Foundations of Artificial Intelligence (FAI) Group

Seminar: Trusted AI Planning (TAIP)

Basics. Seminar, 7 graded ECTS points.

The seminar will be run in a block format. There will be an initial meeting on Monday, April 22, 16:00-17:00. All student presentations will be given on a single day after the end of term.

All meetings will take place in room 3.06, Building E1 1. The seminar language is English throughout.

Supervisors for the seminar are Jörg Hoffmann and Dan Fišer. Additional feedback will be provided by Jan Eisenhut and Marcel Vinzent.

Your task will be to read and understand a piece of research, to write a summary paper in your own words, to give a presentation, and to provide detailed feedback for the paper and presentation of a fellow student.

All email interaction must be pre-fixed with "[TAIP-24]" in the email subject.

No plagiarism. It is Ok (and encouraged!) to use web resources to further your understanding of your assigned topic. However, it is inadmissible to use pieces of such material for your summary paper or presentation. Any plagiarism will result in disqualification from the seminar.

Content. AI Planning is the sub-area of AI concerned with complex action-choice problems. The seminar covers methods supporting trust in algorithmic solutions to such problems. This field of research is very recent, and we cover research conducted in the FAI group. A major concern are neural action policies, i.e., neural networks that map states to actions. While such policies can be very performant, they are fundamentally opaque and come without any guarantees. We cover methods for verifying, testing, and re-training such policies. Furthermore, even for symbolic planning methods explainability is a challenge, and we cover recent work providing explanations in the form of goal conflicts.

Prerequisites. Participants must have successfully completed either an edition of the Artificial Intelligence core course, or of the AI Planning specialized course.

Registration. Is via the central seminar registration system.

Grading. The final grading will be based, in this order of importance, on:

• The quality of your final presentation.
• The quality of your final summary paper.
• The quality of the feedback you provide to your mentee student (see below).
• Your participation in the discussions during the block seminar.

Summary Paper. For the summary paper, you must use this tex template. You are required to read at least 2 related papers, for the related work section. You are allowed to modify the section structure given in the template if, for whatever reason, this is more adequate for the work you are summarizing.

The seminar paper should be about 4 pages long (not counting the literature list, and in the double-column format of the template). This is a rough guideline, not a strict rule. If you need, say, 5-6 pages to do your paper justice then definitely do so.

Schedule and Deadlines (tentative!).

• April 22, 16:00-17:00: Initial meeting. We will give a brief insight into each of the papers.

• April 24: Send a ranked list of the topics you would like to take, by email to Dan Fišer. That is, send something like "area 1.2, area 2.2, area 1.1, area 4.2, area 3.1". Please include into this list all topics that you would be willing to accept. The list must contain at least 5 topics.

  Note that each topic is associated with a mentee student (to whom you will provide feedback, see the following deadlines); and a mentor student (who will provide feedback to you, see the following deadlines). The mentee/mentor assignment will be a "cycle" through each of the topic areas as listed below: within an area with 2 topics, the two students mentor each other; within areas with k>2 topics, the mentor->mentee relation is i->i+1 and k->1. If you want to team up with someone specific, please do state that in your email.

• April 25: Receive your topic. Read the material associated with your topic carefully, and prepare an initial version of your summary paper, using the tex template given above.

• May 16: Deadline for official registration to this seminar (exam registration, Prüfungsanmeldung).

• May 27 - May 31: Make an appointment with your feedback-giver (as listed with each paper) to discuss your paper. The purpose of this meeting is to ensure that you understood the paper correctly, and to ask questions about specific points.

  NOTE: The following deadlines marked with "(ca.)" are meant as a guideline. You are required to do these things, but if you do them 3-4 days earlier or later, that is no problem.

• June 21 (ca.): Send your summary paper to your mentor student (cc supervisor and feedback giver).
• June 28 (ca.): Send feedback regarding the summary paper to your mentee student (cc supervisor and feedback giver).
• July 5 (ca.): Send revised summary paper to your mentor student (cc supervisor and feedback giver).
• July 12 (ca.): Send feedback regarding the revised summary paper to your mentee student (cc supervisor and feedback giver).
• July 19 (ca.): Send presentation slides to mentor student (cc supervisor and feedback giver).
• July 23 (ca.): Send feedback regarding the presentation slides to your mentee student (cc supervisor and feedback giver).
• July 26 (ca.): Send revised presentation slides to mentor student (cc supervisor and feedback giver).
• July 31 (ca.): Send feedback regarding the revised presentation slides to your mentee student (cc supervisor and feedback giver).

• August 5: Send your final summary paper by email to your supervisor.
• August TBA: Give a presentation (20 minutes talk, plus 10 minutes discussion) in the block seminar. Attendance to all talks is required. Please try to stick to the 20 minutes time slot for your talk; it should not be a lot shorter, nor a lot longer.

Topics. Each participant will be assigned one topic, each of which consists of one paper. The overall amount and difficulty of the material associated with each topic is roughly balanced.

Area 1: Policy Verification (supervisor: Jörg Hoffmann)

1. M. Vinzent, M. Steinmetz, and J. Hoffmann, Neural Network Action Policy Verification via Predicate Abstraction, Proceedings of the 32nd International Conference on Automated Planning and Scheduling (ICAPS'22), 2022. Paper: Area 1 Topic 1. Feedback giver: Marcel Vinzent.
2. M. Vinzent, S. Sharma, and J. Hoffmann, Neural Policy Safety Verification via Predicate Abstraction: CEGAR, Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI'23), 2023. Paper: Area 1 Topic 2. Feedback giver: Marcel Vinzent.
3. M. Vinzent, M. Wu, H. Wu, and J. Hoffmann, Policy-Specific Abstraction Predicate Selection in Neural Policy Safety Verification, Proceedings of the Workshop on Reliable Data-Driven Planning and Scheduling (RDDPS), at (ICAPS'23). Paper: Area 1 Topic 3. Feedback giver: Marcel Vinzent.
4. C. Jain, L. Cascioli, L. Devos, M. Vinzent, M. Steinmetz, J. Davis, J. Hoffmann, Safety Verification of Tree-Ensemble Policies via Predicate Abstraction, Proceedings of the Workshop on Reliable Data-Driven Planning and Scheduling (RDDPS), at (ICAPS'24). Paper: Area 1 Topic 4. Feedback giver: Marcel Vinzent.

Area 2: Policy Testing (supervisor: Dan Fišer)

1. M. Steinmetz, D. Fiser, H. Eniser, P. Ferber, T. Gros, P. Heim, D. Hoeller, X. Schuler, V. Wuestholz, M. Christakis, and J. Hoffmann, Debugging a Policy: Automatic Action-Policy Testing in AI Planning, Proceedings of the 32nd International Conference on Automated Planning and Scheduling (ICAPS'22), 2022. Paper: Area 2 Topic 1. Feedback giver: Jan Eisenhut.
2. H. Eniser, T. Gros, V. Wuestholz, J. Hoffmann, and M. Christakis, Metamorphic Relations via Relaxations: An Approach to Obtain Oracles for Action-Policy Testing, Proceedings of the ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA'22), 2022. Paper: Area 2 Topic 2. Feedback giver: Jan Eisenhut.
3. J. Eisenhut, A. Torralba, M. Christakis, and J. Hoffmann, Automatic Metamorphic Test Oracles for Action-Policy Testing, Proceedings of the 33rd International Conference on Automated Planning and Scheduling (ICAPS'23), 2023. Paper: Area 2 Topic 3. Feedback giver: Jan Eisenhut.

Area 3: DSMC and Re-Training (supervisor: Dan Fišer)

1. T. Gros, H. Hermanns, J. Hoffmann, M. Klauck, and M. Steinmetz, Analyzing Neural Network Behavior through Deep Statistical Model Checking, International Journal on Software Tools for Technology Transfer, 2022. Paper: Area 3 Topic 1. Feedback giver: Dan Fišer.
2. T. Gros, D. Höller, J. Hoffmann, M. Klauck, H. Meerkamp, and V. Wolf, DSMC Evaluation Stages: Fostering Robust and Safe Behavior in Deep Reinforcement Learning, Proceedings of the 18th International Conference on Quantitative Evaluation of SysTems (QEST'21), 2021. Paper: Area 3 Topic 2. Feedback giver: Dan Fišer.

Area 4: Explanation (supervisor: Jörg Hoffmann)

1. R. Eifler, M. Cashmore, J. Hoffmann, D. Magazzeni, and M. Steinmetz, A New Approach to Plan-Space Explanation: Analyzing Plan-Property Dependencies in Oversubscription Planning, Proceedings of the 34th AAAI Conference on Artificial Intelligence (AAAI'20), New York City, USA, 2020. Paper: Area 4 Topic 1. Feedback giver: Jörg Hoffmann).
2. R. Eifler, M. Steinmetz, A. Torralba, and J. Hoffmann, Plan-Space Explanation via Plan-Property Dependencies: Faster Algorithms & More Powerful Properties, Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI'20), 2020. Paper: Area 4 Topic 2. Feedback giver: Jörg Hoffmann).
3. A. Siji, R. Eifler, D. Fiser, and J. Hoffmann, Action Policy Explanations in Oversubscription Planning, Proceedings of the International Workshop of Human-Aware and Explainable Planning (HAXP'23), at ICAPS'23. Paper: Area 4 Topic 3. Feedback giver: Jörg Hoffmann).