Automated Planning and Acting

Overview

This lecture presents methods of automated planning and acting, ranging range from deterministic and temporal methods to non-deterministic and probabilistic methods. Next to those methods, this lecture covers decision theory as the basis for probabilistic methods and also take a look at the most recent developments such as provably beneficial acting and human-aware planning.

The course is set up such that no prior knowledge is necessary. Nonetheless, it is beneficial to have a good grasp of search algorithms and a basic understanding of probability theory and logic as certain formalisations might not look as foreign if you are familiar with those.

For the exercise, the plan is to have a seminar-style exercise with each person (or team if the number of participants gets too large) getting a topic to discover on their own and present to their peers. This presentation doubles as the prerequisite for admission to the final exam. The final exam will be an oral exam, bar a different announcement during the first lecture, depending on the number of participants.

Please register in the Learnweb course for more details and material.

Presentation Material

  1. Introduction (pptx, pdf)
  2. Planning and Acting with Deterministic Models (pptx, pdf)
  3. Planning and Acting with Refinement Methods (pptx, pdf)
  4. Planning and Acting with Temporal Models (pptx, pdf)
  5. Planning and Acting with Nondeterministic Models (pptx, pdf)
  6. Standard Decision Making (pptx, pdf)
  7. Planning and Acting with Probabilistic Models (pptx, pdf)
  8. Advanced Decision Making (pptx, pdf)
  9. Human-aware Planning (pptx, pdf)

Literature

  • Automated Planning and Acting, Malik Ghallab, Dana Nau, Paolo Traverso (mainly for topics 1-4, 6)
  • Artificial Intelligence: A Modern Approach (3rd edition, Chs. 16, 17, 21), Stuart Russell, Peter Norvig (mainly for topics 5, 7)
  • Further sources in the form of conference papers / articles on other or more recent developments as marked on slides

All sources should be available (i.e., find-able) online.