Direction of Arrival (DoA) Estimation Under Array Sensor Failures Using a Minimal Resource Allocation Neural Network

This paper presents the use of a minimal resource allocation network (MRAN) for the direction of arrival (DoA) estimation under array sensor failure in a noisy environment. MRAN is a sequential learning algorithm in which the number of hidden neurons are added or removed based on the input data and produces a compact network. The training for MRAN is done under no failure and no noise case and the trained network is then used when there is a failure. Thus, the need for knowing the element and the time of its failure, as required in other methods is eliminated. MRAN's performance is compared with the conventional MUSIC algorithm and also the radial basis function neural network scheme developed by A. H. El Zooghby under normal and failed cases. In normal case, different antenna effects like mutual coupling, nonuniform array and unequal source power have been studied under different signal to noise ratio (SNR) values. Results indicate the superior performance of MRAN based DoA estimation scheme under different antenna effects, failure conditions and noise levels