Call for Papers: ICAPS 2027

The International Conference on Automated Planning and Scheduling (ICAPS) is the premier forum for research on the theory, algorithms, and applications of automated planning and scheduling technology. The 37th edition of the ICAPS conference series will be held in Columbia, SC, USA from June 27, 2027 to July 2, 2027.

In addition to the main track, ICAPS-27 will feature a special track on “Planning Under a Different Name.” Papers submitted to either track will follow the submission and formatting requirements described below. Work that is naturally framed and evaluated within the existing planning and scheduling community, including planning methods that incorporate machine learning or foundation-model components, belongs in the main track. The special track is intended primarily for work rooted in neighboring AI communities that makes a substantive contribution to planning or deliberation but is framed using different terminology, literature, benchmarks, or methodological conventions. Additional scope and evaluation criteria are described on the special track page.

Scope

ICAPS-27 welcomes paper submissions on all aspects of automated planning and scheduling. Planning and scheduling are interpreted broadly to encompass a wide range of decision-making, optimization, and reasoning problems involving the selection, sequencing, coordination, and timing of actions. This includes domain modeling, plan and schedule synthesis, execution and monitoring, failure diagnosis, model repair, and associated learning, representation, and reasoning problems. ICAPS welcomes work on deterministic and stochastic sequential decision-making, under full or partial observability, and with factored or non-factored state representations.

Topics in scope include (but are definitely not limited to):

  • Planning and Scheduling Problems:
    • Classical planning and theoretical foundations of planning
    • Temporal planning, scheduling, and routing
    • Planning under uncertainty, including MDPs, POMDPs, nondeterministic planning, and planning with sensing
    • Planning with incomplete models, incomplete information, and belief states
    • Hierarchical planning
    • Real-time, online, and lifelong planning
    • Multi-agent and distributed planning
    • Motion/path planning, task and motion planning, and planning for hybrid systems
    • Human-aware planning, scheduling, and execution
  • Planning Across the Decision-Making Lifecycle:
    • Knowledge engineering and representation for planning, including reasoning about actions, knowledge, and belief
    • Activity and plan recognition
    • Plan execution, monitoring, and diagnosis
    • Replanning, plan repair, and model repair
    • Planning under execution uncertainty
  • Planning Methods and Computational Techniques:
    • Search methods for planning and scheduling
    • SAT, constraint programming, and model checking
    • Mathematical programming and optimization
    • Local search, evolutionary algorithms, and other heuristic optimization methods
    • Learning-based planning, including planning methods that use or integrate foundation models (LLMs, VLMs, VLAs, etc.)
    • Decomposition, abstraction, and other methods for scalable planning and scheduling
  • Planning Applications and Emerging Directions:
    • Description and modeling of novel application domains
    • Engineering issues in using, deploying, and scaling planning and scheduling techniques
    • User interface design, visualization, or human-system collaboration for a planning and scheduling application
    • Evaluation, testing, and validation of planning and scheduling applications in societal or industrial environments
    • Assessment of the impact of planning and scheduling systems on end users, customers, markets, or society at large
    • Industry/application challenge problems in planning and scheduling (including benchmark instances)

Contributions are welcome in each of the following categories:

  • Theoretical papers, which broaden or improve the set of analytical tools used to study planning and scheduling problems and algorithms. Examples include complexity results, expressiveness, and new theoretical frameworks.
  • Algorithmic papers, which describe novel perspectives and substantial (qualitative or quantitative) improvements for solving planning and scheduling problems. Examples include new optimizations or specializations of existing algorithms, new propagators, and new decomposition approaches.
  • Modeling papers, which describe new representations of planning and scheduling problems and their solutions. Examples include new mathematical frameworks for existing problems, original descriptions of emerging problems, and refinements of existing frameworks for knowledge representation of actions, goals, states, or other rigorously defined concepts.
  • Position papers, which contribute thoughtful critiques or bold new perspectives on the field. Such papers should articulate a clear thesis and support it through appropriate evidence, analysis, or synthesis of prior work. Examples include meta-analysis of research trends, descriptions of new challenge problems suitable for planning and scheduling, historical perspectives and analysis of the field, and technical discussions of various implementation techniques.
  • Tool papers, which describe systems that are useful to and of interest to the planning and scheduling community, and which are built using novel algorithmic and engineering techniques. Examples include: integrated planning systems, model checkers and synthesis tools, libraries for constructing, managing, and transforming representations of planning and scheduling problems, and applications for visualizing, benchmarking, and comparing planners or other types of tools.
  • Empirical and evaluation papers, which provide new scientific insights through careful experimental study of planning and scheduling methods, benchmarks, assumptions, or evaluation methodologies. Such papers should make a substantive contribution beyond applying established methods to additional problem instances.
  • Application papers, which show how planning and scheduling methods can push the envelope for real-world problems.

Papers that do not address problems related to automated planning or scheduling are out of scope and will be rejected without review. Work on reinforcement learning or other forms of sequential decision making is not in scope solely by virtue of addressing sequential decisions; it must make a substantive contribution to planning or scheduling. Where the relationship of the paper to planning and scheduling is not immediately obvious, authors should explain this relationship, and the relevance of the contribution to the ICAPS community, in the abstract and introduction of the paper.

Key Dates

The reference timezone for all deadlines is UTC-12. That is, the deadline has not passed as long as there is still time anywhere in the world.

  • Abstract submission deadline: December 7, 2026, 11:59 PM UTC-12
  • Paper submission deadline: December 14, 2026, 11:59 PM UTC-12
  • Author response period: February 1-4, 2027
  • Notification: February 26, 2027

Author Guidelines

We welcome submissions of both long (8 pages plus additional page(s) for references) and short (4 pages plus additional page(s) for references) papers. Whether the paper is long or short must be indicated at submission time. All papers submitted to ICAPS-27 must be in AAAI Format.

Over-length papers will be rejected without review. An over-length paper is one in which content other than references and the optional ethical impact statement appears on page 9 (long papers) or page 5 (short papers).

All reviewed papers, whether they are in the long or short category, are expected to meet the same high standards set by ICAPS. Contributions will be evaluated according to their nature and submission category. Short papers may be narrower in scope than long papers, yet still make a complete and significant contribution. For example, they may address a highly specific issue or propose or evaluate a small but important extension, a new idea, an empirical finding, or a technical contribution.

All accepted papers will be published in the proceedings and presented at the conference, either orally or as posters.

Papers submitted to ICAPS-27 may not be submitted to other conferences or journals during the ICAPS-27 review period, nor may they be already under review, accepted, or published in other conferences or journals.

Double-blind requirements

Submissions must be double-blind. Authors should omit their names and affiliations and refer to their own prior work in the third person. Acknowledgments and ethics statements must not contain information that identifies the authors.

We discourage authors from posting their manuscripts on arXiv.org or other similar repositories from two weeks before the submission deadline until the author notification deadline to avoid de-anonymizing the paper to potential reviewers.

Reproducibility

ICAPS contributions are frequently empirical, and the community’s ability to build on them depends on how experiments were run. Papers reporting experimental results are expected to state, in the paper or in an appendix:

  • the hardware and operating system used, and the time and memory limits imposed per instance;
  • the exact benchmark set, including how instances were selected and where they can be obtained;
  • the version (release or commit) of every planner, solver, or library used, and all non-default parameter settings;
  • for methods using foundation models, the exact model identifiers and versions, decoding parameters, the full prompts, and the dates on which the models were queried.

Where results are reported on a subset of a standard benchmark set, the selection criteria must be stated and must not depend on the results obtained.

Where a method involves randomness, results must be aggregated over multiple runs, and the number of runs and a measure of dispersion must be reported rather than means alone. Authors should not claim an improvement that falls within the variation they report. Where a claimed improvement is small relative to that variation, authors are expected to support it with an appropriate statistical test over runs.

Comparisons on fixed benchmark sets are a different matter. Standard benchmark sets are not random samples, and instances within a domain are strongly correlated, so tests that treat instances as independent can substantially overstate significance. ICAPS-27 does not require such tests, and reviewers should not ask for them by default. Authors who do report a statistical comparison should state what is being treated as the unit of analysis.

We strongly encourage authors to release code, learned models, benchmark instances, and raw per-instance results, and to include a brief availability statement describing what will be released and where. Links included in submissions must be anonymized; camera-ready versions should use a persistent archive with a DOI.

Reviewers will be asked whether a paper reports enough detail for an independent group to reproduce its main results. Inability to release code for commercial or institutional reasons is not, in itself, grounds for rejection; omitting experimental detail may be.

Ethical / Societal Impact

Authors are encouraged to include a statement of the potential broader impact of their work, including its ethical aspects and future societal consequences. This statement can be included either within the main-body page limit or on the pages otherwise reserved for references. If a paper does not include such a statement but reviewers determine that one is necessary, the authors may be asked to provide a statement during the author response period. The statement will be considered as part of the review process and, for accepted papers, must be incorporated into the camera-ready version.

For inquiries, contact: icaps-2027-organizers@googlegroups.com

ICAPS 2027 Program Chairs

  • Stephen F. Smith, Carnegie Mellon University, USA
  • Jiaoyang Li, Carnegie Mellon University, USA
  • Shahaf Shperberg, Ben-Gurion University, Israel

ICAPS 2027 Conference Chairs

  • Biplav Srivastava, University of South Carolina, USA
  • Forest Agostinelli, University of South Carolina, USA