🚗 Autonomous Navigation Transformer
Transformer-based model designed to convert natural language navigation commands into structured movement sequences.
🧠 Model Details
- Architecture: Decoder-only Transformer
- Hidden Size: 2048
- Layers: 24
- Attention Heads: 16
- Parameters: ~1.3B
- Training Steps: 200k
- Dataset: 120k navigation instruction pairs
📄 Input Format
"Move forward 3 meters and turn right."
📄 Output Format
move(distance=3m) turn(direction=right)
🎯 Intended Use
- Autonomous robots
- Simulation training
- Indoor navigation agents
⚠️ Limitations
- English only
- Structured action output required
- Downloads last month
- 14