1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Enhancement of geometrical predictive handover probability based on coverage sectors
Authors:	Sulaiman Mohammed Alsubaie (P74472)
Supervisor:	Nor Effendy Othman, Dr.
Keywords:	Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia Mobile Internet Protocol Networks Access point
Issue Date:	26-Jan-2018
Description:	Mobile Internet Protocol (IP) handover refers to the seamless communication link change of one node from one access point to another. It is useful for preventing disruption in communication sessions in general, and it has a significant impact on the performance of the vehicular networks for the frequency of an occurrence in particular. This is important for different networks in general and vehicular networks in particular due to the high dependency of different vehicular and intelligent transportation of the internet. Thus, it is highly motivational applications to develop an efficient handover system for vehicular networks. The problematic aspect of vehicular handover is the non-accurate location information that might be provided to the handover because the non-accurate Global Positioning System (GPS) signal in an urban environment especially when the environment is occupied by the tall structures. Hence, it is essential to develop vehicular handover from the perspective of location prediction to assist in correct prediction to the next access point (AP). The two issues of mobile IP handover are in the latency and the possible loss of packets. Most previous studies have concentrated on the architecture aspect of the mobile IP to resolve this problem. Despite the effectiveness of such solutions they do not target directly the latency problem caused by the handover. In this research, a probability based geometrical model is developed for prediction of next AP based on logged information about the history of vehicles mobility in the road with respect to current AP. The methodology is based on dividing the coverage area around each AP to set of sectors and building dynamic probability table about the mobility of the vehicle from one AP at particular sector to the predicted AP. For further improvement in the performance, Kalman filter has been incorporated into each vehicle for accurate prediction of the vehicle location in the coverage zone. Simulation results have proven that our model outperformed the previous models in terms of all evaluation measures of the network performance: Packet Delivery Ratio (PDR), End to End delay (E2E delay), and overhead. The developed approach has shown performance improvements from different perspectives. The developed approach has improved the average PDR over time up to 23.54%. Also, it appears our approach has decreased the overhead by 29.5% over time while the E2E delay over time has reduced up to 87.31%. In term of the number of nodes, the average percentage of improvement in PDR for the developed approach has reached up to 22.32% improvement. Furthermore, the decrease in overhead has reached up to 30.97% for the developed approach and the decrease in the E2E delay has reached 80.97%. Moreover, the average improvement percentage in PDR for the developed approach has reached up to 47.48% in term of velocity. The highest improvement percentage for the developed approach in term of overhead based on velocities was 30.96%. The average decrease in E2E delay for our approach was up to 99,89 %. Furthermore, the developed approach has shown improvement regarding the number of sectors. The improvement percentage in PDR for the developed approach based on the number of sectors has reached up to 22.32%. While the average decrease percentage of overhead was up to 30.96% and the E2E delay has reached up to 80.97%.,Master of Computer Science
Pages:	86
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.