GET-Forschungsseminar Abstracts

Bio-inspired Simultaneous Localization and Mapping

Musa Kazmi, GET Lab

Vortrag: Mi. 09.04.2014, 16:30, Raum P 1.4.17

Abstract:

In robotics, the problem of Simultaneous Localization and Mapping (SLAM) has gained significant importance over the last few years. In this regard, a wide-range of approaches has been developed while emphasizing one or more aspects of it, for instance, loop closure in large-scale environments, data association methods, highly dynamic environments, vision-only localization and mapping, and three-dimensional (3D) maps. The majority of current methods uses probabilistic grounds to tackle these aspects in a strictly Cartesian space unlike bio-inspired techniques which use the biological processes encoded in a mammalian's brain. These criteria of geometric localization of features and the robot itself adds multiple levels of complexity to the problem. This gives a motivation to preserve only the topology of the environment rather than their precise geometry which is how animals do localization and mapping. This research is aimed at learning a semantic network of knowledge representation through cognitive insights for solving data association and loop closure problems which in turn has experimentally shown the potential to deliver robust bio-inspired solutions to the problem of localization and mapping with increased recall-precision.