Type
Text
Type
Dissertation
Advisor
Robertazzi, Thomas | Wang, Xin | Milder, Peter | Rao, Nageswara.
Date
2015-08-01
Keywords
Electrical engineering
Department
Department of Electrical Engineering.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77475
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
In a long-haul sensor network, sensors are remotely deployed over a large geographical area to perform certain tasks, such as tracking the states of one or more dynamic targets. The communication loss and delay inherent over certain types of long-haul links, such as the satellite channels, pose severe challenges to the successful execution of such tasks. First, free and complete information flow over the network is constrained, as a direct result of the lost and delayed messages over the imperfect communication links. Next, the system is often error- and delay-tolerant to a certain extent, but still the underlying tracking tasks are to be finished within a tightly stipulated deadline with outputs, i.e. | the state estimates, that satisfy the system requirement on maximum tolerable error. In addition, due mainly to the limited computational capabilities of the entities within the network, information processing may be inaccurate, inefficient, and time-consuming as well, which in turn negatively affects the final tracking performance. Sensor fusion is a powerful means to aggregate information from multiple sources so that the fused information is expected to possess a much higher quality than that at any of the sensors. However, the above communication and computation constraints over long-haul sensor networks may severely limit the potential of sensor fusion, resulting in reduced fusion gain and degraded estimation performance. In this dissertation, various approaches are proposed to counteract the effects of imperfect communication and computation in long-haul sensor networks, so that the performance of the tracking tasks in the context of sensor fusion can, even under extremely adverse conditions, be guaranteed to meet the system requirements on both accuracy and timeliness. Some of the solutions are message retransmission and retrodiction, information-driven selective fusion, staggered and asynchronous sensing scheduling, information feedback, as well as learning-based fusion design. Extensive analytical and simulation studies specifically for different types of target tracking applications over satellite-based long-haul sensor networks are carried out to demonstrate the major benefits as well as limitations of adopting these approaches. | 181 pages
Recommended Citation
Liu, Qiang, "State Estimation and Fusion in Communication- and Computation-Constrained Long-Haul Sensor Networks" (2015). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3287.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3287