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 and AAAI-17.

• Landmark-Based Plan Recognition (ECAI-16);

• Landmark-Based Heuristics for Goal Recognition (AAAI-17);

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 goalrecognizer1.0.jar -filter experiments/blocks-test/blocks-test.tar.bz2 0

• Option (2): Separated 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 goalrecognizer1.0.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:

• https://github.com/ramonpereira/Landmark-Based-GoalRecognition/tree/obsfacts

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

• https://github.com/ramonpereira/Landmark-Based-GoalRecognition/tree/obsfacts/experiments/factobs

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);