Merging CBR and Neural Networks for SLA-Based Radio Resource Management for QoS Sensitive Cellular Networks

We propose a radio resource management (RRM) approach to guarantee a predefined service level agreement (SLA) with different classes of users, for on-going and in-coming connections in QoS sensitive cellular networks. This approach is based on intelligent agent architecture which gives autonomy to radio network controller (RNC) or base station (BS) in accepting, rejecting or buffering a connection request to manage system capacity. Instead of simply blocking the connection request, different buffering times are allocated to different classes of users based on their SLA. This increases the chances of connection establishment and reduces the call blocking rate extensively. In this paper, we have described a merger of two AI (artificial intelligence) techniques, case based reasoning (CBR) and neural networks (NN). CBR is a library that stores previous successful decisions as cases and allows the agent to deal with current congestion situation efficiently. We have introduced a new concept of training neural network on the records of CBR library to avoid sequential search. In this way, an intelligent solution is proposed to control call blocking rate which is not only accurate but also time efficient