Daniel Polani is Professor of Artificial Intelligence at the School of Computer Science at the University of Hertfordshire, UK. He obtained his PhD in 1996 from the Johannes Gutenberg-University of Mainz, Germany. He was Research Fellow at the Institute for Neuro- and Bioinformatics at the University of Lübeck, Germany, and Visiting Researcher at the University of Texas at Austin, at the Max Planck Institute for Mathematics in the Sciences and the Santa Fe Institute. He has been on the Board of Trustees of the RoboCup Federation for the cadence 2008-2014, Senior Program Committee Member of conferences such as IJCAI, AAMAS, and AAAI. He has organized workshops including the FP7 TRUCE-funded IDeM (Information and Decision-Making) workshop, a workshop on information-theoretic incentives for artificial life at the Artificial Life conference or a mini-symposium and workshops on principled theoretical frameworks for the perception-action cycle at the NIPS (Neural Information Processing Systems) conference. He is editor of the JAAMAS journal, associate editor of Advances in Complex Systems and of Frontiers in Robotics and Artificial Intelligence (Computational Intelligence section), and has served as Guest Associate Editor for PLoS Comp. Biol.
His research focuses on understanding information-theoretic constraints on cognitive processing and action selection and how such processes can emerge from first principles. Any task is subject to various informational constraints. Together with constraints dictated by the make-up of the agent itself, whether biological or artificial, this suggests informationally preferrable ways for agents to interact with their environment, and permits the general quantitative analysis and characterization of the perception-action loop and embodiment via their informational signature on the one side, and the construction of informationally preferred processes for the organization of decision-making, and production of intrinsically driven, emergent behaviours on the other side. The aim is to gain a more fundamental understanding about plausible pathways for the emergence and organization of cognitive and sensorimotor processes in biology through fundamental constraints of physics and the environment on the one side; and to use this understanding to create more flexible, generic and adaptive robotic devices and more natural interactions between humans and robots on the other side.
In the completed FP7 project CORBYS, one goal was to develop a generic cognitive architecture for robot systems; here, amongst other, informational methods were developed to provide a natural route for generic behaviour generation for robots, driven purely by the properties of the sensorimotor loop via the "empowerment" principle. This principle aims at optimizing the informational impedance match of the sensorimotor loop to the agent's environmental "niche", thereby producing "natural" agent behaviours in underspecified situations. In the currently running Horizon 2020 WiMUST project, related principles are employed to address the challenge of managing multivehicle robotic submarine swarms in a robust and flexible way.