Natural-language household robot
A simulation-first pipeline from natural-language commands to checked household-robot behaviors.
Built the starter architecture, protocol schemas and perception-service scaffold.
Shared DSL and message protocol
- Designed a simulation-first path from natural-language commands to checked robot behaviors.
- Built the starter architecture, protocol schemas and perception-service scaffold.
- Result / current state
- The starter includes a shared DSL, message protocol and JSONL logging path.
- Technologies
- Python
- Unity
- FastAPI
- Pydantic
What it does
The project sketches a pipeline from a natural-language instruction to a goal DSL, behavior-tree plan, scene graph, learned skill and explicit safety checks.
My contribution
The starter includes the repository structure, Unity-to-Python protocol notes, shared schemas, perception stub and JSONL logging utilities.
Technical highlight
The protocol separates planning, perception and execution so each stage can be tested independently before real sensors or learned skills are connected.
Results and status
This is a starter architecture, not a finished household robot. The Unity scene, perception and training components remain early-stage.