Dynamical self-organizing maps and behaviour-oriented autonomous robots
This paper proposes a control system for a behaviour-oriented autonomous mobile robot that learns to navigate in its environment via its own sensors from certain start locations to predefined goals without colliding with obstacles. The system consists of two main components: first, simple basic reflexes are used to determine suitable actions for new situations that the robot has not observed before; second, the system uses a dynamical self-organising map (DSOM) to construct a mapping from different sensory situations to appropriate actions according to the outputs of basic reflexes and corresponding sensory situations. The DSOM was developed to fulfill the requirements of real-time operation as well as incremental and on-line learning during navigation. The viability of the proposed control strategy is demonstrated with a simulated NOMAD 200 mobile robot that learns safely and quickly to navigate in its environment, by minimising the need for human guidance during learning and operation.
Bibliographic Reference: Article: Advanced Robotics (1995) : Special issue on Behaviour and Learning
Record Number: 199511155 / Last updated on: 1995-08-23
Original language: en
Available languages: en