A Hybrid Technique for Smooth Handover in Wireless Networks

Abstract

Mobile nodes (MNs) can access the internet through different wireless network interfaces such as wireless fidelity (WiFi), worldwide interoperability for microwave access (WiMAX), and cellular networks like long-term evolution (LTE), fifth-generation (5G) networks, etc. During an ongoing session, if the mobile user moves out of the coverage of one base station (BS) and enters into the coverage of another BS, then his continuous connectivity is maintained using handover. The handover process ensures the seamless switching of MNs among multiple networks without any service degradation. The handover process consists of three phases: handover triggering, network selection, and handover execution. Handover should be triggered at an appropriate time to provide a better quality of experience (QoE) to the mobile customers as well as to avoid mobility-related problems such as unnecessary handovers and handover ping-pongs. Moreover, the handover should be performed with the best available network which can fulfill the requirements of both the user and the system. In this thesis, handover triggering and network selection techniques have been developed to enhance overall network performance. The research work presented in this thesis is divided into four phases: In the first phase, a hybrid predictive handover technique based on long short-term memory (LSTM) and support vector machine (SVM) models has been proposed. A proactive handover technique reduces the handover latency and signaling overhead by predicting handover in advance. The selection of the best network with minimum handover latency provides seamless connectivity to the users. LSTM is used to predict the parameters of MNs such as location coordinates, speed, reference signal received power (RSRP), and reference signal received quality (RSRQ) at the next time step based on their values at previous time steps. The output of LSTM is passed as input to the SVM for the selection of the most appropriate network. The performance of the proposed approach is verified on the MIT human dynamics lab dataset on human behaviors and interactions using positional coordinates and on the 5G dataset using location coordinates along with speed, RSRP, and RSRQ of MNs. The experimental results revealed that the proposed approach has better performance as compared to Naive Bayes and stacked-LSTM approaches in the case of dataset1. The improvement in the validation and testing accuracy of the proposed method is 16% and 18.24%, respectively, as compared to the stacked-LSTM approach. The proposed approach has outperformed the Naive Bayes approach on the whole dataset2 in terms of validation and testing accuracy by 54.72% and 61.22% , respectively. In the second phase, a handover technique based on graph theory and matrix approach (GTMA) and Euclidean distance has been proposed. The handover triggering is performed by computing Euclidean distance among the values of various network parameters such as delay, jitter, packet loss rate, throughput, and price of the current serving network and the best alternative network selected by the network selection technique. If the Euclidean distance is less than the threshold value, then the MN stays connected to the current serving BS otherwise the handover is triggered and the handover request is sent by the MN to the serving BS. GTMA is utilized for ranking the alternative networks and selecting the best available network. GTMA does not explicitly compute the weights of the attributes. So, by employing the GTMA technique, the problem of ranking abnormality is significantly minimized as well as there is a reduction in the number of handovers. GTMA has reduced the number of handovers up to 75.61%, 85.71%, and 66.67% as compared to the traditional multi-attribute decision-making (MADM) methods such as analytical hierarchy process (AHP), a hybrid technique based on AHP and technique for order preference by similarity to the ideal solution (AHP-TOPSIS), and grey rational analysis (GRA), respectively. Using Euclidean distance as a handover triggering technique has further reduced the number of handovers in the case of GTMA and traditional MADM methods for conversational, interactive, streaming, and background traffic types. In the third phase, a key performance indicator (KPI) based technique has been proposed to select the appropriate values of handover control parameters (HCPs) such as handover margin (HOM) and time-to-trigger (TTT). In this technique, the network selection is performed using GTMA. KPI is the weighted average of three performance indicators (PIs) such as handover ratio (HR), handover ping-pong ratio (HPPR), and ranking abnormality ratio (RAR). A dataset is created by computing KPI values corresponding to different values of HOM, TTT, and MN speed in the case of background, conversational, interactive, and streaming traffic types. The optimal values of HCPs are determined by finding the minimum value of KPI corresponding to the different speeds of MNs varying from 10km/h to 120km/h for each traffic type. The simulation results have revealed that GTMA with optimal values of HCPs (GTMA-HCP) has improved the performance in comparison to AHP-HCP, AHPTOPSIS- HCP, and AHP-MOORA-HCP up to 7.76%, 13.48%, 9.96%, and 5.81% in the case of background, conversational, and interactive, and streaming traffic types, respectively. In the fourth phase, the impact of inter-next generation node B (inter-gNB) distance on end-user applications has been analyzed. It has been found that the packet loss, delay, and jitter are directly proportional to the inter-gNB distance whereas throughput is inversely proportional to the inter-gNB distance. An HCP optimization technique has also been proposed based on teacher-learner-based optimization (TLBO) and multi-armed bandit (MAB) framework. In this technique, firstly a dataset of dimensions 3000 ×11 is created using the standalone architecture of the fifth-generation (5G) new radio (NR) library of NetSim. In this dataset, 3000 is the combination of 10 user equipment (UEs) × 10 values of speed × 5 values of HOM × 6 values of TTT. During pre-processing, a new column packet loss is created by subtracting packets received from packets generated. Thus, four columns such as packets generated, packets received, payload generated, and payload received are dropped from the dataset. After pre-processing the dataset, the dimensions of the final dataset are 3000×8 (Speed, HOM, TTT, Destination ID, Packet Loss, Throughput, Delay, and Jitter). Secondly, the weights of the different network parameters such as packet loss, throughput, delay, and jitter are calculated using TLBO. Finally, MAB is used to select the optimal values of HOM and TTT by taking into consideration the speed of the UEs. The performance of the proposed approach is verified by creating three different network scenarios consisting of 30 UEs moving at different speeds. The simulation results have revealed that the proposed approach has improved the overall network performance as compared to different combinations of HOM and TTT namely, HCP1 (HOM=0dB and TTT=40ms), HCP2 (HOM=3dB and TTT=80ms), and Maxspeed approach, respectively in case of scenario-I, II and III. In this thesis, new handover strategies have been proposed to enhance the network performance. The results of the proposed techniques, including the predictive handover method, MADM-based approaches, and the simulation-based MAB technique, have shown that these methods are efficient and outperform the existing techniques. The effectiveness of the proposed approaches may offer valuable insights for addressing the challenges of handover management in wireless networks.