In Radio Access Networks (RAN), fixed configurations result in poor network efficiency. This suboptimal performance is a direct result of user mobility and traffic patterns which changes over time and space. In this paper, we present a framework for a self-optimizing RAN, which adapts Antenna Tilt and Pilot Power according to the current load in the system. The framework uses distributed optimization and network performance-based optimization triggers. Additionally, we introduce the concept of Coupling Matrix, to avoid the traditional global network optimization. For realistic evaluations, we simulated our framework using a moving hot-spot user traffic model. This enables our framework to work under non-uniform traffic distribution with dedicated channels in a WCDMA network. The simulation results show significant performance gains, while ensuring the scalability and reduced complexity of the optimization problem.