horizonrobotics Semantic Navigation Skill
Use this skill for semantic navigation requests.
Preferred Interfaces
HTTP:
POST /api/semantic_nav- JSON body:
{"cmd":"floor,room,object"}
ROS topic:
/chat_loc_pub
Workflow
- Extract the semantic target as floor, room, and object level information.
- Confirm the target is specific enough for the robot to execute.
- Send the request through the documented HTTP or ROS interface.
- Monitor the returned status and stop for clarification if the target is ambiguous.
- Report the semantic target and execution result.
Safety Rules
- Confirm the semantic target is explicit enough to execute.
- Prefer semantic navigation only when the task refers to rooms, floors, or objects.
- If the target is ambiguous, ask for clarification.
- Do not fabricate room names, floor labels, or object identifiers that are not provided.
Example
See assets/sem_nav_examples.sh.