Revision history [back]

Specwise, the structure sensor seems to be very similar to the Kinect/Xtion in terms of resolution, refresh rate, and FOV. It's a shame they didn't include RGB output.

The FOV on the Kinect is they typical limitation for gmapping, since it can't match as many features as a laserscanner, and doesn't map as much area. It's also fairly noisy compared to a typical laserscanner.

You're also missing odometry, which is necessary for gmapping, and in general makes the task a lot harder. You could try hector_slam, which works without odometry, though I've never tested it myself. Feel free to search around for kinect + hector_slam threads, and you'll find that most people don't have a lot of luck with that setup.

Since you asked about architecture, you'd use the pointcloud_to_laserscan nodelet to convert the pointcloud into a laserscan, and then go from there with the standard gmapping/hector_slam tutorials.