Hierarchical Log-Spherical Inference Grid – An Unconventional Approach to Robotic Perception and Action

In this text, an unconventional paradigm for robotic multisensory perception and action will be presented in the form of a generalisation of a framework devised in previous work by the authors, the Bayesian Volumetric Map (BVM). The BVM, a log-spherical inference grid providing an egocentric and probabilistic representation of spatial information, was designed to bridge multisensory perception and actuation by means of a unified framework. The underlying paradigm represents a fundamental departure from traditional outlooks on robotic perception and spatial reasoning, in that it embodies a non-Cartesian and egocentric approach as opposed to the conventional Cartesian and allocentric perspective.