Noticias Noticias

Volver a la página índice

Thesis Defence: Javier Segovia Aguas

Thesis Defence: Javier Segovia Aguas



Javier Segovia Aguas

Program Synthesis for Generalized Planning

Supervisor: Dr. Anders Jonsson



Thesis Brief Description:

My thesis is about Generalized Planning (a new field of Artificial Intelligence), and it is based in the idea that classical planning techniques can be used to find solutions that work for multiple problems. I receive as an input a set of planning problems specified with PDDL language that are compiled into a single planning problem. Then, the compiled problem can be solved with any off-the-shelf classical planner. A solution to this problem is a sequence of actions that program and execute instructions from which algorithm-like solutions are induced. The kind of algorithms generated are called planning programs and finite state controllers. I also show how these formalisms can be extended using previous knowledge by simulating a call stack in planning. Finally, I propose new landscapes for planning that consider alternative tasks, like unsupervised classification of planning instances or generating context-free grammars, showing that "planning" and "learning" AI branches can be connected.

Experience as a PhD Student:

These years as a PhD student have been hard but they worth it. It takes time to start and acquire knowledge in science but everything becomes easier when you are working hand-in-hand with other researchers from one of the best groups in AI. They have really made me grow as a researcher and person. Furthermore, I have been lucky publishing in many conferences from which I learnt to defend oral presentations, scientific discussion and meet very nice people from around the globe. In concusion, I only have words of thanks to administration, researchers and students from UPF that made my life easier and happier the last four years.