costmaps

A costmap is a grid-based representation of the environment in which each cell holds a cost indicating the difficulty of traversing that cell. In robotics, costmaps are used to support motion planning and obstacle avoidance. They are usually two-dimensional arrays with a defined resolution and origin, mapping the robot’s world frame to grid cells.

A costmap is often composed of multiple layers: a static layer representing the known map, an obstacle layer incorporating sensor-derived obstacles, and an inflation or proximity layer that increases cost near obstacles to bias planners away from walls. Some systems use a rolling or local costmap for real-time obstacle avoidance, and a separate global costmap for long-range planning.

Costs are typically expressed as nonnegative integers, with low values for free space, higher values as risk increases, and a maximum value indicating an impassable area. Unknown areas may be assigned intermediate costs or treated as high cost. Planning algorithms consult the costmap to evaluate path feasibility and to prefer routes that minimize total cost while avoiding collisions.

Costmaps are updated from sensor data such as LIDAR, sonar, or cameras, and from map servers. Updates propagate through layers, and the system may use a rolling window to limit memory usage for local/planning. The design aims to provide a fast, modifiable representation that supports reactive re-planning.

In practice costmaps are central to navigation stacks in mobile robotics and autonomous systems, enabling planners to reason about obstacles and safe regions. They trade resolution and update rate against computational requirements, and their effectiveness depends on sensor quality and the accuracy of the underlying maps.