Automated Planning and Scheduling

  • Typ: Vorlesung (V)
  • Semester: WS 18/19
  • Ort:

    Geb. 50.34 Informatik Räume -119 und 301

  • Zeit:

    Mittwochs 14:00 – 15:30 (Raum -119), Donnerstags 14:00 - 15:30 (Raum 301)

  • Dozent:

    Prof. Dr. Peter Sanders
    Dr. Tomas Balyo M.Sc.
    D. Schreiber

     

  • SWS: 2/1
  • LVNr.: 2400026
Voraussetzungen

keine

Lehrinhalt

The course offers an introduction to the methods and techniques used in automated planning and scheduling. The course is focused on classical deterministic planning, i.e., planning in a fully observable deterministic environment. The students will learn how to use automated planners and schedulers and also how they work. The topics covered in the lecture include:

  • applications of automated planning in artificial intelligence
  • formalization of planning problems and the PDDL language
  • computational complexity of planning and scheduling
  • basic state space search algorithms (forwards/backwards search)
  • heuristic search algorithms and planning heuristics
  • plan space planning
  • planning graph and the graph plan algorithm
  • satisfiability based planning
  • hierarchical task network planning
  • classical scheduling approaches
  • constraint-based scheduling
  • planning for virtual agents in computer games
Arbeitsbelastung

2 SWS Vorlesung + 1 SWS Übungen

(Vor- und Nachbereitungszeiten: 4h/Woche für Vorlesung plus 2h/Woche für Übungen; Prüfungsvorbereitung 15h)

Gesamtaufwand: (2 SWS + 1 SWS + 4 SWS + 2 SWS) x 15h + 15h Prüfungsvorbereitung = 9x15h + 15h = 150h = 5 ECTS

Ziel
  • The students will be able to model various planning tasks in the PDDL language and solve them using off-the-shelf planners.
  • The students will understand the approaches used in automated planning and scheduling algorithms, which will allow them to efficiently model and solve real world planning and scheduling problems by selecting the proper algorithms for the given task.

Prüfung

Sie müssen sich vor der Prüfung am Studierendenportal   für diese Veranstaltung anmelden.