Summary
Implemented a full stack where language instructions are parsed, planned, and executed through ROS2 action pipelines and vision-based object perception.
Natural Language Controlled Pick and Place Robot
ROS2-based system that translates user language into executable pick-and-place actions.
$cat case-study.md
What I built
- NLP parsing from commands to task intents.
- ROS2 orchestrator for sequencing perception, grasp, and place actions.
- Stereo perception for object size, color, and depth estimation.
- Retry and verification loops for failed manipulations.
Results
- Reliable command-to-action execution in lab task setups.
- Functional 3D perception pipeline for block handling tasks.
- Further benchmark results can be published after final cleanup.