Global spatial modeling based on dynamics identification according to discriminated static sensations

This article focuses on the problem of identifying and discriminating trajectories (as sequences of situations connected by transition-dynamics) in a mobile robot setup. The continuous time dynamics are segmented and scaled by transition sensations between significantly different static situations. The complementary information out of static attractors and dynamical transitions is fused in a canonical way and under hard real-time constraints. Based on the generated and continuously adapted trajectory models, a global topological model is maintained and attributed suiting the needs of robust qualitative navigation tasks. Neither a global position nor any other global metrical description is generated or employed by this approach, thus it is not meant to fulfil global precision requirements. The presented results describe physical experiments with autonomous robots in unprepared environments.