panda_autograsp.grasp_planners package¶

Submodules¶

panda_autograsp.grasp_planners.gqcnn_grasp_planner module¶

This module uses the BerkleyAutomation GQCNN grasp detector to compute a valid grasp out of RBG-D sensor data.

class panda_autograsp.grasp_planners.gqcnn_grasp_planner.GQCNNGrasp[source]¶

Bases: object

Class for storing the computed grasps. This class is a trimmed down version of the original gqcnn.grasping.policy.policy.GraspAction.

PARALLEL_JAW¶: PARALLEL_JAW Specifies parallel jaw gripper type information

SUCTION¶: SUCTION Specifies suction jaw gripper type information

class panda_autograsp.grasp_planners.gqcnn_grasp_planner.GraspPlanner(model='FC-GQCNN-4.0-PJ', sensor_type='kinectv2')[source]¶

Bases: object

Class used to compute the grasp pose out of the RGB-D data.

cfg¶: autolab_core.YamlConfig The grasp yaml configuration file.

policy_cfg¶: autolab_core.YamlConfig The policy yaml configuration file.

sensor_type¶: str The type of sensor that is used.

gqcnn_model¶: str The GQCNN model that is used.

gripper_mode¶: unicode The gripper that is used.

sensor¶: perception.Kinect2Sensor The sensor object.

grasping_policy¶: gqcnn.CrossEntropyRobustGraspingPolicy The grasp policy.

min_width¶: int The minimum allowed image width.

min_height¶: int The minimum allowed image height.

__init__(model='FC-GQCNN-4.0-PJ', sensor_type='kinectv2')[source]¶

Parameters:	model (`str`, optional) – Name of the grasp detection model, by default DEFAULT_MODEL. sensor_type (`str`, optional) – The name of the RGBD sensor, by default kinectv2.

start()[source]¶: Calls the _start_sensor method.

read_images(skip_registration=False)[source]¶

Retrieves data frames from the sensor.

Parameters:	skip_registration (`bool`, optional) – If True, the registration step is skipped, by default False.
Returns:	`tuple` of `perception.ColorImage`, `perception.DepthImage`, `perception.IrImage`, `numpy.ndarray` – The ColorImage, DepthImage, and IrImage of the current frame.
Raises:	`exceptions.RuntimeError` – If the Kinect stream is not running.

plan_grasp()[source]¶

Samples the image and computes possible grasps.

Returns:	Computed optimal grasp.
Return type:	`GQCNNGrasp`

plan_grasp_bb(bounding_box=None)[source]¶

Samples the image and computes possible grasps while taking into account the supplied bounding box.

Parameters:	bounding_box (`perception.RgbdDetection`, optional) – A bounding box specifying the object, by default None
Returns:	Computed optimal grasp.
Return type:	`GQCNNGrasp`

plan_grasp_segmask(segmask)[source]¶

Samples the image and computes possible grasps while taking into account the supplied segmask.

Parameters:	segmask (`perception.BinaryImage`) – Binary segmask of detected object
Returns:	Computed optimal grasp.
Return type:	`GQCNNGrasp`

execute_policy(rgbd_image_state, grasping_policy, pose_frame)[source]¶

Executes a grasping policy on an RgbdImageState.

Parameters:	rgbd_image_state (`gqcnn.RgbdImageState`) – The `gqcnn.RgbdImageState` that encapsulates the depth and color image along with camera intrinsics. grasping_policy (`gqcnn.grasping.policy.policy.GraspingPolicy`) – Grasping policy to use. pose_frame (`str`) – Frame of reference to publish pose in.

panda_autograsp.grasp_planners.gqcnn_grasp_planner_ros module¶

This is the ROS version of the gqcnn_grap_planner module. This module uses the BerkleyAutomation GQCNN grasp detector to compute a valid grasp out of RBG-D sensor data.

class panda_autograsp.grasp_planners.gqcnn_grasp_planner_ros.GraspPlannerROS(cfg, cv_bridge, grasping_policy, grasp_pose_publisher)[source]¶

Bases: object

Class used to compute the grasp pose out of the RGB-D data.

cfg¶

The grasp yaml configuration file.

Type:	`YamlConfig`

grasping_policy¶

The grasp policy.

Type:	`CrossEntropyRobustGraspingPolicy`

min_width¶

The minimum allowed image width.

Type:	`int`

min_height¶

The minimum allowed image height.

Type:	`int`

__init__(cfg, cv_bridge, grasping_policy, grasp_pose_publisher)[source]¶

Parameters:	cfg (dict) – Dictionary of configuration parameters. cv_bridge (`CvBridge`) – ROS CvBridge. grasping_policy (`GraspingPolicy`) – Grasping policy to use. grasp_pose_publisher (`Publisher`) –

read_images(req)[source]¶

Retrieves the input images from a ROS service request.

Parameters:	req (`ROS ServiceRequest`) – ROS ServiceRequest for grasp planner service.

plan_grasp(req)[source]¶

Grasp planner request handler.

Parameters:	req (`ROS ServiceRequest`) – ROS ServiceRequest for grasp planner service.

plan_grasp_bb(req)[source]¶

Grasp planner request handler.

Parameters:	req (`ROS ServiceRequest`) – ROS ServiceRequest for grasp planner service.

plan_grasp_segmask(req)[source]¶

Grasp planner request handler.

Parameters:	req (`ROS ServiceRequest`) – ROS ServiceRequest for grasp planner service.

execute_policy(rgbd_image_state, grasping_policy, grasp_pose_publisher, pose_frame)[source]¶

Executes a grasping policy on an RgbdImageState.

Parameters:

rgbd_image_state (RgbdImageState) – RgbdImageState from BerkeleyAutomation/perception to encapsulate depth and color image along with camera intrinsics.
grasping_policy (GraspingPolicy) – Grasping policy to use.
grasp_pose_publisher (Publisher) – ROS publisher to publish pose of planned grasp for visualization.
pose_frame (str) – Frame of reference to publish pose in.

Module contents¶

Module containing a number of autonomous grasping solutions.

`gqcnn_grasp_planner`	This module uses the BerkleyAutomation GQCNN grasp detector to compute a valid grasp out of RBG-D sensor data.
`gqcnn_grasp_planner_ros`	This is the ROS version of the `gqcnn_grap_planner` module.