L Introduction. Today's wireless systems are requred to satisfy an increasing demand for coverage, capacity, and service quality. Advanced signal processing techniques are combined with antenna atrays collcepts to produce some promising innovative solutions. Existing wireless systems cannot effectively address problems such as cochannel interference (CCr). Cochannel interference is the most serious limiting capacity fktor in any mobile communication system. As the number of users increase, within a certain region, the likelihood of lnterfenng witb one another increases. In order to solve the CCI problem, first a superresohaion Direction Of Arrival @OA) algorithm is utilized to locate the desired as well as the whannel mobile users. Next, an adaptive array antenna can be used to steer its radiation beam towad the mobilesofinterest[1]audnullstowardthe othersourcesofinterfereace inthesamefirequency slot.Currently, several algorithms can be used to w o r m the direction admg or angle of arrival of signals from mobile users. One drawback of these algorithms is their difficulty of their implementation in d -t i m e because oftheir intensive computational complexity. Neural networks, on the other hand, due to their high-speed computational capability, can yield results in real-time. Moreover, conventional beamformers require highly calibrated autennas with identical element pmpedes. Performance degradation often occurs due to the fact that these algorithms poorly adapt to element failure or other sources of erron. Neural network-based anay antennas do not suffer from this shortcoming.II Neural-network based Direction of Arrival Estimation. Both problems, DOA and null steer& @cam Steering), are approached as a mapping which can be modeled using a suitable artificial neutal network trained with input output pairs. The network is then capable of estimating or predicting outputs not included in the learning phase through generalization. Here, the neural network of choice is the radial basis function neural netwok (RBF") [2], shown in Figure 1 with its input preprocessing and output post -processing sections. For the DOA problem, the anay performs the mapping G R ' +CM fium the space of DO& {e= [e, ,e2 ,-. . , e, ] 1 to thespace of sensor output IS = [ S, , s,, . . -4 1 namely: k=l 0-7803-5639-x/99/