ProcessModelingFeb2009/SimoneTalk
From STLab
Planning and Scheduling with Timelines
Abstract
The need for AI based planning and scheduling tools for controlling complex physical systems is constantly growing in many fields. The design of these applications has followed mainly two distinct directions: on one hand, domain dependent, specific applications; on the other, domain independent, general planning and/or scheduling systems which are customized by changing the domain descriptions. In our work we have followed a direction which lies in between the two, driven by our interests in space related applications. We started from general questions like: what do current P&S applications have in common? What can be re-used (and how) from different planning and scheduling applications without loosing generality or efficiency? Is it possible to design a framework for rapid prototyping of AI P&S systems? Being driven by those questions we have developed a general framework for P&S application deployment, the TRF, Timeline Representation Framework, and an integrated planning and scheduling architecture, called OMPS (Open Multi-Component P&S System). This talk will describe the design principles of P&S systems strongly based on the concept of timelines representing temporal sequences of events in a rather intuitive way. The aim of this approach is to exploit the advantages of both general planning (needed for P&S applications rapid prototyping) and efficient, domain-specific computation (often needed in deployed applications for real world domains).
Short Bio
Simone Fratini is a research scientist at ISTC-CNR. Since September 2002, Fratini has done research focused on constraint-based approaches for modeling and solving planning and scheduling problems. His research interests are on integration of planning and scheduling, AI software architectures for constraint-based time and resource reasoning, Expressive Domain Modeling Languages, Artificial Intelligence Problem Solving Architectures. During his PhD program he has visited for different time periods the Robotics Institute of the Carnegie Mellon University as a research student scholar. In his PhD he has designed OMPS (Open Multi-component Planner & Scheduler) an architecture for integrating planning and scheduling. In 2006-2008 he has been the main designer and developer of the APSI architecture, within the "Advanced Planning&Scheduling Initiative" of the European Space Agency, a reusable software framework for designing and developing space planning and scheduling systems with a timeline based approach.
