Skip to content

CLI reference

Entry points

The installed command and module entry point are equivalent:

online-gcs --help
python -m online_gcs --help

Argument and runtime errors produce a short message and a non-zero exit code.

Minimal run

online-gcs --scene SINGLE_SHELF --iterations 1 --keypoints 2 --prune-interval 0 --seed 42
Option Meaning
--scene SINGLE_SHELF, TWO_SHELVES, or TABLE_THREE_SHELVES
--iterations Number of online queries; must be at least 1
--keypoints Keypoints requested from each fallback path; must be at least 2
--seed Random seed used by supported planner components
--prune-interval Region-pruning interval; 0 disables pruning
--warmstart Build initial IRIS regions before the query loop
--warmstart-seeds Number of warm-start seeds
--smart-keypoints Enable the experimental keypoint selector
--k-shortest-paths Candidate graph paths used for GCS subgraph selection
--verbose Enable additional planner logging

Planner modes

Use only one baseline mode at a time:

online-gcs --scene SINGLE_SHELF --iterations 1 --rrt-only
online-gcs --scene SINGLE_SHELF --iterations 1 --opt-only

--rrt-only runs the sampling-based baseline. --opt-only runs trajectory optimization of fallback paths and therefore depends on solver availability.

Visualization and output

mkdir -p artifacts
online-gcs --scene SINGLE_SHELF --iterations 1 --output artifacts/regions.yaml

Add --visualize to print a local Meshcat URL. Region files and any other generated data belong in ignored results/ or artifacts/ directories. Parallel exploration is available with --parallel --num-workers N; use it only in an environment where multiprocessing and Drake scene construction have been verified.