Heuristics and search algorithms are the two key components of heuristic search, one of the main approaches to many variations of domain-independent planning, including classical planning, temporal planning, planning under uncertainty and adversarial planning. This workshop seeks to understand the underlying principles of current heuristics and search methods, their limitations, ways for overcoming those limitations, as well as the synergy between heuristics and search.
The proceedings are now available here.
All talks are available here:
Schedule (October 21, GMT)
All times are shown in GMT. Visit the official ICAPS schedule to see the global ICAPS schedule in your local timezone.
12:00 – 12:10: Opening Remarks
12:10 – 13:10 Invited Talk: Deep Learning for Generalised Planning (Sylvie Thiébaux)
13:30 – 14:50 Session 1, Chair: Álvaro Torralba
- Learning Search-Space Specific Heuristics Using Neural Network (Liu Yu, Ryo Kuroiwa and Alex Fukunaga)
- Beating LM-cut with LM-cut: Quick Cutting and Practical Tie Breaking for the Precondition Choice Function (Pascal Lauer and Maximilian Fickert)
- Online Saturated Cost Partitioning for Classical Planning (Jendrik Seipp)
- Subset-Saturated Transition Cost Partitioning for Optimal Classical Planning (Dominik Drexler, David Speck and Robert Mattmüller)
16:00 – 17:20 Session 2, Chair: Alberto Camacho
- Investigating Lifted Heuristics for Timeline-based Planning (Riccardo De Benedictis and Amedeo Cesta)
- Generating Data In Planning: SAS+ Planning Tasks of a Given Causal Structure (Michael Katz and Shirin Sohrabi)
- Bounding Quality in Diverse Planning (Michael Katz, Shirin Sohrabi and Octavian Udrea)
- Automatic Configuration of Benchmark Sets for Classical Planning (Álvaro Torralba, Jendrik Seipp and Silvan Sievers)
17:50 – 19:30 Session 3, Chair: David Speck
- Revisiting Dominance Pruning in Decoupled Search (Daniel Gnad)
- On the Optimal Efficiency of A* with Dominance Pruning (Álvaro Torralba)
- Approximate bi-criteria search by efficient representation of subsets of the Pareto-optimal frontier (Oren Salzman)
- An Atom-Centric Perspective on Stubborn Sets (Gabriele Röger, Malte Helmert, Jendrik Seipp and Silvan Sievers)
- Simplified Planner Selection (Patrick Ferber)
The HSDIP workshop is proud to announce that Sylvie Thiébaux will be a keynote speaker for this year’s workshop. Sylvie is a professor of Computer Science at ANU. In the recent past, she was the director of NICTA’s Canberra Laboratories, home to 150 researchers and PhD students, and the Associate Dean Research for the College of Engineering and Computer Science. She is co-Editor in Chief of the Artificial Intelligence journal (AIJ), a AAAI fellow, a former ICAPS president, and associate editor of JAIR.
Submission deadline: July 20
Open discussion: August 18 – September 8
Notification: September 12
Camera-ready copy: October 5
Workshop: October 21
Current HSDIP submissions
For currently submitted papers we will give an initial assessment at the end of March. This will not be in form of an official review, but allows reviewers to ask questions or discuss specific aspects of the work with the authors. The reviewers will actively engage in discussions throughout April. The assessment and discussions are open to the public (but anonymous) and we encourage readers to take part in the discussion. Currently submitted papers are available on OpenReview.
Topics and Objectives
Search guided by heuristics, automatically derived from a declarative formulation of action effects, preconditions and goals, has been a successful approach to domain-independent planning. From the initial success of heuristics based on syntactic relaxations and abstractions, the theory and practice of developing novel heuristics have become more diverse, often borrowing concepts and tools from Optimisation and Satisfiability, and bolder, tackling more expressive planning languages.
In parallel to the increasing maturity of the methods and tools used to derive heuristic methods, important theoretical results have brought around a more clear image of how heuristic methods relate to each other. For instance, it has been shown that classic frameworks for heuristic search as planning can be encoded symbolically and their execution simulated via off-the-shelf satisfiability solvers. Groundbreaking theoretical work has shown how heuristic methods can be grouped into distinct families, depending on whether they can or cannot be shown to dominate or be compiled into each other.
As a result, the formulation of heuristics for domain-independent planning is increasingly being less about describing procedures that exploit specific features in declarative information, and more about describing auxiliary constraints that make apparent those features to off-the-shelf solvers that operate over a logical or algebraic theory that over-approximate the set of valid plans and compute the heuristic estimator.
Last, but not least, there is a growing realization that the search algorithm used can significantly amplify or reduce the utility of specific heuristics. Recent work that highlights the pitfalls latent in well-known search algorithms, also suggests opportunities to exploit synergies between the heuristic calculation and the search control.
The workshop on Heuristics and Search for Domain-Independent Planning (HSDIP) is the 12th workshop in a series that started with the “Heuristics for Domain-Independent Planning” (HDIP) workshops at ICAPS 2007. At ICAPS 2012, the workshop was changed to its current name and scope to explicitly encourage work on search for domain-independent planning.
Examples of typical topics for submissions to this workshop are:
- automatic derivation of heuristic estimators for domain-independent planning
- formal results showing equivalence or dominance between heuristics
- novel heuristic methods dealing with planning with numeric variables and effects, partial observability and non-deterministic action effects
- heuristic estimators for domain-independent planning via procedures or suitably defined encodings of declarative descriptions of planning tasks into Satisfiability or Optimisation
- novel search techniques for domain-independent planning that explicitly aim at exploiting effectively the properties of existing heuristics
- empirical observations of synergies between heuristics and search in domain-independent planning
- challenging domains for existing combinations of heuristics and search algorithms
The HSDIP workshop has always been welcoming of multidisciplinary work, for example, drawing inspiration from operations research (like row and column generation algorithms), convex optimization (like gradient optimization for hybrid planning), constraint programming or satisfiability, or applications of machine learning in heuristic search (e.g., learning heuristics, adaptive search strategies, or heuristic selection).
Please format submissions in AAAI style (see instructions in the Author Kit) and keep them to at most 9 pages including references. Authors considering submitting to the workshop papers rejected from the main conference, please ensure you do your utmost to address the comments given by ICAPS reviewers.
Please do not submit papers that are already accepted for the main conference to the workshop.
Submissions will be made through OpenReview.
The following conditions apply:
- Submissions will be double blind in general and single blind to the area chair.
- The submitted papers, reviews and discussion between authors and reviewers will be public, and all anonymous.
- Discussions between reviewers and organizers will be private.
Every submission will be reviewed by two members of the organizing committee according to the usual criteria such as relevance to the workshop, significance of the contribution, and technical quality. There will be a brief 1 week discussion phase where author and reviewers can interactively engage and discuss the submission and the reviews.
Submissions sent to other conferences are allowed. It is the responsibility of the authors to ensure that those venues allow for papers submitted to be already published in “informal” ways (e.g. on proceedings or websites without associated ISSN/ISBN). In particular, we welcome submissions sent to IJCAI and made sure that the workshop discussion phase does not conflict with the IJCAI rebuttal phase.
The workshop is meant to be an open and inclusive forum, and we encourage papers that report on work in progress or that do not fit the mold of a typical conference paper. Non-trivial negative results are welcome to the workshop, but we expect the authors to argue for the significance of the presented results to alternative lines of research on the topic of choice.
At least one author of each accepted paper must attend the workshop in order to present the paper. Authors must register for the ICAPS main conference in order to attend the workshop. There will be no separate workshop-only registration.
Contact and Inquiries
- Alberto Camacho, Google and University of Toronto, Canada
- Salomé Eriksson, University of Basel, Switzerland
- Daniel Fišer, Czech Technical University, Czech Republic
- Guillem Francès, Universitat Pompeu Fabra, Spain
- Florian Geißer, Australian National University, Australia
- Patrik Haslum, Australian National University, Australia
- Jendrik Seipp, University of Basel, Switzerland
- Silvan Sievers, University of Basel, Switzerland
- David Speck, University of Freiburg, Germany
- Álvaro Torralba, Aalborg University, Denmark