Revision history [back]

1) The model used by costmap_2d is much simpler/faster than gmapping. The costmap relies on localization for accurate position estimates, whereas gmapping keeps track of many more factors to publish a map and a position estimate. The costmap is primarily used when you have already built a static map and want to track your position within it using AMCL, while still performing dynamic obstacle avoidance.

2) Costmap_2d is usually not run by itself (although it can be). It is typically run from within move_base, with its results published as a convenience, not necessarily to be used by other nodes, thus explaining the lack of subscribers.