Research Seminar on "SELFCEPTION: neuro-inspired robotic body perception and action"
"SELFCEPTION: neuro-inspired robotic body perception and action" by Dr. Pablo Lanillos, Chair for Cognitive Systems/TUM.
Abstract: Unlike robots, humans learn, adapt and perceive their body as a unity during interaction with the environment. To enable robots with the same functionalities in complex changing scenarios, perception and action should be revisited. In this talk, I will present the latest results of SELFCEPTION interdisciplinary project , where we are developing computational models for robotic self/other distinction in order to improve interaction under uncertainty. First, I will mathematically describe robotic body perception as a flexible process inspired by Predictive Coding, one of the most promising theories of the inner functioning of the brain. Perception and action becomes two sides of the same process: to continuously approximate reality with the generative models that have been learned by interaction. Basically, the perceived reality is adjusted by minimizing the error between the expected sensation and the observed one. Afterwards, I will discuss the advantages and challenges of this approach through different tests with real multisensory humanoid robots, such as adaptive perception, robust reaching and self-recognition. Finally, I will portrait the future of neuro-inspired embodied Artificial Intelligence for giving meaningful feedback to neuroscience and psychology combining computational modelling with experimental validation in robots.
Short Bio: Pablo Lanillos has a M.Sc. degree in computer engineering and the Ph.D. degree in Artificial Intelligence and Robotics by the Complutense University of Madrid. He is currently a Marie Sklodowska-Curie Fellow leading his EU H2020 funded project SELFCEPTION (www.selfception.eu), at the Institute of Cognitive Systems (ICS) directed by Prof. Gordon Cheng, at the Technical University of Munich (TUM, Germany). Before working in Germany, Lanillos worked in the last 10 years in the Massachusetts Institute of Technology (MIT), in the Polytechnic School of Lausanne, in the ACFR of Sydney University and the Institute for Systems and Robotics, Portugal. He has presented his line of research in several top journals and conferences and recently at the IROS2018 conference, his work was awarded as finalist of better cognitive robotics paper. Besides, he is the vice-president of the Society of Spanish Researchers in Germany (CERFA).
Please note: this talk will be preceded by a 5-min pitch presentation of MSNE student Mohamad Atayi with the title: "Adversarial training of neural encoding models with latent variables"
24 April 2019,16:30h, ICS Karlstraße 45, 6. Floor, room 6009