Resilient tightly coupled INS/UWB integration method for indoor UAV navigation under challenging scenarios.

Based on the high positioning accuracy, low cost and low-power consumption, the ultra-wide-band (UWB) is an ideal solution for indoor unmanned aerial vehicle (UAV) localization and navigation. However, the UWB signals are easy to be blocked or reflected by obstacles such as walls and furniture. A resilient tightly-coupled inertial navigation system (INS)/UWB integration is proposed and implemented for indoor UAV navigation in this paper. A factor graph optimization (FGO) method enhanced by resilient stochastic model is established to cope with the indoor challenging scenarios. To deal with the impact of UWB non-line-of-sight (NLOS) signals and noise uncertainty, the conventional neural net-works (CNNs) are introduced into the stochastic modelling to improve the resilience and reliability of the integration. Based on the status that the UWB features are limited, a 'two-phase' CNNs structure was designed and implemented: one for signal classification and the other one for measurement noise prediction. The proposed resilient FGO method is tested on flighting UAV platform under actual indoor challenging scenario. Compared to classical FGO method, the overall positioning errors can be decreased from about 0.60 m to centimeter-level under signal block and reflection scenarios. The superiority of resilient FGO which effectively verified in constrained environment is pretty important for positioning accuracy and integrity for indoor navigation task. [ABSTRACT FROM AUTHOR]