Integrating knowledge-based systems and neural networks for navigational tasks

摘要

The MONSTER project (Models of Neural-Symbolic Techniques for Extendable Robots) is presented. The project focuses on inherent problems in developing an adaptive system for navigational tasks. The proposed architecture is based on both explicit (symbolic) knowledge of the environment and implicit knowledge encoded in the connections of neural networks. An overview is presented of the current state of the project, pointing out the most relevant aspects in the functional and structural design of the system. Solutions to specific problems related to short-term goals are also discussed.