates a culture vessel, a microscopic observation unit, a mechanism for delivering media and reagents, and an environmental control unit. The control unit inputs disparate data types, such as microscope images and sensor values, into a machine learning model as a common context to continuously determine multiple operations. A key feature is its ability to alter subsequent operations based on the observed state of the cells, rather than simply replaying a fixed schedule. It also includes functions to detect issues like air bubbles in microgravity and perform recovery actions.

Patented Technology 2: An "Operating System for Scientific Experiments" Connecting Different Devices

The "Experimental Module Integrated Control Device" is a higher-level control platform not limited to cell experiments. It connects functionally different experimental modules via a common interface, allowing an LLM to interpret the data from all of them and issue commands. This design allows for the easy addition of new, third-party modules.

Furthermore, it aims to be an edge-cloud cooperative autonomous experiment platform that can switch between high-performance models on the ground and local models on the device depending on the communication status, ensuring experiments can continue even in orbit or remote locations. This is less a piece of software for controlling individual devices and more an "operating system for science" that connects different instruments and data to advance the entire experiment.