Bionically inspired information representation for embodied software agents : realizing neuropsychoanalytic concepts of information processing within the computational framework ARSi10 / by Heimo Zeilinger
VerfasserZeilinger, Heimo Thomas
Begutachter / BegutachterinDietrich, Dietmar ; Barnard, Etienne ; Bruckner, Dietmar
UmfangXII, 151 S. : graph. Darst.
HochschulschriftWien, Techn. Univ., Diss., 2010
Zsfassung in dt. Sprache
Bibl. ReferenzOeBB
Schlagwörter (DE)Bionik / Informationsrepräsentation / Softwareagenten / Künstliche Intelligenz / Multiagentensimulation
Schlagwörter (EN)bionic / information / representation / software agents / artificial intelligence / multi-agent simulation
Schlagwörter (GND)Software / Agent <Informatik> / Informationsmanagement / Bionik
URNurn:nbn:at:at-ubtuw:1-39045 Persistent Identifier (URN)
 Das Werk ist frei verfügbar
Bionically inspired information representation for embodied software agents [9.05 mb]
Zusammenfassung (Englisch)

This work describes the bionically inspired representation of information in a control unit for embodied software agents. It focuses on the first ever realization of neuropsychoanalytic concepts for generating and processing mental data structures in computer science, and compares said approach to establish bionically inspired methodologies. The approach is completely new to Artificial Intelligence and should allow the design of systems following the principles of the human being's mental apparatus, thus enabling them to operate in highly dynamic environments.

An existing decision unit is supplemented with an information representation system composed of an information representation concept, a data storage, and an information management unit. By use of a top-down design approach, the resulting adaptions are introduced into a new model whose implementation in embodied software agents produces the computational framework ARSi10.

The multi-agent framework-based simulator 'Bubble World' is developed as a test-bed with predefined use cases, and the agents' abilities are evaluated through internal and external performance indicators. Internal and external sensor data are mapped to neuropsychoanalytically inspired data structures and used for the decision making process. This allows the agents to interact with their environment while keeping their system resources balanced and thus retaining their functional abilities.