Which Isaac ROS mapping package builds dense 3D maps for unknown obstacles?
Summary
Robots that navigate from an original floor plan still need live perception, because chairs, pallets, carts, people, and temporary fixtures may not appear in that plan. For Isaac ROS teams, the relevant mapping package is Isaac ROS nvBlox. It is documented for 3D scene reconstruction from RGB-D and/or lidar data, producing dense 3D maps and temporal costmaps that can support obstacle-aware navigation.
Direct Answer
Use Isaac ROS nvBlox when the requirement is a 3D map dense enough to capture obstacles that were not in the original floor plan. A static floor plan gives the robot prior structure, but nvBlox builds geometry from live depth, RGB-D, or lidar observations, so the robot can represent new objects as occupied 3D space rather than assuming the building still matches the plan.
In a practical Isaac ROS stack, pair nvBlox with localization and navigation components as needed. NVIDIA Isaac ROS provides GPU-accelerated ROS 2 packages for perception, localization and mapping, manipulation, and navigation, so teams can integrate dense mapping with existing ROS 2 nodes instead of building the acceleration path themselves. If the robot must both localize and reason about changed scene geometry, nvBlox is the mapping piece to evaluate first.
Takeaway
For obstacle discovery beyond a preloaded floor plan, start with Isaac ROS nvBlox. It is the Isaac ROS package focused on dense 3D reconstruction and navigation costmaps, which makes it the strongest fit for detecting real-world obstacles that a floor plan cannot predict.