OctoMap

A 3D probabilistic occupancy map stored hierarchically as an octree, commonly used for robot obstacle avoidance and planning.

OctoMap is an open-source C++ library developed by Armin Hornung, Kai M. Wurm, and colleagues at the University of Freiburg, described in a 2013 paper in Autonomous Robots. It recursively splits space into eight sub-cubes (an octree), and each cell stores a probability of being occupied, distinguishing three states — occupied, free, and unknown — with sensor noise and environment changes corrected gradually through probabilistic updates. Compared with dividing all of space into equal-sized voxels, it subdivides only where there is something to represent, which saves a great deal of memory, and it can be queried at a coarse or fine resolution as needed. Point clouds from a depth camera or lidar can be written directly into it for collision checking; MoveIt’s perception pipeline uses OctoMap to represent obstacles around the robot.

ExampleA depth camera mounted beside a robot arm continuously feeds its point cloud into OctoMap through MoveIt, so the arm automatically plans around clutter newly placed on the table.

Also called

Octree Map, Octree Occupancy Map

Related

Occupancy Grid Map · Voxel · Point Cloud · Truncated Signed Distance Function · MoveIt Motion Planning Framework · Collision Checking

Sources

OctoMap 官方主页 (Chinese)
MoveIt Perception Pipeline Tutorial