No abstract

Two main parameters influence the performance of Handover (HO) in Cellular systems, the Hysteresis (Hys) and the Time-To-Trigger (TTT). In Self-Organized Networks (SON) as has been proposed for LTE, the SON methods configure and optimize these two parameters to optimize HO performance. In this paper, we model the LTE HO procedure and propose an algorithm for self-configuration of the two HO trigger parameters. We model HO performance in terms of HO rate, the Ping-Pong rate, and the Radio Link Failure rate. We observe that the optimum trigger combination can occur anywhere within the parameter space depending on the relative importance given to each of the performance metrics as implemented using a weighting function. This motivates our HO configuration algorithm that searches for optimum parameter values by undertaking optimization one parameter at a time.

Optimal settings for Handover parameters (Hysteresis and Time-to-Trigger) depend on user velocities in the network. The Self-Organization Networks (SON) standard defines the Mobility Robustness Optimization (MRO) use case for the autonomous methods of configuring the parameters in congruence to the mobility pattern. State of the art MRO solutions have relied on expert knowledge, rule based algorithms to search the parameter space; yet it is unwieldy to design rules for all possible mobility patterns in any network. In this work, we present a Q-learning MRO solution, QMRO, which learns the required parameter values appropriate for specific velocity conditions in the individual cells. We compare QMRO against the best static reference configuration (Ref) that is obtained by sweeping the parameter space. Our results show that QMRO is able to learn parameter settings that achieve similar performance to Ref in a realistic network environment where users have dynamically varying velocities.

Cellular radio networks are seldom uniformly loaded. This motivates the need for Load Balancing (LB), as has been defined in the LTE Self-Organization standard. It is expected that on overload, a serving cell (S-cell) initiates LB to transfer some of its edge users to its neighbor cells so called target cells, by adjusting the Cell Individual Offset (CIO) parameter. In this work, we have proposed a reactive LB algorithm that adjusts the CIOs between the S-cell and all its neighbors by a fixed step . Our results show that the best depends on the load conditions in both the S-cell and its neighbors as well as on the S-cell's user distribution. We then propose a Q-Learning (QL) algorithm that learns the best values to apply for different load conditions and demonstrate that the QL based algorithm performs better than the best fixed algorithm in virtually all scenarios.

No abstract

ing network management automation using machine learning based functions called Cognitive Functions (CFs). Thereby, the CFs interact with the environment to learn and decide suitable network configurations to optimize their objectives. To minimize conflicts among the actions of multiple CFs, the CFs send their proposed configurations to a Controller, which in turn computes the final value that optimizes the combined interest of all the CFs. However, simulating real-life CAN deployments is challenging since: (a) the CFs have to reactively and interactively learn on the underlying system, and (b) the Controller must compute the optimal configurations in a dynamic environment with interdependent functions. In this letter, we present a scheme for implementing CAN in a simulation environment highlighting the critical design aspects for generating expected outcomes. Our results validate the proposed implementation design as the desired realistic behavior is obtained from CAN. K E Y W O R D Snetwork management automation, neural networks, self organizing networks INTRODUCTION AND MOTIVATIONCognitive Autonomous Networks (CAN) 1,2 is a promising solution to the requirement of high degree of network automation required in 5G and future networks. CAN increases the degree of autonomy by incorporating various Artificial Intelligence (AI) and Deep Learning (DL) functionalities. A typical mobile network is characterized by several control parameters (CPs) and Key Performance Indicators (KPIs). These CPs are used to modify the characteristics of the network, and the performance of the network can be observed from the KPIs. For example, in a 5G base station (gNB), there are several CPs for example, downlink transmission power (TXP), antenna tilt or remote electrical tilt (RET), and, several observable KPIs for example, downlink average user throughput, radio link failures (RLF), etc. When one (or, multiple) CP(s) is (are) changed, one or multiple KPIs are also affected, for example, downlink average user throughput is affected by TXP and RET. Network performance is optimized when the individual KPIs are optimized. To optimize a KPI, first a specific set of CPs have to be identified, which has an influence over that KPI. After that, values of the CPs in that set have to be adjusted in such a way that the KPI is optimized.In CAN, AI/DL based autonomous functions, called Cognitive Functions (CFs), are used to optimize each KPI. The KPI is called the output of the CF and the set of CPs, which influence that KPI, are called the inputs of the CF. For example,

No abstract

No abstract