This paper presents OctoMap, an efficient probabilistic 3D mapping framework based on octrees. The framework allows for the representation of 3D environments with probabilistic occupancy estimation, explicitly modeling occupied, free, and unknown areas. It uses octrees to efficiently store and update 3D maps while keeping memory consumption low. The framework supports multi-resolution queries and includes a compression method that reduces memory usage by combining coherent map volumes. OctoMap is implemented as an open-source C++ library and has been successfully applied in various robotics projects. The framework is evaluated using real-world datasets and demonstrates efficient and consistent modeling of environments while maintaining minimal memory requirements. The approach is suitable for a wide range of robotic applications, including navigation, mapping, and localization. The framework is memory-efficient, allowing for the representation of large environments and efficient transmission between multiple robots. The implementation supports various platforms and integrates with the Robot Operating System (ROS). The framework is designed to handle sensor noise and dynamic environments, providing a flexible and updatable map structure. The paper also discusses related work in 3D mapping and compares OctoMap with other approaches, highlighting its advantages in terms of efficiency, memory usage, and flexibility.This paper presents OctoMap, an efficient probabilistic 3D mapping framework based on octrees. The framework allows for the representation of 3D environments with probabilistic occupancy estimation, explicitly modeling occupied, free, and unknown areas. It uses octrees to efficiently store and update 3D maps while keeping memory consumption low. The framework supports multi-resolution queries and includes a compression method that reduces memory usage by combining coherent map volumes. OctoMap is implemented as an open-source C++ library and has been successfully applied in various robotics projects. The framework is evaluated using real-world datasets and demonstrates efficient and consistent modeling of environments while maintaining minimal memory requirements. The approach is suitable for a wide range of robotic applications, including navigation, mapping, and localization. The framework is memory-efficient, allowing for the representation of large environments and efficient transmission between multiple robots. The implementation supports various platforms and integrates with the Robot Operating System (ROS). The framework is designed to handle sensor noise and dynamic environments, providing a flexible and updatable map structure. The paper also discusses related work in 3D mapping and compares OctoMap with other approaches, highlighting its advantages in terms of efficiency, memory usage, and flexibility.