ABSTRACT

A routing protocol for mobile ad hoc networks should minimize route setup and maintenance messages, in order to have a low communication overhead. The simulation results show that location stability information can be a useful metric, although it must be used carefully since misinformation about stability patterns is considered quite costly and has negative impact on routing and performance. In order to remove the need for localization and minimize the impact of localization errors, virtual coordinate systems have been proposed as an alternative to geographic location in geographical position based routing algorithms. In order to use Bayesian inference for this purpose, three things are needed-a BN designed to model a path in a Mobile ad hoc network, statistical data to create the joint distributions, and a routing protocol designed to collect the observed variables for each discovered route. The routing metrics used in Bayesian network routing design are presented later.