In Mobile Ad Hoc NETworks (MANETs) autonomous nodes act both as traffic originators
and forwarders to form a multi-hop network.
Out-of-range nodes are reachable through a process called routing,
which is a challenging task due to the constraints of bandwidth and battery power.
Stateless location-based routing schemes have been proposed to avoid complex route
discovery and maintenance, whereby nodes make routing decisions based solely
on the knowledge of their location, the location of their neighbors, and the location of the destination.
Natural routing schemes based on these prerequisites suffer from problems like local maxima or loops.
We mitigate those problems by proposing randomized routing algorithms,
which outperform others in terms of the packet delivery ratio and throughput.
The prerequisite for location-based routing is knowing the location of a node.
Location information is more widely useful anyway for
location-aware applications like security, health care, robotics, navigation etc.
Locating a node indoors remains a challenging problem due to the unavailability of GPS signals under the roof.
For this goal we choose the RSS (Received Signal Strength)
as the relevant attribute of the signal due to its minimal requirements on the RF technology
of the requisite modules. Then profiling based localization is considered that does not
rely on any channel model (range-based) or the connectivity information (range-free),
but rather exploits the context of a node to infer that information into the estimation.
We propose a RSS profiling based indoor localization system, dubbed LEMON,
based on low-cost low-power wireless devices that offers better accuracy than other RSS-based schemes.
We then propose a simple RSS scaling trick to further improve the accuracy of LEMON.
Furthermore, we study the effect of the node orientation, the number and the arrangement
of the infrastructure nodes and the profiled samples, leading us to further
insights about what can be effective node placement and profiling.
We also consider alternate formulations of the localization problem,
as a Bayesian network model as well as formulated in a combinatorial fashion.
Then performance of different localization methods is compared and again LEMON ensures better accuracy.
An effective room localization algorithm is developed, and both single and multiple
channels are used to test its performance. Furthermore, a set of two-step localization
algorithms is designed to make the LEMON robust in the presence of noisy RSS and faulty device behavior.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:AEU.10048/1841 |
| Date | 06 1900 |
| Creators | Haque, Israat Tanzeena |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | 1245935 bytes, application/pdf |
Page generated in 0.0019 seconds