Visual-Inertial SLAM for Unstructured Outdoor Environments: Benchmarking the Benefits and Computational Costs of Loop Closing

2024-08-03

AI Summary

This work addresses the robustness and embedded-system suitability of visual-inertial SLAM (VI-SLAM) for autonomous navigation in dynamic, unstructured outdoor environments. We systematically evaluate seven state-of-the-art open-source VI-SLAM systems—including ORB-SLAM3 and VINS-Fusion—on real-world field datasets. For the first time, we quantitatively characterize the impact of loop closure on the accuracy–computational efficiency trade-off: it reduces absolute trajectory error by 37% on average but increases computational load by 22%–68%. Among all systems, VINS-Fusion and OpenVINS achieve the best overall balance between accuracy and efficiency. We propose a lightweight VI-SLAM evaluation paradigm tailored for embedded platforms, incorporating comparative analysis of multi-modal sensor fusion, relocalization, and global optimization modules. Furthermore, we release the first outdoor loop-closure benchmark dataset alongside fully reproducible evaluation code.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Vision for Robotics & Autonomous Driving
• Natural Language Processing: Safety and Robustness

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Simultaneous Localization and Mapping (SLAM) is essential for mobile robotics, enabling autonomous navigation in dynamic, unstructured outdoor environments without relying on external positioning systems. These environments pose significant challenges due to variable lighting, weather conditions, and complex terrain. Visual-Inertial SLAM has emerged as a promising solution for robust localization under such conditions. This paper benchmarks several open-source Visual-Inertial SLAM systems, including traditional methods (ORB-SLAM3, VINS-Fusion, OpenVINS, Kimera, and SVO Pro) and learning-based approaches (HFNet-SLAM, AirSLAM), to evaluate their performance in unstructured natural outdoor settings. We focus on the impact of loop closing on localization accuracy and computational demands, providing a comprehensive analysis of these systems' effectiveness in real-world environments and especially their application to embedded systems in outdoor robotics. Our contributions further include an assessment of varying frame rates on localization accuracy and computational load. The findings highlight the importance of loop closing in improving localization accuracy while managing computational resources efficiently, offering valuable insights for optimizing Visual-Inertial SLAM systems for practical outdoor applications in mobile robotics. The dataset and the benchmark code are available under https://github.com/iis-esslingen/vi-slam_lc_benchmark.

Problem

• Evaluates Visual-Inertial SLAM systems in unstructured outdoor environments.
• Assesses loop closing impact on localization accuracy and computational costs.
• Analyzes performance for embedded systems in outdoor robotics applications.

Innovation

• Benchmarks Visual-Inertial SLAM systems
• Evaluates loop closing impact on accuracy
• Assesses frame rates on computational load