Sequential detection of DOA in a mobile communication network with directive/adaptive antennas

For hostile communication environments, a non-centralized mobile communication network with directive/adaptive antennas and Global Positioning System (GPS) has been proposed for the purpose of minimizing the radio exposure and making it difficult for an interceptor to detect the signals. This thesis investigates the interceptor's search for the direction-of-arrival (DOA) of the transmitting signal by means of a sequential test and decision when the interceptor also uses a single beam or a double beam directive antenna. Bayes' rule is adopted to recursively update the directional a posteriori probabilities of the transmitter direction, which leads to a sequential decision problem. After making an observation and updating the probability distribution of the DOA, the interceptor decides either the direction it will use for the next search (action) or it stops searching and makes a decision (test).;The dynamic programming algorithm is adopted to obtain the optimal action and test to minimize an objective cost function. This can be set up in many different ways according to different criteria. In this thesis, two objective cost functions are considered: (1) the weighted sum of mean observation number and the average decision error probability and (2) the mean observation number alone with a constraint on the decision error probability.;It is proved that the successive mapping makes some costs converge to the optimal cost and that there exists a stationary optimal policy. To deal with the unknown parameter, the received signal-to-noise ratio (SNR), we propose to take an average when a single beam antenna is used, and to estimate it when a double beam antenna is available.;The costs of a one-step-look-ahead policy and the maximum a posteriori policy are also computed by the dynamic programming algorithm.;Numerical solutions are calculated with a simplified antenna beam shape and they are compared for different policies and different antenna systems. Simulations show that the optimal policy obtained from the simplified model is a good candidate for the original system.