Advisory Board

Dr. Daniel Polani

Daniel Polani, Ph.D. is Principal Lecturer, Department of Computer Science, University of Hertfordshire and Associate Editor of Advances in Complex Systems. His research is currently focused on information-theoretical models to understand properties and possible mechanisms of self-organization for the perception-action loop.
 
Daniel earned his PhD in 1996 at the University of Mainz (Germany) for his work about the Genetic Evolution of Self-Organizing Maps. In 1996–2000, he was a research fellow at the University of Mainz, and in 1997 a visiting researcher at the University of Texas in Austin. In 2000–2002 he worked at the Institute for Neuro- and Bioinformatics at the University of Luebeck (Germany). Since 2002, he is member of the Algorithms and Adaptive Systems Research Groups at the University of Hertfordshire, organizer of the EAL (Embodied Artificial Life) meetings and leader of the SEPIA (Sensor Evolution, Processing, Information and Actuation) group.
 
A central point of his research is the modeling and understanding of complex and self-organizing and -adaptive systems. Since 1996, this research direction is incarnated in two research directions: the study of multiagent systems and of sensor evolution. In multiagent systems, he has done work in the RoboCup framework, on trading agent dynamics and on the emergence of social norms in agent societies. His research on sensor evolution ranges from the conditions driving the emergence of sensorics and their relation to biological systems to fundamental questions about the principles guiding their organization and that of the perception-action loop. In addition, in his work he aims at the structuring of sensomotoric loops from first principles.
 
Daniel is on the Board of Trustees of the RoboCup Federation and reviewer for national and international research councils (e.g. UK, US, Germany, Netherlands) as well as renowned international journals and member of a large number program committees.
 
His research focuses on:

  • Artificial Intelligence: Can we imitate the processes that allow animals and humans to take flexible decisions in a complex and difficult environment? Being able to adapt themselves to different conditions gracefully? Being able to absorb changes in the environmental conditions? Can we do so incorporating learning ability, without compromising generalization ability, without hand-coding all necessary rules into a system? Are there general principles underlying intelligent information processing in living beings which we can exploit without having to resort to specialized solutions that vary from task to task?
     
  • Artificial Life: This question is closely related to Artificial Intelligence. What makes living systems live? What are the mechanisms that allow complexity arise in nature? What role play emergent and self-organizing mechanisms in this context and can this role be formalized? Can understanding of mechanisms for “life” teach us something about how to achieve intelligence in artificial systems?
     
  • Information Theory for Intelligent Information Processing: To put it saucily: information theory is something like the logarithm of probability theory. In early modern times the logarithm simplified multiplication into addition which was more accessible to calculation. Today, information theory transforms many quantities of probability theory into quantities which allow simpler bookkeeping.
     
  • Sensor Evolution: Speaking of information, living beings and artificial agents survive by obtaining information from the environment. Due to the physical limitations of the agents, not all environmental information is gathered; the sensors will concentrate on relevant information that is essential for the goal of the agent.
     
    How is this selection process attained in nature? How can we copy it in artificial systems? How can new channels of environmental information be tapped and exploited by a system that originally accessed a different set of information channels? Can we make use of it in artefacts?
     
  • Collective and multiagent systems: Among the most complex systems are systems of multiple coacting and coevolving agents. These systems are, in general, very difficult to describe and to design properly. Such systems provide a natural high-complexity environment for studies on artificial intelligence and artificial life and are ideal testbeds for theories about social interaction.
Daniel coedited RoboCup 2003: Robot Soccer World Cup VII (Lecture Notes in Computer Science), and coauthored Organization of the Information Flow in the Perception-Action Loop of Evolved Agents, All Else Being Equal Be Empowered, Empowerment: A Universal Agent-Centric Measure of Control, Emergence of Genetic Coding: an Information-theoretic Model, Kernelizing LSPE(λ), Maximization of Potential Information Flow as a Universal Utility for Collective Behavior, Learning RoboCup — Keepaway with Kernels, Least Squares SVM for Least Squares TD Learning, Sequential Learning with LS-SVM for Large-Scale Data Sets, Optimizing Potential Information Transfer with Self-referential Memory, Relevant Information in Optimized Persistence vs. Progeny Strategies, A Simple Modularity Measure for Search Spaces based on Information Theory, and Measuring Informational Distances Between Sensors and Sensor Integration.