الفهرس 
Title : An Improved Approach for Relative Positioning in Mobile Vehicle Ad Hoc Networks using Soft Computing
AbstractOnly 14 pages are availabe for public view
 |  | from | 157 |  |  |
| | | from | 157 | | |
Abstract
Modern vehiclesare equipped with global position systems (GPS) and inertial navigation system. They providethe location information in order to deliver various services to drivers and passengers.Signal blockage, multipath, non-line of sight, propagation delay and noise environment plague GPSespecially in tunnel and urban environments lead to reduce the positionaccuracy. Inertial navigation system (INS) consists of a set of customized sensor nodes such as accelerometers, gyroscopes, compasses, and odometers, are attached to every vehicles to measure their speeds and orientations. The accuracy of such measurements is closely related to the cost of them, systematic errors and random noise errors. These errors are accumulated over passed time. Its accuracy is valid for a short period time.
Typical integration solutions between the GPS as well as INS are unable to achieve the high levels of localization accuracy. They still affect with the accuracy degree of collected measurements. In addition, they consume large processing time according to the length of state vector.
Cooperative localization methods are alternative solutions for above methods. They provide less cost and rich measurements. Vehicles collaborate with each other i.e., called vehicle to vehicle communication (V2V) or with roadside units i.e., called vehicle to infrastructure communication (V2I) to exchange mobility measurements and use them to find more accurate positions. Most modern vehicles are equipped with GPS receiver so they can be considered as mobile roadside units. The estimated vehicle position via V2I communication is higher accuracy than V2V communication. The reasons are returned to the ability of predicting errors in measurements and the nonlinear localization nature. Localization based on V2V communication lead to accumulation errors for a long period.
Data fusion algorithms consider as a partner to enhance vehicle position. Data fusion algorithms give accurate results in case of adjust their parameters, noise covariance and defined the probability distribution that are difficult in the reality.
In this thesis, two localization approachesare proposed in vehicular networks. The first approach is called cooperative localization based on V2I communication and distance information (CLV2I).CLV2I aims to enhance the initial estimated position, handle the nonlinear localization problem and overcome the failure roadside unit. CLV2I assumes estimated method for virtual roadside to handle fault environments. CLV2I uses extended Kalman filter algorithm to handle nonlinear problem. In addition,the general formula of intersection method is utilized to get better vehicles’ initial estimated positions.
The second approach is called fuzzy cooperative localization based on V2I communication and distance information (FCLV2I).FCLV2I aims to adopt the noise in measurement covariance and handle nonlinear localization nature. In addition, Kalman filter is a parametric algorithm. It needs a well estimating for initial state vector, error and noise covariance in process and measurements. Fuzzy intersection method is a fuzzy formula of classical intersection method to get well accurate estimated for initial state vector. Fuzzy inference system is used to update the noise covariance continually. The parameter of fuzzy inference system is tuning by fuzzy clustering based on fuzzy weight expected value algorithm without the needing to expert systems. The number of rules is minimized based on similarity measure.
The experimental results confirm that the two proposed approaches havesupreme accuracy than existing related methods. The position accuracy is evaluated by root mean square errors. The CLV2I approach achieves improvement in position accuracy by 50% and 62% respectively in simple scenario and 28% and 44% respectively in complex scenario compared with other approaches found in the literature. The FCLV2I approach achieves improvement in position accuracy by 75% and 81% respectively in simple scenario and 71% and 77 % in complex scenario compared with other approaches found in the literature.