This article focuses on the problem of balancing between general and specific modelling in complex system design. General models allow for more flexibility regarding the handling of unforeseen situations, while specific models can consider context conditions usually better. Based on the background of mobile robot applications, tasks like navigation, positioning, exploration, homing and others will be investigated regarding their preferences for an appropriate model. A frame called the LISA-model, considering an interaction of global and specific aspects, is proposed.
Due to the fact that the systems discussed here (adaptive, dynamic real-world interactions) are, as a whole, out of the range of analytical methods, real-world scenarios and empirical methods are set up and are discussed.

keywords: mobile robots, world modelling, dynamic indeterministic environments, neural networks, exploration, navigation, self-localization

Paper (pdf) | Paper (ps.gz) | other publications