Automated Guided Vehicle (AGV) technology is critical for advanced manufacturing systems. Autonomous guidance for these vehicle would greatly increase their flexibility and range of application. Current research in the field of neural network control of Automomous Automated Guided Vehicles (AAGVs) has shown that in a structured, hierarchical format, a neural network controller (NNC) can be successfully used for guidance and control. In this paper, a NNC is described developed for an AAGV using a hierarchical structure of highly forward connected, back propagated, recurrent neural networks. Recurrent networks are used to achieve smooth, non-cyclic guidance of the AAGV. The target system is an experimental AAGV equipped with a sonar ranging system and differential-drive propulsion. The vehicle must navigate two dimensional 'mazes' without any prior information about the maze topology. The several different methodologies and algorithms used for training the controller networks are discussed, and the stages in which these different methods are most useful are identified. Finally, we describe how the hierarchical integration of sub-tasks for vehicle control is implemented to provide an overall control system for the AAGV. The controller's navigational abilities are evaluated in a computer simulation involving several floor plans.
Neural network controller for an autonomous guided vehicle
Ein Steuerungs- und Überwachungssystem für ein automatisch geführtes Fahrzeug auf der Basis neuronaler Netzwerke
1992
4 Seiten, 5 Bilder, 13 Quellen
Conference paper
English
A transportable neural network controller for autonomous vehicle following
Tema Archive | 1994
|A Transportable Neural Network Controller for Autonomous Vehicle Following
British Library Conference Proceedings | 1994
|A neural network online controller for autonomous underwater vehicle
Tema Archive | 2007
|Analysis of a Neural Network Lateral Controller for an Autonomous Road Vehicle
SAE Technical Papers | 1992
|British Library Conference Proceedings | 1999
|