LEANING AND RECOGNITION ALGORITHM OF INTELLIGENT AGVSYSTEMstored in the same value. system, Neural Network

1 Suthep Butdee, Thai French Innovation Center, King Mongkut’s Institute of Technology North, Bangkok, 1518 Piboonsongkram Rd. Bangsue, Bangkok, Thailand 10800, [email protected] 2 Anan Suebsomran, Department of Teacher Training in Mechanical Engineering, Faculty of Technical and Education, King Mongkut’s Institute of Technology North Bangkok, 1518 Piboonsongkram Rd. Bangsue, Bangkok, Thailand 10800, [email protected] Abstract  A learning algorithm of a mobility behavior of vehicle is to generally use in AGV system. Learning algorithm of AGV system is proposed in this paper. By finishing the AGV with learning ability, the AGV performs more intelligently and flexibly to apply in factory. Before learning process, the AGV moves directly to destination point by humans teaching. The sensors are used with this system such as vision, odometer, safety switching, lamp etc. These inputs are the information used for training the intelligent AGV system. Before learning process, the AGV moves directly to destination point by humans teaching, when the AGV detects a change of information obtained from sensors either onboard or outboard, the AGV arks for instruction from a trainer to avoid crashing with external obstacles. After being taught, the AGV learns and recognizes to move between station to load and unload the object by its learning experience of job processing. By teaching an AGV with series of instructions until destination point, the robot learns gradually to move to the destination point. The recognition architecture uses the feed forward neural network to training AGV system. The inputs to the recognize network are the sensor patterns derived from encoder sensors, whereas the output action represents the action executed by AGV when the AGV finds the sensor pattern stored in the same value.