Robust, Energy-Efficient, and Scalable Indoor Localization with Ultra-Wideband Technology

Laura Flueratoru

Research output: Book/ReportDoctoral thesisCollection of Articles


Ultra-wideband (UWB) technology has been rediscovered in recent years for its potential to provide centimeter-level accuracy in GNSS-denied environments. The large-scale adoption of UWB chipsets in smartphones brings demanding needs on the energy-efficiency, robustness, scalability, and crossdevice compatibility of UWB localization systems. This thesis investigates, characterizes, and proposes several solutions for these pressing concerns. First, we investigate the impact of different UWB device architectures on the energy efficiency, accuracy, and cross-platform compatibility of UWB localization systems. The thesis provides the first comprehensive comparison between the two types of physical interfaces (PHYs) defined in the IEEE 802.15.4 standard: with low and high pulse repetition frequency (LRP and HRP, respectively). In the comparison, we focus not only on the ranging/localization accuracy but also on the energy efficiency of the PHYs. We found that the LRP PHY consumes between 6.4–100 times less energy than the HRP PHY in the evaluated devices. On the other hand, distance measurements acquired with the HRP devices had 1.23–2 times lower standard deviation than those acquired with the LRP devices. Therefore, the HRP PHY might be more suitable for applications with high-accuracy constraints than the LRP PHY.

The impact of different UWB PHYs also extends to the application layer. We found that ranging or localization error-mitigation techniques are frequently trained and tested on only one device and would likely not generalize to different platforms. To this end, we identified four challenges in developing platform-independent error-mitigation techniques in UWB localization, which can guide future research in this direction.

Besides the cross-platform compatibility, localization error-mitigation techniques raise another concern: most of them rely on extensive data sets for training and testing. Such data sets are difficult and expensive to collect and often representative only of the precise environment they were collected in. We propose a method to detect and mitigate non-line-of-sight (NLOS) measurements that does not require any manually-collected data sets. Instead, the proposed method automatically labels incoming distance measurements based on their distance residuals during the localization process. The proposed detection and mitigation method reduces, on average, the mean and standard deviation of localization errors by 2.2 and 5.8 times, respectively.

UWB and Bluetooth Low Energy (BLE) are frequently integrated in localization solutions since they can provide complementary functionalities: BLE is more energy-efficient than UWB but it can provide location estimates with only meter-level accuracy. On the other hand, UWB can localize targets with centimeter-level accuracy albeit with higher energy consumption than BLE. In this thesis, we provide a comprehensive study of the sources of instabilities in received signal strength (RSS) measurements acquired with BLE devices. The study can be used as a starting point for future research into BLE-based ranging techniques, as well as a benchmark for hybrid UWB–BLE localization systems.

Finally, we propose a flexible scheduling scheme for time-difference of arrival (TDOA) localization with UWB devices. Unlike in previous approaches, the reference anchor and the order of the responding anchors changes every time slot. The flexible anchor allocation makes the system more robust to NLOS propagation than traditional approaches. In the proposed setup, the user device is a passive listener which localizes itself using messages received from the anchors. Therefore, the system can scale with an unlimited number of devices and can preserve the location privacy of the user. The proposed method is implemented on custom hardware using a commercial UWB chipset. We evaluated the proposed method against the standard TDOA algorithm and range-based localization. In line of sight (LOS), the proposed TDOA method has a localization accuracy similar to the standard TDOA algorithm, down to a 95% localization error of 15.9 cm. In NLOS, the proposed TDOA method outperforms the classic TDOA method in all scenarios, with a reduction of up to 16.4 cm in the localization error.
Original languageEnglish
ISBN (Electronic)978-952-03-2938-9
Publication statusPublished - 2023
Publication typeG5 Doctoral dissertation (articles)


Dive into the research topics of 'Robust, Energy-Efficient, and Scalable Indoor Localization with Ultra-Wideband Technology'. Together they form a unique fingerprint.

Cite this