CONTROLAB: integration of intelligent systems for the control of a robot arm

Abstract

CONTROLAB integrates intelligent systems and control algorithms aiming at applications in the area of robotics. This paper focuses on the analysis of the word recognition and the trajectory definition systems considering an application in which a robot arm is commanded by voice to pick up a specific tool placed on a table among other tools and obstacles. Neural network architectures based on the backpropagation and the recursive models are proposed for the implementation of a speaker-independent word recognition system. The robustness of the system using the backpropagation network has been verified in totally uncontrolled environments such as large public halls for the exhibition of new technology products. Experimental results with the recursive network show that a carefully designed network structure is able to overcome the false alarm problem faced by the backpropagation network. The trajectory to be followed by the robot arm is determined through the analysis of image information and the use of the VGRAPH algorithm to avoid obstacles. The algorithm performance is analysed and compared with that achieved by the PFIELD algorithm.