3d navigation in rviz

edit

For a 3D URDF model for rviz/Gazebo of a Create (Roomba without the vacuum) see my repository here.

By supplying this URDF as your robot description in the nav stack your robot will show up as this 3D model in rviz, even if you're just using a 2D simulator like Stage.

However, rviz still won't show a 3D world when using either Stage, Gazebo or even real robots in a real world because the nav stack operates on 2D costmaps built on 2D laser scans. Because of this the costmap will show up in rviz as a 2D map image with a 3D robot superimposed on it (if using a URDF for your robot)

UPDATE: If you want to try some bleeding edge 3D mapping/localization packages you could checkout the 6D-SLAM stuff. You'll have to find a way of simulating a Kinect inside of Gazebo, but I'm pretty sure that has been done already (just not well documented). You'll also need something of a super-computer to run it as Gazebo+Simulated Point Clouds+6D-SLAM = a lot of processing going on. This will give you localization and mapping, but not trajectory/path planning... You'd have to come up with your own 3D path planner.

PS If you're new to ROS, I'd stick with the 2D nav stack until things get a bit more mature.