CIM WELCOMES NEW MEMBER PROFESSOR JOELLE PINEAU
We are most pleased to report that our colleague, Professor Joelle Pineau from the School of Computer Science, joined CIM as a full member in May 2009, after having been a CIM Associate member for four years. Joelle is also a member of Regroupment REPARTI.Professor Pineau is doing work of significant importance to CIM in various facets of Artificial Intelligence and its applications in robotics, machine learning and medical treatments. Her research expertise is broad,ranging from the themes of planning, learning and decision-making, mobile robotics, human computer interaction and adaptive treatment design.
Professor Pineau is co-director of the Reasoning and Learning Laboratory and founder of the SmartWheeler project.
In one of the areas that her research spans, Professor Pineau's group studies planning and learning under model uncertainty and its applications in robotics-human interaction. Most methods for robot planning assume a known model describing the robot's dynamics and perception system. Specifying such a model by hand can be time-consuming and error-prone. Learning techniques can be used to alleviate this problem - however they often require large amounts of data. In certain task domains, such as human-robot interaction, it is not feasible to acquire such large amounts of data.
Professor Pineau studies a new framework for learning and planning under model uncertainty. The method relies on a bayesian formulation of the Partially Observable Markov Decision Process. In this framework, prior knowledge of the domain can be combined with small amounts of data to infer a posterior over model parameters.