Multi-sensor Fusion Towards VINS: A Concise Tutorial, Survey, Framework and Challenges

Abstract

State estimation error occurs due to the uncertainty of sensor measurements. By using multiple sensors fusion technical, the error can be bounded and reduced. The multi-sensor fusion system (MSFS) is a kernel technology to develop a navigation system, in which the simultaneous localization and mapping (SLAM) based on the MSFS is an essential solution for autonomous mobile robots. In this paper, we present a concise study on the MSFS towards the visual-inertial navigation system(VINS), which is imitated the human localization system comprised of inertial sensors and cameras. Firstly, this paper introduces the fundamentals of the MSFS and the inertial sensor-based kinetic model. Secondly, state-of-the-art methodologies and a concise review of VINS are presented. Then, a summary of modern VINS frameworks is indicated to come up with a robust SLAM structure. Finally, the challenges and discussions of MSFS are given.