ORB-SLAM2: an Open-Source SLAM System for Monocular, Stereo and RGB-D Cameras

2016/10/31 by Raul Mur-Artal, Juan D. Tardos · 17 citations
Computer Science · #cs.RO #cs.CV

paper · pdf · doi:10.1109/tro.2017.2705103

Accepted for publication in IEEE Transactions on Robotics

arxiv created 2017/06/19 · arxiv updated 2020/07/24

Abstract

We present ORB-SLAM2 a complete SLAM system for monocular, stereo and RGB-D cameras, including map reuse, loop closing and relocalization capabilities. The system works in real-time on standard CPUs in a wide variety of environments from small hand-held indoors sequences, to drones flying in industrial environments and cars driving around a city. Our back-end based on bundle adjustment with monocular and stereo observations allows for accurate trajectory estimation with metric scale. Our system includes a lightweight localization mode that leverages visual odometry tracks for unmapped regions and matches to map points that allow for zero-drift localization. The evaluation on 29 popular public sequences shows that our method achieves state-of-the-art accuracy, being in most cases the most accurate SLAM solution. We publish the source code, not only for the benefit of the SLAM community, but with the aim of being an out-of-the-box SLAM solution for researchers in other fields.

Cited by

• OVGGT: O(1) Constant-Cost Streaming Visual Geometry Transformer
• Bias-Eliminated PnP for Stereo Visual Odometry: Provably Consistent and Large-Scale Localization
• ToF-Splatting: Dense SLAM using Sparse Time-of-Flight Depth and Multi-Frame Integration
• SmallGS: Gaussian Splatting-based Camera Pose Estimation for Small-Baseline Videos
• Intelligent robotic positioning through AI-enhanced metrology: Integration of standards, sensor fusion, and adaptive calibration
• Multimodal Perception for Goal-oriented Navigation: A Survey
• Pose Optimization for Autonomous Driving Datasets using Neural Rendering Models
• Unreal Robotics Lab: A High-Fidelity Robotics Simulator with Advanced Physics and Rendering
• Imaging for All-Day Wearable Smart Glasses
• An Online Adaptation Method for Robust Depth Estimation and Visual Odometry in the Open World
• Regist3R: Incremental Registration with Stereo Foundation Model
• BendTwin: Robust Dense-to-Sparse Physical Reconstruction with Bending-Aware Differentiable Spring-Mass Models
• ZeroGrasp: Zero-Shot Shape Reconstruction Enabled Robotic Grasping
• PNE-SGAN: Probabilistic NDT-Enhanced Semantic Graph Attention Network for LiDAR Loop Closure Detection
• Knowledge Distillation for Underwater Feature Extraction and Matching via GAN-synthesized Images
• CamC2V: Context-aware Controllable Video Generation
• Embracing Dynamics: Dynamics-aware 4D Gaussian Splatting SLAM