Hippocampal formation-inspired probabilistic generative model

We tackle the challenging task of bridging the gap between the neuroscientific knowledge of hippocampal formation (HPF) and the engineering knowledge of robotics and artificial intelligence. Simultaneous localization and mapping (SLAM) has already been realized in robotics as a basic function for spatial cognition. In this study, we aim to investigate how the SLAM functionality corresponds to the HPF. To this end, a hypothesis based on a literature review is suggested and a direction for its verification is presented, without performing any new simulations. We survey HPF models and various computational ones, including brain-inspired SLAM, spatial concept formation, and deep generative models. Furthermore, we discuss the relationship between the findings of HPF in neuroscience and SLAM in robotics. Thereby, A hippocampal formation-inspired probabilistic generative model (PGM) was constructed using a methodology for constructing a brain reference architecture. We propose an HPF-PGM as a computational model based on a modification of the conventional SLAM model, which is designed to be highly consistent with the anatomical structure and functions of the HPF. By referencing the brain, we suggest the importance of the integration of egocentric/allocentric information from the entorhinal cortex to the hippocampus and the use of discrete-event queues.