Abteilung
Grundlagen der Künstlichen Intelligenz,
Institut
für Informatik,
Universität Freiburg
Abteilung
Grundlagen der Künstlichen Intelligenz,
Institut
für Informatik,
Universität Freiburg
| Exercise Sheet | Handed Out | Due | Model Solution | Common Problems |
|---|---|---|---|---|
| Exercise 1 | Wednesday, October 16 | Monday, October 21 | Model Solution | Common Problems |
| Exercise 2 | Monday, October 21 | Monday, October 28 | Model Solution | Common Problems |
| Exercise 3 | Monday, October 28 | Monday, November 4 | Model Solution | Common Problems |
| Exercise 4 | Monday, November 4 | Monday, November 11 | Model Solution | Common Problems |
| Exercise 5 | Monday, November 11 | Monday, November 18 | Model Solution | Common Problems |
| Exercise 6 | Monday, November 18 | Monday, November 25 | ||
| Exercise 7 | Monday, November 25 | Monday, December 2 | ||
| Exercise 8 | Monday, December 2 | Monday, December 9 | ||
| Exercise 9 | Monday, December 9 | Monday, December 16 | ||
| No exercise for Christmas! | Monday, December 16 | |||
| Exercise 10 | Monday, January 13 | Monday, January 20 | ||
| Exercise 11 | Monday, January 20 | Monday, January 27 | Model Solution | |
| Exercise 12 | Monday, January 27 | Monday, February 3 | Model Solution |
You may work on these assignments and submit them in groups of two
students.
Make sure to clearly indicate both names on your work.
You may write your answers in English or German.
Exercise marks count towards your final grade for this course, which is calculated from exercise marks (20%) and exam marks (80%).
FF is a planning system developed by Jörg Hoffmann that can solve planning tasks specified in the PDDL formalism. You can download ready-to-use versions for Linux (on a PC) or Solaris (on a SUN Workstation such as the ones in the computer pools of building 082).
Alternatively, you can compile FF for another operating system or machine if you download the source package and follow the instructions in the README file. You will need a make tool, a C compiler such as gcc, a parser generator such as yacc or bison, and a lexical analyzer generater such as lex or flex.
To solve planning tasks with FF, you need a domain specification file (e.g. domain.pddl), which defines the predicates and operators of the planning domain, and a problem specification file (e.g. problem.pddl), which defines the objects, initial state and goal states of the planning task. FF is run with the command ./ff -o domain.pddl -f problem.pddl. The solution is printed to screen and to a file named problem.pddl.soln.
The domain specification file and problem specification files for the crazy switches domain used in Exercise 1.1 can be downloaded here.
The domain specification file and problem specification file for the gripper domain used in Exercise 1.2 can be downloaded here.
The initial state and grounded operator definition for the gripper-3 task can be downloaded here.
MBP is a planning system developed at the Centre for Scientific and Technological Research at the Istituto Trentino di Cultura (ITC) by Piergiorgio Bertoli, Alessandro Cimatti, Ugo Dal Lago, Marco Pistore, Marco Roveri, and Paolo Traverso. Among other things, it can solve conditional planning tasks specified in the NuPDDL formalism, which is closely related to the standard PDDL language used by deterministic planning systems. You can download ready-to-use versions for Linux (on a PC) or Solaris (on a SUN Workstation such as the ones in the computer pools of building 082). Source versions are not available; please contact us if the precompiled packages don't satisfy your needs.
Unlike FF, planning tasks for MBP are not necessarily separated into multiple files. For the examples in the exercises, you will only need to specify one input file (e.g. input.npddl). In this case, the easiest way to run MBP is by the command ./run-mbp input.npddl. The solution is printed to the screen; use ./run-mbp input.pddl > output to write it into a file.
helmert@informatik.uni-freiburg.de, 6. Februar 2003