HANDOVER IMPROVEMENT USING NEAR REALTIME RADIO INTELLIGENT CONTROLLER

• Main Author: Haryo Prananto, Baud
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/80807
• Institution: Institut Teknologi Bandung

In cellular communication technology development, especially in LTE and 5G technologies, there are tendencies to use a higher frequency spectrum. While LTE is still limited up to 3 GHz, 5G is already standardized ta o maximum 52 GHz. High-frequency usage is based on spectrum channel availability and the ability of high-frequency spectrum to carry data faster. However, high-frequency spectrum usage results in smaller cell sizes since the wave coverage becomes shorter. Small cell implementation forces operators to deploy more cells to cover an area. This multiple small cell implementation benefits in higher data rate and network capacity. The problem is for high mobility users, such as users in vehicles, must do handover every time moving from one cell to another cell. Handover activity causes delay and data loss potential in high-mobility users. This is because the source base station needs to communicate with other network elements, such as the target base station and MME in the core network, to do coordination for moving the user to another cell. The signaling communication burden will also increase if there are more handover activities. Delay and data loss potential due to handover can be avoided by several means. There are methods to reduce or eliminate handover activities such as Heterogenous Network implementation (where small cells are covered by macro cells at the same location) or Single Frequency Networks (some cells have the same identity and therefore are considered as same cell). There are also methods to reduce delay and increase reliability in every handover activity. Studies show that increasing the handover reliability is better done by choosing the correct target cell to reduce the transmission failure probability after handover. The traditional handover algorithm is proven reliable to determine the target cell in an ideal network condition. However, in non-ideal network cases, this algorithm faces many failures. This research will reduce the delay and increase the handover reliability by utilizing Near Real-Time Radio Intelligent Controller (Near-RT RIC), a new network element based on Open Radio Access Network (O-RAN) standard. In Near- RT RIC, a machine learning algorithm is run to help determine the best target cell during the handover process. Simulations of non-ideal network cases were also done to prove the usability of Near-RT RIC to improve the handover reliability. The reliability is measured in the data transmission success rate after handover. The methods used to determine the target cell in Near-RT RIC are vector autoregression (VAR), Multi-Layer Perceptron (MLP) neural network, dan Long Short-Term Memory (LSTM) neural network. These methods are tested in several non-ideal network simulations and are proven can increase the data transmission success rate compared to the traditional handover algorithm. Based on simulations in this research, it is proven that in a non-ideal condition, a network with coverage holes, the data transmission success rate in a handover case with a traditional handover algorithm only reached 86.2%. That means there is a possibility that the handover process can result in data transmission failure if the target cell is determined by the traditional algorithm. When the target cell was determined by Near-RT RIC, the data transmission success rate increased to 95.3% if using VAR method, 91.9% if using MLP-NN, and 97.6% if using LSTM-NN. However, there are still some challenges on algorithm complexity and data processing speed that may become a room of improvement.