Network Localization and Navigation: Established a framework for cooperative network localization and navigation. Determined the fundamental performance limits, developed distributed filtering algorithms, and designed error mitigation methods using measurement data. Cooperative network localization and navigation can achieve reliable and accurate location-awareness, particularly in harsh environments. Specific contributions include:
- Fundamental Performance Bounds: Put forth the notion of equivalent Fisher information (EFI) and determined the fundamental limits of network localization and navigation. Applied EFI analysis to decompose the contributions from spatial and temporal cooperation into basic building blocks associated with each measurement. Developed a geometrical interpretation for localization information, and derived scaling laws of the network localization performance.
- Scalable and Distributed Algorithms: Developed a scalable, distributed localization algorithm for large cooperative networks using factor graphs and belief propagation. Designed an iterative procedure for nodes to improve their location estimates using belief messages from neighboring nodes. Evaluated the performance of these algorithms based on wideband ranging techniques.
- Generalized Filtering Techniques: Established a general framework for parametric filtering techniques, which enfolds conventional filtering techniques such as Kalman and particle filters, as well as their variants. Introduced a new paradigm called the belief condensation technique that accurately approximates complex statistical distributions, arising in the filtering process, by tractable distributions. Developed filtering techniques for nonlinear/non-Gaussian navigation systems using belief condensation.
- Non Line-of-Sight Ranging Error Mitigation: Developed non line-of-sight (NLOS) mitigation algorithms, using machine learning techniques, which are capable of 1) assessing whether a signal was transmitted in line-of-sight or NLOS conditions, and 2) reducing ranging error caused by NLOS conditions. Performed an extensive measurement campaign on the MIT campus using commercially available FCC-compliant UWB radios. Evaluated the performance of these algorithms using experimental data.