This project implements a semantic-aware navigation system in ROS 2.
Instead of treating the environment as just obstacles and free space, the robot also understands what it is seeing (for example walls, floor, objects) and uses that information while navigating.
The system combines:
- Gazebo simulation
- Nav2 for navigation
- A semantic segmentation node
- Projection of semantic labels into the navigation costmap
This allows the robot to avoid certain types of regions (like obstacles or restricted zones) more intelligently than standard navigation.
- A custom Gazebo world and robot model
- ROS 2 nodes for semantic segmentation and visualization
- A semantic costmap layer that affects path planning
- A launch file to run the full pipeline
---- in progress ----