The third generation systems (e.g. FPLMTS) aim at fulfilling high requirements on flexibility, capacity, quality, coverage and cost. Long before the introduction of the third generation systems their need will be challenged by evolution of the existing cellular systems. The paper describes evolution of the GSM radio network towards a concept built on hierarchical cell structures with high capacity and simple or no frequency planning needed. It is a concept which conforms to the GSM recommendation and that uses random frequency hopping together with adaptive frequency planning. Them authors present solutions for the GSM radio network including control channels as well as the traffic channels. The concept has been verified by system simulations. Problems related to handover and frequency planning are presented and possible approaches are discussed. >