Locomoción de un robot cuadrúpedo basada en redes neuronales artificiales

Abstract

A quadruped robot with locomotion was designed and build based on the artificial neural network perceptron multilayer, to find out if it can keep the stability while walking on uneven terrain. The inductive method was used for the structural design, the internal circuitry, the control algorithm and interface monitoring. The experimental method application allowed verifying the operation of the robot. The component parts of the body and legs of the robot were constructed by 3D printing technique with PLA (Poly Lactic Acid) plastic. The control algorithm for the robot locomotion and monitoring interface data of: ultrasonic sensors, accelerometer and speed, temperature and voltage of each one of the servomotors were developed in MATLAB R2010a. Communication in done wirelessly between the Arduino motherboard plaque and the computer through Xbee devices and as a main power supply two batteries Li-Po were used. By analyzing the data it was found that the quadruped robot shows no significant variation compared with the ideal inclinations, therefore maintains stability despite irregularities and obstacles in the terrain over which it travels. The quadruped robot could be used in industrial processes transporting hazardous materials or substances and exploration of irregular terrain, inaccessible or contaminated.