Skip to content

Latest commit

 

History

History
55 lines (31 loc) · 2.28 KB

README.md

File metadata and controls

55 lines (31 loc) · 2.28 KB
Raw

Landmark-Based-GoalRecognition

Landmark-Based Approaches For Goal Recognition.

  • Goal Recognition Filter using Landmarks;

  • Goal Completion Heuristic;

  • Landmark Uniqueness Heuristic;

These approaches have been published in ECAI-16, AAAI-17, and AIJ.

Usage

  • Option (1): Single tar.bz2 file containing domain, problem (initial state), set of goals, observations, correct goal, threshold value.

Parameters needed: <-filter | -goalcompletion | -uniqueness> <tar.bz2 file> <threshold_value>

java -jar goalrecognizer1.2.jar -filter experiments/blocks-test/blocks-test.tar.bz2 0
  • Option (2): Separate files, e.g., domain, problem (initial state), set of goals, observations, correct goal, threshold value.

Parameters needed: <-filter | -goalcompletion | -uniqueness> <domain.pddl> <problem.pddl> <goals.dat> <observations.dat> <correct_goal.dat> <threshold_value>

java -jar goalrecognizer1.2.jar -goalcompletion experiments/blocks-test/domain.pddl experiments/blocks-test/template.pddl experiments/blocks-test/hyps.dat experiments/blocks-test/obs.dat experiments/blocks-test/real_hyp.dat 0.1

Observations as Facts

Our approaches also deal with observations as facts. To use the recognizers in such mode, please use the branch called obsfacts:

At the following link we have some examples of how we use observations as facts:

There is also an executable file called goalrecognizer-obsfacts.jar.

Dependencies

Our goal recognizer uses the following libs (which are included in lib):

  • jgrapht-jdk1.6.jar (A free Java Graph Library);
  • planning-landmarks2.3.jar (A Landmark Extraction Algorithm based on Ordered Landmarks in Planning);
  • planning-utils2.2.jar (PDDL Parser and Planning data structure from JavaFF);