Given the increasing demand of services on the move, the mobile operators need to increase the capability of 3G systems 
constantly. Sometimes, the only possible solution for improving the system performance is to increase the number of 
transmitters, being this an expensive approach. Besides, in UMTS utilizing more macro-cellular sites does not always entail 
an increasing system capability if they are too near, since mutual interferences can harm the global operation. In these 
cases the only solution is to make use of hierarchical cell structures (HCS), adding micro-cells transmitting at different 
carriers.

This Thesis provides a complementary point of view of the HCS deployment and proposes to take advantage of such deployment 
to improve the UMTS radio access network with the introduction of a new architecture of B-nodes or Multinodes B.

The system to be developed uses UMTS B nodes where several spatially distributed antennas are connected. The Multinode 
antennas cover the service area and form an equivalent to a spatially distributed array. An optimum radio access structure 
for indoor and micro-cellular environments, with high adaptability and flexibility, is achieved when array processing 
techniques, optical links and dynamic resource allocation algorithms are incorporated to the Multinode B.

Advantages of using this configuration, as described along this Thesis, are mainly on improving the capacity of the radio 
access network, minimizing the radioelectric impact and optimizing the resource allocation.

Firstly, in this Thesis it is demonstrated that the new radio access architecture is able to provide an additional spatial 
diversity to the UMTS system, diversity that reduces the power required by the terminals, most of all at micro-cellular 
scenarios.

Regarding dynamic resource allocation (DRA), an efficient scheduler based on Hopfield neural networks (HNN) has been 
proposed to support Quality of Service (QoS) for different services. This new scheme provides an optimum user bit rate 
allocation together with a controlled delay and a high utilization of system resources. The proposed algorithm improves 
all analyzed performance indicator over other conventional algorithms. Moreover, another benefit of using a HNN-based DRA 
algorithm is precisely that its hardware implementation makes possible a real-time running of the algorithm. 

On the other hand, the HNN distributes in an optimal manner the signal transmitted to the user among the antennas included 
in the Multinode B. Thanks to the spatial and the optimal DRA, the power consumption can be divided by five. In addition 
to this reduction in the power, QoS is also improved, decreasing the number of queued users and the average service response 
time.

Finally, in this Thesis it has been proven that the Multinode B system could be implemented in the mean term and the 
implementations costs are lower than other more complex alternatives found in the literature.