A GPU-Based Parallel-Agent Optimization Approach for the Service Coverage Problem in UMTS Networks

In the context of coverage planning and control, the power of the common pilot channel signal determines the coverage area of a network cell. It also impacts the network capacity and thus the quality of service. We consider the problem of minimizing the total amount of pilot power subject to a full coverage constraint. Our optimization approach, based on parallel autonomous agents, gives very good solutions within an acceptable amount of time. The parallel implementation takes full advantage of GPU hardware in order to achieve impressive speed-up. We report the results of our experiments for three UMTS networks of different sizes based on a real network currently deployed in Slovenia.