Talk by Vidal Alcázar Saiz

Speaker:
Vidal Alcázar Saiz

Title:
A personal view on the International Planning Competition:
two unconventional search algorithms and an opportunity for Machine Learning.

Abstract:
In this talk we take a look at the International Planning
Competition, an event in Automated Planning that aims to objectively
compare different approaches against each other and push the state of
the art in the field. In particular, we will introduce our
contribution, two submissions that are characterized for being
unconventional search algorithms. The first one, symbolic search, uses
Binary Decision Diagrams to represent states and actions, and obtained
first and second place in the optimal track. The second one,
Rapidly-exploring Random Trees, was awarded with the "Most Innovative
Planner" award.

Furthermore, we will give an overview of the learning track, a
competition in which the solvers have access to small instances before
facing the real benchmark, with the particularity that the learning
process does not allow any kind of human interaction. This presents
many challenges to Machine Learning practitioners, as their techniques
must be domain independent and able to learn in an automated way.