| dc.description.abstract | Vehicular ad hoc network (VANET) is a novel class of wireless network. Vehicles implemented with on board unit (OBU), which can communicate with each other by vehicle-to-vehicle (V2V) and vehicle to infrastructure (V2R) network architecture. In the V2V network environment, vehicular network topology is changing dynamically all the time with high mobility, the communications occurred in this architecture is intermittent pattern. To solve this problem, numerous researchers have proposed different kinds of VANET routing protocol to enhance the success of packets delivery ratio through choosing the path with high density. After determining the routing path, the choosing of relay nodes usually follow the mobile ad hoc networks (MANETs) routing protocol, that is, it selects the closest node with the destination or consider the power supply and computation ability of node. However, In MANET, mobile nodes are moving in a random and irregular pattern hence, it is very hard to predict their movements. Unlike unpredictable mobility in MANET, vehicles are limited by predefined road segments, traffic rules and driver behaviors. Furthermore, nowadays Vehicular on board units are able to equip a much more powerful vehicle rechargeable battery and a larger volume of communication device than before; therefore, it becomes inappropriate to take power supply volume and computing into account. Therefore, it is unsuitable to determine the relay node by these terms. What’s more important is that the vehicle moves too fast to make the relay node easily to leave the transmission range of sender, as a result, it wastes a lot of network bandwidth for route maintenance and re-discovery. In order to solve the above mentioned problems, this thesis proposed a mobility prediction-based relay selection scheme for stable connections in VANETs which takes both direction of vehicles and routing path into account. It makes packet routing and vehicles keep moving in the same direction, for reducing the link disconnection. This thesis takes GPS errors into consideration when determining the lane position to predict the vehicles mobility, for a even more accurate mobility prediction. Finally, the proposal is examined by the conducted simulation— the simulation results show that the proposed mechanism, on average, can increase connecting time by 25.6% but decrease number of link disconnection by 20.2%; furthermore, the prediction accuracy of connecting time is improved within the absolute error of 5.09%. This proves that in the multi-lanes scenarios, the proposed mechanism can provide even more stable connecting time than others.
| en_US |