الفهرس Only 14 pages are availabe for public view
 |  | from | 173 |  |  |
| | | from | 173 | | |
Abstract
Target tracking is a desired application in wireless sensor networks (WSNs). Tracking includes Target detection, localization algorithm, estimation of next position, tracking protocols, and classification. The desired objective of this study is introducing higher accuracy at low power consumption. The wasted powers in the three main parts of tracking process are improved. These parts are localization algorithm, tracking protocol and filters involved in prediction process. Also introduce the effect of tracking parameters and tuning them to achieve higher accuracy at low power consumption. These main parts have their own survey and new proposed solution. Their main contributions are higher accuracy and stability against noise with minimum power consumption.
The first part introduces an efficient real time localization algorithm with static tracking protocol for multi-target tracking in non-uniform network. This part provides new issues as:
• Scalable architecture for coordinating a non-uniform sensor network for the purpose of maneuvering multi-target tracking.
• The feasibility, minimization of computation and communication overheads by tuning system parameters and understanding the tradeoffs in such systems
• The effect of network parameters on accuracy and communication cost.
• This study introduces graphs and equation can help to estimate the required network specification for certain accuracy and life time.
The proposed localization algorithm combines both multi-dimension scaling and least mean square methods. This combines with the introduce confidence number overcome low connectivity with higher accuracy than previous work. It improves the localization error by 30% as the distance error 10% which is double the previous distance error. Also it can reach low connectivity regions and measure the communication power normalized by mean number of protocol communication.
The second part proposes and evaluates a distributed, energy-efficient, light-weight framework dynamic clustering algorithm for target tracking protocols in wireless sensor networks. Since radio communication is the most energy-consuming operation, this framework aims to reduce the number of messages and the number of message collisions, while providing refined accuracy. In this part the network architecture of Khin Thanda is considered. It is a hierarchical sensor network that is composed of a static backbone of sparsely placed high-capability sensors. That will assume the role of a gateway upon triggered by certain signal events and moderately to densely populate low-end sensors. The tracking protocol is adaptive to the target velocity to ensure high accuracy with minimum power consumption. The study concludes a simple equation for Reporting Frequency including the velocity and time step.
Two elections protocols for Source Node (SN) are proposed. The SN its function is to provide sensor information to their gateway upon request. The study introduces the retransmission rates and their power consumption using NS2 simulator. The introduced study is compared with the adapted protocol of Elham and dynamic clustering of In-Sook Lee. Using NS2 and Matlab to introduce total power consumption for different proposed protocols, total number of messages and also clustering power. The total power reduction is 24%. Taking into consideration the dynamic clustering power the power consumption
The third part, Filtering and Estimation of the next position is introduced for nonlinear models. Two methods are introduced to overcome the difficulty of non-linear model. The first method uses Interacting Multiple Model (IMM). IMM estimator is a suboptimal hybrid filter that has been shown to be one of the most cost-effective hybrid state estimation schemes. The main feature of this algorithm is the ability to estimate the state of a dynamic system with several behavior modes which can “switch” from one to another. The second method uses Second order Extended Kalman Filter (EKF2) which is a single nonlinear filter. Both methods are evaluated by simulation using two scenarios. A comparison between them is evaluated by computing their accuracy, change of operation range and computational complexity (computational time) at different measurement noise. Based on this study which follows the previous studies, IMM provides better accuracy than EKF2 at maneuvering. Tracking maneuvering targets introduce two major directions to improve the Multiple Model (MM) approach: Develop a better MM algorithm and design a better model set. The study shows that the use of too many models is performance-wise as bad as that of too few models. Based on that comparison a new proposed Hierarchal Switching set of IMM (HSIMM) is implemented. HSIMM overcomes the leakage of both classical IMM and Variable IMM. In HSIMM the models is divided into a small number of sets, tuning these sets during operation at the right operating set. HSIMM is tested over three different types of target models. HSIMM introduce reduction in error by 19% for the first part of target kinematics and computation time by 26%. The second part of kinematics VIMM and IMM have minimum x, y position but higher velocity error compared to HSIMM. The computation time is reduced from 0.0073 to 0.0019 clock click for both target scenarios. The computational time is minimum than introduced by IMM and VIMM. HSIMM introduces less error as the noise increase and there is no need for re adjustment to the Covariance as the noise increase so it is more robust against noise and introduces minimum computational time.