TUTORIAL SESSION

Probabilistic Approaches for Artificial Robotic Perception

Dr. Jorge Dias
University of Coimbra, Portugal
jorge@isr.uc.pt

Abstract

In this tutorial it will be presented how Bayesian models can be used to develop artificial cognitive systems that can carry out complex tasks in real world environments. This tutorial summarise the investigations on complex systems and processes on artificial robotic perception provided by the collaboration of multidisciplinary researchers on topics of robotics, engineering, computer science, mathematics, neural science and cognition.
During the tutorial will be presented the mathematical background and the bayesian programming techniques that will allow researchers and PhD students to address the question how sensor data derived from these different sensory modalities could be processed to converge in order to form a coherent and robust perception of the environment. These major topics will be concentrated on techniques and models to:
to represent 3D space within a probabilistic framework;
to combine hierarchically Bayesian models and representations;
to define decision processes based on Bayesian programming and Bayesian models.
This tutorial shows how these three topics are central for the develop of processes of artificial perception based on Bayesian techniques.
During the tutorial it will be presented several examples and case studies applying these techniques to robotics. Many of these examples and applications take inspiration from models of brains of mammals including humans and apply these models in the developments of innovative and self-learning robotic systems.
Contemporary robots and other cognitive artefacts are not yet ready to autonomously operate in complex real world environments and one of the major reasons for this failure in creating cognitive situated systems is the difficulty in the handling of incomplete knowledge and uncertainty. The development of these artificial perception systems focused on multi-modal and multi-sensing integration using computational/statistical models supported by observations of biological systems and experimental evidences obtained by psycho-physical methods/studies.

Objectives

In this tutorial the participants will see how Bayesian models can be used to develop artificial cognitive systems. After a primer on probabilistic inference and an explanation on modelling perception as a process with sensing and action, the attendees will learn how Bayesian techniques could be applied to artificial vision systems. The attendees will have the opportunity to study, and discuss several real examples on the application of these techniques in the robotics domain that includes: artificial perception using vision and sound, human-robot interaction and dialogue, robot haptics.

Timeline

Fundamentals of Bayesian inference [30 min] ;
Representation of 3D space and Sensor Modelling within a probabilistic framework [30 min] ;
Bayesian programming and Modelling for Artificial Robotic Systems [30 min] ;
Hierarchically combination of Bayesian models and representations [30 min] ;
Decision and Control processes for learning and motor control [30 min] ;
Examples, Case-Studies and Discussion [30 min].

Background Knowledge Expected of the Participants

It is expected the participants to have a limited grounding in probabilistic theory. The participants with engineering knowledge on topics of robotics, artificial intelligence, computer science, electrical engineering or signal processing will follow the tutorial easily.

Qualifications of the Instructor(s)

Jorge Dias has a Ph.D. on Electrical Engineering by the University of Coimbra, specialization in Control and Instrumentation. Jorge Dias has research activities in the area of Computer Vision and Robotics and has contributions on the field since 1984. He has several publications in international journals, books, and conferences.
Jorge Dias is teacher from the Department of Electrical Engineering and Computers ( www.deec.uc.pt ) and do research in the Institute of Systems and Robotics (ISR) ( www.isr.uc.pt ) from the University of Coimbra (UC) ( www.uc.pt ). Jorge Dias has been teaching several courses on Computer Vision, Robotics, Automation, he has supervised several Ph.D. students in the field of Computer Vision and Robotics. Jorge Dias was been principal investigator from several research projects and coordinates the Mobile Robotics Laboratory of ISR. Jorge Dias is the currently director of LAS-Laboratory of Systems and Automation (http://las.ipn.pt) from the Instituto Pedro Nunes (IPN) (www.ipn.pt) the technology transfer institute from the University of Coimbra.