Multipath-assisted indoor positioning using radio frequency signals

Event details
Date | 22.10.2015 |
Hour | 14:15 |
Speaker | Dr. Paul Meissner, TU Graz |
Location | |
Category | Conferences - Seminars |
Multipath propagation is considered as the main impairment for radio-based positioning systems. Especially indoors, the large density of strong signal reflections lead to large biases in range estimates. This often leads to a insufficient robustness of the positioning algorithms.
In this talk, we show how deterministic multipath components can be used constructively for positioning instead of treating it as an error source. Using a floor plan and proper geometric and stochastic modeling of the radio signal received by an agent, signal reflections can be treated as additional range measurements to so-called virtual anchors. Performance bounds for the according position error allow for important insights in the role of deterministic multipath for positioning. We present multipath-assisted navigation and tracking (MINT) algorithms and show that the awareness to the uncertainties of the position-related parameters in the radio signal provides both accurate and robust position estimation.
Furthermore, we discuss simultaneous localization and tracking (SLAM) with MINT flavor. We show that a floor plan representation using virtual anchors can be learned online during the tracking of a moving agent. In this manner, still the same level of positioning accuracy as with a known floor plan can be achieved with almost no prior information.
Bio: Paul Meissner received the B.Sc. and MSc. degree (with distinction) in information and computer engineering from Graz University of Technology, Graz, Austria in 2006 and 2009, respectively. He received the Ph.D. degree in electrical engineering (with distinction) from the same university in 2014.
Paul is currently a postdoctoral researcher at the Signal Processing and Speech Communication Laboratory (SPSC) of Graz University of Technology. His research topics include statistical signal processing, localization, estimation theory and radio channel modeling.
In this talk, we show how deterministic multipath components can be used constructively for positioning instead of treating it as an error source. Using a floor plan and proper geometric and stochastic modeling of the radio signal received by an agent, signal reflections can be treated as additional range measurements to so-called virtual anchors. Performance bounds for the according position error allow for important insights in the role of deterministic multipath for positioning. We present multipath-assisted navigation and tracking (MINT) algorithms and show that the awareness to the uncertainties of the position-related parameters in the radio signal provides both accurate and robust position estimation.
Furthermore, we discuss simultaneous localization and tracking (SLAM) with MINT flavor. We show that a floor plan representation using virtual anchors can be learned online during the tracking of a moving agent. In this manner, still the same level of positioning accuracy as with a known floor plan can be achieved with almost no prior information.
Bio: Paul Meissner received the B.Sc. and MSc. degree (with distinction) in information and computer engineering from Graz University of Technology, Graz, Austria in 2006 and 2009, respectively. He received the Ph.D. degree in electrical engineering (with distinction) from the same university in 2014.
Paul is currently a postdoctoral researcher at the Signal Processing and Speech Communication Laboratory (SPSC) of Graz University of Technology. His research topics include statistical signal processing, localization, estimation theory and radio channel modeling.
Practical information
- General public
- Free
Organizer
- Ivan Dokmanic
Contact
- Ivan Dokmanic