A costmap is a fundamental concept in autonomous robotics. It represents the cost (difficulty) of traversing different areas of the map. In the case of a ground robot working on rough ground the cost map could be a 2D map with lower values where the ground it flat and higher where the ground is rough/sloping. The values held in a cost map are usually abstract and don't directly represent any measurement of the world, they are simply used to guide a route planning algorithm to find efficient and safe routes across the ground.