Time: 14/05/20 h. 15:00
PhD Candidate: Pablo Corbalán Pelegrín
Abstract of Dissertation:
Global navigation satellite systems (GNSS) have radically changed business, industry, and society, shaping the way we transport, navigate, and generally live every day. After all these years, however, GNSS location information remains only valuable outdoors, leaving indoor environments where people dwell most of the time without proper localization support. Many technologies and systems have approached this problem including optical, inertial, ultrasonic, and radio-frequency (RF), to name a few; yet the problem remains.
In this thesis, inspired by the indisputable success of GNSS and the re-emergence of ultra-wideband (UWB) radios to the forefront of technology, we aim to change the state of affairs in RF localization by proposing novel clean-slate UWB ranging and localization schemes based on concurrent transmissions. These are generally considered harmful for communication but become a rich source of localization information when combined with knowledge of the channel impulse response (CIR).
Our first novel contribution lies in the concept of concurrent ranging, which allows mobile nodes to simultaneously measure the distance to multiple devices—hereafter, called responders—removing the need for the wasteful long packet exchanges traditionally used for ranging and localization. Different from conventional schemes, which spread responder transmissions over time, we force responders to transmit concurrently and let their signals “fuse” in the wireless channel; the resulting impulse response, as measured by commercial UWB radios, contains all the necessary timing information to extract the desired distance to all responders. This first contribution, however, also serves us to realize the many challenges ahead to unlock the real power of concurrent transmissions for localization.
We address these challenges along the way, starting with Chorus, our second contribution. Chorus exploits an anchor infrastructure that transmits packets concurrently. Mobile nodes listen for these transmissions and measure from the CIR the time difference of arrival (TDoA) of the concurrent signals, privately computing their own position at a high rate using hyperbolic localization. This reverse TDoA scheme, although simple in concept, is extremely powerful in that it enables passive self-localization of infinitely many targets at once, a feature largely missing in the RF literature. In Chorus, we address the difficult challenges to reliably detect and identify the signal from the different responders. Yet, the limited transmission precision of commercial UWB transceivers constrains the many benefits of Chorus.
In this context, we i) contribute a model to ascertain the impact of the transmission uncertainty on concurrent transmissions, and ii) address the issue with a compensation mechanism that fine-tunes the local oscillator frequency of responders while they prepare to transmit, allowing us to simultaneously tackle the impact of clock drift on distance estimation. We demonstrate in our evaluation that with this compensation mechanism we can schedule transmissions with < 1 ns error, removing the need to share timestamps to precisely measure distance. We rebuild concurrent ranging around this mechanism, obtaining decimeter-level ranging and localization at a fraction of the cost of conventional schemes. These results turn concurrent ranging into an immediately applicable technique that new systems can now exploit, benefiting from a different set of trade-offs hitherto unavailable. Further, the TX compensation mechanism can be directly applied to Chorus, similarly making fast and accurate passive self-localization a tangible reality.
We continue our endeavor with a systematic characterization of the conditions under which UWB concurrent transmissions succeed to provide reliable ranging and communication across different complex channels. The results we put forth empower developers to fully exploit concurrent transmissions in their designs, potentially inspiring a new wave of ranging, and also communication, primitives that can bring to UWB the same striking benefits found in low-power narrowband radios.
The thesis is completed by looking at other challenges preventing the wide adoption of UWB localization systems, namely, large-scale operation, energy efficiency, and the complexity to install anchor deployments. We tackle these aspects in the last part of the thesis with three additional contributions. First, we propose Talla, a TDoA system that provides seamless large-scale localization for many tags across cells of time-synchronized anchors. Secondly, we fuse UWB ranging with odometry information and build an uncertainty model that only triggers new UWB estimates if and when needed, reducing consumption and channel utilization while satisfying the application-specific demands in terms of accuracy. And thirdly, we build state-of- the-art mechanisms to automatically compute the positions of all anchors deployed across large areas based on ranging information, facilitating anchor network deployment for the many UWB-based real-time location systems (RTLS) to come.
Overall, this thesis changes the landscape of UWB localization with a new set of potentially disruptive schemes and systems that exploit the peculiar benefits of concurrent transmissions and that consequently, redefine the trade-offs of the technology.
Contact: ict.school [at] unitn.it (ICT International Doctoral School)