Modern AI Planning

Lectures:

26h.

Recommended for

Graduate and undergraduate students with an interest in AI. ECSEL students are most welcome.

The course was last given:

New course.

Goals

The course aims to give an introduction to the planning problem in AI, and to the many techniques that have recently been developed to deal with it.

Prerequisites

Some previous experience with logic is useful. The knowledge provided by any basic AI course should suffice.

Organization

The course consists of a number of lectures/seminars and a lab course.

Beyond that, the organization of the course is left open-ended; details concerning the schedule and the presentation of the course material (such as whether there should be teacher-given lectures or student presentations or both) will be decided by teacher and course participants together at the start of the course.

Contents

The planning problem in AI. Search and basic planning algorithms (forward-chaining, regression, partial-order). Graphplan and heuristics. SAT and CSP encodings. Domain analysis. Abstraction and goal ordering. Hierarchical planning and domain-dependent search control. Extensions to uncertanity and metric time.

Literature

A bibliography, with help of which articles will be selected by the teacher and the participants.

Teachers

Patrik Haslum, (one or, if possible, two) guest lecturers.

Examiner

Patrick Doherty

Schedule

Spring 2001.

Examination

A set of assignments, consisting mostly of experimenting with existing planning systems.

Credit

3-5 credits, depending on assignments.