From Non-deterministic Planning to Agent Planning Programs and Goal Recognition
Speaker:
Prof. Sebastian Sardina
Data dell'evento:
Giovedì, 23 November, 2023 - 12:00
Luogo:
Aula Magna
Contatto:
patrizi@diag.uniroma1.it
Abstract: Fully Observable Non-deterministic (FOND) is an elaboration of classical planning that has attracted much attention in recent years. It has been linked to the long-standing program synthesis problem and has been used as a way to solve other types of problems such as probabilistic planning, LTL synthesis, and generalized planning. In FOND planning, solutions are no longer linear sequence of actions, but structures with conditionals and loops. In this talk I will give an overview of three problems in the areas of AI knowledge representation for reasoning about dynamic systems. I will start with the problem of AI planning with actions that have non-deterministic effects and whose solutions rely on some sort of fairness assumptions in the environment. I will then present the problem of automatically synthesising plans for complex programs build from many declarative goals. Finally, I will discuss how AI planning can be used to provide a crisp approach to goal/intention recognition, the inverse of planning that aims to understand the goal of an agent from its behaviour. The aim of the talk is not to provide a detailed description or analysis of any of these problems, but rather a general overview that could potentially trigger discussion and possible collaborations.
Bio: Sebastian Sardina is a Professor at RMIT University within the School of Computing Technologies. He completed his PhD and M.Sc in the Cognitive Robotics Group at the University of Toronto (Canada) and, before that, a BSc in Computer Science at South National University in Bahia Blanca (Argentina). Sebastian's research falls in the intersection between knowledge representation for reasoning for action and change, automated planning, and intelligent agents. In a nutshell, his research seeks better representation models and algorithms for programming intelligent controllers operating in complex and dynamic environments. Beyond research, he is also interested in bringing Computational Thinking (and programming) to the community, particularly to children and youth; by delivering hands-on workshops and talks to students and educators. He has recently been part of the panel at VCAA conducting the study design review for the new Algorithmics (HESS) VCE program.