Synthesis of Distributed Cognitive Systems: Interacting Maps for Sensor Fusion

Abstract

Event Abstract Back to Event Synthesis of Distributed Cognitive Systems: Interacting Maps for Sensor Fusion Cristian Axenie1* and Jorg Conradt1 1 Technische Universität München, Electrical Engineering and Information Technology, Germany Environmental interaction is an important aspect in the life of every physical entity, which impacts its internal state and allows the acquisition of new behaviors. A prerequisite for such interaction is multimodal sensory fusion with the goal of providing a consistent internal representation of the external world. In technical systems, such a representation is obtained by fusing noisy data from various available sensors, which combined generate a richer representation compared to individual use. State-of-the-art technologies apply probabilistic methods for combining prior and observed information, but these techniques are computationally inefficient for a large number of sensors. In our project we propose extending an existing initial model of interacting maps for sensory interpretation with the focus on multimodal sensory fusion. The core aspect here is to design and implement a neuro-biologically inspired method for real-time distributed interpretation of sensory stimuli in a mobile robotic setting: a distributed graphical computing system with inter-merged information storage and processing that allows efficient parallel reasoning. This network architecture consists of interconnected heterogeneous software units, each encoding and processing a different feature about the environment that is represented in a local map. A map here is defined as a pixel based representation of some perceived or inferred aspect of the current environment, in which each entry contains a value of a certain type (e.g. a scalar of image brightness or a 2D vector of local optic flow). Such extracted pieces of environmental knowledge interact by mutual influence to ensure overall system coherence: the values represented in each map slowly drift in the direction that minimizes disagreement between the two sides of an assumed inter-map relationship through mutual influence. Although by system design each map specializes in representing a certain feature of the environment, the required relations between maps for such influence are established through a learning process: the true relations between maps are a priori unknown, but exist hidden within the observed real-world data. Therefore, our system needs to infer those relations over time from correlated sensory signals. Starting from the proposed distributed processing architecture, our main research investigates self-constructing and adapting internal relations between maps (structural plasticity) that reflect the underlying relations between multimodal sensory information streams. We investigate optimized network distributions on available computing resources without an explicit a priori network topology provided by a systems designer. This new architecture reflects distributed information processing as known to occur in brains, and ensures a fast, robust and scalable computational architecture appropriate for real-time real-world robotic applications. Acknowledgements The foundations of this work have been laid during workshops such as the Telluride and the CapoCaccia Cognitive Neuromorphic Engineering Workshop. Funding and inspiration for application scenarios has come from the DFG Cluster of Excellence “CoTeSys” and the Muinch Bernstein Center for Computational Neuroscience. Keywords: network computation, self-growing systems, sensory fusion Conference: Bernstein Conference 2012, Munich, Germany, 12 Sep - 14 Sep, 2012. Presentation Type: Poster Topic: Sensory processing and perception Citation: Axenie C and Conradt J (2012). Synthesis of Distributed Cognitive Systems: Interacting Maps for Sensor Fusion. Front. Comput. Neurosci. Conference Abstract: Bernstein Conference 2012. doi: 10.3389/conf.fncom.2012.55.00077 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 22 May 2012; Published Online: 12 Sep 2012. * Correspondence: Mr. Cristian Axenie, Technische Universität München, Electrical Engineering and Information Technology, Munich, 80333, Germany, cristian.axenie@tum.de Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Cristian Axenie Jorg Conradt Google Cristian Axenie Jorg Conradt Google Scholar Cristian Axenie Jorg Conradt PubMed Cristian Axenie Jorg Conradt Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.