Foundations of Artificial Intelligence (FAI) Group
divider line

   ← Back to FAI group members

divider line Picture broken :/

Pascal Lauer

Computer Science Department
Foundations of Artificial Intelligence (FAI) Group
Campus E 1.1, R. 3.14
Universität des Saarlandes
66123 Saarbrücken, Germany

Mail:

divider line

Hi, I am Pascal. In 2020 I received my B.Sc. at Saarland University. I continued my studies at Saarland University and obtained my M.Sc. in 2022. I am now a PhD student in a Dual PhD arrangement between Saarland University (under Prof. Dr. Jörg Hoffmann) and Australian National University (under Prof. Dr. Pascal Bercher). As per ANU regulations, I also have two additional co-supervisors: Assoc. Prof. Dr. Álvaro Torralba and Assoc. Prof. Dr. Daniel Fišer. Álvaro has been a mentor since my B.Sc. thesis, and his support continues to this day. I also had the opportunity to work with Dan as undergrad in Saarland. While co-supervisor roles are often just a formality, I continued to work closely with both Álvaro and Dan, in addition to Jörg and Pascal, and I deeply appreciate the support and guidance from all of them.

I work in the field of Automated Planning, which focuses on developing general problem solvers for decision-making problems. Here, a user defines a world state and specifies ways (called actions) to modify the state consecutively. The planner then generates a sequence of actions (a plan) to achieve the goal condition set by the user. Many successful planners rely on heuristic search. But most heuristic search approaches use a propositional (grounded) formulation of actions, even though users typically describe problems in a more compact first-order logic (lifted) form. Converting lifted to grounded can be computationally infeasible.
My PhD research analyzes this issue and explores ways to mitigate it. I identify use-cases where lifted approaches are provably beneficial and develop heuristics that operate directly on lifted representations. This overcomes limitations of propositional methods. Here, the solver makes the best use of the description the user provides. So, rather than solely optimizing the solver to its limits, an alternative approach is to examine how a problem should be modeled to make solving easier. I have recently started exploring this perspective as well, as I believe considering both perspectives is crucial to achieving the best performance.
Overall, by focusing on a user provided knowledge representation, my research closely aligns with general problem-solving paradigms from fields like logic or operations research. Many advancements from these areas directly benefit my work. I hope that in the future my research will, in turn, contribute to these fields as well.

divider line

Publications:

Google Scholar DBLP

Conference Papers:

Workshop Papers:

divider line

Teaching:

The roles listed were all teaching positions in at Saarland University in Germany. In Germany, a teaching assistant is more like an Australian course convener, responsible for creating materials and handling organizational matter, while the professor gives the lectures. A student teaching assistant mainly helps with tutorials. The supervising student teaching assistant role is a rare position between the two mentioned roles and involves helping organize the workload for large tutor groups.

divider line

Academic services:

Reviewing:

I was a reviewer for the following journal(s): I was a PC member at: I subreviewed for:

Thesis Supervision:

I supervise(d) the thesis of:

Other Activities:

divider line

Imprint
Data protection