Learning through Interaction

Grounding AI in Real-World Experiences

Multi-Agent Collaboration

Collaborative Innovation

Ethical Development

We develop AI systems where communication emerges naturally from the interactions between agents. This approach mirrors the evolution of human language, enabling AI.

Our AI agents are designed to operate within physical or simulated environments, learning from direct interaction with their surroundings. This embodiment ensures a deeper understanding of context and enhances the agents’ ability to perform complex tasks.
We create dynamic environments where multiple AI agents coexist and interact, fostering collaboration and competition. This approach helps develop robust, adaptable AI systems capable of managing diverse, real-world scenarios.
At MicrocosmAI, we believe in the power of collaboration. By partnering with academic institutions, research organizations, and industry leaders, we accelerate innovation and create cutting-edge solutions that push the boundaries of AI technology.
We prioritize ethical considerations in all aspects of our AI development. By embedding transparency, fairness, and accountability into our processes, we ensure that our AI systems operate responsibly and positively impact society.
What is MicrocosmAI_Embodied Multi-Agent Environments
What is MicrocosmAI

What is MicrocosmAI?

MicrocosmAI is pioneering advancements through research and development in genuine machine cognition and human-level understanding for embodied artificial agents. Our agents demonstrate emergent behaviors such as communication and cooperation, which arise from the complex dynamics of their interactions with other agents and their simulated physical environments. We achieve this by leveraging artificial neural models to drive decision-making in multi-agent systems, focusing on collaborative tasks that demand grounded and context-aware communication.

Who is MicrocosmAI_

Who is MicrocosmAI?

MicrocosmAI is a research initiative comprising enthusiastic researchers and students from all over the world. We’re dedicated to advancing the field of AI with our research and development in the open-source domain, building frameworks for training multi-agent systems in simulated physical environments. On our blog, we also highlight especially innovative student projects emerging from our teaching at the University of Osnabrück. Additionally, we participate in public events focused on science communication to engage both technical and non-technical audiences, fostering understanding and transparency in AI.

Our Mission

Embodied AI

Design embodied AI agents that
learn and adapt by interacting with
simulated physical environments
to develop decision-making skills.

Multi-Agent Simulations

Create diverse environments and
custom AI models, where agents
can cooperate or compete in
complex problem-solving tasks.

Emergent Behavior

Analyze how fundamental
principles of intelligence
and adaptability arise from
simple multi-agent dynamics.

Open Source & Research

Commit to open-source solutions
for science, openly share research,
and accelerate innovation and
collaboration in the AI community.

Educate Next Generations

Inspire and educate a new
generation of AI practitioners
with hands-on learning and
innovative student projects.

Science Communication

Effectively communicate the
impact and potential of our
research to both technical and
non-technical audiences.

Latest Insights and Updates

EBIMAS Student Project Presented at Osnabrück University’s 50th Anniversary

On October 26th, Osnabrück University celebrated its 50th anniversary with an open day, offering...

A Review of Nine Physics Engines for Reinforcement Learning Research

Table of Contents Abstract We present a review of popular simulation engines and frameworks...