CONTROLAgents & automation
JEV Drone Simulation in Automata Room
Automata Room@automataroom
0:57This project tests an AI model's ability to reason about the physical world and translate that reasoning into robotic behavior by controlling a drone within a simulated warehouse environment.
SignalThe workflow: JEV → understands the scene → generates the program → controls the drone → executes inside the simulation.
We just added a drone to Automata Room and started testing JEV in a physical style simulation environment.
The goal is simple: give an AI model not just a prompt, but a robot, a room, spatial constraints, and executable actions.
In this experiment, JEV is controlling a Skydio X2-style drone inside a warehouse environment. The agent can reason about altitude, obstacles, room geometry, and movement before executing the behavior.
The workflow:
JEV → understands the scene → generates the program → controls the drone → executes inside the simulation.
This is where Automata Room gets interesting.
We’re not just testing whether an LLM can write code.
We’re testing whether an AI model can reason about the physical world and turn that reasoning into robotic behavior.
Next: more complex environments, more objects, and harder autonomous tasks.
Giving AI a world to reason about.