Professor Wheeler Ruml
How do self-driving cars figure out what to do? How should an autonomous airplane fly to reach its destination fastest, use the least amount of fuel, and avoiding bad weather? How should a butler robot decide how to go about setting the table for dinner?
In this seminar-style class, we'll read scientific papers about planning and decision-making algorithms (the part of AI concerned with deciding what to do). You will perform a research project on any topic you want that relates somehow to planning, decision-making, autonomy, or combinatorial search. The idea is to learn what research is like by actually doing some of it. The exact topics covered will be determined by the interests of the participants. Student from prior years are welcome to take the class again and either extend their previous work or choose a new topic. Examples of possible topics include:
The main student deliverables will be brief written responses to the papers each week and a substantial paper at the end of the semester reporting on the research project. Projects can be done in teams or individually. Projects can focus on general-purpose planning algorithms or topics more specific to robotics. Projects that relate specifically to robotics can be implemented in simulation or on actual robots.
You should already be a fluent programmer in a high-level language (like Java, C++, Scala, or Rust), understand common data structures (like trees, hash tables, and heaps), and be familiar with basic complexity analysis (big-O notation). Prior exposure to algorithms (eg, CS 758/858) and either artificial intelligence (eg, CS 730/830) or robotics (eg, CS 733/833) is helpful but not required. Students outside of CS are welcome but should probably check with me to ensure they have the background to succeed in the class.
Meeting time and place: Last time I checked, for Fall 2024, we were scheduled to meet Tuesdays and Thursdays 2:10-3:30pm in Hamilton-Smith 344. Let me know if you are interested in the class but can't make it at those times and we can try to find a time that works for everyone.
The class is usually offered every fall semester.
Tue Aug 27: We'll talk about what topics we might want to focus on for this semester. Bring ideas! If you are new to reading papers, you might want to review How to Read a Paper and The Task of the Referee.
Fri Dec 8 noon-2pm: poster presentations
Thu Dec 7: final presentations
Tue Dec 5: guest lecture on marine autonomy by Exail
Thu Nov 30: paper drafts due, those not in seminar present status updates
Tue Nov 28: how to write a paper, give a talk, and design a poster
[ Thu Nov 23 is Thanksgiving ]
Tue Nov 21: TAMP (IJRR, 2022)
Thu Nov 16: SM-Type(h) (ICAPS, 2022)
Tue Nov 14: movies, status updates
Thu Nov 9: birds in boots (AIIDE, 2021)
[ Tue Nov 7 is a Friday schedule ]
Thu Nov 2: LTS+CM (IJCAI, 2023)
Tue Oct 31: interim presentations
Thu Oct 26: potential search (ICAPS, 2011). if you are interested, see also DPS (SoCS-16)
Tue Oct 24: movies, status updates
Thu Oct 19: ILDS (AAAI, 1996)
Tue Oct 17: movies, status updates
Thu Oct 12: VRP (J Heuristics, 1995)
Tue Oct 10: movies, status updates
Thu Oct 5: empirical methodology: Hooker (J Heuristics, 1995) and Ruml (SoCS, 2010)
Tue Oct 3: monobead (ICAPS, 2022)
Thu Sep 28: AlphaZero and supplement (Science, 2018)
Tue Sep 26: problem presentations
Thu Sep 21: D*lite (AAAI, 2002)
Tue Sep 19: the LKH TSP heuristic (EJOR, 2000)
Thu Sep 14: alternating heuristics (extended TR version of an ICAPS-10 short paper)
Tue Sep 12: learning heuristic error online (ICAPS, 2011)
Thu Sep 7: MCPP (RAL, 2023)
Tue Sep 5: rectangle search (see email for link)
Thu Aug 31: ARA* (NIPS, 2003)
Tue Aug 39: We'll talk about what topics we might want to focus on for this semester. Bring ideas! If you are new to reading papers, you might want to review How to Read a Paper and The Task of the Referee.