Class for constraints on the visibility relationship between a sensor and a target. More...

#include <kinematic_constraint.h>

Inheritance diagram for kinematic_constraints::VisibilityConstraint:

Public Member Functions
virtual void	clear ()
	Clear the stored constraint.
bool	configure (const moveit_msgs::VisibilityConstraint &vc, const robot_state::Transforms &tf)
	Configure the constraint based on a moveit_msgs::VisibilityConstraint.
virtual ConstraintEvaluationResult	decide (const robot_state::RobotState &state, bool verbose=false) const
	Decide whether the constraint is satisfied in the indicated state.
virtual bool	enabled () const
	This function returns true if this constraint is configured and able to decide whether states do meet the constraint or not. If this function returns false it means that decide() will always return true -- there is no constraint to be checked.
virtual bool	equal (const KinematicConstraint &other, double margin) const
	Check if two constraints are the same.
void	getMarkers (const robot_state::RobotState &state, visualization_msgs::MarkerArray &markers) const
	Adds markers associated with the visibility cone, sensor and target to the visualization array.
shapes::Mesh *	getVisibilityCone (const robot_state::RobotState &state) const
	Gets a trimesh shape representing the visibility cone.
virtual void	print (std::ostream &out=std::cout) const
	Print the constraint data.
	VisibilityConstraint (const robot_model::RobotModelConstPtr &model)
	Constructor.
Protected Member Functions
bool	decideContact (const collision_detection::Contact &contact) const
	Function that gets passed into collision checking to allow some collisions.
Protected Attributes
collision_detection::CollisionRobotPtr	collision_robot_
	A copy of the collision robot maintained for collision checking the cone against robot links.
unsigned int	cone_sides_
	Storage for the cone sides.
double	max_range_angle_
	Storage for the max range angle.
double	max_view_angle_
	Storage for the max view angle.
bool	mobile_sensor_frame_
	True if the sensor is a non-fixed frame relative to the transform frame.
bool	mobile_target_frame_
	True if the target is a non-fixed frame relative to the transform frame.
EigenSTL::vector_Vector3d	points_
	A set of points along the base of the circle.
std::string	sensor_frame_id_
	The sensor frame id.
Eigen::Affine3d	sensor_pose_
	The sensor pose transformed into the transform frame.
int	sensor_view_direction_
	Storage for the sensor view direction.
std::string	target_frame_id_
	The target frame id.
Eigen::Affine3d	target_pose_
	The target pose transformed into the transform frame.
double	target_radius_
	Storage for the target radius.

Detailed Description

Class for constraints on the visibility relationship between a sensor and a target.

Visibility constraints are the most complicated kinematic constraint type. They are designed to test whether a visibility relationship is maintained in a given state. For the visibility relationship to be maintained, a sensor must have an unimpeded view of a target - this is useful, for instance, if one wants to test whether a robot can see its hand with a given sensor in a given state. The mechanism that is used to test the constraint is a combination of geometric checks, and then a collision check that tests whether a cone that connects the sensor and the target is entirely unobstructed by the robot's links.

The constraint consists of a sensor pose, a target pose, a few parameters that govern the shape of the target, and a few parameters that allow finer control over the geometry of the check. There are two general ways that this constraint can be used. The first way to leave max_view_angle and max_range_angle in the VisibilityConstraint message as 0.0. In this case, the only check that is performed is the collision check against the cone. This check assumes that the sensor can see in all directions - the orientation of the sensor has no effect on the creation of the cone. It is important to note that the sensor and the target poses must not result in collision - if they are associated with sensor frames or robot links they must be specified to be outside the robot's body or all states will violate the Visibility Constraint. For an example of a visualization of a visibility constraint (produced using the getMarkers member function), see this image:

Visibility constraint satisfied as cone volume is clear

The cone is shown in red, and the arrows show normals. This visibility constraint uses a sensor pose (the narrow end of the cone ) of the PR2's narrow_stereo_optical_frame, except shifted along the Z axis such that the pose is outside of the robot's head and doesn't collide. The cone is allowed to collide with the actual sensor associated with the header frame, just not with any other links. The target pose is a 5 cm circle offset from the l_gripper_r_finger_tip_link, again so as not to collide - again, the cone can collide with the target link, but now with any other links. In the indicated state the cone is collision free and thus satisfies the collision component of the visibility constraint. In this image, however, the right arm intersects the cone, violating the visibility constraint:

Visibility constraint violated as right arm is within the cone

Note that both the target and the sensor frame can change with the robot's state.

The collision check essentially asks whether or not the volume between the sensor pose and the target pose are clear, but that's only one aspect of visibility we may care about. The visibility constraint also allows for some geometric reasoning about the relationship between the sensor or the target - this allows information such as approximate field of view of the sensor to be factored into the constraint. To use this reasoning, the sensor_view_direction of the field should first be set - this specifies which axis of the sensor pose frame is actually pointing. The system assumes that the the sensor pose has been set up such that a single of the axes is pointing directly out of the sensor. Once this value is set correctly, one or both of the max_view_angle and the max_range_angle values can be set. Setting a positive max_view_angle will constrain the sensor along the specified axis and the target along its Z axis to be pointing at each other. Practically speaking, this ensures that the sensor has sufficient visibility to the front of the target - if the target is pointing the opposite direction, or is too steeply perpindicular to the target, then the max_view_angle part of the constraint will be violated. The getMarkers function can again help explain this - the view angle is the angular difference between the blue arrow associated with the sensor and the red arrow associated with the target.

Max view angle is evaluated at 0.0

Max view angle evaluates around pi/4

Max view angle evaluates at pi/2, the maximum

Sensor pointed at wrong side of target, will violate constraint as long as max_view_angle > 0.0

If constraining the target to be within the field of view of the sensor is the goal, then the max_range_angle can be set. The range angle is calculated as the angle between the origins of the sensor and the target frame - the orientations are unimportant. In the above images, the range angle is always 0, as the target's center is directly lined up with the blue arrow. In this image, however, the view angle is evaluated at 0.0, while the range angle is evaluated at .65.

Range angle is high, so only sensors with wide fields of view would see the target

By limiting the max_range_angle, you can constrain the target to be within the field of view of the sensor. Max_view_angle and max_range_angle can be used at once.

Definition at line 749 of file kinematic_constraint.h.

Constructor & Destructor Documentation

kinematic_constraints::VisibilityConstraint::VisibilityConstraint ( const robot_model::RobotModelConstPtr & model )

Constructor.

Parameters:

[in] model The kinematic model used for constraint evaluation

Definition at line 648 of file kinematic_constraint.cpp.

Member Function Documentation

void kinematic_constraints::VisibilityConstraint::clear ( ) [virtual]

Clear the stored constraint.

Implements kinematic_constraints::KinematicConstraint.

Definition at line 654 of file kinematic_constraint.cpp.

bool kinematic_constraints::VisibilityConstraint::configure	(	const moveit_msgs::VisibilityConstraint &	vc,
		const robot_state::Transforms &	tf
	)

Configure the constraint based on a moveit_msgs::VisibilityConstraint.

For the configure command to be successful, the target radius must be non-zero (negative values will have the absolute value taken). If cone sides are less than 3, a value of 3 will be used.

Parameters:

[in] vc moveit_msgs::VisibilityConstraint for configuration

Returns:: True if constraint can be configured from vc

Definition at line 670 of file kinematic_constraint.cpp.

kinematic_constraints::ConstraintEvaluationResult kinematic_constraints::VisibilityConstraint::decide	(	const robot_state::RobotState &	state,
		bool	verbose = `false`
	)		const `[virtual]`

Decide whether the constraint is satisfied in the indicated state.

Parameters:

[in]	state	The kinematic state used for evaluation
[in]	verbose	Whether or not to print output

Returns:

Implements kinematic_constraints::KinematicConstraint.

Definition at line 917 of file kinematic_constraint.cpp.

bool kinematic_constraints::VisibilityConstraint::decideContact ( const collision_detection::Contact & contact ) const [protected]

Function that gets passed into collision checking to allow some collisions.

The cone object is allowed to touch either the sensor or the header frame, but not anything else

Parameters:

[in] contact The contact in question

Returns:: True if the collision is allowed, otherwise false

Definition at line 997 of file kinematic_constraint.cpp.

bool kinematic_constraints::VisibilityConstraint::enabled ( ) const [virtual]

This function returns true if this constraint is configured and able to decide whether states do meet the constraint or not. If this function returns false it means that decide() will always return true -- there is no constraint to be checked.

Implements kinematic_constraints::KinematicConstraint.

Definition at line 771 of file kinematic_constraint.cpp.

bool kinematic_constraints::VisibilityConstraint::equal	(	const KinematicConstraint &	other,
		double	margin
	)		const `[virtual]`

Check if two constraints are the same.

For visibility constraints this means that:

The types are the same,
The target frame ids are the same
The sensor frame ids are the same
The cone sides number is the same
The sensor view directions are the same
The max view angles and target radii are within the margin
The sensor and target poses are within the margin, as computed by taking the norm of the difference.

Parameters:

[in]	other	The other constraint to test
[in]	margin	The margin to apply to all values associated with constraint

Returns:: True if equal, otherwise false

Implements kinematic_constraints::KinematicConstraint.

Definition at line 743 of file kinematic_constraint.cpp.

void kinematic_constraints::VisibilityConstraint::getMarkers	(	const robot_state::RobotState &	state,
		visualization_msgs::MarkerArray &	markers
	)		const

Adds markers associated with the visibility cone, sensor and target to the visualization array.

The visibility cone and two arrows - a blue array that issues from the sensor_view_direction of the sensor, and a red arrow the issues along the Z axis of the the target frame.

Parameters: