An Unscented Information-Form Finite-Memory Approach for Simultaneous Localization and Calibration With Application to Uncrewed Aerial Vehicles

초록

Uncrewed aerial vehicle (UAV) localization is typically performed through sensor fusion to leverage the complementary strengths of multiple sensors. However, establishing reliable localization redundancy with a single sensor remains fundamentally important for system reliability. In particular, external forces during flight, such as collisions, wind effects, or payload changes, induce significant discrepancies between control commands and the actual motion, making robust localization an inherently challenging task. This study explores a simultaneous localization and calibration (SLAC) algorithm for UAVs that improves localization accuracy by directly estimating the unknown external disturbances and compensating for their effects. A finite-memory SLAC (FM-SLAC) algorithm uses only recent information from a finite interval and has shown strong robustness in two-dimensional localization for mobile robots. However, in three-dimensional (3D) UAV localization, it exhibited numerical instability during the computation of the large matrix inversions included in the filter gain. To overcome this drawback, this study proposes a novel unscented information-form finite-memory approach for SLAC (UIFM-SLAC). By employing an unscented transformation (UT), an unscented finite-memory SLAC (UFM-SLAC) is developed to avoid the calculation of the Jacobian matrix and the inverse of large matrices required in the conventional FM-SLAC. Furthermore, an information form is integrated into the framework to establish the UIFM-SLAC, which reduces dependence on initial estimates, thereby eliminating numerical instabilities in high-dimensional models and mitigating accuracy loss caused by the linearization of nonlinear dynamics. Experimental comparisons with standard and state-of-the-art algorithms for robust UAV localization were conducted, and the results demonstrated the superior accuracy and robustness of the UIFM-SLAC. Note to Practitioners-The UIFM-SLAC algorithm proposed in this paper is designed for precise 3D localization of UAVs in indoor environments or confined outdoor areas, and is characterized by robustness to payload variations and collisions. We consider a wireless sensor network that provides range measurements for UIFM-SLAC; however, the method is not limited to UWB-based systems and can be applied to any sensing system that can provide at least four simultaneous range measurements between fixed anchors and a tag mounted on the UAV. Compared with vision sensor-based localization approaches, the proposed approach is insensitive to illumination conditions, operates stably even during high-speed maneuvers and sharp turns, and entails a lower computational burden.

키워드