introlab/rtabmap
Rtabmap
This project is a comprehensive library and toolkit for simultaneous localization and mapping, designed to construct three-dimensional environment models while tracking device position. It functions as a robotics perception framework that processes data from RGB-D, stereo, and lidar sensors to enable autonomous navigation and spatial awareness.
The system distinguishes itself through its focus on long-term mapping and global consistency. It employs a sophisticated loop-closure detection engine and graph-based pose optimization to identify previously visited locations and eliminate cumulative drift errors. By utilizing memory-managed keyframe selection, the software balances computational load, allowing it to maintain large-scale spatial databases and merge multiple independent mapping sessions into a unified, consistent structure.
Beyond core mapping, the toolkit provides extensive utilities for data management, including tools for editing point clouds, inspecting database files, and exporting models into standard 3D formats. It supports both real-time operation and offline analysis, offering command-line utilities for replaying recorded sensor data to test different mapping parameters or evaluate odometry algorithms.
The software includes a graphical interface for monitoring mapping processes and visualizing graph structures in real time. It is designed for integration with standard robotics middleware and supports containerized deployment to ensure consistent execution across various hardware architectures.
Features
- Robotics Perception Frameworks - Processes visual sensor data to enable autonomous navigation, loop closure detection, and global pose graph optimization.
- Simultaneous Localization and Mapping - Constructs 3D environment models while simultaneously tracking device position using visual and depth data.
- Spatial Keyframe Managers - Maintains a long-term spatial database by selectively storing and retrieving keyframes to balance computational load and accuracy.
- Loop Closure Detection - Detects previously visited locations to correct spatial drift and ensure global map consistency.
- RGB-D Reconstruction - Converts raw RGB-D or stereo camera streams into detailed three-dimensional point clouds and textured meshes.
- SLAM and Mapping - Performs simultaneous localization and mapping using RGB-D, stereo, or lidar sensors to build 3D environment models.
- Camera Stream Integration - Interfaces with depth-enabled camera hardware to stream visual data for real-time environment mapping.
- Pose Graph Optimizations - Refines spatial maps by optimizing pose graph constraints to eliminate cumulative trajectory drift.
- Robotics Middleware - Facilitates communication between robotic components and navigation stacks through standard middleware frameworks.
- Spatial Mapping Memory Managers - Limits the number of stored locations used for loop closure and optimization to maintain consistent real-time performance during large-scale mapping tasks.
- Mapping Environment Visualizers - Provides a graphical interface to monitor real-time sensor data and observe the construction of 3D environments.
- Keyframe Selectors - Balances computational load by selectively storing and retrieving spatial keyframes for long-term mapping.
- Visual Odometry Systems - Tests and compares different visual or geometric odometry techniques using live camera feeds or pre-recorded sensor data.
- Visual Odometry and SLAM - Evaluates and tests geometric algorithms to track device movement and refine spatial positioning using camera feeds.
- Global Geometry Optimizers - Corrects map drift across sessions by detecting additional loop closures and performing bundle adjustment to refine global structure.
- Depth Sensor Calibrators - Computes distortion models for depth sensors by comparing captured data against generated 3D maps.
- Map Merging - Combines multiple independent mapping sessions into a unified database with gravity constraint alignment.
- Sensor Data Recordings - Captures raw image and depth streams from connected cameras into a structured database format for later analysis.
- SQLite Storage - Serializes mapping sessions and sensor data into local SQLite database files for persistent storage.
- 3D Model Exporters - Converts captured point cloud data into standard formats like PLY or PCD for external use.
- Point Cloud Processors - Captures, cleans, and exports spatial data into standard formats like PCD, PLY, and OBJ.
- Stereo Vision Reconstruction - Processes stereo image pairs, video files, or pre-recorded databases to construct 3D environments and perform localization.
- Static Map Localization - Uses a pre-existing environment database to determine the current position of a robot without building new map data.
- Incremental 3D Mapping - Constructs three-dimensional spatial representations by processing input from RGB-D or stereo camera sensors.
- Sensor Data Abstraction Layers - Provides a unified interface to decouple physical sensor hardware from mapping and processing logic.
- Asynchronous Processing Pipelines - Executes resource-intensive mapping tasks in background threads to maintain real-time system responsiveness.
- Mapping Session Managers - Merges multiple mapping sessions by identifying loop closures or resumes previous sessions to continue building maps.
- Robotics Processing Pipelines - Processes visual data through modular stages including camera input, odometry estimation, and map construction.
- Multi-Session Optimizers - Merges and optimizes multiple mapping sessions into a single consistent database to expand coverage and correct drift.
- Data Stream Recording and Replay - Replays recorded sensor data streams to test mapping parameters and recompute odometry.
- Mapping Process Monitors - Provides real-time visual feedback and statistical analysis of the internal mapping state during active operation.
- Spatial Graph Visualizers - Displays consolidated map and graph structures from multiple mapping sessions.
- SLAM and Mapping - RGB-D graph SLAM with loop closure detection.
