Towards Control and Coordination in Cognitive Autonomous Networks

Introduction of Artificial Intelligence (AI) and Machine Learning (ML) in mobile networks helped in achieving a great degree of automation through Cognitive Autonomous Networks (CAN). In CAN learning based functions, called Cognitive Functions (CF), adjust network control parameters to optimize specific Key Performance Indicator (KPI), which are the CF's objectives. The CFs work in same level in the hierarchy sharing the same resources, which very often introduces an overlap among their target control parameter adjustment, i.e., at one point of time, multiple CFs may want to change a single control parameter albeit by different degrees or to different values depending on their respective level of interest in that parameter. Correspondingly, a Controller is required in CAN to coordinate the sharing of the parameter among the independent CFs according to those varying interests. In this paper we study two of such existing Controllers and how disadvantages of one leads to development of the other. We study both the Controllers in a simulation environment and compare the performance and time complexity between them.