IMU-Centric Multi-Sensor Fusion via Factor Graph Optimization: A Review of Robust Navigation and Adaptive Weighted Fusion

Abstract

Precise and dependable positioning is needed for a lot of different areas such as self-driving cars, robot, and autonomous systems especially in complicated or not good situations. To get rid of the limitations of monolithic sensor architecture, the utilization of multiple sources has emerged as an important method and FGO is getting more recognized as a good framework for estimating states because it has global coherence and flexible structure. In this paper, we do a full examination of the IMU-centered multimodal integration frameworks in the FGO setup involving IMU-GNSS, IMU-vision, IMU-LiDAR-vision and IMU-4D radar. Examining all the pros and cons of each kind of sensing, especially looking into the way that working together makes it more reliable and precise. Furthermore there is also an extra classification of adaptive integration protocols referred to as model-based vs data-based it shows about what the protocol does when a failure or variance in the sensor occurs. According to the survey it is a major way of strengthening systems Adaptive Weighted Fusion (AWF) but what currently exists remains mostly sensor based and disjointed. At last this paper talks about remaining troubles and possible examination route, it stresses the need of consistent adaptive structure can deal different sensors in practical surroundings.