Precise UWB-Based Localization for Aircraft Sensor Nodes

In this work, an indoor positioning system (IPS) is introduced to overcome the tedious task of configuration of sensor nodes in an aircraft. Our positioning system is based on a ultra- wideband (UWB) commercial off-the-shelf (COTS) system, which was selected because of its fine resolution in time. In the first part of the work, time of flight (ToF) and multilateration algorithms are implemented and evaluated in two and three dimensional scenarios. Our measurement results show an accuracy below 10 cm in line-of-sight (LOS) conditions. However, when experiments are held inside a cabin mock-up under the presence of non-line-of-sight (NLOS) condition, the accuracy gets significantly worse. To overcome this issue, we introduce a artificial neural network (ANN)-based localization approach in the second part of the work to enhance the localization accuracy using raw channel impulse response (CIR) data provided by the localization system. We first illustrate that our approach is able to distinguish between LOS/NLOS conditions, with an accuracy of more than 85%. We then demonstrate that our ANN can also be trained to directly predict the localization of an object. Our experiments show that the localization error is reduced by approximately 70% resulting in 12.3 cm on average, in comparison with the time-based approach which has 43 cm error for the same measurement setup.