Heimo Zeilinger

Cognitive Automation—Survey of Novel Artificial General Intelligence Methods for the Automation of Human Technical Environments

Automation, the utilization of control and information technologies for reducing the need for human intervention in the production process is about to meet Cognition-the science concerned with human thinking-and related sciences. More and more processes require analysis and insights that allow controlling them beyond the mere execution of rules and beyond prefitted controllers in order to automatically keep them within the desired conditions. Automatic and flexible decision making based on challenging conditions such as increasing amounts of information, lacking prior knowledge of data, incomplete, missing or contradicting data, becomes the key challenges for future automation technologies.

Simulation Results for the ARS-PA Model

The project ARS (Artificial Recognition System) develops a novel approach to decision making in the domain of building automation systems. Concepts from neurology, psychology, and psychoanalysis are used for this approach. We give a short overview about this decision making concept, the simulation environment, and results from simulation experiments. These simulation runs consist of several autonomous agents grouped to two teams which are competing for restricted resources. ARS deals with agents as an intermediate step towards buildings. For some resources cooperation among team mates is necessary. One team is using the ARS approach including concepts like emotions, drives, and desires. Further, it uses a value system inspired by social levels to stimulate cooperation among the team mates. The other team-as a reference-is realized by using a simple rule based decision making approach. The simulation results show an improvement in team-survival due to the introduction of social levels.

Intelligent streetlight management in a smart city

The intention of a Smart City concept is among other things to increase the energy efficiency of a city. Typically addressed topics are energy management or urban traffic management. Up to now Information and Communication Technology (ICT) measures within a Smart City to increase energy efficiency in public lighting have not been considered in detail. The paper presents an approach of an intelligent streetlight management system being an integral part of a Smart City platform. Field bus technology is used at the field level. Control and monitoring strategies are implemented as web services in the central software. Interoperable interfaces to other parts of the platform are specified. Real-life use cases at test sites in Austria demonstrate the benefits of the system: increased energy efficiency without compromising public safety.

Cognitive decision unit applied to autonomous biped robot NAO

The novel approach to use meta-psychology — the theoretic foundation of psychoanalysis — as archetype for a decision making framework for autonomous agents was realized in simulations recently. In addition, multiple studies showed the capability of a robot to sense and interact in its environment. This work fills the gap between sensing, environmental interaction and decision making by grounding these topics with an agents internal needs using the concepts of meta-psychology. The bodies of typical agents are equipped with internal systems which can generate bodily needs — for example the urgent need for food. As proof-of-concept we implemented this concept on a simulated agent as well as on a physical real humanoid biped robot to additionally proof the concept within a fully controlled simulated environment. The use of the common humanoid robot platform NAO, which has 25 degrees of freedom and biped locomotion, enforced us to deal with complex situations and disturbed sensor readings. NAO provides various internal sensors like engine temperature or battery level as well as external sensors like sonar or cameras. An implemented visual marker detecting system allowed us to detect objects in the surrounding environmental, representing food or energy sources. We show, how it is possible to use the psychoanalytically inspired framework ARS to control a real world application, the robot NAO.

Extending a smart wheelchair navigation by stress sensors

This paper discusses the adaptation of a novel bionic approach of AI to an Assisted Powered Wheelchair (APW) controller, in order to facilitate its driving by people with severe impairments. The control unit offers three modes of operation and this proposal refers to the second control that assists the wheelchair user in driving tasks like passing a door or avoiding obstacles. The users psychic condition is considered in the APW navigation decision and it is determined by a stress detection system composed by electrocardiogram and skin conductance sensors. The control unit has to deal with environmental and stress data and must trigger actuating commands. The prototype implementation will be done within the platform STAGE and afterwards it will be embedded in a real powered wheelchair.

Bionic Inspired Decision Making Unit Model for Autonomous Agents

Modern intelligent control systems for building automation, ambient living or surveillance control have to deal with upcoming requirements like increasing complexity of sensor data, or supporting human beings at their jobs. Therefore the system has to decide on actions taken in case of unforeseen situations. The most complex as well as effective decision making unit that is known, is represented by the human mind. Therefore it seems to be feasible to emulate the human mind in order to design a system for approximating its abilities. While in current Artificial Intelligence a number of generally accepted models, which rely on psychological or neurological theory, are known this paper will introduce a different approach. As a result of the interdisciplinary work of engineers and psychoanalysts a functional model of the human mind has been designed on the base of psychoanalytical theory. As result this work describes the first version of this model in detail and gives an overview on the psychoanalytical theory it relies on.

Perceptive learning — A psychoanalytical learning framework for autonomous agents

Reacting to unpredictable situations is one of the main problems in technical automation. This article introduces two developed framework for decision units for embodied autonomous agents. The models are compared to an existing architecture, the BDI architecture. The basis of the described model is using psychoanalytical concepts. Drives, emotions and wishes are the basics that build a reaction due to an existing situation. A concept of strategy planning, using the Freudian wish has been described. Using a concept of thinking of the humans psyche that develops during childhood, the difference between thing presentations and word presentations is shown and based upon that, a possible learning strategy is elaborated.

Increasing energy efficiency with traffic adapted intelligent streetlight management

Street lighting consumes a non inconsiderable amount of energy. First promising approaches include the deployment of LED based lights to reduce energy consumption. The paper presents a further step of increasing energy efficiency by implementing traffic adapted intelligent management strategies of the luminaries. The idea is to include a communication module into each streetlight that exchanges data via a field aggregated point device with a streetlight management center being an integral part of a traffic management system. Information of the current traffic situation is delivered by traffic sensors and applies traffic adapted control of the luminaries. Consequently, additional energy savings and CO2 reductions are possible without decreasing road safety because a traffic dependent level of luminosity can be provided if needed.

Bionically inspired information representation module

In artificial intelligence, various theories on knowledge representation have gained ground. While logic descriptions, semantic nets, frames and associative networks are used for the implementation, the question which theory is best to be used as archetype for the knowledge model is controversially discussed. Regarding the use of biological principals, neural nets and psychological inspired theories are applied to technical concepts. Both approaches show disadvantages regarding their engineering usage. This article presents an information representation module that is developed for a psychoanalytical inspired reasoning unit. The module and its advantages in contrast to other approaches are discussed and tested in a game of artificial life. Defined data structures and the psychoanalytic background are explained. A use case is shown which helps to understand the principals behind the model.

Bionic inspired information representation for autonomous agents

In AI, different bionic inspired theories on memory have gained ground. On the one hand neuronal networks are often used to realize learning systems. On the other hand psychology influences engineers by introducing the models of implicit and explicit memory. While neurology leaves out the question how mental functions emerge out of the brain structure, psychology classifies information but does not make a statement on its structure. This article presents a psychoanalytic inspired approach on information representation introduced within the ARS project. A short overview on the ARS model is given and the psychoanalytic theory behind information representation is presented. The developed technical concept is introduced and a use case for testing the model is defined.