Patrik Haslum

Patrik Haslum is an AI planning researcher at the Australian National University who focuses on admissible heuristics and plan quality optimisation. As a PhD student at Linköping University, he wrote "Admissible Heuristics for Optimal Planning" with Héctor Geffner in 2000, and the paper received the 2010 ICAPS Influential Paper Award. He co-authored the merge-and-shrink abstraction paper in the Journal of the ACM in 2014, served as program co-chair of ICAPS 2015 in Jerusalem, and wrote a book on the Planning Domain Definition Language with Nir Lipovetzky, Daniele Magazzeni and Christian Muise, published in 2019. He is an Associate Professor in the Intelligent Systems cluster of the ANU School of Computing.

Now at: Australian National University

Timeline

2000

Admissible heuristics with Geffner

In 2000, as a PhD student at IDA, the Department of Computer and Information Science at Linköping University, he writes "Admissible Heuristics for Optimal Planning" with Héctor Geffner, a guest researcher there.
2005–2009

Refining the h^m heuristic family

In 2005, he and Blai Bonet and Héctor Geffner introduce the additive h^m heuristic and the constrained pattern database heuristic for domain-independent planning. In 2009, writing from the Australian National University, he shows in an ICAPS short paper that the step from h^1 to h^m can be made by changing the planning problem.
2010

An ICAPS Influential Paper Award

ICAPS gives its 2010 Influential Paper Award to "Admissible Heuristics for Optimal Planning", the 2000 paper he wrote with Héctor Geffner while a PhD student at Linköping.
2014

Merge-and-shrink abstraction

In 2014, he, Malte Helmert, Jörg Hoffmann and Raz Nissim publish "Merge-and-shrink abstraction: A method for generating lower bounds in factored state spaces" in the Journal of the ACM. The paradigm represents a more general class of abstractions than pattern databases and, applied to planning, yields heuristic functions competitive with the state of the art.
2015

ICAPS program chair and BDPO2

In 2015, he serves with Shlomo Zilberstein as program chair of the 25th International Conference on Automated Planning and Scheduling in Jerusalem, listed with NICTA and ANU. With Fazlul Hasan Siddiqui, he publishes "Continuing Plan Quality Optimisation" in the Journal of Artificial Intelligence Research, presenting BDPO2, a system that keeps improving plans by replacing parts of them with better subplans.
2018

State constraints in planning

In 2018, he, Franc Ivankovic, Miquel Ramirez, Dan Gordon, Sylvie Thiébaux, Vikas Shivashankar and Dana S. Nau publish a JAIR paper that extends classical AI planning with systems of state constraints. The work adapts relaxation-based admissible heuristics to the extended formalism and targets problems such as planning-based control of power networks.
2019

A book on PDDL

In 2019, he, Nir Lipovetzky, Daniele Magazzeni and Christian Muise publish "An Introduction to the Planning Domain Definition Language" in the Synthesis Lectures on Artificial Intelligence and Machine Learning. The book covers PDDL, a formal knowledge representation language designed to express planning models.
As of October 2026

Associate Professor at ANU

He is an Associate Professor in the Intelligent Systems cluster of the ANU School of Computing and an HDR Co-Convener, with research on AI planning, plan quality optimisation, and AI for decision support. In 2026, he is a co-author of "I Always Told My Mom That Order Is Overrated: Unordered HTN Planning is in PSPACE and Models Problems Beyond STRIPS" at ICAPS 2026.