probabilistic_grasp_planner::ProbabilisticGraspPlanner Class Reference

#include <probabilistic_grasp_planner.h>

Public Member Functions
void	plan (const std::string &arm_name, object_manipulation_msgs::GraspableObject &graspable_object, std::vector< object_manipulation_msgs::Grasp > &grasps, bool visualize_results, bool prune_grasps)
	ProbabilisticGraspPlanner (ObjectsDatabasePtr database)
void	setMarkerPublisher (object_manipulator::GraspMarkerPublisher publisher, object_manipulator::GraspMarkerPublisher debug_publisher, object_manipulator::GraspMarkerPublisher *rank_publisher)
Private Member Functions
void	appendMetadataToTestGrasps (std::vector< object_manipulation_msgs::Grasp > &input_list, std::vector< GraspWithMetadata > &output_list, const object_manipulation_msgs::GraspableObject &graspable_object)
void	clusterGrasps (std::vector< GraspWithMetadata > &input_list, std::vector< GraspWithMetadata > &cluster_rep_list)
ObjectRepresentation	getObjectRepresentationFromCluster (const object_manipulation_msgs::GraspableObject &request_object, std::string arm_name)
ObjectRepresentation	getObjectRepresentationFromDatabaseObject (const household_objects_database_msgs::DatabaseModelPose &model_with_score, std::string arm_name)
void	populateRepresentationsList (std::vector< ObjectRepresentation > &representations, std::string arm_name, const object_manipulation_msgs::GraspableObject &request_object, const bool enable_cluster)
void	printGrasps (const std::vector< GraspWithMetadata > &grasps)
void	printRepresentations (const std::vector< ObjectRepresentation > &representations)
void	pruneGraspList (std::vector< GraspWithMetadata > &grasps, const double threshold)
void	visualizeGrasps (const std::vector< GraspWithMetadata > &grasps, object_manipulator::GraspMarkerPublisher *grasp_marker_pub, bool color_by_rank)
Private Attributes
ObjectsDatabasePtr	database_
object_manipulator::GraspMarkerPublisher *	debug_grasp_marker_publisher_
	GMP which shows all grasps and colors based on relative probability.
object_manipulator::GraspMarkerPublisher	debug_postprune_grasp_marker_publisher_
	GMP for grasps after pruning and clustering but before second pass.
object_manipulator::GraspMarkerPublisher	debug_precluster_grasp_marker_publisher_
	GMP for grasps after pruning but before clustering.
object_manipulator::GraspMarkerPublisher	debug_preprune_grasp_marker_publisher_
	GMP for grasps before pruning.
object_manipulator::GraspMarkerPublisher *	grasp_marker_publisher_
	GMP which shows the grasps returned by the planner, colored by relative probability compared to the best grasp.
ros::NodeHandle	nh_
object_manipulator::GraspMarkerPublisher *	rank_grasp_marker_publisher_
	GMP which colors grasps based on their rank in the sorted list.

Detailed Description

Definition at line 43 of file probabilistic_grasp_planner.h.

Constructor & Destructor Documentation

probabilistic_grasp_planner::ProbabilisticGraspPlanner::ProbabilisticGraspPlanner ( ObjectsDatabasePtr database ) [inline]

Creates a new ProbabilisticDatabasePlanner with a connection to the database

Definition at line 114 of file probabilistic_grasp_planner.h.

Member Function Documentation

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::appendMetadataToTestGrasps	(	std::vector< object_manipulation_msgs::Grasp > &	input_list,
		std::vector< GraspWithMetadata > &	output_list,
		const object_manipulation_msgs::GraspableObject &	graspable_object
	)			`[private]`

The "tool point" is roughly in the middle of the object enclosed by the grasp, so roughly 13cm from wrist

Definition at line 524 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::clusterGrasps	(	std::vector< GraspWithMetadata > &	input_list,
		std::vector< GraspWithMetadata > &	cluster_rep_list
	)			`[private]`

TODO: this line adds non-cluster grasps to the result list so that we can see all grasps with cluster reps highlighted. Remove this to actually filter out non-cluster rep grasps.

Definition at line 559 of file probabilistic_grasp_planner.cpp.

ObjectRepresentation probabilistic_grasp_planner::ProbabilisticGraspPlanner::getObjectRepresentationFromCluster	(	const object_manipulation_msgs::GraspableObject &	request_object,
		std::string	arm_name
	)			`[private]`

Creates an object representation based on the cluster, complete with a grasp retriever and probability computer.

Definition at line 295 of file probabilistic_grasp_planner.cpp.

ObjectRepresentation probabilistic_grasp_planner::ProbabilisticGraspPlanner::getObjectRepresentationFromDatabaseObject	(	const household_objects_database_msgs::DatabaseModelPose &	model_with_score,
		std::string	arm_name
	)			`[private]`

Creates an object representation based on a recognition result, complete with a grasp retriever and probability computer.

Definition at line 356 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::plan	(	const std::string &	arm_name,
		object_manipulation_msgs::GraspableObject &	graspable_object,
		std::vector< object_manipulation_msgs::Grasp > &	grasps,
		bool	visualize_results,
		bool	prune_grasps
	)

Given a graspable object describes the seen point cloud, plans which grasp is most likely to successfully manipulate the underlying object. At the end of the computation, the grasps vector stores a list of grasps sorted by probability of successful manipulation. Optionally will publish the grasp markers for visualization.

TODO: switch this to a std::list, we don't need random access

note that grasps should now be very small

Definition at line 80 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::populateRepresentationsList	(	std::vector< ObjectRepresentation > &	representations,
		std::string	arm_name,
		const object_manipulation_msgs::GraspableObject &	request_object,
		const bool	enable_cluster
	)			`[private]`

Given an initial GraspableObject containing representations from different sources (database fits, clusters, etc), populates a list of Representations each labeled with their type and only storing one representation of the real object (for example, only one DatabaseModelPoseWithScore or only one PointCloud)

Definition at line 276 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::printGrasps ( const std::vector< GraspWithMetadata > & grasps ) [private]

Definition at line 502 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::printRepresentations ( const std::vector< ObjectRepresentation > & representations ) [private]

Definition at line 485 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::pruneGraspList	(	std::vector< GraspWithMetadata > &	grasps,
		const double	threshold
	)			`[private]`

Definition at line 256 of file probabilistic_grasp_planner.cpp.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::setMarkerPublisher	(	object_manipulator::GraspMarkerPublisher *	publisher,
		object_manipulator::GraspMarkerPublisher *	debug_publisher,
		object_manipulator::GraspMarkerPublisher *	rank_publisher
	)			`[inline]`

Sets the grasp marker publishers that can optionally be used by this Planner. If the publishers are null, the planner won't try to publish anything.

Definition at line 137 of file probabilistic_grasp_planner.h.

void probabilistic_grasp_planner::ProbabilisticGraspPlanner::visualizeGrasps	(	const std::vector< GraspWithMetadata > &	grasps,
		object_manipulator::GraspMarkerPublisher *	grasp_marker_pub,
		bool	color_by_rank
	)			`[private]`

Publishes the grasps to rviz with specific namespaces for grasps belonging to each different model, colored with green being the best grasp and red being the worst.

Definition at line 446 of file probabilistic_grasp_planner.cpp.