Integrated VIO Positioning System Based on Nonlinear Factor Graph Optimization

Today, positioning has attracted special attention as one of the fundamental aspects of various systems. Over time, the navigation system's error, especially in the Inertial Navigation System (INS), has significantly increased. This error increase can lead to serious issues in navigation processes, especially on long routes. To ensure the high accuracy and stability of navigation on long routes, the use of advanced and effective auxiliary systems is necessary. The goal of this article is to present an innovative method for positioning by combining camera and inertial sensor data to enhance accuracy in this vital process. This method is particularly important in self-driving vehicles, domestic robots, and search and rescue robots operating in enclosed environments or areas without active GPS coverage. Consequently, there is a crucial need for a real-time positioning system for these systems. In this article, we have implemented a real-time Visual-Inertial Odometry (VIO) navigation system on a PC platform, which performs significantly better than the INS system alone. In this system, cameras are used to correct the accumulated INS error, and the fusion of cameras and inertial data is done using a graph. To validate and evaluate the system, practical data collected by the system are used, demonstrating that the proposed VIO system exhibits more than a 90% improvement compared to INS.