ObjectRecognitionSkill.action

#################
# Goal definition

# The perception pipeline supports several operation modes.
# The SETUP modes are for preloading the objectModel before 3D perception.
# -> Useful to reduce perception time -> for example, while a robot arm is
#    moving to the scan position, the pipeline can be loading and
#    setting up the objectModel data.
# The option to save the filtered point cloud is useful for exporting a
#  filtered and segmented cluster for later usage as reference point cloud.
# -> For example, you can take a scan, move a gripper to the grasp pose and
#    then send a goal to the perception pipeline to save the cluster that is
#    closest to the TF associated with that gripper
#    (using the goal field tfNameForSortingClusters).
# -> Moreover, you can encode the grasp pose as the coordinate system of the
#    reference point cloud by setting the filteredPointcloudSaveFrameId to
#    the gripper frame_id.

# Perception modes:
# SETUP
# |-> The objectModel will be loaded if it was not cached in the previously goal.
# SETUP_WITHOUT_CACHING
# |-> The objectModel will be loaded even if it was previously cached.
#     Useful to reload a model that was changed after the caching goal.
# SAVE_FILTERED_POINTCLOUD
# |-> The filtered point cloud will be saved with the name specified in objectModel
#      in the folder reference_pointclouds_database_folder_path
#      (specified in the object_recognition.launch).
# PERCEPTION
# |-> The pipeline will perform 3D perception to estimate the 6 DoF pose of the
#     object specified in objectModel.
#     This is the default mode (when operationMode is empty).
string operationMode

# The reference point cloud objectModel can be:
# 1 - .ply file name, present in the reference_pointclouds_database_folder_path
#     (specified in the object_recognition.launch).
# 2 - Full path to the .ply file.
# The reference point cloud must have uniform point density.
# |-> If you use CAD models, check the documentation about mesh tessellation
#     for ensuring that the point cloud has uniform point density.
string objectModel

# TF name for sorting clusters based on the distance metric.
string tfNameForSortingClusters

# Cluster to select after they are sorted by the configured metric.
# For example, if the euclidean_clustering module was configured to use
# MaxClusterSizeSorter, the index 0 will be for the largest cluster,
# the index 1 will be for the second largest cluster, ...
int32 clusterIndex

# When using Principal Component Analysis (PCA), a custom flip axis can be
# specified to ensure repeatability when estimating the initial 6 DoF pose.
geometry_msgs/Vector3 pcaCustomXFlipAxis

# When using the SAVE_FILTERED_POINTCLOUD operation mode, the point cloud
# can be transformed into a specific TF frame id.
# |-> Useful to encode the grasp pose of a gripper as the coordinate
#     system of the reference point cloud.
string filteredPointcloudSaveFrameId

# Estimative for the 6 DoF pose of the camera (base_link_frame_id),
# specified in the object_frame_id coordinate system that is
# associated with the objectModel specified above in this ROS message.
# It will only be used if the sum of the pose quaternion values is not zero.
# When the initial pose is specified in the goal, the initial pose estimation modules
# will not be used (such as PCA and feature matching).
geometry_msgs/Pose initialPose
---

###################
# Result definition

# 6 DoF pose estimated for the objectModel
geometry_msgs/PoseStamped pose

# Goal aborted -> 0
# Goal preempted -> Less than 100
# Goal succeeded -> 100
int32 percentage

# String containing one of the following:
# |-> action_server_name + ": Succeeded"
# |-> action_server_name + ": Aborted | Reason: " + message
# |-> action_server_name + ": Preempted"
string skillStatus
---

#####################
# Feedback definition
int32 percentage
string skillStatus