Goal Recognition with Path Signature:

This repository provides the code and dataset from the paper: Online Goal Recognition using Path Signature and Dynamic Time Warping

Install requirements

Symk Top-K planner: https://github.com/speckdavid/symk

PPDL parser: https://github.com/pucrs-automated-planning/pddl-parser

Install requirements:

pip install -r requeriments.txt

Running the Experiments

All experiments are run using the script compute_experiments.py, located in the ./Continuous directory.

Example (Continuous Domain):

python3 ./Continuous/compute_experiments.py -method gprs -par 10 -m 0.0 -p 0.0 -k 5

All experiments are run using the script compute_experiments.py, located in the ./Discrete directory.

Example (Discrete Domain):

python3 ./Discrete/compute_experiments.py -method gprs -m 0.0 -p 0.0 -k 5

Script Arguments

Argument	Description
-method	(str) Method name to use: • gprs: Goal Recognition with Path Signature • gprs_dtw: GRPS with Dynamic Time Warping
-par	(int) Number of parallel processes to use (e.g., 10)
-m	(float) Merge threshold. Use 0 to disable merging
-p	(float) Prune threshold. Use 0 to disable pruning
-k	(int) Top-K parameter: how many top-ranked trajectories to consider (e.g., 1, 15)

---

Output Format

After running the script, a .csv file containing the experiment results will be generated and saved in the results/ directory.

• The CSV file includes performance metrics (e.g., PPV) for each problem and parameter combination used in the experiment.

---