SJTU-ViSYS/Ground-Fusion

What it solves

Ground-Fusion is designed to provide robust and accurate localization and mapping for ground vehicles using low-cost sensors. It specifically addresses "corner cases"—challenging environments where individual sensors (like cameras or GNSS) might fail or provide degraded data—ensuring the vehicle can maintain its position without drifting.

How it works

The system uses a factor graph to tightly integrate data from four different sources: RGB-D images, inertial measurements (IMU), wheel odometry, and GNSS signals. To handle various starting conditions, it employs three distinct initialization strategies (stationary, visual, and dynamic). It also includes built-in mechanisms to detect and handle sensor anomalies or degradation in real-time to maintain system stability.

Who it’s for

This project is intended for robotics developers and researchers working on ground vehicles that require reliable SLAM (Simultaneous Localization and Mapping) without relying on expensive lidar sensors.

Highlights

• Multi-Sensor Fusion: Tightly integrates RGB-D, IMU, wheel odometry, and GNSS.
• Robustness: Specifically engineered to handle visual, wheel, and GNSS challenges (corner cases).
• Flexible Initialization: Offers three tailored methods for starting the system depending on the vehicle's state.
• Dense Mapping: Capable of producing real-time dense color maps of the environment.