3D UWB Positioning System on Data Processing by ANN

A three-dimensional positioning system based on the use of an ultra-wideband pulse signal processed by an artificial neural network is presented. The positioning is performed by receiving a pulse signal radiated by three spatially spaced ultra-wideband log-periodic antennas excited by pulse signals of different durations. An artificial neural network recognizes from which antenna and at what angle the signal was received. Numerical modeling of the radiation of a log-periodic antenna under impulse excitation makes it possible to obtain the time dependencies of the radiated waves and, accordingly, data for training a deep artificial neural network. An algorithm for pre-processing time dependencies for further training of a deep fully connected artificial neural network is developed. The structure of an artificial neural network for classification of time dependencies by the angle of observation for three ultrawideband antennas with different durations of the excitation Gaussian nanosecond signal is constructed. The possibility of determining the area of space where the receiving system can be located after receiving pulse waves from all three log-periodic antennas is demonstrated. The proposed system can be used for local positioning of various moving objects, such as GPR.