In this research project it has been proposed the study and the development of an Artificial Neural Network (ANN) in order to implement it in programmable logic devices such as FPGAs (Field Programmable Gate Arrays) using a hardware description language, to control a prototype of a mobile robot. There are many studies in the literature that apply neural networks, so its interest in their domain and application. Articles about Neural Networks are mostly implemented in sequential machines, mainly due to lower cost. On the other hand, the evolution observed in the last decade with programmable logic devices, including FPGAs, reduced prices and increased capacity, are an excellent alternative to the implementation of Neural Networks with multitasking parallel execution in hardware. Therefore one can improves the speed of Artificial Neural Network processing in the training phase, without damaging its efficiency when compared to the same system software. Nowadays FPGAs have sufficient logic elements to a Neural Networks implementation, and an important resource that is using a hardware description language (HDL), making the project easier to change and providing portability between development devices / software. It is intended to synthesize an Artificial Neural Network using a Cyclone II FPGA family-ALTERA for the control a prototype of a mobile robot.
News published in Agência FAPESP Newsletter about the scholarship: