Implementación de una red neuronal artificial tipo SOM en una FPGA para la resolución de trayectorias tipo laberinto

This article presents the design and implementation of a system of control through the use of artificial neural networks (ANN) with unsupervised learning type SOM (self-organizing maps) or Kohonen Maps assembling in a reprogrammable Gate Array (FPGA) Spartan 6, with the purpose to determine the effects that has such a system in the behavior of a mobile robot in the resolution of labyrinth trajectories. For this reason, it was necessary to carry out an exploration stage, in which the robot collects the information of the environment through its sensors, sending the acquired data to a computer using wireless communication (communications modules XBEE), in order to carry out the training of the RNA, obtaining synaptic weights clusters that are required for the mobile will be able to fulfill its objective. These processes are controlled and displayed by a graphical user interface or GUI developed in Matlab, called ZiggyLab interface, allowing the user to monitor the mobile as it evolves their learning in the different environments in which they will mobilize. Finally, we used a platform may be modified in order to fit different mazes, and carry out the tests that consisted in developing the different stages already mentioned, and this way of analyzing the results which show the learning of the mobile robot at the time of resolving a path labyrinth as the main effect on their behavior, due to the implementation of the system.