Abstract
It is a challenge to build robust simultaneous localization and mapping (SLAM) system in dynamical large-scale environments. Inspired by recent findings in the entorhinal–hippocampal neuronal circuits, we propose a cognitive mapping model that includes continuous attractor networks of head-direction cells and conjunctive grid cells to integrate velocity information by conjunctive encodings of space and movement. Visual inputs from the local view cells in the model provide feedback cues to correct drifting errors of the attractors caused by the noisy velocity inputs. We demonstrate the mapping performance of the proposed cognitive mapping model on an open-source dataset of 66 km car journey in a 3 km × 1.6 km urban area. Experimental results show that the proposed model is robust in building a coherent semi-metric topological map of the entire urban area using a monocular camera, even though the image inputs contain various changes caused by different light conditions and terrains. The results in this study could inspire both neuroscience and robotic research to better understand the neural computational mechanisms of spatial cognition and to build robust robotic navigation systems in large-scale environments.
Highlights
One of the fundamental challenges to integrate robots into our daily life is to endow with the ability of spatial cognition
Our cognitive mapping model is run on a personal computer with 3.4 GHz six-core Intel i7 processor and 64 GB memory
We demonstrated that our cognitive mapping model using only visual sensory information could successfully build a coherent topological map of largescale outdoor environment from an open-source dataset of 66 km urban car journey
Summary
One of the fundamental challenges to integrate robots into our daily life is to endow with the ability of spatial cognition. Over the past two decades, many efforts have been exerted to develop truly autonomous robot navigation systems that are able to simultaneously localize, map, and navigate without human interventions in domestic home and workplace. This line of research has been recognized as the problem of Simultaneous Localization and Mapping (SLAM) (Smith and Cheeseman, 1986; Bailey and Durrant-Whyte, 2006; Durrant-Whyte and Bailey, 2006; Thrun and Leonard, 2008) and constitutes a significant area of robotic research. Numerous variants of filtering algorithms have been proposed to estimate the position and the heading direction of the robot in large environments. Filtering algorithms provide beautiful mathematical frameworks for SLAM, they are in lack of robustness, since bad data association (i.e., the matching between perception and internal representation) can often lead them into irreversible divergence, impairing their application
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
