Automata

Embodied AI Meets the Garden

An open research platform where autonomous construction vehicles learn on real terrain — in a 4×8 ft bed of soil, sand, or gravel. The same scale as a raised garden or greenhouse bed: dig, grade, haul, and level earth where the rubber meets the dirt.

Automata brings robotics and AI into the garden shed and the greenhouse. Robots prepare soil, shape terrain, and work alongside planting and growing experiments — while learning physics and coordination. Not a toy: a serious foundation for research, education, and hands-on certification.

Why This Matters

Most AI today is disembodied. Language models reason in text but have no physical understanding of the world. Yet intelligence evolved through interaction with matter — digging, moving, building, shaping terrain.

Gardening is one of the oldest ways we shape the world: till the soil, make a bed, plant, tend, harvest. Automata asks: What happens when AI learns by shaping the physical world? In a raised bed or a test arena, robots learn to move earth — and AI learns how the world pushes back.

System at a Glance

• Physical terrain — 4×8 ft bed (greenhouse- and garden-bed scale), soil/sand/gravel, optional irrigation
• Autonomous robots — RC excavators, trucks, dozers; ESP32-S3, micro-ROS
• Perception — overhead gantry, RGB/depth, terrain mapping
• AI planning — ROS2 server (Jetson), coordination, deployment
• Simulation — NVIDIA Isaac Sim for training and sim-to-real transfer

Concept: terrain bed as robotics sandbox — same footprint as a garden bed.

Support the Platform

Open source, extensible, modular. We’re building Automata for labs, schools, and researchers. Your support funds hardware kits, documentation, and certification curricula.

OpenCollective GitHub Sponsors Hardware kits

Product Tiers

From a single raised-bed kit for makers and gardeners to a full research rig with multiple robots and perception — same soil, same scale.