Project Info
Inspiration
AxisGen was inspired by how tools like LangChain made LLM workflows easier to structure, chain, and deploy. We wanted to bring a similar idea to Physical AI: a layer that turns a simple beginner-level robot task prompt into a broader set of simulation-ready prompts. Robot deployment should not only be accessible to experts with custom simulation knowledge. AxisGen aims to make robotic deployment more accessible for any customized task by helping users quickly generate the diverse scenarios needed for robust training and evaluation.
What it does
AxisGen takes one robot-task prompt and expands it across environmental axes such as lighting, object placement, clutter, camera viewpoint, distractors, task difficulty, robot embodiment, and safety constraints. Instead of producing one fixed scenario, it generates a body of structured prompts that can be used to create more diverse simulation datasets for robot learning.
How we built it
We built AxisGen as a prompt expansion pipeline for Physical AI simulation. We designed structured templates that preserve the core task while systematically changing environmental variables. We used Nebius Physical AI, Isaac Lab planning, Python, and prompt-generation logic to connect a simple input task to many simulation-ready scenario descriptions.
Challenges we ran into
The biggest challenge was balancing variation with consistency. If the generated prompts change too much, they no longer represent the original task. If they change too little, the dataset is not diverse enough. We also had to think carefully about which environmental axes actually matter for robot robustness and how to format the prompts so they are useful for simulation.
Accomplishments we're proud of
We are proud that AxisGen turns a single task idea into a scalable set of useful scenario variations. Instead of focusing on one specific robot task, we created a general system that can support many customized tasks. Our proof of concept showed that prompt expansion can become a practical layer in making Physical AI dataset generation more accessible.
What we learned
We learned that prompt structure matters a lot for robotics. A good robot-learning prompt needs more than a task description; it needs environment details, object variation, safety constraints, and success conditions. We also learned that robustness can start before training, at the data-generation and scenario-design stage.
What's next
Next, we want to connect AxisGen more directly with Isaac Lab so generated prompts can automatically become simulation configurations. We also want to add scoring for prompt quality, generate larger scenario batches, export datasets into LeRobot or GR00T-compatible formats, and test how much environmental-axis variation improves downstream robot model performance. Long term, we want AxisGen to help make robotic deployment easier for anyone building a customized robot task.
This project did not link a GitHub repository.
Analysis
No indexed repository for this project, so there are no commit stats to show.
Technology
- PythonUnchecked
No repository was indexed for this project, so these Devpost claims have not been checked against code.
AI coding agents
No repository was indexed, so agent usage could not be checked.
Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.
Codebase size
No repository was indexed, so there is no codebase to measure.
This project did not link a GitHub repository, so there is nothing to diagram.
This project did not link a GitHub repository, so its feature claims have not been checked against code.
Export this project's context (description, README, evidence, key source files) to chat with an AI agent elsewhere.